Traveler Response to Transportation System Changes Handbook, Third Edition: Chapter 16, Pedestrian and Bicycle Facilities (2012)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

effects and anticipates them at the outset within the impact analysis research and model design. A number of the newer studies listed in Tables 16-82 and 16-83 illustrate movement in this direction, albeit mostly in an exploratory context. An additional example of controlling for attitudes and self-selection, beyond those in Tables 16-82 and 16-83, uses exercise instead of travel as the metric being investigated. It is covered in the “Public Health Issues and Relationships” subsection of the “Related Information and Impacts” sec- tion (see Handy, Cao, and Mokhtarian in Table 16-123 and accompanying discussion). In the final model developed in that research, attitudes were found significant and were included. However, none of the neighborhood preferences—set forth as indicators of likely self-selection—proved to be. Thus this particular investigation found self-selection not a significant factor at all, although attitudes toward physical activity were (Handy, Cao, and Mokhtarian, 2007). The larger body of relevant research does tend to show, however, that individual preferences are a factor deserving serious attention when seeking to encourage walking, cycling, or other active travel for either transportation and environmental or public health purposes. It would also seem that the phenomenon of neighborhood-preference matches and mismatches needs to be taken into account in context with possible undersupply of alternative development that offers compact, mixed-land-use, active-transportation-accessible, pedestrian/bicycle friendly neighborhoods. Such undersupply may nullify or reverse the attenuation of alternative-development land-use effects estimated in much of the self-selection research. This would render alternative development even more beneficial than state-of-the-art modeling incorporating self-selection effects might sug- gest. In other words, undersupply of alternative development may be creating pent-up demand leading to such housing being snapped up by those particular home-seekers most anxious to accommodate, to the fullest extent, pre-existing preferences for active transportation and mini- mization of VMT. If so, self-selection may be actually adding to public benefits where the supply of alternative development is failing to meet the demand. RELATED INFORMATION AND IMPACTS The first three subsections of the following collection of related NMT information cover the amounts and characteristics of pedestrian and bicycle trips at national, state, regional, and facility levels. The next two subsections examine examples of travel mode shifts with opening of new shared use NMT or bike lane facilities, and the amount of time required for usage of new facilities to stabilize or mature and thus become established. Overview NMT safety information and com- parisons are then provided. They are followed by a subsection on public health issues, impacts, and relationships, concluding with an “Adult and Child Public Health Relationships Summary” of NMT facility improvement and program effects. This health impacts summary parallels the “Traveler Response Summary” offered as part of the chapter’s introductory “Overview and Summary.” The final two “Related Information and Impacts” subsections address traffic, energy, environmental, economic, and equity impacts. Extent of Walking and Bicycling Data on current non-motorized transportation (NMT) trip making are instructive both as context and in their own right. The context provided is particularly helpful for scaling the impacts of traveler and recreational responses to NMT facilities and programs in terms of their relative impact on the uni- verse of travel or physical activity. 16-297

An important first step, however, is to have a clear understanding of NMT trip accounting peculiar- ities involving mode share definitions and coverage. Many standard regional and national house- hold transportation data sources identify and count NMT travel only if it is the “prime” or primary mode, in other words, only if an NMT mode is used exclusively for the entire trip from an origin to a separate destination. Walking or bicycling for access to or egress from other modes may or may not be picked up in any particular household travel survey. Even if obtained, the NMT identification often becomes lost in the typical trip accounting process, showing up only in special tabulations. Additional important background on these data issues is found in the “Analytical Considerations” discussion in this chapter’s “Overview and Summary” section (See “Analytical Considerations”— “National and Regional Non-Motorized Transportation (NMT) Data”—“Modal Definitions for Multi-Modal Trips”). Definitions, with examples, of prime-mode share, sub-mode share, and mode of access share are provided in the introduction to the “Pedestrian/Bicycle Linkages with Transit” subsection of the “Response by Type of NMT Strategy” section. The review that follows distinguishes between NMT as the primary or exclusive mode (“prime mode”) and NMT as a feeder and distribution mode for public transit (“mode of access”). The almost complete lack of regional and national data on NMT travel for access to modes other than public transit results in global “feeder and distribution mode” data being restricted to transit access (Agrawal and Schimek, 2007). While walking to and from private vehicles is more prevalent, such walk trips tend to be quite short. They are of case-specific interest for major parking facilities and instances where safety issues are involved, and in general for urban and suburban central business districts (CBDs), business/commer- cial strips with on-street parking (Schneider, 2011), and campuses with remote or peripheral parking. Out of a total of 48.6 billion primary mode and transit-linked walk trips per year in the United States in 2009, 16 percent represent walking in connection with transit use. Relative to 4.1 billion U.S. primary mode bicycle trips, not including transit-linked bike trips (Kuzmyak et al., 2011), a rough estimate developed below under “Extent of Bicycling” suggests that perhaps another 10 per- cent or so may represent bicycling in connection with transit use. Eight out of ten (78.7 percent) of adult respondents to the 2002 National Survey of Pedestrian and Bicyclist Attitudes and Behaviors reported walking, running, or jogging outdoors at least once for no less than 5 minutes during the last 30 summertime days. This percentage represents approximately 164 million U.S. pedestrians age 16 years or older. The bicycling equivalent was one out of four (27.3 per- cent), representing approximately 57 million adults who rode a bicycle (NHTSA and BTS, 2002). These statistics are perhaps the more notable for the adults who essentially did not walk or bicycle at all. Extent of Walking The U.S. Census Bureau’s American Community Survey (ACS) was deemed by researchers for the peri- odically issued Bicycling and Walking Benchmarking Report to be, within its limitations, the prefer- able source of large-area NMT travel data for the United States. This determination came because the survey is taken throughout the year, on a continuing basis, and has a sample size that lends itself to city-specific analysis. Its crucial limitations are that it covers only trips to and from work (commute trips), by persons 16 years of age and older, focuses only on the usual travel mode, and may be pre- sumed to subsume most NMT trips linked to motorized modes within motorized-prime-mode cate- gories such as auto or transit (Thunderhead Alliance, 2007, McGuckin and Srinivasan, 2005). Each of these limitations is disadvantageous, to some degree, when attempting to examine overall NMT use. The 2007 ACS found that 2.8 percent of trips to and from work were made exclusively by walking (Alliance for Biking & Walking, 2010). This may be compared with the percentage of trips for all 16-298

travel purposes made exclusively by walking that have been derived from the U.S. Department of Transportation’s National Household Travel Survey (NHTS).61 These walk-only mode share per- centages, inclusive of weekend travel, were 8.7 percent in 2001 and 10.1 percent in 2009 (Alliance for Biking & Walking, 2010, Kuzmyak et al., 2011). The 2007 ACS-based 2.8 percent finding for the nationwide commute trip walk-only share becomes 4.8 percent if applied only to the 51 largest cities. Less dramatically, the 2001 nationwide walk percentage of 8.7 percent for all persons and travel purposes becomes 11.0 percent if the same 2001 NHTS-based calculation is restricted to U.S. metropolitan areas in which the largest cities are located (Alliance for Biking & Walking, 2010). Year 2007 walk shares on a statewide basis, for travel to-and-from work, range from highs of 8.4 per- cent in Alaska and 6.3 percent in New York State down to 1.3 percent in Alabama. The 11 states with walk commute shares of 4 percent or greater are all in the north except for Hawaii. Perhaps surpris- ingly, none contain large metropolitan areas except for New York City and Honolulu. The 10 states with the lowest walk commute shares are all in the southeast (if broadly defined to include Oklahoma and Texas) and together constitute the U.S. states with walk commute mode shares below 2 percent. Walk-to-work shares for the three cities at the top, Boston, Washington, DC, and New York City, range from 10.3 to 13.3 percent. Oddly, Boston has the highest and New York has the lowest of these shares (Alliance for Biking & Walking, 2010), undoubtedly an artifact of Boston’s tightly drawn municipal boundary and the inclusion of all of New York City’s boroughs within its city limits. The prevalence of walking is higher the shorter the trip. In the 2001 NHTS adults (18 or more years of age) making trips of 1 mile or less were found to have a 27 percent walk mode share, over 3 times the 8.7 percent walk mode share for all persons, purposes, and trip lengths. The walk mode share for children (5 to 15 years of age) making school trips of 1 mile or less was 36 percent (Committee on Physical Activity, Health, Transportation, and Land Use, 2005). For related information on walk trip length distributions, and associated characteristics and influences, see “Characteristics of Walking and Bicycling Overall”—“Trip Distance and Duration” below, and also “Trip Factors”— “Walk Trip Distance, Time, and Route Characteristics”—“Walk Trip Speeds and Lengths” in the “Underlying Traveler Response Factors” section. The NHTS is a particularly useful national source, not only because it covers travel for all purposes by persons of all ages, but also because it has data on both walk-only trips and transit-access walk trips that are at least partially internally consistent. NHTS surveys have collected their information on trips primarily from trip diaries covering all modes of travel (Agrawal and Schimek, 2007, Liss et al., 2003). The 2001 survey diaries covered a sample of 64,000 households62 and the 2009 survey diaries covered a sample of 150,000 households (Kuzmyak et al., 2011). Table 16-85 provides an overall compilation of walk trip totals, proportions, distance, and duration from the 2001 and 2009 surveys. 16-299 61 Throughout discussions of NHTS results, 2009 data are used when the particular calculations and assess- ments were available in consistent or preferred format without need for original research, and 2001 infor- mation is used where they were not, or for comparison. Over this period, walk-only mode shares grew 13 percent, while bicycle shares grew somewhat less. In many instances it is not yet known to what extent relative relationships may have shifted. In examples examined (such as proportion of walk trips involving transit access/egress, which rounds to 16 percent in both 2001 and 2009) shifts appear to be minor. 62 Almost 63 percent of the 2001 NHTS sample was composed of “add-on” surveys arranged to provide larger sample sizes for nine regional study areas (Alliance for Biking & Walking, 2010). Differential expansion fac- tors allow representation of national data (Liss et al., 2003), but the underlying national sample was smaller than implied by the total.

Three survey and analysis protocols affecting trip data in Table 16-85 are important to understand, especially given that the protocols differ from historic metropolitan trip-based survey and analy- sis procedures. For walking in connection with a trip via transit, the walk access to transit and the walk egress from transit are conflated into a single one-way walk trip (the walk component of a “transit trip”). For walking from and back to home for recreation or exercise (a “circular” trip), the activity is included and is broken into two trips, one out to the farthest point from home and one back (Agrawal and Schimek, 2007, Kuzmyak et al., 2011). Lastly, the trip data cover all 7 days of the week, weekdays and weekend days (McGuckin and Srinivasan, 2005). The 42.3 billion U.S. annual walk trips identified for 2001 in Table 16-85 represent 153 walk trips per person per year, or 10.4 percent of all person trips. Of these, 35.4 billion were walk-only and 6.9 billion were walks associated with transit use. Mean and median distances of walk-only trips were 0.62 and 0.25 miles, respectively, while corresponding mean and median travel times were 10.0 and 16.4 minutes, respectively (Agrawal and Schimek, 2007). The large differences between the means and medians indicate trip length distributions skewed toward short trips but including significant numbers of fairly long trips. Between the 2001 and 2009 NHTS surveys, the absolute number of U.S. annual walk trips increased by 15 percent to 48.6 billion. Walk trips per person increased by 12 percent to 172, and the walk trip share of all trips increased by 13 percent to 11.8 percent. Only the share of walk to/from tran- sit trips relative to all trips stayed the same, at 1.7 percent (Kuzmyak et al., 2011). This is the first increase in overall U.S. walking activity in over 30 years demonstrated with comparisons thought to be sound (see Table 16-87 including Note A). It is quite likely the first increase, excepting possi- 16-300 Walk Trip Parameter Walk Only Walk to/from Transit Total Walk Trips 2001 2009 2001 2009 2001 2009 Trips per Year (Billions) 35.4 B 41.0 B 6.9 B 7.6 B a 42.3 B 48.6 B Trips per Person per Year 128 145 25 27 153 172 Share of All Trips 8.7% 10.1% 1.7% 1.7% 10.4% 11.8% Mean Distance 0.62 mi. 0.70 mi. n/a n/a n/a n/a Median Distance 0.25 mi. n/a n/a n/a n/a n/a Mean Travel Time 16.4 min. 14.9 min. 13.8 min. b n/a 16.0 min. c n/a Median Travel Time 10.0 min. n/a n/a n/a n/a n/a Notes: a Calculations of 2009 walk to/from transit trips include only those transit trips for which walking was the travel mode for both access and egress (Kuzmyak et al., 2011). The walk to and the walk from transit service together register as only 1 walk trip in this tabulation. b Revised calculation supplied by the corresponding author of the paper. Computed as the sum of averages of 6.34 minutes spent walking to transit and 7.44 minutes spent walking from transit to the final destination (Schimek, 2008). Since some transit access/egress does not utilize the walk access mode, the access plus egress mean would be slightly lower if computed on the basis of the average transit trip, but probably more or less the same if computed in the manner of the 2009 travel times (see Note A). c Weighted average calculation by Handbook authors. Sources: Derivation from 2001 NHTS by Agrawal and Schimek (2007), modified per Schimek (2008), and from the 2009 NHTS by Kuzmyak et al. (2011). Table 16-85 Number, Proportion, Distance, and Duration of U.S. Walk Trips in 2001, 2009

ble short-term 1970s gas-crisis responses, since the gasoline rationing of World War II. Other 2001 versus 2009 comparisons are displayed in Table 16-85. Details on distance and duration distribu- tions are provided below in the “Characteristics of Walking and Cycling Overall” subsection. There is, however, one category of walk trips known to have continued to diminish in terms of mode share between 2001 and 2009. This category is that of schoolchildren, age 5 to 18, traveling to and from school. (Data are presented in Table 16-91, with accompanying discussion, at the end of the upcoming “Extent of Bicycling” presentation.) All purposes of travel are represented in the combined weekday and weekend-day statistics of Table 16-85. If walk-only trips are divided into utilitarian trips and recreation/exercise trips, using 2001 data for completeness, the latter (mean trip length 1.0 mile) are found to average twice as long as utilitarian trips (mean trip length 0.5 miles). Recreation/exercise trips account for roughly 1/4 of all trips, but given their longer length, equate to about 1/2 the U.S. national distance walked. Mean travel times of people walking to and from transit in connection with an average single tran- sit trip were 13.8 minutes, only 16 percent less than the mean for walk-only trips (Agrawal and Schimek, 2007, Schimek, 2008, McGuckin and Srinivasan, 2005).63 The NHTS has questions about number of walk and bicycle trips per week that are separate from the survey day trip inquiries (Clifton and Krizek, 2004). The results for walking are presented in Table 16-86. They show that the majority of walk trips by U.S. residents are actually made by roughly 1/4 of the population. Of the people queried in the 2001 NHTS, 84 percent reported no walk trips in their daily routine. Table 16-86 indicates that 35 percent reported no walk trips at all in the preceding week in 2001, dropping to 32 percent in 2009. The median number of walk trips shifted from two per week to three in round numbers. The mean based on survey-day responses increased from 2.9 per week in 2001 to 3.3 per week in 2009. Some 91 percent averaged only one walk trip or less per day during the week in 2001, dropping to 86 percent in 2009 (Weinstein and Schimek, 2005, Kuzmyak et al., 2011). 16-301 63 CDC research (Besser and Dannenberg, 2005) can be used to derive, on the basis of two transit trips per tran- sit user per day (Agrawal and Schimek, 2007), a mean walk time per transit trip of 12.2 minutes. See Note B to Table 16-85 for a likely explanation of much or all of the difference.

Inspection of Table 16-86 shows that the percentages of people reporting zero trips up through four trips per week all decreased by roughly 10 percent between 2001 and 2009. The percentage report- ing six trips per week stayed the same, while the percentages reporting five, seven, and eight or more trips a week all increased. This provides substantiation, from a separate NHTS trip-recall line of questioning, that walking did increase between 2001 and 2009 in the United States. Table 16-87 illustrates the steadily downward U.S. trend of walk-only trips, as a percentage of total trips, in the latter part of the 20th Century. It also illustrates, along with both Tables 16-85 and 16-86, the modest reversal of this trend in the first decade of the 21st Century. Table 16-87 covers walk mode shares over time for all trip purposes combined and for the work commute alone. Bicycle-only shares are included and, because most transit trips involve substantive walking (and some involve bicy- cling), transit shares are also listed. These data are assembled from the Nationwide Personal Transportation Study (NPTS), the predecessor survey to the NHTS, along with the 2001 NHTS itself and also the U.S. Census decennial surveys and American Community Survey for 2009. 16-302 2001 NHTS 2009 NHTS Walk Trips per Week Percent of Persons Cumulative Percentage Percent of Persons Cumulative Percentage 0 35% 35% 32% 32% 1 7 41 6 38 2 11 52 10 48 3 11 63 10 58 4 7 70 6 64 5 7 77 8 72 6 3 80 3 75 7 11 91 11 86 8+ 9 100 13 100 Source: Derived from 2001 NHTS by Agrawal and Schimek (2007) and from 2009 NHTS by Kuzmyak et al. (2011). Table 16-86 Number of Walk Trips Reported for the Preceding Week, 2001 and 2009

There is a methodological enhancement, and thus inconsistency, between the NPTS and NHTS sur- veys that must be taken into account. Survey protocol changes were made for the 2001 NHTS sur- vey that were designed to capture previously unreported walk trips. It is felt that these changes, and not a shift in trends, were the primary cause of the 2001 increase in reported walk mode share (Hu and Reuscher, 2004). The work-purpose data from the U.S. Census probably provides a bet- ter indicator of 1990 to 2000 trends, indicating continued decline in walking up to that point. The 2009 ACS results, methodologically similar to the U.S. Decennial Census with a margin of error of ±0.1 percent, indicate a 2000 to 2009 increase in bicycling to work, stability in walking to work, and an increase in transit use with its associated walking (U.S. Census Bureau, 2011). These work commute outcomes are not in conflict with the NHTS all-trip-purposes results for the 8 years, but suggest that the increase observed in overall walk-only trips may mainly be the result of more walking for non-work utilitarian purposes and/or recreation and exercise. A metropolitan area perspective is provided by the National Capital region. Weekday trip mode shares increased in the Washington, DC, metropolitan area as a whole—between 1994 and 2007/08—by 1.6 percentage points for walk-only (from 7.8 to 9.3 percent), 0.2 percentage points for bicycle (from 0.5 to 0.7 percent), and 0.7 percentage points for transit (from 5.6 to 6.3 percent) (Griffiths, 2010). Metropolitan area research from the other coast provides a start at answering the question of how much walking occurs as a result of parked-car egress and access, along with other short walks 16-303 Travel Mode 1969/70 1977 1980 1983 1990 1995 2000/01 2009 All trip purposes Bicycle n/a 0.7% — 0.8% 0.7% 0.9% 0.9% 1.0% Walk-only n/a 9.3 — 8.5 7.2 5.4 8.6 a 10.1 b Transit c 3.2% 2.6 — 2.2 2.0 1.8 1.6 1.9 Work purpose trips Bicycle n/a — 0.5% — 0.4% — 0.4% 0.6% Walk-only 7.4% — 5.6 — 3.9 — 2.9 2.9 Transit 8.9 — 6.4 — 5.3 — 4.7 5.0 Notes: All-trip-purposes shares are from the NPTS/NHTS. Work-purpose shares are from the U.S. Decennial Census except for 2009, which are from the American Community Survey (ACS). In multi-year columns, the odd-numbered year pertains to NPTS/NHTS all-purpose shares and the even-numbered year pertains to U.S. Census work-purpose shares. a The 1995-2001 NPTS/NHTS increase in all-trip-purposes walk-only shares relates primarily to survey methodology changes, designed to capture previously unreported walk trips (Hu and Reuscher, 2004). The work-purpose data from the U.S. Census provide a better indicator of trends in the 1990-2000 decade. b Taken from Table 2-1 of Kuzmyak et al. (2011). c Transit mode shares are included as an approximate indicator for the substantive walking that occurs in connection with most transit travel. The 1969 NPTS-derived transit share is adjusted for comparability, compensating for lack of NMT trips in the original survey. Sources: NPTS results for 1969, 1977, 1983, 1990, and 1995; NHTS results for 2001; U.S. Census results for 1970, 1980, 1990, and 2000 as reported in Pucher and Renne (2003); NHTS results for 2009 as reported in Kuzmyak et al. (2011); and U.S. Census Bureau ACS 2009 (2011), with 1969 transit share adjustment by the Handbook authors. Table 16-87 NMT Mode Shares for All Trip Purposes and Work Purpose Trips, 1969–2001

within trip destination areas that may or may not be picked up in surveys like the NHTS. The research obtained completed face-to-face interviews with about 1,000 patrons of 20 retail pharmacy stores located throughout the midsection of the San Francisco Bay Area. The stores were in San Francisco proper, the San Mateo County portion of Silicon Valley, Contra Costa County, Berkeley, Oakland, and suburban Alameda County. Complete details were obtained on all trips within the home-to-home tour that included the pharmacy. No time limit was placed in the tour definition on intermediate stops, such that—for example—a tour including a pharmacy stop on the way home from work would include all travel between home and work, from and to the work location, and between work and home. The primary modes used for the whole tour averaged, among tours, 21.3 percent walk, 2.2 percent bike, 9.9 percent transit, and 66.6 percent auto (Schneider, 2011). For comparison, the 2009 NHTS obtained NMT shares for the San Francisco-Oakland-San Jose Metropolitan Statistical Area (MSA) of 14.3 percent walk, 1.9 percent bicycle, and 3.0 percent walk to/from transit (Kuzmyak et al., 2011). This is a reasonable degree of consistency considering the relatively auto-oriented nature of MSA areas not covered by the pharmacy surveys, along with other basic differences. In contrast to the 21.3 percent walk primary-mode share for tours intercepted at pharmacies, 51.9 percent of these same tours involved walking between stops or along a street at some time during the tour. Walking under one roof between stops was not counted, and neither was walk- ing between a parked car and the adjacent dwelling or destination building. The primary mode used on trips within shopping districts and corridors was, respectively, 65.2 and 72.8 percent walk. Total distance by mode for the entire tour averaged 4.5 percent walk. Total travel distance inter- nal to shopping districts and corridors was, respectively, 54.6 and 67.5 percent walk (Schneider, 2011). These findings, while not differentiating in the aggregate between transit access, auto access, and purely walking trips, begin to give measure to the larger role of walking relative to the nar- row perspective imposed by primary-mode analytical protocols. Extent of Bicycling Bicycling, as observed in the United States of the late 20th and early 21st Centuries, “accounts for a minute percentage of Americans’ overall trips” (MacLachlan and Badgett, 1995) and, for adults at least, has been characterized as a “fringe mode” and “rare behavior” (Krizek and Johnson, 2006). The 2009 ACS found that 0.6 percent of trips to and from work were made exclusively by cycling (U.S. Census Bureau, 2011). The comparable figure for all purposes of weekday and weekend travel, by persons of any age, is 1.0 percent for trips made exclusively by cycling as derived from the 2009 NHTS. Other bicycle trip statistics for 2009 corresponding to the walk trip statistics of Table 16-85 are 4.1 billion bike trips per year, 14.5 bike trips per person per year, 2.3 miles mean bicycle trip distance, and 19.4 minutes mean travel time (Kuzmyak at al., 2011). Trend data on bicycling assembled from the NPTS and NHTS do not exhibit the methodological inconsistencies associated with the 2001 changes to walk trip surveying methodology. Together, the NHTS and NPTS indicate that U.S. nationwide bicycling mode shares have been relatively sta- ble over one-third of a century, with perhaps a very slight increase from about 0.7 percent in 1977 to 1.0 percent in 2009. Bicycling mode shares over time for all trip purposes and for the work com- mute were included above in Table 16-87. Despite the consistency over time of the NPTS/NHTS national data for bicycling, there is cause to be cautious in working with these small percentages, especially when comparing across surveys with different types of methods. For example, a study commissioned by Los Angeles County found 16-304

2.4 percent of all trips in that county to be by bicycle, while the 2001 NHTS-based estimate was 1.0 percent. Three 21st Century Los Angeles County commute mode share estimates are virtually identical, however: 0.61 percent from the 2000 U.S. Census and 0.59 percent from the 2005 ACS (Thunderhead Alliance, 2007), remaining essentially the same in the 2007 ACS. A second example is even more notable. The 2001 NHTS-based bicycle-share estimate for all city and county of San Francisco trips was 0.93 percent. A city-commissioned study with “a larger sample size and more robust methods” found an all-trip-purpose bicycling mode share of 6 percent. The 2000 Census identified a 1.98 percent work-trip mode share while the 2007 ACS found a 2.52 percent share (Alliance for Biking & Walking, 2010). There is apparently less difference between all-purpose bicycle usage rates for the nation as a whole compared to large metropolitan areas than is the case for walking. The nationwide 2001 cycling percentage of 0.90 percent for all travel purposes shifts only to 0.94 percent if the NHTS- based calculation is limited to U.S. metropolitan areas in which the 50 largest cities are located. However, the 2007 ACS-based 0.5 percent commute trip cycling share nationwide becomes 0.8 per- cent when restricted to the 51 largest cities. Cycling shares on a statewide basis for 2007 travel to and from work range from highs of 1.9 percent in Oregon and 1.4 percent in Montana to 0.1 per- cent in Alabama, Arkansas, and Tennessee. The 12 states with bicycle commute shares of 0.7 per- cent or greater are all in the west, including all continental states bordering on the Pacific and other individual states as far east as Wisconsin. The 12 states with the lowest bicycle commute shares, all 0.2 percent or less, are in the southeast if extended to include Texas and West Virginia. One additional state, Rhode Island, is in the 0.2 percent category (Alliance for Biking & Walking, 2010). In the 2001 NHTS, adults (18 or more years of age) making trips of 5 miles or less had an 0.6 per- cent bicycle mode share. The bicycle mode share for children (5 to 15 years of age) making school trips of 2 miles or less was 1.5 percent (Committee on Physical Activity, Health, Transportation, and Land Use, 2005). The fact that the adult 5-mile-or-less share is below the bicycle share for all persons, purposes, and trip lengths is almost certainly the result of inclusion of children in the global statistic and/or analytical issues such as the small bicycle trip sample size. It is amply demonstrated on the basis of trip length distributions that bicycle use must fall off faster with increasing trip length than is the case for trip making overall. Information on bicycle trip lengths is found below under “Characteristics of Walking and Bicycling Overall”—“Trip Distance and Duration” and also in the “Underlying Traveler Response Factors” section (see “Trip Factors”— “Bicycle Trip Distance, Time, and Route Characteristics”—“Bicycle Trip Speeds and Lengths”). Although NHTS data are theoretically available to make an adjustment to include bicycling as a mode of access to public transit, the small sample size would be an issue. The national data pre- sented here do not include as bicycle trips those bike trips made in conjunction with transit use. Tables 16-88 and 16-89 together provide summarized 1996–98 survey data covering bicycling shares for access to and egress from public transit routes in 14 U.S. cities, shown along with com- parable information for walk access/egress and other access/egress modes. The overall bicycle access/egress shares, averaging from 0.6 to 1.4 percent depending on the end of the trip in ques- tion (McCollom Management Consulting, Inc., 1999), suggest that inclusion of transit access/ egress cycling would increase the estimate of 0.9 percent of all 2001 U.S. trips being by bicycle to a total of 1.0 percent at most (see Table 16-87 above). This crude circa 2000 estimate, made starting with the 4.7 percent transit share presented in Table 16-87 for 2001 and applying bicycle access shares from Tables 16-88 and 16-89, can also be taken to suggest that bicycle trips taken in conjunc- tion with transit amounted to on the order of 1/10 of 1 percent of all U.S. trip making. An update to 2009 would increase these percentages by a tiny fraction, but the result would still be a small proportion when viewed globally. 16-305

16-306 Access Mode All 14 Systems Systems Classified by Size Multi-Modal Systems Small Medium Large Bus Rail Bicycle 0.6% 0.3% 0.8% 0.3% 0.3% 0.8% Walk 67.0 84.8 61.9 62.6 74.0 47.7 Auto Driver 9.6 2.0 10.7 13.5 4.0 20.5 Auto Passenger 3.4 1.6 3.9 4.2 2.5 5.5 Bus/Train 19.3 11.3 22.7 19.5 19.1 25.2 Notes: Small-sized systems (0 to 500,000 service area 1997 population) include Grand Rapids, MI, Kenosha, WI, and Lincoln, NB. Medium-sized systems (500,000 to 1,500,000 service area 1997 population) include Austin, TX, Buffalo, NY, Portland, OR, and Sacramento, CA. Large-sized systems (1,500,000 or more service area 1997 population) include Chicago, IL (Chicago Transit Authority bus and HRT only), and Pittsburgh, PA. Multi-modal (bus and urban rail) systems include Buffalo, Chicago, Pittsburgh, Portland, and Sacramento. Results shown are for the bus or rail components as indicated. Auto passenger includes passenger drop-off and passengers in cars parked. Bus/Train represents surveyed riders who started on another transit route and transferred to the route being surveyed. Source: McCollom Management Consulting, Inc. (1999). Table 16-88 Access Mode Share Percentages from Home to Bus and Rail Routes (and Return) for 14 U.S. Systems, 1996–1998 Egress/Access Mode All 14 Systems Systems Classified by Size Multi-Modal Systems Small Medium Large Bus Rail Bicycle 1.4% 3.5% 1.1% 0.4% 0.5% 1.0% Walk 73.0 78.0 70.4 73.9 72.3 69.9 Auto Driver 2.6 1.1 3.3 2.5 1.5 4.9 Auto Passenger 2.7 2.9 2.7 2.5 2.6 2.6 Bus/Train 20.3 14.5 22.5 20.8 23.2 21.5 Notes: Notes same as Table 16-88 except Bus/Train represents surveyed riders who completed their trip on another transit route and transferred from the route being surveyed. Source: McCollom Management Consulting, Inc. (1999). Table 16-89 Non-Home Egress/Access Mode Percentages for the Bus and Rail Routes of 14 U.S. Systems, 1996–1998

Walking in connection with transit service occurs in greater magnitudes. Tables 16-88 and 16-89 indicate that—across 14 systems—67 percent of transit route riders starting from (or returning to) home reached (or left) the transit route on which they were surveyed by walking. At non-home ends of trips, 73 percent of transit egress or access was by walking (McCollom Management Consulting, Inc., 1999). If one subtracts out transfer passengers who reported their access or egress mode as being bus or train, and normalizes the remaining access/egress percentages, it can be seen that on a system basis 83 percent of riders of the 14 systems starting from home walked to the tran- sit system, and 92 percent of system riders leaving or accessing the system away from the home used the walk mode. (Corresponding figures for bicycling are 0.7 percent and 1.8 percent.) This walk access/egress information undergirds the findings in the preceding “Extent of Walking” dis- cussion concerning the importance of transit riding to understanding of the total amount of walk- ing in the United States. The overall frequency of bicycling among the U.S. public has been explored in a number of sur- veys. Those conducted over the past 15 years show that in general, as the defined time frame increases, so does the number of people who report cycling during that time period. There is exten- sive variation across geographic areas. The 2001 NHTS found a range of 0.2 to 2.4 percent of per- sons bicycling during their survey day across the various Metropolitan Statistical Areas covered, with a range of 4.5 to 12.7 percent cycling sometime during a week. Rodale Press surveys in 1992 and 1995 found 16.6 to 21.2 percent to have cycled during a month, compared to 27 percent over the summer of 2002 as determined by BTS, and 37 to 46 percent over a full year as found by Rodale. Minnesota DOT in 2003 found that 1/2 the population in their state never cycled (Krizek et al., 2007). Table 16-90 categorizes 2009 NHTS respondents nationwide by the number of bicycle trips taken per week. Of all respondents, 13 percent were found to have bicycled at least once in the pre- ceding week (Kuzmyak et al., 2011). 16-307 Bike Trips per Week Percent of Persons Cumulative Percentage Bike Trips per Week Percent of Persons Cumulative Percentage 0 87% 87% 5 1% 98% 1 4% 91% 6 0.4% 98% 2 3% 93% 7 1% 99% 3 2% 95% 8+ 0.8% 100% 4 1% 97% Source: Derived from 2009 NHTS by Kuzmyak et al. (2011). Table 16-90 Number of Bicycle Trips Reported for the Preceding Week, 2009 NHTS Looking specifically at school children from age 5 to 18, bicycling to school as a percentage mode share has stayed close to the range of 1/2 to 1 percent established in 1977 and 1983, the first two survey years it was separately measured. Data in Table 16-91 demonstrate that the major shift has been in walking for school access, which declined from 22.5 percent in 1977 to 9.5 percent in 2009. Even more dramatic is the change from 1969 to 2009, which can be measured only in total walking and bicycling access to school. That percentage, for children, plummeted from over 40 percent in 1969 to 10 percent in 2009, a huge concern for public health practitioners and a significant contrib- utor to school-area congestion and automotive pollution (Moudon, Stewart, and Lin, 2010, Kuzmyak et al., 2011).

Characteristics of Walking and Cycling Overall A number of tabulations of pedestrian and bicycle trip and trip-maker characteristics were pre- sented in the “Underlying Traveler Response Factors” section, in support of examining influences on NMT choices. In addressing “Characteristics of Walking and Cycling Overall,” these tabulations will be referred to as appropriate. A summary perspective is provided here, along with additional data displays. The focus of this overall-characteristics subsection is on describing the nature of walk- ing and cycling, primarily in the United States, along with presenting related insights and informa- tion potentially useful in NMT evaluation and design. The focus of the earlier “Underlying Traveler Response Factors” section is on how the manifestations of walking and cycling, such as NMT trip generation, mode choice, route choice, and time-of-day choice, are affected by the characteristics of the environment, the trips, and the trip makers. Trip Distance and Duration Lengths of walking and bicycling trips are governed by the location of activities and the interplay of the corresponding travel desires with the locations of NMT facilities. Also having important roles are the purposes of the desired trips and the character and quality of the NMT facilities available. Trip distance and duration findings also vary according to the research design. Information derived from the NHTS conforms with many aspects of standard metropolitan transportation planning prac- tice, the bulk of the data being derived from daily trip diaries which accept “no trips” (NMT or other- wise) as a legitimate survey response. As previously indicated, 84 percent of the 2001 NHTS survey respondents reported no walk trips on their survey day. The acceptance and recordation of no NMT trips, after probing to make sure none were overlooked, means that the results should parallel what one would expect on a typical day in the United States. In contrast, surveys that ask about the most recent walk or bike trip—while they may provide needed information on infrequent trip making— overweight and thus overemphasize infrequent trips. Results may thus be skewed, as characteristics of infrequent trips may be different from trips made on a frequent, regular basis. Table 16-92, derived from the 2009 NHTS, provides a rough but presumably reliable overall picture of walk trip distance and duration distributions in the United States. The results are somewhat lumpy, as a result of self-reported survey limitations. Trip durations tend to be self-reported in round numbers, thus the disproportionate percentages of trips in the 5, 10, 15 and 30 minute categories. The walk trips in this tabulation exclude transit access trips. The mean walk-only distance calculated was 0.70 miles, but the median was 4 blocks (about 0.45 miles), indicating a skewing of the mean by a lesser number of fairly long trips. Similarly, the mean travel time was 14.9 minutes, with a median 16-308 Travel Mode 1969 1977 1983 1990 1995 2001 2009 Bicycle n/a 1.0% 0.5% 1.0% 1.1% 0.8% 0.7% Walk n/a 22.5 14.5 18.2 10.6 12.1 9.5 Total NMT 40.7% 23.5% 15.0% 19.2% 11.7% 12.9% 10.2% Notes: Includes children ages 5 to 18. Source: NPTS results for 1969, 1977, 1983, 1990, and 1995, and NHTS results for 2001, as reported in Moudon, Stewart, and Lin (2010). NHTS results for 2009 as reported in Kuzmyak et al. (2011). Table 16-91 NMT Percent Mode Shares for Child Transportation to School, 1969–2009

value of 10 minutes. Some 23 percent of walk-only trips were a mile or more in length, while 13 per- cent were 30 minutes or more in duration (Kuzmyak et al., 2011). 16-309 Walk Trip Distance Walk Trip Duration Blocks/Miles a Frequency (percent) Cumulative Frequency Time Frequency (percent) Cumulative Frequency < 5 min. 16% 16% 1 block 16% 16% 5 min. 16 32 2 blocks 17 33 6-9 min. 6 39 3 blocks 7 40 10 min. 16 54 4 blocks 15 55 11-14 min. 4 58 5-8 blocks 22 77 15 min. 16 74 1 mile 11 88 16-29 min. 13 87 1.1 to 2 miles 9 97 30 min. 6 93 > 2 miles 3 100 > 30 min. 7 100 Notes: a It is assumed that 9 blocks are equal to 1 mile. Source: Derived from 2009 NHTS by Kuzmyak et al. (2011). Table 16-92 Distribution of U.S. 2009 Walk-Only Trips by Distance and Duration A Centers for Disease Control and Prevention (CDC) analysis of the 2001 NHTS made from the public health perspective found that Americans who used public transit spent a median time of 19 minutes daily walking to and from transit. This statistic is a total for all transit access and egress during the day (Besser and Dannenberg, 2005), most frequently (but not always) four walk segments daily (Agrawal and Schimek, 2007). If one treats walk access to and walk egress from a single transit trip as one walk trip, as commonly done in analysis of NHTS data, the median transit-linked walk time becomes 9.5 minutes. Additional discussion of this CDC research is provided further on under “Public Health Issues and Relationships” (see “Baseline Walking and Bicycling Activity” and also Table 16-123). Table 16-93, the bicycle equivalent of Table 16-92, draws bicycle trip distance and duration distri- butions from the 2009 NHTS. It has the same limitations as described for Table 16-92. The mean bicycle-only distance calculated was 2.3 miles, but the median was 1 mile. As with walking, this differential indicates a skewing of the mean by a lesser number of long trips, only more so in the case of bicycling. The mean bicycle travel time was 19.4 minutes, with a median value of 15 min- utes. Of all bicycle trips, 12 percent were 30 minutes or more in duration, virtually the same as for walking, but 26 percent were 2 miles or more in length (Kuzmyak et al., 2011). The typical bicycle trip takes 30 to 50 percent more time than its walk-trip counterpart, but covers 2.2 to 3.3 times as much ground. The greater disparity between the mean and median for distance than for time sug- gests wide variation in bicycling speeds.

Reporting of the 2002 summer survey performed by NHTSA and BTS also provides comparative walk and bike trip distance data, but from a different perspective. This survey utilized a variant of the “most recent trip” inquiry methodology in that it recorded active transportation data for the day (within the last 30 days) of most recent walking or bicycling activity. It also differed from the NHTS by defining a trip from home and return “with no real destination” and no stops as a single trip, not separate trips to and from a farthest point (NHTSA and BTS, 2002). Both of these survey differences are thought to increase reported trip distances, and certainly affect means and related computations in some way. In any case, Table 16-94 presents the summertime trip length distributions obtained. This tabulation serves to again illustrate the larger geographic market potentially served by bicycling as compared to walking. It also again illustrates the preponderance of trips that are short, relative to the “most recent day” survey means of 1.2 miles for walk trips and 3.9 miles for bike trips. The “most recent day” median walking distance was slightly over 0.5 miles, while the comparable median cycling distance was somewhat under 2.0 miles (NHTSA and BTS, 2002). Comparison of the “most recent day” mean and median walk and bicycle trip distances with those from Tables 16-92 and 16-93 does indicate that the recorded trip lengths, particularly the means, are indeed longer than obtained from trip diary surveys using the NHTS protocol. 16-310 Bicycle Trip Distance Bicycle Trip Duration Blocks/Miles a Frequency (percent) Cumulative Frequency Time Frequency (percent) Cumulative Frequency < 5 min. 9% 9% 1 block 9% 9% 5 min. 13 22 2 blocks 10 19 6-9 min. 6 28 3 blocks 5 24 10 min. 16 44 4 blocks 7 31 11-14 min. 3 47 5-8 blocks 15 46 15 min. 18 65 1 mile 11 57 16-29 min. 14 78 1.1 to 2 miles 17 74 30 min. 9 88 > 2 miles 26 100 > 30 min. 12 100 Notes: a It is assumed that 9 blocks are equal to 1 mile. Source: Derived from 2009 NHTS by Kuzmyak et al. (2011). Table 16-93 Distribution of U.S. 2009 Bicycle-Only Trips by Distance and Duration

Trip Purposes Utilitarian walking and bicycling trips are, overall, usually made for the same reasons as motor- ized utilitarian trips. Common utilitarian purposes include going to work, shop, or school (and returning home), or to obtain medical/dental care, conduct personal business, eat a meal, or visit an entertainment venue. The major differences in trip purpose distributions involve recreational or exercise purposes, which are much more prevalent among NMT trips. Recreational and exer- cise trips may be for reasons of enjoyment, physical fitness, or general health. Table 16-69 of the “Underlying Traveler Response Factors” section (see “Trip Factors”—“Trip Purpose”), presents walk and bike mode shares for work and various non-work trip purposes. Such mode choice proportions, applied to overall trip-making by purpose, give absolute numbers of walk and bike trips. This allows calculation of the distribution of walk and bicycle trips among trip pur- poses. Such distributions may, however, be calculated directly from survey results. Table 16-95, derived from the 2009 NHTS, presents such an examination for 10 trip purposes. It covers work trips and six other utilitarian purpose categories, two recreation/exercise categories, and one miscellaneous “other” category, plus trips with unspecified purposes. “To home” trips from the trip diaries were allocated according to the travel purpose at the trip origin. The results given in Table 16-95 include not only walk and bicycle trip proportions by trip purpose category, but also a separate purpose dis- tribution for walk trips to/from transit stops and stations. In addition, distances and durations are provided for each NMT mode and purpose, except walk to/from transit (Kuzmyak et al., 2011). 16-311 Walk Trips Bicycle Trips Trip Distance Range Frequency (percent) Cumulative Frequency Frequency (percent) Cumulative Frequency 0.25 miles or less 26.9% 26.9% — — 0.26 to 0.5 miles 19.6 46.5 — — 0.5 to 1.0 miles 20.7 67.2 — — Subtotal, 1.0 miles or less 67.2% 67.2% 38.6% 38.6% 1.1 to 2.0 miles 18.0% 85.2% 18.5% 57.1% 2.1 to 5.0 miles — — 23.8 80.9 5.1 to 10.0 miles — — 11.8 92.7 More than 10.0 miles — — 7.3 100.0 Subtotal, more than 2.0 miles 14.8% 100.0% 42.9% 100.0% Note: See discussion in text above of methodological limitations. Source: NHTSA and BTS (2002) with elaboration by the Handbook authors. Table 16-94 Attitudinal Survey 2002 Trip Lengths On Most Recent Day Walked/Biked

Of all 2009 walk-only trips, 61 percent were in trip purpose categories primarily associated with utilitarian travel, and 37 percent were in categories primarily associated with recreation or exer- cise.64 Walk to/from transit trips had an even higher utilitarian travel proportion, at 83 percent, with over 6 times the percentage of work commute trips. Of transit access/egress trips, 30 percent were to and from work, 11 percent were to and from school (or school-related library trips or place- of-worship-related trips), 17 percent were for shopping, and 15 percent were for medical/dental or other personal business. Bicycle trips were the most oriented to recreation or exercise, with cycle trips being 50 percent utilitarian, 49 percent recreation or exercise, and 1 percent miscellaneous. 16-312 Proportions by Purpose (percent) Average Trip Length (miles) Average Travel Time (minutes) Trip Purpose Walk Only Transit Access a Bicycle Walk Only Bicycle Walk Only Bicycle Utilitarian Trips To/from work 4.5% 29.8% 10.9% 1.0 3.8 16.2 21.2 Work-related business 1.7% 3.6% 1.8% 1.1 3.3 14.0 21.7 School/religion-related 8.6% 10.9% 6.0% 0.6 1.6 14.5 15.2 Shopping, buy goods/gas 14.7% 16.6% 9.8% 0.6 1.3 12.7 14.0 Visit friends/relatives 8.7% 7.5% 13.0% 0.6 1.0 11.7 13.9 Medical/dental 0.9% 5.4% 0.2% 0.7 2.2 16.1 26.0 Other personal business b 21.5% 9.4% 8.2% 0.5 1.4 11.2 15.5 Recreation/Exercise Trips Rest, relaxation, vacation 1.9% 0.8% 2.1% 0.8 2.4 22.5 21.0 Other social/recreational c 35.4% 11.0% 47.3% 0.8 2.6 18.3 22.5 Miscellaneous Trips Other 1.4% 3.6% 0.1% 1.2 2.3 13.1 16.0 Unspecified 0.8% 1.5% 0.8% 0.8 2.7 22.0 25.7 Trips by all Purposes 100% 100% 100% 0.7 2.3 14.9 19.4 Total Trips (millions) 40,962 7,647 4,082 40,962 4,082 40,962 4,082 Notes: a Walk to/from public transit stop/station (transit access/egress). b Includes family/personal business, buy services, day care, grooming, pet care/dog walk, transport someone, wedding/funeral, attend civic meeting, social event, get meal/snacks. c Includes social/recreational; exercise (walking, jogging, etc.); and some purposes normally considered as utilitarian, such as go out for entertainment, play sports, visit public place, social event, get/eat meal/coffee/snacks. The NHTS covers all trips by persons of all ages (Liss et al., 2003) on all 7 days of the week, weekdays and weekend days (McGuckin and Srinivasan, 2005). The 7-day-a-week coverage lowers work trip percentages relative to those seen in weekday-only surveys and tabulations. Sources: Derived from 2009 NHTS by Kuzmyak et al. (2011). Table 16-95 Proportions, Distance, and Duration of Walk and Bike Trips by Trip Purpose 64 Multiple NMT trip purposes/motivations were not addressed in the NHTS. If a survey respondent had mul- tiple purposes/motivations for a particular walk or bike trip, it was implicitly up to him or her to choose which single purpose/motivation to report. (In the “Overview and Summary,” see “Analytical Considerations”— “Trip Purpose Versus Motivation” for further exploration of NMT trip purpose identification issues.)

Mean distances were not available for walk to/from transit trips. Walk-only mean distances ranged from 0.5 to 0.7 miles for school, shopping, visiting friends and relatives, health services, and other personal business categories. Recreation/exercise walk trips averaged 0.8 miles to the farthest point reached in terms of distance from the origin. Work, work-related, and miscellaneous- other walk-only trips were in the range of 1.0 to 1.2 miles (Kuzmyak et al., 2011). Tabulations from the 2001 NHTS suggest that median walk-only trip distances run about 1/2 of mean distances for shopping, errands, personal business, and recreation/exercise trips, and 1/3 or less of mean dis- tances for work and school trips (Agrawal and Schimek, 2007). Bicycle trip distances run 2 to 4 times as long as walk-only trips, with mean distances ranging from 1.0 one-way miles for visiting friends and relatives to 3.8 miles for work trips, as seen in Table 16-95. There is less difference between walk and bicycle trips when viewed from the perspective of time expended, although travelers tend to allocate somewhat more time to bicycle trips. Both walk and bicy- cle trip time duration means lie within the 11-to-15-minute range in the case of school, shopping, visit friends and relatives, and other personal business categories. Among the longer trip time averages are 16 minutes walk and 21 minutes bike for work trips, 16 minutes walk and 26 minutes bike for med- ical/dental-purpose trips, and 18 minutes walk and 22 minutes bike to the farthest point reached dur- ing recreation/exercise trips (Kuzmyak et al., 2011). User Characteristics Prevalence of walking and cycling trips, like motorized trips, is influenced not only by the type and proximity of activities and the facilities for travel, but also—and strongly so—by the socio-economic characteristics of the trip making population. Walking and bicycling rates and characteristics are influ- enced by gender, age, income, auto ownership, education, and ethnicity. They are also affected by individual caution, proficiency (especially for cycling), physical capability, and attitude. Global relationships of user characteristics to NMT trip making have been examined under “User Factors” in the “Underlying Traveler Response Factors” section, illustrated with tabulations by user category, primarily on the basis of U.S. average active transportation mode shares from the 2001 or 2009 NHTS. Mode share tabulations are not quite the same as data on absolute numbers of trips, because differential trip generation rates also affect numbers of trips. For example, lower income people and the elderly tend to make fewer trips, so a relatively higher mode share exhibited by one of these groups may be damped down in terms of actual trips in the category. Such circumstances are noted, where important, in the “Underlying Traveler Response Factors” discussions. In the “User Factors” subtopic, the primary tables and discussions of interest to a global understanding of user characteristics are the following: • Gender: Table 16-72 under “Gender,” and the accompanying development of indications that the lesser bicycling of females in the United States is balanced by more walking. • Age: Table 16-74 under “Age,” and the discussions of NMT activity decline with the onset of adulthood and then aging, aside from increased walking for recreation and exercise by seniors, and the magnitude of walking and bicycling in childhood. • Income: Tables 16-75 and 16-76 under “Income,” respectively providing both NMT mode shares by income and income distributions by mode, along with the indication that while walk- only trip activity and bicycle trip activity vary only moderately with income, walk-transit trip making is much more prevalent in low income households. 16-313

• Auto Ownership: Table 16-78 under “Automobile Ownership,” and the demonstration that active transportation use is several times higher in households without cars. • Education: Table 16-79 under “Education,” and accompanying analyses indicating that while the least educated have the highest NMT mode shares, once factors such as housing patterns (including densities and neighborhood walkability) and auto ownership are accounted for, the proclivity to walk for exercise and utilitarian purposes increases with education. In addition, the “Underlying Traveler Response Factors” section also addresses the user charac- teristics/factors of ethnicity (see “User Factors”—“Ethnicity”), caution and proficiency (see “Other Factors and Factor Combinations”—“Security and Safety”), and attitude (see “Other Factors and Factor Combinations”—“Attitudes and Modal Biases”). The facility-specific “Facility Usage and User Characteristics” coverage in the next subsection provides further insight, but as always, must be used with caution when making extrapolations from individual sites to other or larger areas and applications. As will be demonstrated, there is considerable variation among areas, facility types, and particular locations. Facility Usage and User Characteristics The tables and discussion that follow offer a selection of the information encountered on NMT facility traffic and usage patterns, and also on characteristics of facility travel purposes and of the facility users themselves. This is presented in a manner as specific to individual types of facilities as possible. The approach is in contrast to pedestrian and bicyclist characteristics data located in the earlier “Underlying Traveler Response Factors” section, which is focused more on national or other broad-based perspectives. There is also facility usage and user characteristics information located in the facility-specific sub- sections of the “Response by Type of NMT Strategy” section, and in individual case studies. That information is primarily from “after” studies done following facility implementation. Users of this chapter interested in facility-specific data should check both the following presentation and the applicable “Response by Type of NMT Strategy” and “Case Studies” topics. Frequency of Facility Usage by Facility Type Analysis of usage distribution among motorized transportation facilities, such as freeways versus arterials, is typically based on traffic and passenger count data. Count information on NMT usage of different transportation facilities is, however, totally inadequate for estimating usage distribu- tion among NMT facility types. Such analyses must be based on reports by pedestrians and bicy- clists on how they themselves have traveled. The 2002 national survey on pedestrian and bicyclist attitudes and behaviors provides one such source of information. Its reporting of facility types used is based on the most recent walk or bike trip in 30 days by survey respondents. The type of facility identified is that most used during the trip (NHTSA and BTS, 2002). Table 16-96 presents the results. 16-314

The information in Table 16-96 indicates that 51 percent of walk trips take place on facilities specifically constructed for pedestrian or other NMT use—sidewalks and paths/trails. Another 41 percent of walk trips take place on roads. For pedestrians, roads would normally be considered an inferior facility type, even though local streets are undoubtedly heavily represented in the “Paved roads, not on shoulders” category. Facility appropriateness is harder to identify in such a straightforward manner in the case of bicycle trips. Appropriate on-street bicycle routes and facilities are, for adults at least, likely safer than riding on most sidewalks. (This finding is explored further in the upcoming “Safety Information and Comparisons” subsection.) Table 16-97 supplements the bicycle facility use information of Table 16-96 by introducing results for five U.S. urban areas from the Nonmotorized Transportation Pilot Program Evaluation Study. Covered are “reference trips” selected for each survey respondent by the inter- viewer. With multiple responses allowed, the five-area total sums to 140 percent. A rough normaliza- tion can be obtained, by dividing through by 1.4, for comparison with Table 16-96. The distribution becomes remarkably similar when this is done, except that bike lane use appears to be over twice as prevalent in the five-area sample. This exercise suggests that a little over one-half of bicycling occurs on the more ideal facilities—local streets, bike lanes, and bike paths. In more global terms, barely over one-half of walk trips and bike trips occur on facilities that may be readily presumed suitable. 16-315 Facility Type “Most Used” for Walk Trips “Most Used” for Bike Trips Sidewalks 45% 14% Paved roads, not on shoulders 25 48 Shoulders of paved roads 8 13 Bicycle lanes on roads — 5 Bicycle paths/walking paths/trails 6 13 Unpaved roads 8 5 Grass or fields 5 — Other 3 2 Total (all facility types) 100% 100% Source: NHTSA and BTS (2002). Table 16-96 Facilities Used for Most Recent Walk/Bike Trip per Attitudinal Survey 2002 Facility Type Columbia, Missouri Marin Co., California Minneapolis , Minnesota Sheboygan, Wisconsin Spokane, Washington Five-Area Total Sidewalk 28% 16% 15% 10% 20% 18% Local street 42 49 48 41 46 45 Busy street 28 25 28 30 26 28 Bike lane 11 27 18 14 14 17 Bike path 13 27 32 10 18 19 Rural road 4 14 0 13 10 8 Other 8 2 0 4 10 5 Sample Size 72 51 60 70 50 303 Note: Multiple responses allowed in Pilot Program survey. Columns total to more than 100%. Source: Krizek et al. (2007). Table 16-97 Facilities Used for Bicycle “Reference Trip” in Pilot Program 2006 Baseline Survey

Neither of the facility-use distributions presented above separately identifies use of bicycle boule- vards. This reflects the small number of such facilities nationwide and the fairly recent recogni- tion, beyond a few “early adopter” localities, of this facility type as more than a niche application. Portland, Oregon, has a number of bicycle boulevards. Table 16-67, in the “Underlying Traveler Response Factors” section (See “Trip Factors”—“Bicycle Trip Distance, Time, and Route Characteristics”—“Bicycle Route Choice”) provides trip mileage distributions among Portland’s facility types for bicycle-only utilitarian trips, based on GPS-based survey measurements. Adding in exercise and “loop” bicycle trips, the distribution becomes 22 percent on arterials with no bike lane, 27 percent on low traffic streets with no bike lane or bicycle boulevard provisions, 26 percent on streets with bike lanes (550 miles available), 9 percent on bicycle boulevards (30 miles avail- able), 14 percent on off-road, shared use trails (130 miles available), and 2 percent other (Dill and Gliebe, 2008). The percentage of bicycle miles of travel attracted to bicycle boulevards in Portland is remarkably high relative to the comparatively small survey-year extent of such facilities. Sidewalks and Streets in Suburbs and City Neighborhoods Typical Suburban- and City-Neighborhood Pedestrian and Bicycle Volumes. Given the prevalence of low to moderate pedestrian volumes on neighborhood sidewalks, volume information is generally obtained only at points where land use or activity concentrations cause a buildup. Table 16-98 provides illustrative circa 2002 pedestrian intersection volumes at a wide variety of San Francisco Bay Area loca- tions selected by local authorities for their importance on account of critical location within the NMT infrastructure, crash history, or other concerns. Three out of nine counties are selected for presentation here. They are Napa County, predominantly rural with many vineyards, but also containing small towns and expanding exurban development; Santa Clara County, the heart of Silicon Valley and mostly suburban in nature; and San Francisco City and County, the most dense of the three primary urban cen- ters of the region. The counts were taken on Tuesdays, Wednesdays, and Thursdays only, from 7:00 to 9:00 AM and 4:00 to 6:00 PM. They have been expanded by a factor of 2.5 to give a very approximate estimate of daily volumes. This factor was developed by analogy with recent Bay Area 24-hour vehicle counts and was intended for bicycle count expansion (Wilbur Smith Associates, 2003). The Handbook authors have taken the liberty of applying the factor to the pedestrian counts as well to give a rough feel for daily NMT activity. Intersection counts such as these, in the case of four-legged intersections (which constitute the vast majority), represent approximately twice the average pedestrian or bicycle volume on the individual intersecting streets, and four times the average individual sidewalk volume in the case of pedestrians (assuming there are sidewalks on both sides of both streets). It is interesting to note that, with one excep- tion, all of the suburban intersections with daily NMT volume estimates exceeding 1,000 are located in the heart of traditional rural or railroad-suburb downtowns with 19th or early 20th century roots. Thus they technically violate the “outside CBDs” restriction. The one exception—California Avenue and Escuela Street in Mountain View—is central to an area of low-rise apartments, with ethnic gathering spots, just beyond the tighter early 20th Century residential street grid. At roughly 2,700 daily pedestrian and bicycle crossings, the pedestrian volumes at the Mountain View site are about one-third the volumes at the San Francisco neighborhood intersection of Geary and Divisadero Streets, also covered in Table 16-98. Exploration of this circumstance is instructive. While the neighborhood residential densities are likely roughly similar, the Geary and Divisadero intersection features a greater mix of land uses plus intersecting high-frequency bus routes. Both of these characteristics are generally deemed indicators of higher pedestrian volume likelihood. 16-316

Pedestrian activity effects of mixed land use have been addressed in the “Response by Type of NMT Strategy” section (see “Pedestrian/Bicycle Friendly Neighborhoods”) and in Chapter 15, “Land Use and Site Design,” there under “Response by Type of Strategy”—“Diversity (Land Use Mix)”—“Accessibility, Entropy, and Other Measures” and also “Land Use Mix and Transit Use.” 16-317 Jurisdiction Intersection Area Type and Adjacent Land Use Intersection Legs with Sidewalks 7 to 9 AM 4 to 6 PM ~ 2002 Daily Intersection NMT VolumePeds Bikes Peds Bikes Napa County (Two surveyed rural intersections [Oakville and unincorporated County] are omitted, having no sidewalks or pedestrian traffic) Am. Canyon SR 29 @ American Canyon Suburban: shopping center, vacant land 2 out of 4 5 2 4 6 40 Calistoga Lincoln St. (SR 29) @ Washington Suburban: retail, eateries 4 out of 4 263 9 738 38 2,600 County Dry Creek @ Orchard Rural: vineyards, fields 0 out of 3 15 6 0 25 100 Napa Lincoln Ave. @ Jefferson St. Suburban: high sch., gas, retail, vacant 4 out of 4 65 27 56 39 500 Napa 1st @ School Rd. Suburban: retail, bank, city hall 3 out of 3 133 10 382 41 1,400 St. Helena Main (SR 29) @ Adams Suburban: retail, ofc., bank, gas/auto 4 out of 4 106 5 365 25 1,300 Yountville Finnell @ Yountville Rural: town hall, homes, vineyard 3 out of 3 96 9 39 29 400 Santa Clara County (Silicon Valley proper: Three surveyed intersections — in the East Bay [Milpitas] and South County [Morgan Hill, Gilroy] — are omitted) Campbell Bascom @ Hamilton Suburban: retail, gas 4 out of 4 30 64 71 59 600 Cupertino Stevens Creek Blvd. @ De Anza Suburban: bank, civic center, gas 4 out of 4 67 23 108 41 600 Mtn. View California St. @ Escuela Ave. Suburban: residential apartments 4 out of 4 589 104 307 92 2,700 Palo Alto Foothill Expwy. @ Page Mill Suburban: fields, office building 4 out of 4 1 63 8 82 400 Palo Alto University @ Emerson Suburban: retail, restaurant 4 out of 4 295 80 557 42 2,400 San Jose San Fernando @ 7th Urban: CBD (details n/a) 3 out of 3 631 20 674 39 3,400 San Jose Santa Clara @ Montgomery Urban: arena, parking lots 3 out of 3 114 18 111 32 700 Santa Clara El Camino Real @ Railroad Suburban: storage, police, auto rental, etc. 4 out of 4 34 20 45 23 300 Santa Clara Homestead Rd. @ Kiely Blvd. Suburban: retail, gas 4 out of 4 107 23 121 27 700 San Francisco City/County (The surveyed intersections do not include any in the core financial/retail district north of or along Market Street) San Francisco 3rd St. @ Howard Urban: convention ctr., theatre, hotel, ofc. 4 out of 4 2,227 n/a 2,698 n/a 12,000 (ped only) San Francisco Embarcadero @ Washington Urban: urban waterfront (details n/a) 3 out of 3 318 115 516 181 2,800 San Francisco Seventh @ Folsom Urban: CBD fringe (details n/a) 4 out of 4 810 207 789 151 4,900 San Francisco Geary @ Divisadero Urban: apartments, retail, garage, chapel 4 out of 4 1,157 n/a 1,436 n/a 6,500 (ped only) San Francisco Ocean @ Geneva Urban: college, residential, gas, firehouse 4 out of 4 266 n/a 323 n/a 1,500 (ped only) Notes: The Handbook authors have taken the liberty of applying the study’s bicycle expansion factor (peak periods x 2.5 = daily) to both bicycles and pedestrians, in order to provide rough, order-of-magnitude approximations of the daily NMT intersection volumes. Source: Wilbur Smith Associates (2003), daily volumes and missing urban land-use/sidewalk data estimated/supplied by Handbook authors. Table 16-98 Illustrative Intersection Pedestrian and Bicycle Volumes from Selected San Francisco Bay Area Counties The San Francisco Bay Area intersection count data, as can be seen in Table 16-98, includes both pedestrians and bicycles. The counts were structured as if all bicycles would be on the street (Wilbur Smith Associates, 2003), but there may have been some cyclists approaching the intersec- tions using a sidewalk.65 Suburban and City-Neighborhood Pedestrian and Bicyclist Trip Purposes. A study in Texas pro- duced more limited volume examples but offers the advantage of some information on walking and bicycling trip purposes in such areas. Eight suburban and neighborhood locations, mostly 65 In some jurisdictions riding a bicycle on the sidewalk is prohibited.

intersections, were selected for study on the basis of having supportive bicycle and pedestrian facilities. Two each were in College Station, Austin, Houston, and Dallas. Looking first at College Station and Austin, three streets had bike lanes and attracted weekday 12-hour bicycling volumes of 73 to 161 bicycles, averaging 115. The three cross streets in these locations carried seven to 314 bicycles, averaging 110. Overall bicycle trip purposes obtained for the three intersections in a rudimentary survey were 14 percent recreation, 53 percent work, 28 percent school, and 5 percent personal, shopping, and other. Pedestrian volume along the six streets ranged from two to 92 per- sons, averaging 41. Overall trip purposes for walkers at the three intersections were 25 percent recreation, 20 percent work, 25 percent school, 20 percent personal, and 10 percent other. The fourth location, Loop 360 in Austin, was an outlier both statistically and geographically. Survey responders among the 62 counted cyclists all reported recreational activity. No survey returns were obtained from the six pedestrians. Each of the four locations in Houston and Dallas featured a trail, either stand-alone or in conjunc- tion with a street or arterial. Weekday 12-hour bicycle volumes on the four trail corridors ranged from 111 to 346, averaging 205. The two intersecting streets which were counted had only 17 bicy- cles total, so the bicycle trip purpose information for these four sites is nearly trails-only. Overall bicycle trip purposes were 53 percent recreation, 44 percent work, and 3 percent personal and other. Pedestrian volumes along the four trails, including joggers, ranged from 67 to 626, averag- ing 235. On the two cross streets, pedestrian counts were 26 and 108. Overall pedestrian trip pur- poses were 65 percent recreation, 4 percent work, 6 percent personal, and almost 25 percent other (Hottenstein, Turner, and Shunk, 1997). The recreation trip purpose was thus in the majority for these Houston and Dallas trail-dominated sites, but in the minority for the College Station and Austin intersections where roadways with bicycle lanes along with undifferentiated mostly-two- lane cross streets both played major roles. On the other hand, the trail recreational trip proportions identified in this particular study are generally lower than found in trail studies that include week- end usage, several of which are examined below under “Off-Road Shared Use Paths.” Mixed-Use Suburban and City-Neighborhood Pedestrian Volume Variations. One set of the very limited published or presented data on non-CBD sidewalk volume temporal patterns covers monthly variations on University Avenue in San Diego. This east-west arterial, on the opposite side of Balboa Park from the downtown, is a non-radial, cross-town facility. Mixed-use development fronting the side- walk at the count location reflects extensive small-shop commercial use. Morning peak-hour pedes- trian volume, circa 2008, was in the 76 to 225 range (Jones, 2009). Percentages of annual volume by month are provided in Table 16-99. No strong seasonal pattern stands out, with May the highest month at 10.2 percent of yearly volume, and September the lowest with 6.2 percent. San Diego is of course known for its year-round moderate and relatively dry climate. 16-318 Month Percent Month Percent Month Percent Month Percent January 8.0% April 8.0% July 9.6% October 8.2% February 8.2 May 10.2 August 7.4 November 8.2 March 8.8 June 9.0 September 6.2 December 8.2 Note: Basis for November percentage is 2007-08 average, December is 2007, all other months are 2008; all percentages are normalized to 100% for 12 months. Source: Jones (2009), with percentage values scaled from the presentation graphic and normalized by the Handbook authors. Table 16-99 Monthly Variation, University Avenue Sidewalk, San Diego

A suburban CBD weekday hourly variation example is provided in the “Case Studies” section (see Table 16-128). As indicated there, it is not known whether the substantial differences in peaking among counts were caused by land use differences on opposite sides of the studied arterial crosswalks, pre- and post-Christmas shopping pattern differentials, count protocol differences, or some combination of these factors. The 13-hour counts, two per east-west crosswalk at one intersection, ranged from 400 to 2,100 pedestrians per crosswalk. Peak hour volumes ranged from 80 to 280, with individual crosswalk highest-peak hours starting at times ranging from noon to 5:00 PM. (See “More—Volume Variability” in the “Special Mini-Studies in Montgomery County, Maryland” case study.) Sidewalks and Other Provisions in Major Central Business Districts Central Business District Pedestrian Volume Characteristics. City centers, and other major activ- ity centers with large employment, are environments associated with substantially higher volumes of pedestrian trips. Major proportions of person-trips within metropolitan CBDs are made by pedestrians. Circa 1970 it was estimated that 55 percent of morning peak period person trips in Midtown Manhattan were by pedestrians, with the figure increasing up to 70 percent during the noontime and afternoon peaks (Pushkarev and Zupan, 1975). The trips are usually short, typically less than a few blocks. They mainly reflect movements from parking and transit terminals to places of work, between stores (and offices) in the retail core, and other building-to-building trips—often for eat-meal purposes. Many walk trips in high-density areas are trips that would be made by auto- mobile in environments where activities are more dispersed (Levinson, 1972). Pedestrian travel within central areas is highly concentrated in the retail and commercial cores. Major internal travel movements take place between relatively few areas, usually within the retail shopping area. Pedestrian flows typically correlate closely with land value profiles. The decline in land values as one moves out from major intersections mirrors the patterns of people walking along the street (Berry, 1967). Pedestrian volumes are far more localized than transit or automobile passenger flows. Data collected in a series of comprehensive CBD pedestrian surveys circa the 1970s may be dated insofar as absolute values are concerned, but still illustrate the degree of localization very clearly. For exam- ple, while 10-hour sidewalk volumes along State Street in Chicago’s Loop between Madison and Washington Streets at one time exceeded 50,000 persons, sidewalk volumes were only 11,000 per- sons between Lake Street and Wacker Drive three blocks up the street, and below 7,000 persons 5 blocks to the west on Franklin Street. Daily crosswalk volumes exceeded 20,000 persons in Seattle’s core area but dropped to 3,000 persons within two blocks beyond the core. Midday pedestrian traffic volumes on Fifth Avenue in New York of about 50,000 per hour were paralleled by volumes averaging 10,000 along the Avenue of the Americas and 5,000 persons per hour on Eighth Avenue, 2,000 feet to the west. Fifth Avenue noontime and evening peak pedes- trian volumes declined rapidly north of 57th Street. The peak-hour pedestrian volumes along Washington, Summer, and Tremont Streets in the heart of Boston proper approached 6,000 per- sons near major subway entrances but dropped to fewer than 2,000 persons within a few blocks. Along Wilshire Boulevard in Beverly Hills, 6-hour pedestrian volumes dropped from 3,000 to 300 within a few blocks. Similar phenomena were recorded in Philadelphia and Dallas (Pushkarev and Zupan, 1971, Wilbur Smith and Associates, 1970, Levinson, 1982). Lunchtime 3-hour counts taken in 2002 in Columbus, Ohio, illustrate the same pedestrian volume con- centration phenomenon, but in a medium-size city. Counts from 11:00 AM to 2:00 PM on two individ- ual sidewalk sections immediately alongside the State House showed 500 and 800 pedestrians. Across 16-319

from the State House and within 1/2 block, eight sidewalk counts along office and commercial frontage ranged from 1,200 to 2,700 pedestrians each, averaging 2,100 pedestrians. Further away, up to three blocks, eight additional lunchtime sidewalk counts ranged from 400 to 1,400, averaging 900. Once beyond 1/2 block of the State House, the drop-off to the north, east, and south appeared to diminish. (No counts were made beyond 1/2 block to the west.) Three 11-hour counts, all within 1/2 block of the State House at prime locations, ranged from 6,100 to 6,900 pedestrians, averaging 6,500. At these locations, the 3-hour midday pedestrian traffic aver- aged 38 percent of the corresponding 11 hour volumes. Downtown Columbus, in 2002, had 475 stores with $761 million in annual sales, 75,000 office workers, 20,000 transit rider arrivals daily, and some 30,000 students about 1/2-mile to the east (Capital Crossroads, 2003). Table 16-100 displays a selection of 11-hour two-way counts made in central Minneapolis in 2002, showing the variations by hour in levels of sidewalk and Skyway pedestrian activity. The sidewalk counts were taken along the Nicollet Transit Mall. Note that on the east-side sidewalk, the counts one block apart during the busy noon hour differ by a factor of two, underscoring the points made earlier about the very localized nature of pedestrian traffic flows. The 8th to 9th Street block was the busiest on the mall in 2002, with a combined east and west side volume of 23,600 pedestrians. The pedestrian volume was up 69 percent over the prior year as the result of store openings (Bruce, 2002a and c). The Nicollet Mall sidewalk counts show a minor influence of commute hours, but a strong relation- ship to noon-hour activity, and also what may be presumed is a reflection of afternoon shopping activity. The Skyway count made just west of the Nicollet Mall exhibits a similar pattern, but with commute flow influences barely discernible. This is to be expected, as the sidewalks serve persons leaving or accessing their bus stops, while the Skyways are one storey removed. Caution should be applied in comparing the sidewalk and Skyway counts, as the sidewalk counts were on warm, sunny, September days and the Skyway counts were on cool late October days with mixed weather. The last-listed count in Table 16-100 is for a Skyway toward the northwest perimeter of the Skyway system, close to I-394 parking and transit facilities. The pedestrian commute flow can be clearly seen in the hourly flow distribution. All these 11-hour Minneapolis counts are averages for 2 days except for the last 2 hours of this last-listed Skyway segment. The counts on the omitted day were affected by a special event between 4:00 and 6:00 PM. Compared to the data shown, the special event increased the 4:00 to 5:00 PM tally by 34 percent and the 5:00 to 6:00 PM tally by 48 percent, providing a classic example of the importance of special events and other exogenous circumstances in the understanding and recording of NMT flows (Bruce, 2002a). Table 16-101 provides aggregated hour-of-day distributions for pedestrians and for bicyclists entering and leaving the Minneapolis CBD, separately by direction of flow. These data are from the 2003 Cordon Count. The CBD cordon counts are taken periodically on or about September 10th and obtain a 12-hour tally of vehicles and people entering and leaving the core area broken down by 15-minute intervals and travel mode. The pedestrian hour-of-day distributions are quite different from those of any other travel mode, exhibiting a major midday peak, even though measured along the periphery of the CBD. The bicycle hour-of-day distributions are roughly similar to the distributions for all travel modes combined when examined on an hourly basis. Bicycles exhibit a sharper “peak-of-the-peak,” however, with 5.1 percent of the inbound flow between 7:45 and 8:00 AM, and 5.9 percent of the outbound flow between 5:00 and 5:15 PM (SRF Consulting Group, Inc., 2003). Additional Minneapolis NMT count information is provided in the “50 Years of Downtown NMT Facility Provisions—Minneapolis” case study. Also, Table 16-8 in the “Pedestrian Zones, Malls, and Skywalks” subsection (see “Pedestrian Skywalks”—“Skywalk Impacts on Walking”) illustrates the 16-320

interplay of noontime Twin Cities Skyway use versus use of parallel crosswalks. Observed Skyway vol- umes slumped in summer and rose in winter, volumes on the parallel crosswalks did the opposite, and the total of the two stayed within plus or minus 10 percent throughout all 12 months of the year (Heglund, 1980). This outcome suggests that, with a weather-protected option provided, even a rigor- ous northern climate such as found in Minneapolis and St. Paul can have overall business-area walk- ing levels unaffected by season. 16-321 Hour Beginning: 7 AM 8 9 10 11 12 PM 1 2 3 4 5 PM Total Nicollet Mall, 6th-7th Sts., East side 386 320 193 204 655 982 587 491 420 509 470 5215 Tues. 9/10/02 and Tues. 9/17/02 7.4% 6.1% 3.7% 3.9% 12.6% 18.8% 11.2% 9.4% 8.1% 9.8% 9.0% 100% Nicollet Mall, 6th-7th Sts., West side 435 366 241 290 888 1231 807 674 733 1082 1013 7757 Wed. 9/4/02 and Tues. 9/17/02 5.6% 4.7% 3.1% 3.7% 11.4% 15.9% 10.4% 8.7% 9.5% 13.9% 13.1% 100% Nicollet Mall, 8th-9th Sts., East side 535 457 489 513 1085 1984 1294 832 624 771 728 9310 Wed. 9/4/02 and Tues. 9/17/02 5.7% 4.9% 5.3% 5.5% 11.7% 21.3% 13.9% 8.9% 6.7% 8.3% 7.8% 100% Skyway, 9th St., West of Nicollet 614 962 693 941 2151 3299 1864 1172 945 981 825 14445 Tues. 10/22/02 and Thu. 10/31/02 4.3% 6.7% 4.8% 6.5% 14.9% 22.8% 12.9% 8.1% 6.5% 6.8% 5.7% 100% Skyway, 1st Ave., South of 6th St. 700 710 293 139 229 228 240 223 321 640 a 777 a 4497 Wed. 10/23/02 and Thu. 10/30/02 15.6% 15.8% 6.5% 3.1% 5.1% 5.1% 5.3% 5.0% 7.1% 14.2% 17.3% 100% Notes: All counts are 2-day averages unless footnoted. a (Count is 10/23/02 only). Percentages are calculated on the basis of 11-hour totals. Source: Bruce (2002a and c). Table 16-100 Weekday Hour-of-Day Patterns of Pedestrian Traffic on Nicollet Transit Mall and Skyways of Minneapolis, 2002 Hour Beginning: 630 AM 730 830 930 1030 1130 1230 PM 130 230 330 430 530 PM Total Pedestrians, Inbound 1554 2788 1763 1308 1149 1952 2020 1307 1210 1200 1313 882 18446 Percentage by Hour 8.4% 15.1% 9.6% 7.1% 6.2% 10.6% 11.0% 7.1% 6.6% 6.5% 7.1% 4.8% 100% Pedestrians, Outbound 600 1109 906 932 933 1613 1905 1314 1355 1766 2665 1711 16809 Percentage by Hour 3.6% 6.6% 5.4% 5.5% 5.6% 9.6% 11.3% 7.8% 8.1% 10.5% 15.8% 10.2% 100% Bicyclists, Inbound 170 331 247 173 141 138 158 149 172 187 192 161 2219 Percentage by Hour 7.7% 14.9% 11.1% 7.8% 6.4% 6.2% 7.1% 6.7% 7.8% 8.4% 8.6% 7.3% 100% Bicyclists, Outbound 91 116 114 110 118 115 120 166 204 272 411 267 2104 Percentage by Hour 4.3% 5.5% 5.4% 5.2% 5.6% 5.5% 5.7% 7.9% 9.7% 12.9% 19.5% 12.7% 100% Notes: Sum of cordon counts taken at 32 non-ramp stations on September 10, 2003. Percentages are calculated on the basis of 12-hour totals. Source: SRF Consulting Group, Inc. (2003), with hourly-on-the-half-hour sums of 15-min. counts, and percentages, by the Handbook authors. Table 16-101 Weekday Hour-of-Day Patterns of Pedestrian and Bicycle Traffic Entering and Leaving Minneapolis CBD, 2003 Central Business District Pedestrian Trip Purposes and Modal Linkages. Pedestrian trip pur- poses reported in comprehensive studies conducted in 1970 in downtown Seattle are shown in Table 16-102. Although the numerical values in this dated information should be treated with cau- tion, the relationships remain instructive. Work and commercial or personal business trips accounted for over one-half of the pedestrian trip total. Shopping activity accounted for 30 percent of all pedestrian trips even though shopping trips comprised only an estimated 15 percent of all CBD person-trip destinations made by all modes of travel (Wilbur Smith and Associates, 1970).

There are, of course, major variations in CBD pedestrian trip purposes by time of day. During the AM peak period, most trips originate at home and manifest themselves as walk trips between a parking place or transit stop and a place of CBD employment. The reverse predominates in the PM peak. At noontime, in strong contrast, the main pedestrian movements are between offices and places such as restaurants. To support a circa 1970 Midtown Manhattan circulation study two Regional Plan Association office building surveys were supplemented with nearly 4,400 interviews conducted at 22 sites represent- ing six different land uses. The results confirmed the midday importance of walk trips to and from offices. Between 12:00 Noon and 2:00 PM destinations of walk trips from offices were restaurants (53 percent), other offices (21 percent), retail establishments (9 percent), residences (5 percent), and various other land uses (12 percent) (Lemer, Bellomo, and Liff, 1972). Trip purpose information tends to mask the importance of linkages with other travel modes as a major component of CBD walking activity. In addition, survey data identifying the extent to which observed pedestrians are making trips in connection with use of some other primary mode, as compared to utilizing the walk mode exclusively, is quite limited. Such data as do exist indi- cate that parking space-to-building or transit stop-to-building walk trips are usually more preva- lent than walk trips between buildings. In downtown Seattle circa 1970, for example, 56 percent of all 7:00 AM to 7:00 PM weekday pedestrian trips were to or from transportation facilities and 44 percent were inter-building trips. The transportation facility percentage was comprised of 39 percent automobile parking and 17 percent transit stops and terminals (Wilbur Smith and Associates, 1970). What is evident from such relationships is that the locations of parking facilities and transit termi- nals are necessarily a major influence on the patterns and volumes of walking trips within a CBD. While major changes in Seattle’s transportation system may have altered the percentages some- what, the order of magnitude relationships would still be valid for any typical large-city core area. A 21st Century summertime off-peak weekday and Saturday survey in Toronto on Bloor Street side- walks between the Bathurst Street and Spadina Avenue subway stations did determine what the usual mode of travel to the area was for pedestrians interviewed. It was 46 percent walk, 12 percent bicycle, 32 percent public transit, and 10 percent motor vehicle (Sztabinski, 2009). A similar survey in New York City’s SoHo district, on Prince Street sidewalks between Broadway and 6th Avenue, found that modes utilized by the interviewees to arrive in the area that day were 29 percent walk, 5 percent bicycle, 54 percent rail transit, 2 percent bus, 9 percent taxi or livery service, and 9 percent private motor vehicle (Schaller Consulting, 2006). 16-322 Purpose Percent Purpose Percent Work 24.1% Shopping 30.8% Commercial Business 12.3 Eat-Drink 5.8 Personal Business 17.6 Social-Recreational 2.4 Sales and Service 2.2 Other 4.8 Note: Downtown Seattle has had extensive office tower development since collection of this data. Source: Wilbur Smith and Associates (1970). Table 16-102 Reported Purposes of Seattle CBD Pedestrian Trips, 7:00 AM–7:00 PM, 1970

These two surveys only indirectly address the question of whether the observed walk trips were walk-only or multimodal. Also important is that central Toronto and Lower Manhattan are areas of intensive public transit service. Obviously a survey of this type in an area not similarly served would produce different results. Two quite different circa-1980 perspectives on the proportion of persons using downtown sidewalks for accessing and egressing bus service come from on-street pedestrian malls in Portland, Oregon, and Minneapolis, Minnesota. In Portland, pedestrian volumes on a pair of one-way, transit-mall streets were estimated to be 75 percent bus riders (Dueker, Pendleton, and Luder, 1982). In Minneapolis, most pedestrians interviewed on Nicollet transit mall sidewalks in 1977 were there for shopping (57 percent), pleasure (42 percent), or because of their work (24 percent). Only 16 percent were there because it was their bus stop location (5 percent were on the sidewalks for other reasons— multiple answers were allowed) (Edminster and Koffman, 1979). Background on these transit malls is found in the “Response by Type of NMT Strategy” section (see “Pedestrian Zones, Malls, and Skywalks”—“Pedestrian Zones and Malls”—“Transit Malls”) and also in the “50 Years of Downtown NMT Facility Provisions—Minneapolis” case study. On-Street Bicycle Facilities Perhaps the least well represented of NMT facility types among published bicycle volume compila- tions are on-street bicycle facilities. The “after” counts of before-and-after studies of bike lanes and bicy- cle boulevards provide a source, however. A number of such counts are presented or cross-referenced in the “Bicycle Lanes and Routes” subsection, within the “Response by Type of NMT Strategy” section (see “Bicycle Lane Implementation”—“Before-and-After Counts and Surveys” and also “Bicycle Lane Variations, Bicycle Boulevards, and Other Signed Bicycle Routes”—“Bicycle Boulevards”). Reported 1-hour bicycle volumes in California after bicycle lane implementation range from just over 80 on Fell Street in San Francisco (one-way, PM peak, bicycles in lane only) (Chaney, 2005) to an average of a little over 500 on Anderson Road in Davis (two-way, an average of one AM and two PM peak hours, all bicycles anywhere on the street or sidewalk) (Lott, Tardiff, and Lott, 1979). An issue with “after” counts of before-and-after studies is that there may be substantial subsequent growth. St. Kilda Road in Melbourne, Australia, was carrying about 75 bicyclists in the AM peak hour a year after bike lane implementation, but that number had grown to some 500 bicyclists 10 years after opening (Davies, 2007). In terms of all-day counts, Oriental Boulevard in Brooklyn was, 1 year after bicycle lane installation, carrying just over 100 bicycles and other human-powered wheeled vehicles in 11 weekday hours (Chaney, 2005). Six bicycle lanes in downtown Toronto had annual average weekday bicycle volumes ranging from 570 to 1,900, an overall average of 1,230 bicycles, roughly 2 years after implementation (Macbeth, 1999). Extrapolated 24-hour bicycle volumes on the various bicycle-boulevard-like segments of the Bikeway system in Vancouver, BC, Canada, ranged from about 40 to almost 1,100 daily (Chaney, 2005). Bicycle volumes on the Bryant Street bicycle boulevard in Palo Alto, California, fall in the mid to upper part of this range (Ciccarelli, 2010), but the Vancouver range is exceeded by the 2008 volume of 1,900 bicycles reported for the Lincoln-Harrison bicycle boulevard in Portland, Oregon (Alta Planning + Design, 2009a). Counts assembled in 1997 for the Palo Alto, California, Bicycle Transportation Plan help illuminate, in terms of bicycle volumes handled, where on-street routes and facilities fit in the total spectrum of urban 16-323

bicycle facilities. Palo Alto is a university town with a long-standing bicycle-friendly reputation, but is also part of the larger Silicon Valley environment. The 1990s population of Palo Alto was on the order of 60,000 residents. The 1990 U.S. Census journey-to-work Palo Alto bicycle share was 5.8 percent, com- pared to 1.4 percent for Santa Clara County as a whole and 0.4 percent for the nation. Taken together, the bicycle counts suggest that all types of facilities, on-street and off-street, have a significant role to play, and that location is a crucial factor in determining bicycle facility volumes. A brief summary of the counts illustrates the point: • The highest bicycle usage—making allowances for count duration—was reflected in an 8-hour count of 830 bicycles at the intersection of the Bryant Street bicycle boulevard and an arterial with bicycle lanes, both on-street facilities in the heart of the city.66 • Two other counts in the 800 to 900 range, actually higher but obtained in 12-hour counts, were at grade separations serving arterials with and without bicycle lanes and piercing barriers separating Stanford University from much of the city. • Four 8- and 12-hour bicycle counts in the 400 to 600 range included two exclusive bicycle/ pedestrian bridges over creeks (one at the end of the bicycle boulevard and one serving a mix of bicycle lanes, routes, and paths); one arterial intersection adjacent to Stanford with bike lanes on one leg; and one intersection of bike lanes and a path. • Ten counts in the 200 to 400 bicycle range included one grade separation, seven intersections of streets and arterials with and without bicycle lanes or signed routes, and two exclusive bridges over creeks, one serving a bicycle route and the other a path. • Finally, the two lowest volumes were both on bicycle lanes bordering the baylands at the edge of the city (City of Palo Alto, 2001). Thus the highest and lowest Palo Alto bicycle volumes were encountered on designated on-street bike facilities, with about every imaginable facility-type combination in between. As noted elsewhere, trip purpose information for bicyclists utilizing bicycle lanes is extremely scarce. A 7:00 to 9:00 AM weekday bike lane survey in the Seattle CBD found 97 percent of survey respondents to be making a work, school, or other utilitarian trip (Niemeier, Rutherford, and Ishimaru, 1995b). The Texas surveys summarized under “Suburban and City-Neighborhood Pedestrian and Bicyclist Trip Purposes” found 84 percent of College Station and Austin bicycle trips to have work and other utilitar- ian purposes, over a 12-hour weekday period, among a population of cyclists roughly evenly split between users of urban streets with bike lanes and users of undifferentiated streets (Hottenstein, Turner, and Shunk, 1997). There is presently little basis on which to extrapolate what the trip purpose mix for bicycle boulevards might be. The apparent attractiveness of such facilities for less experienced bicyclists and females (Dill and Gliebe, 2008) supports speculation that the purpose distributions would lie between the mixes attracted by bicycle lanes and off-road shared use paths. Volunteer responders to an on-line survey, among residents fronting on the SE Salmon Street bicycle boulevard in Portland, Oregon, reported their 16-324 66 An independent report of a 1997 8-hour count on Bryant Street gives 385 bicycles (Ciccarelli, 2010). This sug- gests that intersection counts reported in the Bicycle Transportation Plan (City of Palo Alto, 2001) are totals for all movements in the intersection, on all streets and bicycle facilities involved.

three top bicycling destinations (on or off the bicycle boulevard) were social/recreational (82 percent), shopping/errands (61 percent), and work (59 percent) (VanZerr, 2010). Off-Road Shared Use Paths The most extensive and readily available store of contemporary NMT volume characteristics and facility user information is that pertaining to off-road, shared use paths. Even so, its comprehen- siveness and consistency don’t begin to approach that for motorized travel modes and facilities. The National Bicycle and Pedestrian Documentation Project, for example, was just initiated in 2002 and as of early 2009 remained a volunteer effort with no source of funding and no resources for quality assurance or control (Jones, 2009). A selection of off-road shared use path volume and facil- ity user characteristics information is presented here. Path Volume and Usage Patterns. Table 16-103 illustrates the monthly patterns of traffic observed on San Diego and Indianapolis paths, and gives Indianapolis temperatures for perspective. These data are placed in context of season and climate with six other paths and path groupings in the “Underlying Traveler Response Factors” section (See “Environmental Factors”—“Natural Environment”—“Combined Walk and Bike Seasonal Effects” and Table 16-63). The Gilman Bike Path in San Diego is in a freeway and active-railroad transportation corridor in the suburbs, the Strand Bike Path interconnects beachfront urbanization and parkland (Jones, 2009), the Monon Trail is a rail trail through the heart of Indianapolis (Indiana University, 2001), and the other Indianapolis trails are an assortment of rail trails, riverside trails, and a canal towpath.67 Day of week volumes and percentage distributions are provided for the same two San Diego paths in Table 16-104, along with percentage distributions for the Terry Hershey Park Trail in Houston. The Houston facility is a riparian greenway trail. Of particular interest are the separate volume and percentage distributions for pedestrians and bicyclists on San Diego’s Strand Bike Path. In this one instance, pedestrian traffic is much more variable day-by-day than bicycle traffic. On Saturday, the peak day, the volume of 620 bicycles was 2.8 times the volume of 220 pedestrians. On Wednesday, the low traffic day, the volume of 390 bicycles was 9.8 times the count of 40 pedestrians (Jones, 2009). Table 16-105 presents weekday combined volume distributions by hour of day for nine mostly urban/suburban paths, and weekend volumes for eight, utilizing nine- and eight-path averages. In the weekday data, two outliers tend to balance out, and the average is a good representation of the dominant pattern of extended morning peak, noontime peak, and more sizeable afternoon/evening peak. Showing Manhattan separated out serves to display a slightly different pattern with an earlier morning peak and a particularly sharp peak at 6:00 to 7:00 PM, a phenom- enon seen—with variations in the timing—on a little over half the paths. In the weekend data, the non-urban recreational trail and the Bosque Trail in Albuquerque are both outliers, with heavy morning usage disproportionate to light afternoon usage. Together they tend to warp the average. Manhattan separated out serves to represent the majority of the paths better than does the eight- path weekend-day average. 16-325 67 Trail descriptions in this discussion rely both on the cited sources and on supplemental trail web-search results.

Wednesday, June 13, 2007, hourly variation data for two San Diego paths (not previously intro- duced) exhibit quite different pedestrian and bicyclist activity patterns over a 24-hour period. Walkers and cyclists were separately identified using active infrared counting technology (Jones, 2009). The results dramatically illustrate how much NMT facility usage characteristics can vary by sub-regional location and orientation. One of these two paths, the Rose Canyon Bicycle Path, serves a low density area with few destina- tions of its own and has a commuter orientation. Significant pedestrian activity (more than two per hour) was limited to the 7:00 AM to 11:00 AM period, with essentially no activity in the heat of the afternoon. The 21 pedestrians counted all day, 6 percent of path traffic overall, had secondary peak hours starting at 7:00 AM (14 percent) and 7:00 PM (10 percent) that may or may not have been related to commuting activity. The main peak for walkers was 10:00 to 11:00 AM (24 percent). Significant bicyclist activity occurred between 5:00 AM and 9:00 PM. Bicycle traffic, at 327 cyclists, was 94 percent of the path traffic total of 348. A morning peak occurred between 7:00 and 9:00 AM (12 percent each hour) while the evening peak was sharper, concentrated between 5:00 and 6:00 PM (12 percent). 16-326 U.S. Location/Description Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. Southwest (San Diego, CA) Gilman Bike Path 4.8% 5.2% 7.0% 7.2% 9.0% 7.6% 15.8% 12.8% 10.0% 6.0% 7.2% 7.4% Strand Bike Path 6.2 6.4 9.4 8.8 8.8 10.2 14.4 7.0 7.6 8.2 7.0 6.0 Midwest (Indianapolis, IN) 30 Indianapolis locations 2.0 3.6 4.6 9.8 11.0 13.0 14.8 15.8 13.0 6.6 3.8 2.0 4 locations - Monon Trail 3.6 3.4 5.8 12.0 11.2 12.6 13.0 12.6 10.6 6.8 4.6 3.8 Indianapolis temperature ( F) 36 36 38 58 64 70 74 76 66 56 50 38 Notes: Basis for San Diego November percentages is 2007-08 averages, December is 2007, all other months are 2008; all San Diego percentages are normalized to 100% for 12 months. No years given for Indianapolis counts. Source: Jones (2009), with percentage values scaled from presentation graphics, and normalized (San Diego only), by the Handbook authors. Table 16-103 Monthly Patterns of Traffic on Shared Use Paths for All Users Combined (Pct. by Month of Annual Traffic) U.S. Location/Description Mon. Tues. Wed. Thurs. Fri. Sat. Sun. 7-Day Volume Southwest (San Diego, CA) Gilman Path, Week of 7/23/07 320 (10%) 420 (13%) 410 (13%) 430 (13%) 380 (12%) 690 (21%) 590 (18%) 3,240 (100%) Gilman Path, Week of 7/30/07 310 (8%) 440 (12%) 500 (13%) 460 (12%) 620 (16%) 760 (20%) 730 (19%) 3,820 (100%) Strand Bike Path, Pedestrians 60 (8%) 70 (9%) 40 (5%) 70 (9%) 130 (17%) 220 (29%) 180 (23%) 770 (100%) Strand Bike Path, Bicyclists 400 (12%) 420 (13%) 390 (12%) 410 (12%) 460 (14%) 620 (19%) 600 (18%) 3,300 (100%) South Central (Houston, TX) Terry Hershey Park Trail 16% 11% 12% 11% 15% 18% 17% 100% Notes: All users combined, except as noted otherwise. No dates given for Strand Bike Path or Terry Hershey Park Trail counts. Source: Jones (2009), with scaling from presentation graphics, and calculation of percentages (San Diego only), by the Handbook authors. Table 16-104 Day of Week Patterns of Traffic on Shared Use Paths (All-User Daily Volumes and Percent by Day of Weekly Use)

A major contrast is provided by the second path, the Mission Beach Bicycle Path, an urban water- front facility that is recreational in focus with many destinations. Here the pedestrian traffic of about 870 walkers was 43 percent of the counted path traffic total of roughly 2,020. Significant pedestrian activity occurred between 6:00 AM and 10:00 PM. The morning pedestrian peak hour started at 7:00 AM (10 percent), followed by a secondary noon hour peak (7 percent) and a broad afternoon and evening peak with its apogee at 5:00 to 6:00 PM (12 percent). Significant bicycling activity ran from 5:00 AM to midnight and the 24-hour total was about 1,160 cyclists, 57 percent of path traffic. The traditional morning peak hour was minor, starting at 7:00 AM (6 percent). The more dominant morning peak hour started at 11:00 AM (9 percent), and after dropping by barely more than 1/3, the bicycle traffic peaked again at 6:00 PM (9 percent) (Jones, 2009). Additional trail count information along with data on peak-hour timing and percentages are given for six Indiana trails in the case study “Six Urban, Suburban, and Semi-Rural Trails—Indiana Trails Study” found near the end of this chapter. (Table 16-135 describes the trails and Table 16-136 gives the count and peaking information.) The Monon trail of Indianapolis exhibited the sharpest all- user peak of any encountered in the literature, 17.9 percent between 6:00 and 7:00 PM weekdays in September, 2000, and 19.4 percent between 5:00 and 6:00 PM in October when the days were shorter (Indiana University, 2001). Path User Mode Distributions. Table 16-106 gives proportions walking, running or jogging, cycling, or in-line skating, based on trail traffic observations or survey results for six different paths or groups of paths. For a majority of the paths, information was also obtained on other or associ- ated activities, as listed in the “Other” column. The Hennepin County (Minneapolis area) survey obtained frequency of use information. With that data they found that 2 percent of summertime trail users reported extensive on-trail skiing activity, obviously in the winter, with another 12 per- cent reporting occasional skiing (Hennepin County, 2005). 16-327 Hour Beginning: 6 AM 7 8 9 10 11 12 PM 1 2 3 4 5 6 7 8 9 PM Weekday Hourly Patterns Nine U.S.-location average 25% 5% 7% 75% 7% 65% 75% 45% 5% 65% 8% 11% 105% 65% 3% 2% Manhattan separated out 4 8 85 55 6 55 6 3 5 65 75 95 12 8 3 2 Weekend Hourly Patterns Eight U.S.-location average 05 25 7 9 10 10 9 65 7 75 8 8 6 4 3 2 Manhattan separated out 05 15 55 85 95 95 95 65 75 85 85 85 65 45 3 2 Notes: Nine-location weekday average included one each from the five New York City boroughs, one from Licking County in Ohio (July), the Monon Trail in Indianapolis (October), the Terry Hershey Park Trail in Houston (May-Oct.), and a recreational area trail with no urban anchor (Outerbanks in North Carolina). Eight-location weekend average included the New York City boroughs, the Bosque Trail in Albuquerque, the Terry Hershey Trail (May-Oct.), and the Outerbanks trail. No dates given other than the months indicated. Source: Jones (2009), with percentage values scaled from presentation graphics by the Handbook authors. Table 16-105 Weekday/Weekend Hour-of-Day Patterns of Traffic on Shared Use Paths (All-User Pct. by Hr. of 16-Hour Use)

The path traffic distributions in Table 16-106 were obtained from classification counts and surveys that were apparently all taken at on-trail intercept points. As discussed in the “Overview and Summary” (see “Analytical Considerations”) and in the “Indiana Trails” case study, this analytical approach produces a “traffic” rather than a “user-visit” perspective. It under-emphasizes trail use for shorter trips such as walk trips, trips for non-work utilitarian purposes, and probably trips by women, younger children, and elders, while over-emphasizing longer trips (more likely to be inter- cepted) such as seen with adult bicycling. (The Mission Beach Bicycle Path discussed above is the type of situation where it can be readily imagined that a tally of users, in contrast to traffic, would show pedestrian activity dominance rather than a bicyclist majority.) Only the six-trail “Indiana Trails Study” described in the next-to-last case study obtained a true user perspective by observing and surveying persons as they entered or exited the trails in the course of beginning or ending their trail visits. In addition, the three-trail Hennepin County survey may have come closer than others, by very explicitly allowing only one survey response per person. Even with the methodological precaution, the Hennepin County trail traffic was reported to consist of a quite high 85 percent cyclists (Hennepin County, 2005). NMT mode share counts made during fieldwork seem to validate this high proportion. The other trails and trail groupings listed in Table 16-106 had trail traffic cyclist percentages ranging from 66 down to 38 percent. In contrast, users of the six studied Indiana trails exhibited cyclist percentages that, aside from 77 percent on one of two semi-rural trails, ranged from 40 down to 23 percent (see Table 16-138). Walkers exceeded 50 percent of Indiana trail users on both the Indianapolis Monon Trail (51 percent) and the Greenfield Pennsy Trail (54 percent) (Indiana University, 2001), while among all 13 trails encompassed by Table 16-106, walking as a component of path traffic exceeded 50 percent only on the Blackstone Valley bike path (51 percent walk) in Rhode Island (Gonzales et al., 2004) and the Shoal Creek Trail (52 percent walk) in Austin, Texas (Shafer et al., 1999). Two sets of survey results are available for the Monon Trail in Indianapolis, one obtained with an interview survey of user visits (persons entering and exiting the trail) and the other obtained as a survey of traffic (persons observed at points on the trail itself). This makes for an instructive com- parison, set forth in Table 16-107 for four activity-types and also gender. Some caution is needed in evaluating the Table 16-107 results differentials, in that the user survey was in 2000 and the traf- fic observations were in 2004, the user survey was in July-August and the traffic observations were in June–July, the 7-1/2 mile trail of 2000 had been extended by 2004 (Indiana University, 2001, 16-328 U.S. Location/Description Walkers Run/Joggers Cyclists In-line Skaters Other 3 Hennepin Co. (Minneapolis) trails, urb./sub. 6% 3% 84% 7% 0.2% (also see text re. skiing) 4 Rhode Island trails, suburban/towns/rural 32 7 49 12 — Capital Crescent Trail (MD-DC), inner suburban 34 15 41 7 3% infants in strollers W&OD Trail, Northern Virginia, sub./exurban 16 16 66 3 (2% with pets, 1% with strollers) 3 Texas Trails, Houston, Austin, urban/sub. 32 29 38 1 0.1% other Iron Horse Trail, S. F. East Bay, exurban 27 9 51 13% incl. other 0.1% equestrians (W&OD also) Notes: Trails are predominantly rail-trails except for alignments via riparian greenways in Houston and Austin, Texas. Traffic proportions were obtained from classification counts except for use of short-form on-trail surveys in Hennepin County, Rhode Island, and Texas. Source: Hennepin County (2005), Gonzales et al. (2004), Maryland-National Capital Park and Planning Commission (2001), Bowker et al. (2004), Shafer et al. (1999), East Bay Regional Park District (1998). Table 16-106 Shared Use Path Traffic Proportions by Type of Activity (Weekday and Weekend Combined)

Lindsey et al., 2006), and the trail locations used may not have been entirely comparable. However, the differences seen in results are sufficiently sizable, and in the expected direction, that survey timing or survey station location “noise” falls short as a credible explanation. The higher male proportion, the much higher bicycle proportion, and the lower proportions of other trail activities, including a much lower proportion of walking, can all be clearly seen in the main-line trail traffic observations as com- pared to the trail user-visit interview results. 16-329 Monon Trail Survey Type Walk Run Bicycle Skate/Other Male Female 2000 Trail User-Visit Survey 51% 13% 23% 13% 46% 54% 2004 Trail Traffic Observations 19 11 61 10 57 43 Note: The 2000 user visit survey data is from statistically controlled interviews of users beginning or ending trail use, not the mail back survey (Indiana University, 2001). Sources: Indiana University (2001), Lindsey et al. (2006). Table 16-107 Comparison of Monon Trail 2000 User-Visit Interview Survey Results and 2004 Traffic Survey Observations for Activity Type and Gender Trail activity mix may vary over time. This is demonstrated and discussed in the case of the Burke- Gilman/Sammamish River (B-G/SR) Trails in greater Seattle, in connection with Table 16-17 within the “Response by Type of NMT Strategy” section (see “Shared Use, Off-Road Paths and Trails”—“Shared Use Path Implementation”—“Seattle Urban/Suburban Trails”). Path Mode-of-Access Distributions. Substantial off-road shared use path facilities and other NMT features attractive in their own right, such as major NMT bridge facilities, generate facility-access trips that typically include motorized access along with NMT access modes. Auto access is some- times appreciable enough that parking availability becomes a concern and may actually influence patterns of use (Lindsey et al., 2006). Table 16-108 presents mode of access shares over time for the B-G/SR Trails, and also for four additional U.S. trails or groups of trails. All of these data are from surveys of persons on the trail itself (user traffic) and thus likely over-represent the characteristics of longer trips. Note the decline in auto access from 1985 on the B-G/SR Trails. In the previously cross- referenced “Seattle Urban/Suburban Trails” discussion, it is hypothesized that this decline may be the result of expanding alternative trail options, available for recreation and exercise, afforded by a growing trail system. In the “Indiana Trails” case study, data provide trail mode of access from a user-visit perspective for six Indiana trails (see Table 16-138). Not much studied is the obvious relationship between choice of shared-use path access mode and distance from home to a path. On the Monon Trail in Indianapolis, just under 1/2 of trail users reported a 0-1 mile distance to the trail (the next questionnaire choice being 2-4 miles), and just under one-half reported use of NMT modes for trail access, as compared to driving to the trail. The 85th percentile trail user lived 5-8 miles from the trail (Indiana University, 2001).

Path and Trail Purposes of Use. Table 16-109 provides trip purpose information for the five paths or groups of paths among those listed in Table 16-106 that surveyed this information. The Minneapolis-area Hennepin County trails exhibit the highest commute share, at 10 percent, even with the “Other” purpose having been inflated by inclusion of “Multiple [purpose] Responses” (Hennepin County, 2005). The other notable deviation from the typical is seen with the San Francisco East Bay’s Iron Horse Trail, which on its former railroad alignment, passes through or close to both historic-small-town business and modern shopping centers. As covered in Note G of the table, “Retail” and “Restaurant” trips together comprise an unusually high 16 percent of reported trip purposes (East Bay Regional Park District, 1998). The other three trails have more typical primary purpose distributions, insofar as can be seen from the data, joining examples such as the six Indiana trails covered in the “Indiana Trails” case study (see Table 16-138). 16-330 Location/Timing Walk/Run Bicycle Skate Bus/Metro Auto Other B-G/SR Tuesday, 1985 a 11.8% 54.2% n/a n/a 34.0% 0.0% B-G/SR Tuesday, 1990 20.9 58.0 n/a n/a 20.8 0.2 B-G/SR Tuesday, 1995 12.6 62.2 n/a 1.5% b 21.9 1.8 B-G/SR Saturday, 1985 3.3 58.7 n/a n/a 38.0 0.0 B-G/SR Saturday, 1990 13.2 53.5 n/a n/a 32.8 0.6 B-G/SR Saturday, 1995 11.0 52.1 n/a 2.5% b 31.7 1.7 3 Minneapolis area trails 8% 75% 3% n/a 13% n/a 4 Rhode Island trails d c 14.8 25.1 1.2 n/a 58.1 0.8% W&OD Trail, No. VA 15 38 n/a 2% 44 1 3 Houston/Austin trails 43.4 e 33.3 0.5 0.3 21.8 0.6 Notes: a B-G/SR stands for the combined Burke-Gilman/Sammamish River Trails, in northeast urban and suburban greater Seattle. All counts were on or close to May 20 (Moritz, 2005b). Data for 2000 are not shown because of a large non-response to the trail-access question. b On Tuesday: Comprised of 1.0% bike-on-bus, 0.5% bus/foot. On Saturday: 1.9% bike-on- bus, 0.6% bus/foot. In 2000, these proportions had roughly doubled on average, while in- line skating, recorded for the first time, accounted for less than 1/2 of 1% of all access. c See Table 16-106 for descriptions of the last four listed trails. d Short-form on-trail survey results. Results from long-form mail-back survey were similar. e Composed of 24.5% walk and 18.9% run/jog. Sources: Moritz (2005b), Hennepin County (2005), Gonzales et al. (2004), Bowker et al. (2004), Shafer et al. (1999) Table 16-108 Mode-of-Path-Access Proportions for Shared Use Paths, All Facility Users

One other item of special interest that comes from the surveys reported on in Table 16-109 pertains to the three Texas trails. Almost all of the 14 percent three-trail work-based out-and-back trips noted in table Note E took place on Houston’s Buffalo Bayou Trail, representing 42 percent of the users of that one trail. This trail, with employment sites nearby, was found to have become “a very popular midday and after work jogging circuit.” Indeed, 9 percent of the users of this trail were accompanied by business associates (Shafer et al., 1999). Travel purpose distributions can shift over time, as illustrated by time series data for the Seattle- area Burke-Gilman/Sammamish River Trail. On that trail the weekday work/school commute proportion has grown from 10 percent in 1985 to over 30 percent in 2005 (Moritz, 2005a and b). A full discussion of this instance is presented in connection with Table 16-18 in the “Shared Use, Off- Road Paths and Trails” subsection of the “Response by Type of NMT Strategy” section (see “Shared Use Path Implementation”—“Seattle Urban/Suburban Trails”). Path and Trail User Characteristics. Individual trail traffic gender distributions are reasonably consistent with national data for all types of facilities, with only Virginia’s W&OD Trail exhibiting a non-conforming result. In Hennepin County, Minnesota, 62 percent of trail traffic was found to be male, and 37 percent female (Hennepin County, 2005), a distribution one might expect given the bicycle-use dominance of the three Minneapolis area trails studied. The split was closer on the four Rhode Island trails, 56 percent male and 44 percent female overall, with 54 percent female on 16-331 U.S. Location/Description Exercise Recreation Commute Other 3 Hennepin Co. (Minneapolis) trails, urb./sub. 60% 8% a 10% 22% b 4 Rhode Island trails, suburban/towns/rural c 76 42 4 2 W&OD Trail, Northern Virginia, sub./exurban 7 d 84 d 6 3 3 Texas Trails, Houston, Austin, urban/sub. 90 e — 4 6 Iron Horse Trail, S. F. East Bay, exurban — 64 f 1 35 g Notes: Information obtained from intercept surveys and interviews. a Consists of 4% “Enjoy Scenery,” 3% “Socialize,” and 1% “Walk Pet.” b Consists of 6% “Shop/Errands,” 3% “Meet Family/Friends,” 1% “School,” 10% “Other/ Multiple Responses,” and 2% “No Response.” c Multiple responses were allowed in response to the Rhode Island trip purpose question. d The survey combined “Fitness” with “Recreation” (84%). “Training” (7%) has been placed in the “Exercise” column. e “Exercise” and “Recreation” were not explicitly identified. The 90% value entered in the “Exercise” column consists of 76% home-based out-and-back (a.k.a. “loop” or “round”) trips and 14% work-based out-and-back trips. f “Exercise” was not a survey form option. Trips taken for exercise were typically assigned by respondents to the “Recreation” category, but in some cases interviewers determined that “individuals had actual destinations and had chosen to use the trail because of the opportunity for exercise.” g Consists of 9% “Retail,” 8% “Restaurant,” 6% “Friends,” 4% “Park/Recreational Facility,” 3% “Another Town,” 3% “School,” 1% “Other,” and less than 1% “BART” (rail transit). Source: Hennepin County (2005), Gonzales et al. (2004), Bowker et al. (2004), Shafer et al. (1999), East Bay Regional Park District (1998). Table 16-109 Purpose of Use of Shared Use Paths (Weekday and Weekend Combined)

the Blackstone Valley bike path where use for walking dominated. A similar relationship among trails was found on the three Texas facilities, with females at 48 percent on the Shoal Creek Trail where walking dominated, while the three-trail average essentially matched the Hennepin County experience with a 63/37 split of males versus females (Gonzales et al., 2004, Shafer et al., 1999). Both trails in the suburbs of Washington, DC, had fairly even gender distributions. The split was 53/47 for the Capital Crescent Trail in Maryland and 47/53 for the W&OD trail in Virginia, with the tilt toward use by females on the W&OD trail found despite the low 16 percent-walker trail- traffic mix (Maryland-National Capital Park and Planning Commission, 2001, Bowker et al., 2004). Data from the Indiana studies, the only instance encountered of truly surveying users as compared to trail traffic, shows close gender splits on all except the rural, bicycle-dominated Cardinal Greenway rail-trail radiating out from Muncie. (See Table 16-136 in the “Indiana Trails” case study.) Path users include persons of all ages, but results from the two studies that reported age by even increments—the Hennepin County survey and the W&OD trail evaluation—suggest that the most extensive use of trails is made by middle-aged adults in the 45-54 age bracket (more than 1/4 of all users). Adults in the 35-44 age bracket are next most prevalent, followed by adults in the 25 to 34 age bracket (Hennepin County, 2005, Bowker et al., 2004). Different age-bracket survey specifications make it difficult to generalize, but it appears that seniors—surveyor identified or reporting age 65 and above—compose only 6 to 15 percent of trail traffic on the Hennepin County, Rhode Island, Capital Crescent, W&OD, Iron Horse, and Indiana trails listed in Table 16-106 (or covered in Table 16-137) (Hennepin County, 2005, Gonzales et al., 2004, Maryland-National Capital Park and Planning Commission, 2001, Bowker et al., 2004, East Bay Regional Park District, 1998, Indiana University, 2001). The corresponding average is about 8 percent seniors. Children and adolescents are even less consistently identified, if covered at all. Two trails used a “less than 15” definition for the younger set and two trails included children and most or all teenagers. Into the “less than 15” category, the Rhode Island study placed 19 percent of trail traf- fic and the Capital Crescent survey placed 8 percent (Gonzales et al., 2004, Maryland-National Capital Park and Planning Commission, 2001). Into the broader category, the Hennepin County survey placed 3 percent and the Iron Horse Trail study placed about 20 percent (Hennepin County, 2005, East Bay Regional Park District, 1998). The average for younger users was thus about 12 per- cent of total traffic. The relative use of paths by children and young adults appears to be less than the relative use of the walking and bicycling modes nationwide by the same age groups. The oppo- site appears to hold for those of middle age.68 Much as middle-aged groups are heavily represented among path users, with a tilt toward older middle-aged users, so are middle-income groups with a tilt toward upper incomes. The proportions of Indiana trail survey respondents reporting household incomes between $40,000 and $80,000 annually in year 2000 dollars were in a tight range between 45 and 51 percent for 6 urban, suburban, and semi- rural trails. Higher income respondents comprised 16 to 21 percent except for 33 percent in Indianapolis, and lower income respondents made up 33 to 39 percent of respondents except for 22 per- cent in Indianapolis (Indiana University, 2001). (See Table 16-137 in the “Indiana Trails” case study.) Fewer middle-income and more upper-income survey respondents were encountered on 3 Texas trails, with 22 percent reporting annual incomes below $40,000 annually, 33 percent between 16-332 68 This and similar comparisons with U.S. national data to follow are drawn on the basis of national perspec- tives offered in the “Underlying Traveler Response Factors” section. See the applicable discussions and tables within the “User Factors” subsection.

$40,000 and $80,000, and about 45 percent reporting higher incomes (Shafer et al., 1999). Still higher incomes were reported by respondents to the W&OD Trail survey in upper-income Northern Virginia suburbs of Washington, DC. There the survey sample exhibited an average household annual income of $98,600 and almost 25 percent reported an income in excess of $120,000 (Bowker et al., 2004). Clearly path user makeup reflects the communities through which a trail passes. It also appears that trail users tend to fall more in the middle-income and/or upper-income cate- gories than U.S. pedestrians and bicyclists in general. Paths appear to attract educated users, although the proportion with college degrees varies markedly according to path location. Indiana trail surveys in Greenfield and Portage found about 1/3 of trail users to be college graduates, while trails in the other 4 Indiana areas studied had over 1/2 of users holding college degrees. Almost 80 percent of Monon Trail users in Indianapolis reported college degrees (Indiana University, 2001). Responders to the mail-back surveys on the 3 studied Texas trails had college degrees in 85 percent of all cases, over 1/2 accompanied by advanced degrees (Shafer et al., 1999). These findings are consistent with national-level research results indicating that propensity to walk, especially for exercise and recreational purposes, is highly correlated with educational attainment (Agrawal and Schimek, 2007). National bicycling versus education relationships are not as well studied. All path studies that have investigated path user racial composition and ethnicity report low pro- portions of minorities on the paths, but the Hennepin County study makes special note of it. Their surveys—backed up by surveyor observations—found non-white persons to represent less that 20 percent of Midtown Corridor trail users, even though people of color are in the majority in adjoining Minneapolis neighborhoods. Overall responses for the 3 trails surveyed indicated a dis- tribution of 92 percent white, 5 percent non-white, and 3 percent no race identified (Hennepin County, 2005). Survey responses for the W&OD trail in Northern Virginia came 85 percent from whites, 2 percent from blacks, 4 percent from Hispanics, 6 percent from persons of Asian ethnic- ity, and 1 percent from Native Americans, with the question unanswered by 2 percent (Bowker et al., 2004). Corresponding data from the 3 Houston and Austin trails was 87 percent white, 3 per- cent black, 6 percent Hispanic, and 3 percent other, on average (Shafer et al., 1999). Surveyor-recorded racial makeup information on Indiana Trails ranged from 86 percent white, 10 per- cent black, and 4 percent Hispanic in Ft. Wayne to 98 percent white, 1 percent black, and 1 percent Hispanic in Greenfield. The potential for survey bias was highlighted by survey responses that came back 94 and 100 percent white, respectively, for the trails in these two cities. Similar discrepancies were noted in all areas (see Table 16-137 in the “Indiana Trails” case study), despite high survey mail-back rates (Indiana University, 2001). There are a number of possible reasons besides path use propensity differentials for the low numbers of minorities reported on paths, as well as for other phenomena such as somewhat low usage by lower-income persons. Possible explanations include not only survey response biases but also lesser mileage of paths in crowded urban environments. Travel Behavior Shifts Improvements to pedestrian and bicycle infrastructure have been shown in preceding sections to attract additional facility usage in most cases. When pedestrians or bicyclists are attracted (or repelled) by improvements (or degradations), shifts among travel modes take place along with some occurrences of new or redirected trips (or trips forgone). Two more or less fully developed empiri- cal investigations of such effects have been encountered, both from cities in Australia. Both studies focus on peak-period weekday travel. They illustrate phenomena and relationships not readily dis- cernible from “before and after” volumes and modal percentages, as instructive as those may be. 16-333

Prior or Alternative Modes of New Facility Users Radial Off-Road Paths in Melbourne. The Melbourne evaluation of substitute modes of path users provides the more straightforward example of travel behavior shift research findings, because the results are not complicated by multi-mode trip recombinations. Its reliance on path bicyclist percep- tions of travel alternatives makes it perhaps less robust, however, than if actual prior-activity data had been practicable to obtain. The survey utilized 12 intercept locations with count-based controls. An impressive 77 percent survey response rate was achieved for the one in seven sample of bicyclists using Melbourne off-road paths radial to the CBD. The path orientation and the 7:00 to 10:00 AM Monday timing of the survey resulted in a low proportion of recreational trips. Trip purposes were determined to be work (85 percent), university (3 percent), school (2 percent), recreation, (8 percent), and other (2 percent) (Rose, 2007). The work trip dominance must be considered in interpreting the results. Respondents were asked, in the self-administered questionnaire, about how their travel behavior would be different if the path they were riding on had not been built. Table 16-110 summarizes the responses. The very small proportion who indicated they would not make the trip without the bicy- cle facility (1 percent), and the modest proportion indicating they would alter their destination (4 per- cent), represent responses constrained by the fact that work and school trips (90 percent of the total) cannot readily be adjusted in that manner. Nevertheless, the reporting that some of these responses would occur is highly instructive. The researchers postulate that the size of the proportion respond- ing that they would continue to bicycle without destination change (75 percent) might be smaller if the responders were fully aware of the limitations of alternative routes (Rose, 2007). Nevertheless, a 20 percent mode shift in itself is not insubstantial. The implication is that the bicycle facilities have led to 25 percent more peak period cycling in Melbourne’s radial corridors (or 26 percent accounting for new trips) than would be occurring without the paths. 16-334 Travel Selection Had Bicycle Facility Not Been Built Percent Combined Categories Would still cycle, changing route only 75% Continue to bicycle Would still cycle, but to a different destination 4 79% Would change mode to car driver 7 Change travel mode Would change mode to car passenger 1 20% Would change mode to public transportation 12 Forgo trip Would not make the trip 1 1% Source: Rose (2007). Table 16-110 Melbourne Cyclist Responses on Their Travel Behavior Had Their Path Not Been Built Goodwill Pedestrian and Bicycle Bridge in Brisbane. The Brisbane example is more unique, but it importantly demonstrates the extent and complexities of travel shifts that can occur under specific sets of circumstances. The preceding “Facility Usage and User Characteristics” discussion notes that the CBD locations of parking facilities and transit terminals are a major influence on the patterns and vol- umes of walking trips within a downtown. The Goodwill Bridge experience demonstrates that placing a major new NMT link-up in amongst parking facilities, transit terminals, and the CBD destinations they could or do serve can produce a major perturbation of the pedestrian (and bicycle) trip patterns and volumes. The Goodwill Bridge experience is first introduced in this chapter under “Pedestrian/Bicycle Systems and Interconnections” (see “River Bridges and Other Linkages”). A full description is pro- vided there. In brief, the Goodwill pedestrian/bicycle bridge crossing of the Brisbane River was

opened in October, 2001. It for the first time provided an NMT connection between the south end of the CBD—and adjoining university—to South Bank residential and mixed-use areas, cheaper automobile parking, an additional south-of-river commuter railroad station, and key express bus stops. The post-graduate student research that reported on the induced travel behavior shifts explicitly obtained information on bridge user travel behavior before and after the new facility was in place, albeit relying on survey respondent recall of prior travel choices (Abrahams, 2002). It also obtained other insightful bridge user information already reported on in the earlier discussion. The Goodwill Bridge analysis did not investigate travel behavior shifts other than changes in mode. Possible attraction of new trips and changes in trip destination choice were not inquired about in the sur- vey. It appeared that persons whose circumstances in the prior condition did not fit with the options pre- sented in the survey’s previous travel mode question, tourists included, tended to answer “other.” The survey anticipated that the prior trips of some bridge users would involve more than one mode, and allowed survey respondents to give multiple answers concerning prior modes. The analysis then used the responses, and also information solicited on prior use of the upstream Victoria Bridge, to divide users who had originally walked or bicycled and who had retained that same NMT mode (shifting routes to the Goodwill Bridge) from those who had changed their travel mode in response to the new bridge. The first section of Table 16-111 shows the results of that analysis for commuters. The second Table 16-111 note does the same for non-commuters. Overall, 58 percent of Goodwill Bridge users were continuing to use their previous NMT mode, including 40 percent previously crossing on the Victoria Bridge, and 42 percent were persons who had changed mode. Among user subgroups, 52 percent of commuter pedestrians walking the bridge had changed modes, 19 percent of commuter bicyclists were likewise mode-changers, 34 percent of non-commuter walkers were mode-changers, and 26 percent of non-commuter cyclists had changed their NMT mode compared to their pre-bridge trips (Abrahams, 2002).69 Mode changers who reported only one prior mode were separated out for further analysis, though only in the case of commuters, providing the single-prior-mode information for commuter mode changers used to develop the second section of Table 16-111. Note that the table is constructed so that mode-changer subcategories nominally total to 100 percent, requiring inclusion of an entry for mode changers whose prior trip had involved more than one mode. The prior means of travel for Goodwill Bridge commuters involved more than one mode for 17 percent of walkers and the same for cyclists. The number of two-or-more mode trips did not increase much when bridge use was introduced in the case of bicycle commuters, but it tripled in the case of pedestrian commuters.70 The previously cross-referenced “River Bridges and Other Linkages” discussion provides addi- tional detail and context concerning the multimodal activity observed after bridge opening. 16-335 69 Goodwill Bridge trip diversion and mode shift estimates presented here reflect adjustment by the Handbook authors for differential pedestrian versus bicyclist survey response rates. The researcher reported survey response rates of 25 percent for walkers and 50 percent for cyclists (Abrahams, 2002). Despite the derivation of these rates without benefit of survey control counts, when these percentages are used to normalize the split between walkers and cyclists, the results are highly reasonable. (The result is 82 percent walkers and 18 percent cyclists, compared to weekday all-day cyclist percentages from counts taken 3 months previous— shown in Table 16-23—that range from 16 to 22 percent bike and have a weighted average of 19 percent.) Consequently, the Handbook authors elected to normalize the pedestrian and bicyclist results on the basis of the response rates given, for purposes of reporting trip diversions and mode shifts. 70 The determination as to what constituted a separate mode requiring identification on the survey response was effectively left up to each responder. This approach probably led to some fuzziness in the handling of walk trips linked to motorized modes, which may in turn have affected the reliability with which the gain in trips of more than one mode can be calculated.

Table 16-112 augments Table 16-111 by displaying and giving percentages for the full array of prior modes reported by Goodwill Bridge Mode Changers. As can be seen in both tables, commuter rail, bus, and auto are together prominent in the prior mode arrays. Because of the gain in trips of more than one mode, it is likely that a number of the mode changes affected only the downtown and cross-river end of the trips involved. For example, a commuter who had driven into the Brisbane CBD might now drive only to less expensive parking in the South Bank, and walk the remainder of the distance via the Goodwill Bridge. Given the peculiarities of train station placement, the rough equivalent could be happening for trips via rail as well (Abrahams, 2002). 16-336 Percent of Commuters Within Category Mode Retention Versus Change Bridge Commuter Walkers Bridge Commuter Cyclists Prior and current mode same 48% 81% Prior mode different (Mode Changer) 52 19 Percent of Mode Changers Within Category Prior Mode of Mode Changers Bridge Commuter Walkers Bridge Commuter Cyclists Train only 23% 17% Bus only 26 33 Ferry only 6 0 Taxi only 0 0 Car only 20 6 Walk only — 22 Bike only 3 — Other (single mode) 6 6 More than one mode 17 17 Notes: Adult and late-teen school trips were designated as commuter trips. (Bridge users under 18 were not surveyed.) For Goodwill Bridge non-commuter walkers, 66% had the same prior and current mode and 34% were mode changers. For non-commuter cyclists, 74% had the same prior and current mode and 26% were mode changers. Source: Abrahams (2002), with elaboration by the Handbook authors. Table 16-111 Goodwill Bridge Mode Change Record for Commuter Walkers and Cyclists Prior Modes Reported Commuter Walkers Commuter Cyclists Non-Commuters Train 27% 22% 7% Bus 42 33 22 Ferry 7 11 3 Taxi 2 0 3 Car 31 11 45 Walk — 33 6 Bike 7 — 3 Other 7 11 16 Note: See Table 16-111 (including notes) for proportions of surveyed bridge users who had retained their same mode and were thus not classified as “Mode Changers.” Source: Abrahams (2002), with elaboration by the Handbook authors. Table 16-112 Prior Mode Use Reported by Mode Changers (Multiple Responses Allowed)

Mode Shares “Before and After” Mode shares obtained before and after infrastructure improvements provide less precise insights on shifts than actual prior mode data, because they do not give explicit information on what users of a new mode were doing previously, except possibly by inference. Such data are nevertheless quite useful, and in some instances increases in use of particular modes together with decreases for others provide a basis for judging what users attracted to newly enhanced modes were doing pre- viously. Information of this type is provided, if available, in the “Response by Type of NMT Strategy” section. A fully comprehensive presentation of before and after mode shares is exemplified by Table 16-6, in the “Pedestrian Zones, Malls, and Skywalks” subsection (see “Pedestrian Zones” under “Pedestrian Zones and Malls”). This example presents study area and overall CBD mode shares for all primary modes before and after Downtown Crossing pedestrian zone implementation in Boston. Another such tabulation, this one illustrating mode shifts over time in response to a city-wide bicycle and pedes- trian facility and transit enhancement program for the entire city of Boulder, Colorado, is found in Table 16-44 of the subsection “NMT Policies and Programs” (see “New World Program Examples”— “Denver, Colorado”). Given that individual pedestrian and bicycle programs are often too incremen- tal to produce discernible areawide changes in other travel modes, these types of before and after mode share presentations are somewhat rare, making travel behavior shift evaluations of the types accomplished in Melbourne and Brisbane all the more valuable. Time to Establish Facility Use Transportation options that have been in place for some time reflect usage that is “stabilized,” “matured,” or “established.” Similarly, usage forecasts are typically prepared with travel demand estimation models and techniques keyed to travel behavior that reflects established travel choices and patterns. Experience from motorized transportation shows, however, that usage of a new facil- ity or service will have to build up to stabilized levels of usage over time. Prospective users have to find out about the new travel option and its advantages to their particular situation. The oppor- tune time for initiating use may not come at once. The result is lower usage during the initial months of option availability (Pratt and MWCOG, 1987). The same need for time to establish use applies to pedestrian and bicycle facilities. Lower initial use may be incorrectly interpreted as a sign of poor investment or design, or failure. Motorized Transportation and NMT Experience Compared Documented time-series volume-of-usage data for new facilities is scarce, so it is useful to examine what is available for NMT in the context of motorized transportation findings. Table 16-113 summa- rizes examples of motorized transportation experience, with emphasis on movement of people as con- trasted to vehicular traffic. The last three rows present available NMT experience. The experience is expressed, in addition to “Months to Stabilize,” in terms of the percentage of stabilized usage that is observed in the first month or so after opening, the first year, and the second year. Note that “estab- lished,” “stabilized,” or “matured” should not in this context be taken to necessarily infer a flat usage plateau with long-term permanence. It may be the reaching of a steady growth that parallels secular trends, such as population growth. The established pattern may also be disrupted by events such as fuel price changes, facility expansion, or competition from a new facility. The “Months to Stabilize” figures in Table 16-113 are developed based on visual examination of trend data and are identified on the basis of the first substantial period of stabilized ridership if there is more than one such period. 16-337

The motorized transportation and NMT facility usage outcomes presented in Table 16-113 are dis- played in order of increasing numbers of months required after opening for stabilization of usage to occur. The ordering is very instructive. First are averages for two commuter rail line implementations and for three new sections of Metro (heavy rail transit). Both systems are in the metropolitan Washington, DC, area. Ridership on these urban rail lines stabilized in about 2 years or less—consid- erably less in the commuter rail examples (Pratt and MWCOG, 1987, Parsons Brinckerhoff et al., 1994). In the case of new urban rail lines, many riders have already been using public transportation, and for them the switch to rail is highly sensitive to travel time and convenience. Cost differentials and socio- economic factors play just a minor role (Pratt, 1971). Those shifting from auto commuting are making a more substantial change, but have the added impetus of parking cost savings at the destination, con- sequential in the Washington, DC, examples and most areas with rail systems. Attitudes toward new rail systems tend to be positive or neutral and their use is not generally looked upon askance. The shift to urban rail use is thus comparatively fluid, reflected in the relatively short 6- to 26-month usage mat- uration times observed on the five individual lines. 16-338 Service or Facility Use as a Percentage of Stabilized Use Months to StabilizeInitial Use 1st Year Use 2nd Year Use VRE, Northern Virginia, Manassas and Fredericksburg commuter rail lines 56% 80% 106% 8 Washington Metrorail Yellow and Red (NW mid and outer segments) Lines 42% 66% 94% 20 Prince William County, VA, OmniLink demand responsive bus lines (5 lines) 17% 51% 89% 26 1960s Los Angeles and Long Island jobs- access bus routes (1 route each) 28% 54% 84% 27 Houston North, Katy, Northwest, and Southwest HOV lanes 30% 60% 80% 48 Melbourne St. Kilda Road bike lanes n/a 16% 19% 84 Seattle Burke-Gilman/Sammamish River Trails (Tuesday data) n/a 37% n/a 90 Seattle Burke-Gilman/Sammamish River Trails (Saturday data) n/a 78% n/a 90 Notes: Available data for Washington Metrorail Orange Line (east), one additional Los Angeles jobs-access bus route, one additional Long Island jobs-access bus route, and four additional Houston and Dallas HOV lanes are not included either because of almost immediate stabilization (2 cases) or because of lack of stabilization during the study period (5 cases). All percentages and numbers of months (except Washington Metrorail) are approximations by the Handbook authors based on graphed (transit) or tabulated (NMT) time series data. Means for each type of motorized transportation are simple averages. The OmniLink values are based on the total ridership on five routes, two of which did not open until the 5th month. The 1960s jobs access bus route percentages are based not on use per se, but instead on the inverse of deficit per passenger (as a surrogate). Sources: Washington Metrorail – Pratt and MWCOG (1987), Virginia Railway Express (VRE) – Parsons Brinckerhoff et al. (1994), and TCRP Report 95 figures and tables as follows: OmniLink (Prince William County, VA) – Chapter 6, “Demand Responsive/ADA,” Figure 6-5; 1960s jobs access bus routes – Chapter 10, “Bus Routing and Coverage,” Figure 10-1; Houston HOV lanes – Chapter 2, “HOV Facilities,” Figure 2-4; Melbourne St. Kilda bike lane – Table 16-114 (below); Seattle Burke-Gilman/Sammamish River Trails – Table 16-17 (see “Response by Type of NMT Strategy” — “Shared Use, Off-Road Paths and Trails” — “Shared Use Path Implementation” — “Seattle Urban/Suburban Trails”). Table 16-113 Motorized and NMT Facility Usage Maturation Experience

Next in Table 16-113 are a demand responsive bus system in outer Washington suburbs of Northern Virginia and bus routes designed to connect outlying jobs with urban pockets of unemployment in Los Angeles and on Long Island. In these examples, there were no previous viable transit connections, so all riders had to have made substantial travel adjustments. In addition, users of the jobs-access bus routes had to secure employment at the newly accessible locations. These various accommodations lead to less fluid travel changes, and usage stabilization on the three systems/lines assessed required 1-1/2 to 3 years. (See “Service Development and Time Lag” in the “Related Information and Impacts” sec- tion of Chapter 6 and “Traveler Response Time Lag” in the corresponding section of Chapter 10). The final motorized transportation entry in Table 16-113 is for usage by bus riders and carpool pas- sengers of HOV lanes in the Houston area. Here much more was involved than simply travel mode shifts alone. For potential bus passengers to even be in a position to consider use, bus routes via the HOV lanes had to be established, involving time-consuming steps for the transit agency. Carpool and vanpool users of the HOV lanes had to make ridesharing arrangements in order to affect mode shift decisions. HOV lane implementations thus offer an example of facility usage growth in the presence of barriers that must be overcome before shifts can occur. The individual HOV lanes studied had widely varying rates of usage maturation ranging from 2 to over 6 years. Although the initial use per- centage averages are not out of line with the urban bus averages included in Table 16-113, the aver- age time span before stabilization for the four HOV facilities covered was higher, on the order of 4 full years. (See “Time to Establish Ridership and Use” in the “Related Information and Impacts” section of Chapter 2, “HOV Facilities,” for further information). The small amount of data available for NMT facilities indicates that movement toward a stabilized usage level is even slower than the average for HOV lanes. The weekday time series data for the St. Kilda Road Bike Lanes in Melbourne, Australia, and the Burke-Gilman/Sammamish River Trails in Seattle, Washington, suggest a very gradual response to bicycle or bicycle and pedestrian facility availability. The better part of a decade is apparently required to reach stabilized usage levels. Barriers which must be overcome to make use of NMT facilities for utilitarian travel, most common on week- days, have been discussed in the “Underlying Traveler Response Factors” section. They range from potential user safety concerns to workplace dress codes (Goldsmith, 1992). Confounding externalities, such as facility extensions and end-to-end conjoining of the two pathways in the case of the Seattle trails, may also be lengthening the time for usage stabilization. In any case, the choice to use NMT facilities is clearly a “sticky” one, not fluid, typically taking substantial elapsed time. The weekend data from the Seattle trails, also entered in Table 16-113, hint that usage buildup for week- end walking and bicycling may be less gradual, although the observed 7-1/2 year time to reach fully stabilized usage is the same for the observed Saturdays as for the observed Tuesdays. Weekend usage is more heavily oriented toward exercise and recreation. A significant portion of the walking and bicy- cling involved may simply reflect a shift in exercise or recreation venue, likely combined with increased activity, with smaller components representing outright initiation of exercise or change in mode of exer- cise or active recreation. Further examination of the St. Kilda Road Bike Lanes and the Burke- Gilman/Sammamish River Trails experiences is provided or cross-referenced below. Melbourne St. Kilda Road Bike Lanes An AM peak hour count of St. Kilda Road bicyclists has been made almost every year since 1 year prior to the 1993 opening of bicycle lanes on this major traffic route oriented to Melbourne City. Bicyclist injuries per year have also been tracked for St. Kilda Road. Table 16-114 provides these data and also a “hazard ratio” calculation carried out as described in Note A of the table. The bicyclist count data are the basis for the St. Kilda Road Bike Lane facility usage maturation entry in Table 16-113. 16-339

AM peak bicycle volumes grew nearly sevenfold in the first 12 years of St. Kilda Road bicycle lane operation, from 66 in 1993 to 459 in 2005. The process has been very gradual, however, as discussed already. Possible reasons for the substantial jump in usage between 1994 and 1996 are not provided or speculated on in the available documentation. Of special interest is the stabilization of reported bicy- clist injuries even as bike lane usage continued to dramatically expand. This beneficial effect has been attributed to the bicycle lane installation. A review of experience with the Australian state of Queensland’s Main Roads cycling policy concludes, “The evidence [that in Table 16-114] broadly sug- gests that creation of a complete cycle network through implementation of the Main Roads cycling pol- icy will have a positive impact on cycling mode share and safety” (Davies, 2007). The hazard ratio in the final column of Table 16-114 is designed to better illustrate the safety trend. After an apparent doubling of hazard in the first year of bike lane installation, the hazard per bicyclist quickly dropped back to the before-bike-lanes level. Then, 5 years out from opening of the lanes, it dropped further over time as lane usage increased. These data not only seem to demonstrate the long- term safety benefit of the bike lanes, but also can be interpreted to support the “safety in numbers” hypothesis examined in the upcoming “Safety Information and Comparisons” subsection. Seattle Burke-Gilman/Sammamish River Trails The Seattle Burke-Gilman/Sammamish River Trails, like the St. Kilda Road bike lanes, did not reach an apparently stabilized level of usage until 7 or so years after facility implementation. Usage evolution has been quite complex and has been made more difficult to assess by a number of con- founding NMT facility changes. The Burke-Gilman Trail component was extended subsequent to 16-340 Year Injuries per Year Cyclists in AM Peak Hazard Ratio a 1991 (before) 4 n/a — 1992 (before) 3 42 1.0 1993 (lanes opened) 11 66 2.3 1994 5 76 0.9 1995 7 n/a — 1996 10 130 1.1 1997 9 154 0.8 1998 10 160 0.9 1999 11 n/a — Early 2000 17 416 0.6 Late 2000 11 382 0.4 2002 19 318 0.8 2003 7 511 0.2 2004/2005 b 7 459 0.2 Note: a This hazard ratio, computed by the Handbook authors from data in the preceding columns, is reported injuries per year divided by the counted number of bicyclists in the AM peak and normalized to a value of 1.0 for the 1992 “before-bike-lanes” condition. b 2004 injuries (cyclists n/a) and 2005 AM peak cyclists (injuries n/a). Source: Davies (2007) with elaboration (see Note A). Table 16-114 Melbourne St. Kilda Road Bike Lane Volumes and Injury Rates over Time

the initial 1980 count. Five years following the point selected as representing the reaching of sta- bilized usage for purposes of Table 16-113, the Burke-Gilman and Sammamish River Trails were joined. Usage recorded in the following 5-year count was up markedly, higher than seen 5 and 10 years afterward. Early use of the trails was 90 percent for recreation and exercise, but the work/school commute and other utilitarian purpose proportions have grown, sharply at first, and steadily thereafter. These and other usage observations based on time-series data are detailed and inter- preted in the “Response by Type of NMT Strategy” section (see “Shared Use, Off-Road Paths and Trails”—“Shared Use Path Implementation”—“Seattle Urban/Suburban Trails”). Safety Information and Comparisons There is a substantial body of pedestrian- and bicycle-oriented references, manuals, toolkits, instruc- tional materials, and research on causes of traffic crashes and preventative or remedial facility designs and modifications to employ in response (Nabors, et al., 2007, Nabors, et al., 2008, Harkey and Zegeer, 2004, Campbell et al., 2004). Safety issues are not a central focus of this “Traveler Response to Transportation System Changes” Handbook. Nevertheless, it is difficult to avoid the subject when addressing walking and cycling, since many actual or potential non-motorized travelers cite perceived safety as a major factor in their active transportation decisionmaking. In addition there is concern that crash outcomes may detract from the health benefits of walking and cycling. Accordingly, while this “Pedestrian and Bicycle Facilities” chapter hews to the “Traveler Response” Handbook practice of leaving facility design analyses and recommendations to other publications, this subsection does highlight safety information bearing on the overall risk of walk- ing and cycling and the suitability of pedestrian and bicycle provisions. Selected NMT safety indi- cators and relationships are provided. Available comparisons with other developed countries and among facility types that contribute to the total picture of program and facility effectiveness are presented. Discussion of what is known about the effects of perceived safety on NMT travel choices is, however, found within the earlier “Underlying Traveler Response Factors” section (see “Other Factors and Factor Combinations”—“Security and Safety”). Effects on NMT travel choices of per- sonal safety concerns about physical attacks and crime are addressed in that same subsection. Pedestrian and Bicyclist Safety Highlights Pedestrian and cyclist crash, injury, and mortality rates are at levels that may be viewed as a glass half full or a glass half empty. The rates are low enough that they should not be used to dissuade people from walking or cycling for active transportation or exercise and enjoyment. They are high enough to provide strong impetus for safety improvements. The crash data summarizations pro- vided here are intended to help in scaling the magnitude and nature of NMT traffic safety risks. Crash Rates from Various Perspectives. Almost all comprehensive U.S. crash rate statistics per- tain only to crashes that in some way involve a motor vehicle operating on a public highway. Unless otherwise identified, all statistics presented here fall into that category. Pedestrian and bicy- cle crashes that do not happen on a public street or highway, or do not involve conflict between the pedestrian or cyclist and a motor vehicle, are rarely documented in conventional transporta- tion crash statistics. U.S. motor vehicle crashes in 2004 included 12.7 motor vehicle occupant fatalities per 100,000 pop- ulation, 1.6 pedestrian fatalities per 100,000, and 0.3 bicyclist fatalities per 100,000. These are, how- ever, measures of “population burden” rather than risk. To produce an estimate of risk requires a 16-341

measure of exposure. Traffic exposure measures that have been used include distance traveled, time duration of travel, and number of trips (Beck, Dellinger, and O’Neil, 2007). Distance traveled is the traditional traffic engineer’s measure of crash exposure, although not necessar- ily the most appropriate for NMT. Since NMT trips are relatively slow and short, particularly in the case of walk trips, use of this measure produces the highest NMT crash rate estimates relative to other travel modes. Distance-based pedestrian fatality rates have been estimated, with 2001 data, at 140 fatal- ities per billion kilometers walked (22 per 100 million miles). This is over 23 times the distance-based 6 fatalities per billion kilometers rate for auto occupants. Corresponding rates for bicyclists are 72 fatal- ities per billion kilometers cycled (12 per 100 million miles), or 12 times the rate for auto occupants (Pucher and Dijkstra, 2003). Use of time of travel as the exposure measure is one option, little used in the United States, for addressing the inability of a distance-traveled exposure measure to account for the large speed dif- ferentials between NMT and motorized travel. One U.S. analysis has been prepared using time of travel, drawing on 2001 Fatality Analysis Reporting System (FARS) and National Household Travel Survey (NHTS) fatal crash and travel data. It found a walking risk of 4.94 pedestrian deaths per million hours of walking as compared to 2.90 auto occupant deaths per million hours of auto travel, indicating a risk of walking 1.7 times that of traveling by auto. Bicycling fatality rates were not examined. The research also, as one illustration of application of the technique, found—in terms of fatalities per unit of time—that walking is as safe as driving or riding in an auto during the daytime hours of 7:00 AM to 5:00 PM. The other side of the coin is, however, that walking was measured to be 5 to 8 times as risky during other hours as traveling by auto. The researcher notes that since the trans- portation infrastructure does not much change according to time of day, the explanation must lie in other factors such as light conditions, pedestrian and auto occupant behavior including inebri- ation, traffic flows, enforcement levels, and availability of emergency services (Chu, 2003). Use of number of person trips as the exposure measure is a more common approach to NMT crash rate computations and comparisons designed to compensate for NMT versus motorized travel speed differentials. A CDC research effort utilized 1999 through 2003 FARS and National Automotive Sampling System—General Estimates System (GES) crash statistics together with per- son trip estimates based on the 2001 NHTS to produce the annualized fatal and non-fatal injury rates summarized in Table 16-115 for various travel modes (Beck, Dellinger, and O’Neil, 2007). Table 16-115 illustrates that on this basis making a trip by walking or cycling is 1.5 and 2.3 times as likely to result in a fatality, respectively, as taking a trip in an auto or other private passenger vehicle. Walking or bicycling is 39 and 25 times safer compared to riding a motorcycle, respectively, in terms of fatalities per trip. 16-342

The pedestrian and bicyclist crash rates per trip were also analyzed by gender and age group. Roughly speaking, males are on the order of twice as prone to crashes as females. The fatal injury rate for male cyclists stands out as being almost four times the equivalent rate for female cyclists (27.6 versus 7.2 per 100,000,000 trips). Walking and bicycling fatal-injury rates increase steadily with age. Combined fatal and non-fatal injury rates, however, peak at ages 15-24 and then decline with age. The increasing fatality rate in the face of an overall decline in the injury crash rate after age 24 is consistent with vehicle crash studies showing that fatality increase with age is primarily attributable to the greater fragility of older persons rather than decline in crash avoidance (Beck, Dellinger, and O’Neil, 2007). Crash estimates roughly comparable to those presented for the general population in Table 16-115 have been developed by a select TRB committee for student travel during school hours, accepted as an approximation of travel to and from school. FARS and GES fatal and non-fatal injury data from 1991 through 1999 were combined to average out infrequently occurring incidents, and rates were determined on the basis of trip data from the 1995 NPTS. The estimates are “confounded by inconsistent and incomplete data,” but deemed sufficient to make gross comparisons of relative risks among school travel modes. One notable issue is that while bus access pedestrian crashes (such as when crossing the street to get on a bus) are counted as bus crashes for school-bus trips, that is not the case for “other bus” (mainly transit bus) trips (Committee on School Transportation Safety, 2002). The committee’s findings are summarized in Table 16-116. Walking or cycling to school, in terms of fatalities per trip, involves more risk than being driven by an adult but is safer than traveling in a vehicle driven by a teenager. Student bicyclists are more prone to crashes than student pedestrians. 16-343 Fatal Injuries Non-Fatal Injuries Total Injuries Mode of Travel Rate per Exposure Indexed to Auto Rate Rate per Exposure Indexed to Auto Rate Rate per Exposure Indexed to Auto Rate Passenger Vehicle 9.2 1.0 803.0 1.0 812.2 1.0 Motorcycle 536.6 58.3 10,336.6 12.9 10,873.2 13.4 Pedestrian 13.7 1.5 215.5 0.3 229.2 0.3 Bicyclist 21.0 2.3 1,461.2 1.8 1,482.2 1.8 Bus 0.4 <0.05 160.8 0.2 161.2 0.2 Other Vehicle 28.4 3.1 1,020.6 1.3 1,049.0 1.3 Overall Rate 10.4 – 754.6 – 765.0 – Notes: Unit of exposure is person trips in hundreds of millions (100,000,000 trips). Based on 1999 through 2003 FARS and GES crash statistics together with trip estimates derived from expanding the 2001 NHTS. Includes only pedestrian and bicyclist crashes involving a motor vehicle operating on (or entering/leaving) public streets or highways. Sources: Beck, Dellinger, and O’Neil (2007), with injury totals and indexing by the Handbook authors. Table 16-115 Annualized Injury Rates per 100 Million Person Trips by Mode of Travel

All of these crash statistics involve estimation of the exposure measure. None are as accurate for pedestrians and cyclists as for motorized forms of transportation. Studies of emergency room records suggest that there is significant undercounting in official crash statistics of pedestrian and bicycle injuries. One study found that even in the category of bicycle/motor-vehicle crashes only two-thirds of events serious enough to entail emergency room treatment were recorded in State motor-vehicle crash records. Another factor contributing to the undercounting of pedestrian and bicycle injuries is that an estimated 64 percent of pedestrian injuries and 70 percent of bicycle injuries did not involve a motor vehicle. Moreover, some 53 percent of pedestrian injuries and 31 percent of bicycle injuries occur in non-roadway locations including sidewalks, parking lots, and off-road paths (Turner et al., 2006). At least the physics of velocity and mass suggest that the injuries not involving a motor vehi- cle are probably less likely to be in the serious injury or fatal category. An analytical problem that is especially important in the case of pedestrians is hinted at in the report on risks of travel to school referred to above. Public bus access pedestrian crashes, even when crossing the street to get on a public (non-school) bus, are counted as pedestrian crashes for school children (Committee on School Transportation Safety, 2002). The fact is, the same account- ing applies in the case of most adult crash statistics, and not just in the case of public buses, but for all forms of urban rail transit as well. At the same time, the trips involved are identified as public transit trips. This means that transit access walk trips, and also transit access bike trips, are not in the denominator (exposure measure) but are in the numerator (number of crashes) of NMT crash rate statistics. Indeed, a comparable problem occurs in the case of walk trips associated with driving, such as walking from an off-site parking facility to one’s place of work. The result is that most NMT crash rates must be to some unknown but significant degree overstated, a rate over- statement certain to be magnified in cities with substantial transit usage and off-site parking, each 16-344 Fatal Injuries Non-Fatal Injuries Total Injuries Mode of Travel Rate per Exposure Indexed to Adult-Driver Auto Rate Rate per Exposure Indexed to Adult-Driver Auto Rate Rate per Exposure a Indexed to Adult-Driver Auto Rate Passenger Ve- hicle driven by: Adult 1.6 1.0 490 1.0 490 1.0 Teenager 13.2 8.2 2,300 4.7 2,310 4.7 Pedestrian 4.6 2.9 310 0.6 310 0.6 Bicyclist 9.6 6.0 1,610 3.3 1,620 3.3 School Bus 0.3 0.2 100 0.2 100 0.2 Other Bus 0.1 0.1 120 0.2 120 0.2 Overall Rate 3.5 – 650 – 650 – Notes: Unit of exposure is person trips in hundreds of millions (100,000,000 trips) by students during normal school hours. Based on 1991 through 1999 FARS and GES crash statistics and 1995 NPTS data. Includes only pedestrian and bicyclist crashes involving a motor vehicle operating on (or entering/leaving) public streets or highways. a “Total Injuries” rounded to nearest 10 in conformance with “Non-Fatal Injuries” data. Sources: Committee on School Transportation Safety (2002), with “Total Injuries” and indexing by the Handbook authors. Table 16-116 Annualized Injury Rates per 100 Million Schoolchild Person Trips by Mode

with their associated walking. This issue potentially pertains whether the exposure measure is miles/kilometers, trips, or hours, although the hours-based fatal crash rate research presented above (Chu, 2003) carefully assigned access and egress times to the walk mode. Finally, two overall observations may be drawn from the various crash rate data. On the one hand, pedestrian and bicycle fatality rates are high by any measure, clearly demonstrating need for system-wide improvements to increase NMT safety. European data presented below in the “Foreign Versus U.S. Safety Comparisons” show that this can be done. On the other hand, walk- ing and cycling injury rates based on trips made or hours spent in the activity are not so much higher (if any higher) than automobile occupant injury rates as to suggest advisability of avoiding walking or cycling for any age group. The NMT risks—particularly for pedestrians—are not hugely different, on a per trip or per hour basis, than the private mode of travel risks generally accepted by the U.S. populace as a fact of life. Most Prevalent Crash Causes. Alcohol/drugs are significantly involved in NMT/motor-vehicle crashes, and darkness appears to play a major role as well. Almost one-half (48 percent) of fatal pedes- trian-vehicle crashes in 2009 involved alcohol, most often on the part of the pedestrian. Involvement was 43 percent calculated on the basis of intoxication defined as a blood alcohol concentration (BAC) of 0.08 grams per deciliter or higher. Of these, 6 percent involved both an intoxicated pedestrian and an intoxicated driver, 29 percent involved intoxicated pedestrians and sober drivers, and 8 percent involved intoxicated drivers and sober pedestrians. For fatal bicycle-vehicle crashes, the bicyclist and/or the motor-vehicle operator was deemed intoxicated (BAC ≥ 0.08) in 1/3 of all cases. Almost 1/4 of the cyclists killed were themselves intoxicated (NHTSA, 2010). Since 2003, there has been a modest absolute and relative decline in intoxication involvement in fatal pedestrian crashes, particu- larly on the part of drivers, but not much change in the case of fatal bicycle crashes (NHTSA, 2010, Turner et al., 2006). Studies of pedestrian incidents in 2003 determined that 64 percent occurred at night, whether or not intoxication was involved. Of fatalities involving pedestrians under 16 years of age, 65 percent occurred between 3:00 PM and 7:00 PM. Almost 1/3 of fatal bicycle crashes took place between 5:00 PM and 9:00 PM. Pedestrian and bicyclist activity is quite likely high during late afternoon and early evening hours, increasing exposure, and it may be presumed that reduced visibility is also a factor. Late hours, and also weekends, are the times most associated with alcohol intoxica- tion (Turner et al., 2006). In judging the safety of walking and bicycling as a travel mode or form of exercise for children or adults, it is important to consider what crash and mortality rates might be if calculated separately for sober pedestrians and cyclists or if calculated only for daylight hours prior to the evening rush. The per-hour pedestrian fatality crash rate calculations reported above, showing walking during the daylight hours of 7:00 AM to 5:00 PM to be as safe as traveling by auto, give strong indication of daytime sober-walking safety. Speed is a critical factor in fatal crashes, with injury severity strongly dependent on impact speed. An all-new, large data set (previous data were mostly decades old) composed of 490 German pedestrians aged 15 to 96 years suffering injuries from head-on crashes with automobiles has been analyzed. Children under 15, crashes involving SUVs and trucks, crashes involving persons lying on the street, and crashes with no injury reported were not included. The sample was drawn from the German In-Depth Accident Study (GIDAS), and data preparation included an adjustment for underreporting of minor-injury crashes based on comparison with the German national statistics on pedestrian crashes. Both age and speed were found highly significant predictors of fatal events but the results presented here are from a simplified model including only speed. The predictive 16-345

logistic curve takes an “S” shape, starting to move upward more sharply above 20 to 30 km./hour (12 to 19 miles per hour [mph]) and reaching 50 percent probability of death at a little more than 75 km./hour (47 mph). The study found the fatality risk in injury crashes to be approximately 1.5 percent chance of death at 30 km/hour (19 mph). Risk more than doubles at 40 km/hour (25 mph) and more than doubles again at 50 km./hour (31 mph), such that the risk at 50 km./hour is 5 times the risk of pedestrian death at 30 km./hour (Rosén and Sander, 2009). An earlier study which reworked previously eval- uated crash records from Great Britain allows expansion of the GIDAS-based analysis to children. It derived separate curves similar to the GIDAS-based analysis for both children up through 14 years of age and adults 15 through 59 years of age. The results for both children and adults reach 50 per- cent probability of death at 70 to 75 km/hour (44 to 47 mph) (Davis, 2001). This is the same result as obtained in the GIDAS-based analysis without adjustment for underreporting of minor-injury crashes. It thus seems reasonable to assume that the newer relationships can safely be used as a basis for understanding childhood risks as well as adult risks. Both studies found the elderly to be especially prone to fatal crashes. The earlier study developed a predictive curve specific to injury crashes involving persons age 60 and above. It showed 50 per- cent probability of death at 45 to 50 km/hour (28 to 31 mph) for this age group, compared to 70 to 75 km/hour (44 to 47 mph) for younger ages (Davis, 2001). Eight crash-type categories encompass two-thirds of all pedestrian crashes involving a motor vehi- cle on, or entering or leaving, public streets or highways. These eight categories are fairly evenly distributed in prevalence, each accounting for between 7 and 11 percent of all reported vehicle- pedestrian crashes. Most frequently resulting in serious or fatal injuries (1/3 to 1/2 serious/fatal) are midblock dart/dash, other midblock, intersection dash, other intersection, and walking-on- roadway crashes. Only slightly less frequently resulting in serious or fatal injuries (1/5 to 1/4 seri- ous/fatal) are crashes of the not-in-roadway/waiting-to-cross, vehicle turn/merge, and backing vehicle categories. The eight most common conflict types for bicycle-vehicle crashes cover just over one-half of all cycling crashes involving a motor vehicle. Each account for 4 to 10 percent of reported crashes. Four of these crash types each result in serious or fatal injuries about 1/4 of the time. They are ride- out at stop sign, ride-out at residential driveway or alley, motorist left turn into cyclist, and cyclist left turn into same-direction motorist. Of all crashes in these four categories, 2/5 are of the ride- out at stop sign type. The other four most common bicycle-vehicle crash types incur serious or fatal injuries at the rate of very roughly one in ten reported incidents. These crash types are other cyclist- ride-out-at-intersection (one in six serious/fatal), motorist facing a stop sign, motorist mid-block drive out, and motorist right turn.71 16-346 71 These pedestrian and bicycle crash proportion and severity conclusions are drawn by the Handbook authors directly from the notes to Figures 3-4 through 3-19 in Turner et al. (2006), the original source being an FHWA study of 8,000 pedestrian and bicycle crashes in five states. Only included are pedestrian and bicyclist crashes involving a motor vehicle operating on, or entering or leaving, public streets or highways. Thus omitted in this data are all trail crashes except those occurring at road crossings and intersections. Data provided fur- ther on under “Facility Type Safety Comparisons” suggest that falls and collisions with fixed objects are highly prevalent NMT crash types not covered by motor-vehicle-conflict crash data.

Foreign Versus U.S. Safety Comparisons Pedestrian and bicycle crash fatalities have each declined in the United States over the past quar- ter century or more. However, accompanying trends suggest that—at least until the 2000–2009 decade—the declines may be mostly related to reductions in walking and bicycling activity. For example, reduction in child cycling has been posited as the underlying cause of the U.S. cyclist fatality decline. An Insurance Institute investigation is reported to have found that adult cyclist fatalities actually increased from 302 in 1976 to 560 in 1997 (Pucher and Dijkstra, 2000). Only since 2000 has a decline in NMT fatalities been seen (NHTSA, 2010) in conjunction with walking and bicycling rates that have climbed slightly or remained stable. (For 2001 and 2009 walk and bike share statistics see Tables 16-85 and 16-87 in the “Extent of Walking and Bicycling” subsection).72 In contrast, the pedestrian and cyclist fatality rates in countries such as the Netherlands, Denmark, and Germany are not only much lower today than rates in the United States, they have also dropped much faster over time (Pucher and Buehler, 2008a). This is important because of questions about transferability of results. Contentions that fundamental conditions in European countries are so unique as to explain current NMT safety rate differences are difficult to positively refute. Different rates of safety improvement, in contrast, are less readily linked to inherent differences among coun- tries aside from government policy emphasis. Moreover, cogent explanations linked to deliberate Dutch, Danish, and German policies and programs have been offered for the European fatality reduc- tions. Such explanations lead to corollary conclusions that application of similar approaches in coun- tries such as the United States could bring comparable fatality reductions. Specific safety improvement approaches applied in the Netherlands and Germany include (Pucher and Dijkstra, 2000, Pucher and Buehler, 2008a and b, Lusk et al., 2011): • Pedestrian system enhancements—such as pedestrian zones, well-lit sidewalks on both sides of streets, and zebra crosswalks. • Bike paths and lanes—mileage doubled in the Netherlands and Germany between the mid- 1970s and the mid-1990s, with recent emphasis more on design enhancements such as bike lane buffering and physical separation to create cycle tracks. • Traffic calming—of most streets in residential neighborhoods, including substantial use of 30km/hour (19 mph) speed limits. • Intersection modifications—such as special bike lane and stop bar arrangements, colored pave- ment guidance, and pedestrian and cyclist traffic signal provisions including activation, phases, and timing. • Secure bike parking—positioned for safety, with guarded parking areas at key locations. 16-347 72 U.S. pedestrian fatalities were 4,763 in 2000 and 4,092 in 2009. Bicyclist fatalities were 693 in 2000 and 630 in 2009. During the same time span, however, the proportion of total vehicular crash fatalities that were pedes- trian fatalities increased from 11 to 12 percent and the proportion of total fatalities that were bicyclist fatal- ities increased from 1.7 to 1.9 percent (NHTSA, 2010). In drawing inferences about the raw fatality numbers, it must be kept in mind that this was a decade that ended with higher gasoline prices, higher unemploy- ment, and a dip in VMT.

• Education and training—grade school classroom instruction and on-facility cycling lessons, mandated driver training, and motorist license exam testing of pedestrian and bicyclist crash avoidance skills. • Traffic regulations and enforcement (strict compared to the United States)—motorists are basi- cally assumed responsible in crashes involving child, elderly, or disabled pedestrians or cyclists even in cases of jaywalking or other unsafe behavior. Table 16-117 compares contemporary cycling fatality rates among North American and European countries, expressed in deaths per 100 million kilometers cycled, using country by country averages for the years 2002 through 2005. 16-348 Country Rate a Country Rate a United States 5.8 Germany 1.7 Italy 3.5 Denmark 1.5 United Kingdom 3.0 Sweden 1.5 Canada 2.4 Netherlands 1.1 France 2.0 Note: a Cycling fatalities per 100,000,000 kilometers cycled. Source: Pucher and Buehler (2008a). Table 16-117 Cycling Fatality Rates in North America and Europe, 2002–2005 Facility Type Safety Comparisons Many safety comparisons can potentially be made among NMT facility types and options. The fol- lowing discussion is limited, however, to three comparisons of particular interest to broad NMT plan- ning decisions. They are the safety of off-street versus on-street bicycle facilities, cycling on sidewalks versus on-street in traffic, and cycling in cycle tracks compared to other on-road situations. Cycling Safety on Off-Street Versus On-Street Bicycle Facilities. The relative safety of cycling on sep- arate facilities versus cycling on streets and roads in traffic is a subject that has engendered much con- troversy (Pucher, 2001). The controversy is not aided by the focus of the primary crash statistics sources on only those crashes involving a motor vehicle using, entering, or leaving a public highway. One investigation that does provide a slice of data on the relative safety of different types of facilities is the December 1996 survey of League of American Bicyclists (LAB) members. A 20 percent sample was drawn proportionate to state population, and the 1,956 valid survey returns represent a 42 per- cent useable response rate. The survey population was not intended to be representative of all cyclists. The LAB-member respondents were adult or older teen (average age 48), largely male (80 percent), and fairly experienced, as may be inferred from their average annual cycling distance of 4,670 km. (2,900 miles). Included in the survey instrument was a question about serious crashes, defined as involving at least $50 in property damage or medical expense. Of respondents, 29 percent reported having had a crash in 1996, including falls and striking fixed objects, while 9 percent reported having a serious crash.

A relative danger index (RDI) was calculated as the fraction of bicycle crashes reported for a par- ticular facility type divided by the fraction of kilometers ridden on that facility type. An RDI of over 1.0 thus indicates a facility type where the rate of crashes is higher than the overall crash rate, on all types of facilities, for the survey population. The lowest crash rates were for on-street bike lanes (RDI=0.41), followed by signed on-street bike routes (RDI=0.51), major streets without bike facilities (RDI=0.66), minor streets without bike facilities (RDI=0.94), shared use paths (RDI=1.39), and off-road/unpaved paths (RDI=4.49). The highest crash rates were on “other” facilities, mostly sidewalks (RDI=16.34). Sidewalks are thus shown to be by far the most unsafe type of facility for a LAB-member cyclist (Moritz, 1998). This is a finding that when generalized to the adult cycling population is mostly but not universally corroborated in the “Cycling Crashes on Sidewalks Versus Other Facilities” discussion. The LAB survey and two earlier studies agree that on-street bike lanes and signed on-street bike routes taken together have the lowest crash rates, with RDIs around 0.5. On the other hand, the detailed LAB survey results showed crashes least likely to be serious on off-road/unpaved paths, shared use paths, and minor streets without bike facilities (Moritz, 1998). The result, for other than sidewalks, is less difference among facility types in terms of serious crashes. In round numbers, calculation of serious crash RDIs from the reporting of LAB-member survey results shows them to be 0.6 for on-street bike lanes, a little over 0.9 for all other on-street conditions, 1.3 for shared use paths, 2 for off-road/unpaved paths, and 19 for “other” facilities, mostly sidewalks. There is an element of “apples and oranges” non-comparability encountered in trying to compare crash rates for on-street facilities with off-street facilities. Relative to on-street facilities, shared use paths attract more child cyclists—including cyclists in training—and relatively fewer “hard-core” experienced cyclists. They also serve pedestrians, in-line skaters, and the occasional baby con- veyance, wheelchair, and dog-walker with or without leash. If even a subset of the resultant user mix were to be introduced on-street, the on-street crash rates would likely be altered for the worse, even for the adult bicyclist component. (See “More—Off-Street Versus On-Street NMT User Mix” in the “Special Mini-Studies in Montgomery County, Maryland” case study, including Table 16- 129, for a direct off-street versus on-street user mix comparison.) A safety study of three mixed-use trails in Connecticut’s Farmington Valley, also based on survey respondent recall, utilized surveys handed to all trail users 18 years of age and older. The bicyclist crash rate was 150 per million miles of travel. Falls were included, constituting 63 percent of all crashes, and were found to be about as often associated with an injury as collisions. Analysis indicated that cyclists on the trails incur three times the crash rate of on-trail pedestrians but only roughly one- half the crash rate of in-line skaters. In light of the relatively small sample size, the crash rate differ- ence between cyclists and in-line skaters was not statistically significant, but the crash rate differences between pedestrians and wheeled users were. Slightly less than one in five of all crashes occurred at trail intersections with roads. The trail with the highest overall crash rate was the most heavily used and had the most intersections, while the trail with the lowest rate had the smallest percentage of wheeled users and the fewest crossings (Aultman-Hall and LaMondia, 2005). Whereas the Farmington Valley study did not examine relative crash rates among facility types, there has been a highway agency study in Boulder, Colorado, which did so. Described at the end of the “Sidewalks” discussion below, it found shared-use “side path” safety to be at least as good as the safety of Boulder’s on-street bicycle facilities (Roskowski et al., 2010). Cycling Crashes on Sidewalks Versus Other Facilities. Designation of sidewalks as bikeways has generally been identified as a practice to be avoided if possible for safety reasons. Cycling on side- walks may be desirable for children traveling at low speeds, but for the general population most 16-349

data suggest it presents greater risk compared to other cycling environments. Safety issues on side- walks include conflicts with other people, poles, and sidewalk furniture, and most importantly, conflicts with driveway, alley, and street intersection vehicular traffic. Cyclists on sidewalks are placed in particular danger at traffic conflict points by their relatively fast movement in directions not allowed on adjacent traffic lanes and their inability to act like vehicles in intersections. Either of these circumstances can engender motorist surprise and confusion (Turner et al., 2006). Researchers in a 1979 Eugene, Oregon, study found the crash rate on the three sidewalk bicycle-route sections to be close to 3 times higher than on the city’s signed or striped bicycle lanes. Similarly, a 1974 Palo Alto, California study found that while only 15 percent of bicycle travel occurred on streets with sidewalk bicycle paths, about 70 percent of the bicycle-motorist collisions occurred on such streets (Zehnpfenning et al., 1993). A 1995 route choice and fall-and-collision survey taped to handlebars of bicycles parked at employ- ment areas throughout Ottawa and in central Toronto, Ontario, Canada, confirmed substantially higher event rates (falls and crashes per-kilometer bicycled) for sidewalks.73 The commuter cyclist respondents who used sidewalks did so mainly along major roads. It was determined that cyclists who choose to use sidewalks have higher event rates than non-sidewalk cyclists even when on roads and to some extent on paths. These “sidewalk cyclists” (bicyclists who used sidewalks) thus had higher event rates in general than other cyclists. Sidewalk cyclists reported higher helmet use rates, suggesting more caution, and bicycled fewer miles. The researchers posited that the side- walk cyclists in Ottawa and Toronto, where sidewalk bicycling is by no means encouraged, are on the whole less skilled and perhaps themselves more “dangerous” and in need of training (Aultman-Hall and Adams, 1998). A somewhat different understanding of sidewalk bicycling risks was developed in a follow-up Palo Alto study of bicycle-motor vehicle collisions at intersections. Bicycle crash statistics were obtained from police reports for 1981 through 1990. Bicycle/motor-vehicle collisions accounted for 314 out of the 371 crashes for which substantially complete reports were available. Bicycle obser- vations and counts were taken in May 1987 at intersections where 92 of the 233 intersection crashes had occurred, in order to establish a basis for exposure rate calculations by cyclist and behavior category. Only the bicycle/motor-vehicle collisions were analyzed, and only those occurring where the exposure counts were taken were carried into the final analysis. The final sample was 89 crashes with information on all four key variables. Relative crash risk factors for different demo- graphics and circumstances were calculated, indexed to a value of 1.0 as the average risk factor for all cyclists and situations analyzed (similar to the RDI described above for LAB members). While cyclists 17 years of age and under had a relative risk of 1.0 when cycling on sidewalks, the same as for all cyclists and situations overall, cyclists 18 years of age and older had a risk factor of 2.4. Contributing factors may have been ongoing safety education in the school system for younger cyclists and the faster speed of older cyclists. A key determinant was direction of travel. Cyclists of all ages had a risk factor of 0.7 when cycling on sidewalks in the same direction as vehicular traffic in the adja- cent lane, but incurred a risk factor of 3.0 when traveling counter to the flow of adjacent traffic. The study authors note a lack of success in Palo Alto’s attempt to enforce one-way bicycle travel on cer- tain sidewalks. Overall, bicycling on sidewalks was associated with a risk factor of 1.4 as compared to 0.8 for cycling on roadways, a statistically significant sidewalk to roadway risk ratio of 1.8 (Wachtel 16-350 73 The survey not only allowed evaluation of fall and collision events not covered by police records, it also found that none of the 82 sidewalk falls and 32 sidewalk collisions recorded by survey respondents had been reported to police. Only two might likely have shown up on medical records.

and Lewiston, 1994). (Compare these factors to the RDIs for LAB members presented above, which show even higher relative danger in sidewalk cycling by adults [Moritz, 1998].) A review of 682 year 2000 police reports on crashes involving bicycles in Phoenix, Arizona, tends to confirm that bike riding on sidewalks is much more risky in the direction of travel opposite of vehicular flow in the adjacent travel lane. (Exposure measures and rates were not developed.) Of pre-crash bicyclist riding positions and directions, 5.9 percent involved sidewalk bicycling “with traffic” and 22.6 percent involved sidewalk bicycling “against traffic.” Bicyclists aged 11–20 were most frequently involved in bicycle-vehicle crashes overall, twice the number in the next highest 10-year age increment (Cynecki, 2011). While the literature on safety issues and crash rates for bicycling on sidewalks is obviously not exten- sive, and some questions about underlying risk factors are raised by the Ottawa and Toronto Research, only one study has been encountered that takes serious issue with the proposition that sidewalk cycling tends to be relatively hazardous on average. Noting a basic intersection conflict similarity between sidewalks used for cycling and Montreal two-way cycle tracks, the Montreal cycle track safety study described below took a critical look at the Wachtel and Lewiston analyses in Palo Alto. The Montreal study authors report finding that the Palo Alto evaluation was limited to intersection crashes. They further report that when non-intersection crashes are included, the relative risk for side- walk cycling versus on-street is lowered from 1.76 (corresponding to the 1.8 value given above and statistically significant) to 1.07 (not statistically significant) based on the Palo Alto data, and that side- walk cycling in the same direction as the closest traffic lane becomes almost twice as safe as in-traffic cycling (Lusk et al., 2011). It would appear that the relative danger of sidewalk cycling cannot be regarded as an issue that has been completely resolved. Boulder, Colorado, has “side paths” along roadways among its inventory of shared-use paths. Built to various standards, some operate more satisfactorily than others. An array of signage, coloration, and geometric design retrofit strategies has been applied to improve conditions. Safety is also thought to be enhanced by the relatively large number of bicyclists, who thus make themselves “expected” users. The Colorado Department of Transportation has conducted an analysis of pedestrian- and bicy- cle-related crashes showing the side paths to have crash rates no higher than the on-street bicycle sys- tem, which consists of a mix of streets with bike lanes, signed bike routes, and bikeable shoulders (Roskowski et al., 2010). Cycle Track Versus Other On-Road Cycling Safety. Conventional bike lanes are generally credited with enhancing safety (Moritz, 1998, Cynecki, 2011, Cambridge Community Development, 2011) although safety conclusions have been drawn in significant measure from studies of bicycle behavior rather than differential crash rate analysis (Harkey, Stewart, and Rodgman, 1996). A few specific exam- ples of safety gains with bike lane implementation have been presented, in conjunction with traveler response information, in the “Response by Type of NMT Strategy” section (see “Bicycle Lanes and Routes”). Time series bicycle crash risk data for the St. Kilda Road bike lane in Melbourne, Australia, is tabulated in the preceding “Time to Establish Facility Use” subsection (see “Melbourne St. Kilda Road Bike Lanes” including Table 16-114 from Davies, 2007). An important inference which may be made from that information is the occurrence of a hazard increase in the 1 year following St. Kilda Road bike lane implementation—somewhat more than a doubling—followed by a return to previous bicyclist hazard levels and then, after 4 or 5 years, a decline into much lower levels of risk concurrent with increased bike lane use. The issue of whether on-street cycle tracks offer similar or more safety advantages over bicycling in mixed traffic flow is of particular concern as U.S. cities consider their introduction. Looking to European examples, the 29,000 kilometers of cycle tracks in the Netherlands have been credited— 16-351

along with other initiatives (listed above)—for the very low bicyclist injury rate in that country. After the Netherlands, next in degree of cycle track deployment is Denmark (Lusk et al., 2011). A before-and-after study of cycle tracks and intersection treatments in Copenhagen included 1,000 interviews, 1,500 counts, and analysis of 8,500 crashes. The bike-lane and cycle-track traffic stream in Copenhagen includes 5 percent mopeds. Contrary to surveyed Copenhagen bicyclists’ percep- tions and North American experience, installation of bike lanes was accompanied by increases in crashes of 5 percent and in injuries of 15 percent, with the risk falling disproportionately on bicy- clists and moped riders. Corresponding combined crash and injury statistics for cycle tracks indi- cate a 9 to 10 percent increase in crashes and injuries. Unlike the case with bicycle lanes, the safety decrease associated with cycle tracks occurred entirely at intersections, with the stretches in between intersections exhibiting a 10 percent reduction in crashes and a 4 percent decline in injuries (Jensen, Rosenkilde, and Jensen, 2007). The cycle track risk in Copenhagen has fallen primarily on pedestrians, bicyclists, and moped rid- ers navigating intersections. Their crash experience has led Copenhagen to experiment with dif- ferent treatments in and approaching intersections, some of which hold promise. The study authors also judge that gains in health from increased physical activity induced by the cycle track and bicycle lane system are producing gains that “are much, much greater than the losses in health resulting from a slight decline in road safety” (Jensen, Rosenkilde, and Jensen, 2007). The role of mopeds in the mix was not examined, and neither was the newness of the studied bike lane or cycle track installations reported. There have been many concerns about transferability of overall European experience to the North American context. To address these concerns, and learn more about cycle track safety, an interna- tional team—lead by the Harvard School of Public Health—took advantage of Montreal’s well established system (and comprehensive records) for a comparative cycle track safety analysis. The local emergency medical response database was used as the primary bicycle injury record source. Police crash records were employed for missing information such as direction of cyclist travel. For comparison, a “reference street” (or streets) was selected for each of the six cycle tracks studied, gener- ally a parallel street. The reference streets had no special bicycle facilities. Historic bicycle volume records were available for the cycle tracks. These bicycle volumes were adjusted/extrapolated to the reference streets on the basis of 2-hour counts taken simultaneously on each pairing of a cycle track with its reference street(s). The relative risk of injury (RR) for the cycle tracks was computed as the ratio of injuries to bicyclists on each cycle track divided by the corresponding ratio for its reference street(s). Statistically significant results were obtained for three of the cycle tracks, with RR values ranging from 0.32 to 0.48, indicating that bicycling on these particular cycle tracks was 2 to 3 times as safe as cycling on their reference streets. Including results not statistically significant, the range of RR values for the six studied cycle tracks becomes 0.32 to 1.18. The RR value for all studied cycle tracks combined was 0.72, indicating that bicycling was 39 percent safer on the six cycle tracks than bicy- cling in mixed traffic on the reference streets. These favorable results were obtained despite the fact that Montreal cycle tracks are two-way facilities, not as safe as one-way facilities according to Dutch guidelines, and lack parking setbacks at intersections as recommended by the Quebec Ministry of Transport (Lusk et al., 2011). A cycle track and a pair of buffered bike lanes in Portland, Oregon, were not in place long enough, when studied, for crash statistics analysis. A survey showed that a majority of surveyed users felt safer. On the SW Broadway cycle track, which replaced a bicycle lane configuration, the propor- tion of bicyclists who elected to cycle in mixed traffic rather than on the bicycle facility itself fell 16-352

from 12 to 2 percent. The buffered bike lanes were placed on a one-way couplet, SW Stark and State Streets, that had not had bicycle lanes previously. The Portland study made video observations of intersections along the new facilities. Analysis of the videos identified nearly one in 10 of all inter- actions between cyclists and pedestrians as potentially unsafe. Survey respondents included high proportions of motorists and bicyclists expressing confusion about relevant traffic regulations (Monsere, McNeil, and Dill, 2011). The latter findings are hardly definitive with regard to safety impact but do suggest need for further education of the public. Other Traffic Safety Issues and Findings Following are some pedestrian and bicyclist safety topics likely to be of special concern to practitioners focused on encouraging active transportation but cognizant of the need for awareness of safety implica- tions. Note that schoolchild safety information was included above within the “Pedestrian and Bicyclist Safety Highlights” discussion (see “Crash Rates from Various Perspectives” including Table 16-116 and associated text). Street Crossing Safety. One street crossing safety issue in particular deserves special mention here, namely, pedestrian safety within marked crosswalks. Although there are many uncertainties in the existing research, it does appear that an unfortunate trade-off exists with respect to crosswalk mark- ing under certain problematic conditions. These conditions involve uncontrolled at-grade crossings of roadways (i.e., with no traffic signal or stop sign protection for the crosswalk) in the presence of substantial traffic volumes and multiple travel lanes. Marked crosswalks may attract some additional pedestrian activity (see “Street Crossings”—“Crosswalks and Traffic Controls”—“Pedestrian Crossings” in the “Response by Type of NMT Strategy” section), but where the described problem- atic multi-lane traffic conditions exist, they apparently do so at the risk of pedestrian-vehicle crash rates that are higher to a statistically significant degree. A review of 11 intersection studies from 1965 to 2005 determined that nine of the 11 found higher pedestrian-vehicle crash rates in the presence of plain painted crosswalks than in situations where legal crosswalks (projections of intersecting sidewalks or roadsides) remained unmarked. Most of these studies did not differentiate between controlled and uncontrolled intersections (Chu, Guttenplan, and Kourtellis, 2007). A Federal Highway Administration (FHWA) study published in 2005, however, did make this dif- ferentiation, looking only at uncontrolled crossings—2,000 of them in 30 U.S. cities within 17 states. It also differentiated by number of lanes, presence or lack of a raised median, average daily traffic (ADT), and posted speed. The study examined 5 years of pedestrian crash experience and devel- oped exposure measures at 1,000 marked crosswalks along with 1,000 unmarked matched com- parison sites. Statistical analysis indicated that posted speed limits did not significantly affect the prevalence of crashes, although 43 percent of crashes at posted speeds of 35 mph and above were fatal or involved serious injury, as compared to 23 percent at sites with lower posted speeds. No difference in crash rates with and without crosswalk markings was found for two-lane streets. However, at crossings of the wider, busier arterials—generally those with ADTs over about 12,000 vehicles per day, or 15,000 in the case of roadways with raised medians—crash rates were found to be several times higher in the presence of plain marked crosswalks than without them. Midblock crossings were included in the analysis, and seemed to adhere to the same overall crash experience patterns (Zegeer et al., 2005, Chu, Guttenplan, and Kourtellis, 2007). Results are summarized in Table 16-118. 16-353

Effects of crosswalks per se on crash severity that were isolated in the 2000-crossings study were not statistically significant, but there appeared to be more fatal and serious injury crashes in marked as compared to unmarked crosswalks on multilane roadways. This may be related to the greater tendency of elderly persons to choose marked crossings over unmarked crossings, noted in the “Street Crossings” subsection cross-referenced above (Zegeer et al., 2005, Chu, Guttenplan, and Kourtellis, 2007). On the other hand, one Florida study of midblock crosswalk crashes under conditions of darkness identified lesser severity for those crashes occurring in marked crosswalks (Chu, Guttenplan, and Kourtellis, 2007). Among the various uncertainties surrounding these research findings is the question of what causes the observed crash rate relationships. At first it was hypothesized that the cause is a false sense of security on the part of crosswalk users. Subsequent observational studies of pedestrian “looking behavior” when crossing two- and three-lane streets found that pedestrians in marked crosswalks had as good or better crossing behavior than those in unmarked crosswalks. More recently, in a paired-crosswalk study of six locations, multiple factors have become evident. The studied sites were all in San Francisco, Oakland, or Berkeley; one was two-lane, one was three-lane, and the others involved four-or-more lane roadways. At the two sites where looking behavior dif- ferences were significant, both multi-lane locations, pedestrians using the unmarked crosswalk looked both ways more than those using the marked crosswalk. Pedestrians using the unmarked crossings were more likely to run (significant at four locations), and waited for longer gaps in traf- fic (significant at five locations). Prompt yielding of right-of-way by vehicles was more prevalent at marked crosswalks (significant at all six locations), yet average pedestrian exposure to vehicles was also higher in the marked crosswalks (significant at five locations). The observation deemed most critical, statistically significant at three of the four sites with four-or- more lanes, involved incidence of multiple threats. In this scenario the approaching vehicle in the lane nearest to the pedestrian yields and a vehicle traveling in the same direction in the next lane over, the view of which is now obstructed by the yielding vehicle, does not. The occurrence of multiple threats was higher with the marked crosswalks than with the unmarked crosswalks at these three sites and showed a similar trend at the other four-or-more lanes site as well. Thus four-or-more lane marked crossing users not only exhibited ordinary caution as compared to the extraordinary caution of unmarked crossing users, they were also exposed by the higher yielding behavior (the legal response) to more multiple-threat situations. Multiple-threat scenarios produce the most common type of vehicle-pedestrian crash at uncontrolled intersections (Mitman, Ragland, and Zegeer, 2008). 16-354 Roadway Type Number of Lanes Average Daily Traffic Crosswalk Crash Rate Significant Difference of Number SitesMarked Unmarked No Median 2 All 0.12 0.12 No 914 No Raised Median 3-8 12,000 0.17 0.25 No 260 No Raised Median 3-8 12,000 - 15,000 0.63 0.15 Yes 149 No Raised Median 3-8 > 15,000 1.37 0.28 Yes 417 Raised Median 3-8 15,000 0.17 0 No 87 Raised Median 3-8 > 15,000 0.74 0.17 Yes 173 Note: Crash rates are given in units of vehicle-pedestrian crashes per million pedestrian crossings. Source: Zegeer et al. (2005). Table 16-118 Marked Crosswalk Versus Unmarked Crosswalk Crash Rate Comparisons

In the context of this marked-crosswalk dilemma, it has been observed that “doing nothing for established [pedestrian] demand is not sound public policy” (Chu, Guttenplan, and Kourtellis, 2007). Noting that many U.S. agencies “have elected to remove marked crosswalks at uncontrolled intersections or have shown resistance to installing them in the first place,” it has been advised that “[s]uch an approach does not address the safety and mobility needs of pedestrians” (Mitman, Ragland, and Zegeer, 2008). Cutting back on attractive street crossing opportunities certainly detracts from built environment ingredients found important in this chapter to encouragement of all forms of active transportation. Engineering response has been placed on modifying traffic signal warrants for more emphasis on pedestrian needs and augmenting crosswalks on busy arterials at uncontrolled locations with addi- tional treatments. These treatments range from geometric features such as median refuge islands to “active when present” traffic control devices/beacons (Fitzpatrick et al., 2006). Surveys and focus groups have identified substantial motorist and pedestrian confusion concerning pedestrian right- of-way laws, a confusion that is worse in the case of unmarked crosswalks. Recommended educa- tion and enforcement countermeasures have included signage to encourage careful looking behavior, improved state driver license testing and driver manual coverage of pedestrian right-of-way law, and enforcement designed to educate, such as highly publicized “stings” (Mitman, Ragland, and Zegeer, 2008). Transit Access Safety. Transit passengers are almost inevitably pedestrians for some part of their trip. Access points for transit, such as bus stops or rail stations, are typically areas with higher lev- els of pedestrian activity than other locations. Further, the nature of bus service normally requires each passenger to make at least two street crossings as part of every round trip. Street crossings also can be more dangerous at bus stops than at other intersections, as stopped buses can impede sight lines. Boarding and alighting passengers may try to cross at a point without a crosswalk, par- ticularly if they are about to miss the bus. The majority of transit passenger pedestrian crashes and fatalities occur at locations without pedestrian signals, either at an intersection or a mid-block loca- tion. The nature of streets on which transit is typically operated, characteristics of transit service and vehicles, the increased level of pedestrian activity at transit stops, and the high use of transit by recent immigrants unfamiliar with U.S. traffic norms, all contribute to high pedestrian crash rates near transit access points (Burnier, 2005, Nabors, et al., 2008). Pedestrian crashes while accessing transit may not be a significant factor in travel decisions,74 but they are increasingly recognized as a major safety concern. The effect of transit service and transit stops on pedestrian safety has been studied, but reliable and complete data about pedestrian crashes is diffi- cult to assemble. Several studies have shown that vehicle-pedestrian crashes are more likely to take place in areas with high levels of transit service. In a study of crashes in Baltimore, 78 percent of pedes- trian crashes were found to have occurred in areas with high transit accessibility (Burnier, 2005). This observation reflects both the increase in pedestrian activity and the increased risk in these areas. TCRP Report 125: Guidebook for Mitigating Fixed-Route Bus-and-Pedestrian Collisions, made further strides in determining the hazards to pedestrians accessing transit by categorizing reports of collisions involv- ing interaction of pedestrians and buses according to “bus action.” The study found that 34 percent of 16-355 74 While transit-related pedestrian crashes may be “below the radar” for individual adult travel decisions, the potential danger for children has certainly been a factor in the choice to maintain or introduce separate yel- low school bus transportation systems in preference to use and adjustment of local public transit services for school access.

the crashes reviewed occurred while the bus was turning and 25 percent occurred when the bus was at or near a stop. Both the data and the perception by transit agency staff underscore the importance of safe crossings for pedestrians accessing transit and of addressing pedestrian safety during bus turn- ing maneuvers (Pecheux et al., 2008). Safety in Numbers. At least six published analyses of crash statistics have developed empirical evidence that rates of collisions between pedestrians or cyclists and motor vehicles are lower in areas with higher amounts of non-motorized travel. The findings do not imply that the absolute number of crashes would be lower, although that has been observed in one reported instance. The “safety in numbers” pattern seems to hold across countries, states, cities, and specific intersections, and across time periods. An explanation offered is that motorists apparently drive more cautiously when greater numbers of walkers and cyclists are in evidence (Jacobsen, 2003, Victoria Transport Policy Institute, 2007, Alliance for Biking & Walking, 2010). A corollary deduction may be drawn “that shifts from driving to nonmotorized modes can reduce total per capita crash risk.” (Victoria Transport Policy Institute, 2007). Some caution is required in drawing conclusions from analyses such as these, insofar as environ- mental factors that support increased walking and cycling may also contribute to increased safety, introducing questions of causality (Thunderhead Alliance, 2007). Conclusions may also be affected by the apparent exclusion from some or all of the safety calculations, for lack of complete data, of pedestrian and bicyclist crashes not involving motor vehicles—such as falls and fixed- object crashes. Two of the studies estimated mathematical relationships, which proved similar, between NMT vehi- cle crashes and various measures of non-motorized travel activity. Both aggregate cross-sectional data and time series data were employed. For all cases except those involving time-series data, results con- verged around a relationship indicating that crash totals increase with the 0.4 power of the measure of pedestrian or bicyclist activity.75 Again excepting the time-series data, the values obtained in the two studies ranged from 0.13 to 0.67 for the exponent across nine sets of circumstances, four involv- ing walking and five involving bicycling. Individual aggregate values were computed for a dataset of 68 California cities; a dataset of 47 Danish towns; Hamilton, Ontario, Canada; Gothenburg, Sweden; and datasets of eight to 14 European countries. All showed NMT crash rates to be less in the presence of more walking or bicycling activity, with diminishing numbers of additional NMT crashes associ- ated with incrementally higher active transportation volumes. The time series data results, from the United Kingdom and the Netherlands, and all pertaining to bicycling, were more varied. Overall, however, they supported the inverse relationship between crash rates and walking or bicycling activity. Data from the Netherlands actually showed a decrease in absolute numbers of fatalities with increasing cycling (Jacobsen, 2003). The time-series observations may well have been more subject to exogenous factors such as global shifts in traffic conditions and safety programming. 16-356 75 A power (exponent) of less than 1.0 indicates that the increase in injuries is less than a 1:1 linear relationship with non-motorized travel activity—an inverse relationship between the crash rate and the measure of walk- ing or cycling. An exponent of 0.4 indicates, for example, that a community with twice as much walking can expect to have just 32 percent more total injuries. Taking into account the amount of non-motorized activ- ity, an individual pedestrian’s risk in the example city with twice as much walking would be 66 percent of the risk of an individual pedestrian in the city with less walking (Jacobsen, 2003).

Public Health Issues and Relationships Health is defined by the World Health Organization as “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity.” Health is enhanced by physical activ- ity directly and also indirectly through avoidance of excess body weight associated with inactivity. Inadequate physical activity is one of four primary risk factors for obesity, along with poor nutrition, caloric intake in excess of calories expended, and genetic predisposition. Transportation infrastructure and land use arrangements, especially the provision or lack of elements supportive of non-motorized travel, affect an individual’s options for physical activity. Consequently, “Health should be an impor- tant consideration in transportation decisions” (Dannenberg, 2004 and 2005). Some idea of the separate albeit related dangers of excess weight and inactivity may be garnered from the World Health Organization estimate that 2.8 million deaths worldwide annually result from overweight and obesity whereas “physical inactivity is (separately) responsible for an addi- tional 3.2 million deaths” (de Nazelle et al., 2011). Each year an estimated 200,000 to 300,000 pre- mature deaths occur in the United States as a result of physical inactivity (Heath et al., 2006). Persons who are capable of exercise yet participate in no leisure time physical activity exhibit a mortality hazard ratio of 1.6 over an average of 4 years of follow-up. This hazard ratio value is esti- mated relative to reporting any leisure time physical activity (hazard ratio of 1.0) in the 1997–2000 National Health Interview Surveys (NHIS). It is adjusted for sociodemographic variables, health behaviors, chronic diseases, and serious psychological distress (Pratt, 2009).76 Earlier analysis with the 1987 NHIS found direct medical costs to be higher for inactive persons whether they were grouped by smokers or non-smokers, presence or lack of physical limitations, or gender. When stratified by gender and age, the only categories where active individuals had higher direct medical expenses than corresponding sedentary persons were adolescent/young-adult males (age 15–24) and elder males (age 75 and older). U.S. annual direct medical costs resulting from lack of adequate physical activity were estimated in the same study at $330 to $1,053 in 1987 dollars per able person age 15 and above per year. On the basis of the lower $330 figure, this equated to $76.6 billion annually in 2000 dollars (Pratt, Macera, and Wang, 2000). This estimate is given further context in the subsection on “Economic and Equity Impacts” to follow. The Surgeon General in 1996 recommended 30 minutes or more of at least moderate physical activ- ity daily for adults. The standard has more recently been refined by the U.S. Department of Health and Human Services (HHS) into minimum recommendations for able adults of 150 minutes a week of moderate-intensity or 75 minutes a week of vigorous-intensity physical activity, preferably spread throughout the week. Suitable activity lasting at least 10 minutes at a stretch counts toward the minimum. The recommendation for children and adolescents is at least 60 minutes of activity daily. Brisk walking is a common standard for moderate physical activity. A walking speed of 3 to 16-357 76 Leisure time activity is not the ideal measure for use in the context of walking and cycling, which can be for utilitarian purposes as well as leisure, but the leisure time emphasis is imposed by the design of the NHIS. The NHIS offers the advantage of a nationally representative sample of non-institutionalized U.S. adults that is large enough to allow segregating out persons incapable of physical activity. (Such persons have an esti- mated hazard ratio of 2.3 or greater [Pratt, 2009].) A hazard ratio may be thought of as an expression of rela- tive probability. A hazard ratio of 1.6 in the context given suggests that an exercise-capable adult who engages in no leisure time physical activity is 60 percent more likely to die at any given time than someone of similar age, sex, and circumstances who engages in any such activity.

4 mph (15 to 20 minutes per mile) qualifies. Bicycling qualifies, with health benefits far exceeding risks from traffic injuries. Walking or cycling as part of a daily commute or for any utilitarian pur- pose counts as much as leisure activity (Besser and Dannenberg, 2005, Committee on Physical Activity, Health, Transportation, and Land Use, 2005, Department of Health and Human Services, 2008, Pucher, Dill, and Handy, 2010, de Nazelle et al., 2011).77 Examination of relationships between the built environment on the one hand and physical activity, obesity, and health on the other is a relatively new field of research. The causal link is well established between physical activity and health. It is the connection between the built environment and adequate physical activity that is less well understood. Features of the built environment that can play a role in increasing physical activity range from public parks and readily accessible gymnastic facilities to pedestrian and bicycle facilities and access to public transportation (Committee on Physical Activity, Health, Transportation, and Land Use, 2005, Besser and Dannenberg, 2005). Baseline Walking and Bicycling Activity CDC analysis of the Behavioral Risk Factor Surveillance System (BRFSS) national survey for 2001 found 45 percent of the U.S. adult population to be meeting the recommended physical activity guidelines, with 26 percent of adults deemed inactive. Corresponding findings for 9th to 12th grade adolescents were 69 percent meeting guidelines and 10 percent inactive. There is preliminary indi- cation of a moderate upward trend. BRFSS data for 2009 indicate 49 percent of adults to be meeting minimum guidelines. The BRFSS surveys have shown walking to be a dominant form of physical activity, but activity definition changes have prevented drawing a more definitive conclusions (Committee on Physical Activity, Health, Transportation, and Land Use, 2005, Centers for Disease Control and Prevention, 2011). The Nonmotorized Transportation Pilot Program Evaluation Study performed a 2006–2007 baseline “before” analysis, covering the five pilot program communities. It affords an indicator of how much present day walking and bicycling activity alone contributes to meeting the Surgeon General’s recom- mendations. A word of caution is, however, in order. The activity comparison drawn upon here was presented by the Pilot Program researchers as a demonstration of survey results reasonableness. Using the analysis to draw activity contribution conclusions is almost certainly not an originally intended use of the data. Pilot Program findings were drawn from a five-area self-administered survey of adults that obtained a 15 percent response rate, with 34 percent of eligible respondents completing a follow-up interview or web survey. Elaborate sample weighting procedures keyed to the U.S. Census were applied to com- pensate for follow-up survey respondent differences relative to year 2000 community demographics and also overrepresentation of non-auto commuters, bicyclists in particular. The sample weights served to expand the results to the five study-area populations. The number of final weighted sam- ples totaled just under 1,380 overall (Krizek et al., 2007). Residents of the city of Minneapolis and surveyed portions of Marin County, California, were found overall to engage in walking and cycling at greater frequencies and for longer durations than 16-358 77 These recommendations derive from findings that these minimum degrees of physical activity are associ- ated with substantial health benefits, but should not be taken to imply either that more is not better or that less is useless. Various non-transportation physical activities also qualify (Physical Activity Guidelines Advisory Committee, 2008).

in the city of Columbia, Missouri; Sheboygan County, Wisconsin; and Spokane County, Washington. Combined 2006–2007 walking/cycling frequency and duration results for the five communities are provided in Table 16-119. Also provided is roughly parallel national moderate and vigorous activ- ity information from BRFSS public data files. The questions used by the two surveys were similarly structured, although one deals with walking and cycling while the other addresses all forms of mod- erate and vigorous exercise. The pilot study results are from the weighted samples described above, while the BRFSS results are taken from very large numbers of unweighted samples (Krizek et al., 2007). 16-359 Five-Community Pilot Project National BRFSS Walk Bike Moderate Vigorous Days per week engaging in activity for at least 10 consecutive minutes 0 days 16.5% 80.1% 18.5% 58.4% 1 day 3.7% 3.8% 3.3% 7.1% 2 days 8.3% 4.5% 8.5% 9.0% 3 days 16.5% 5.7% 16.4% 11.0% 4 days 9.1% 2.0% 10.2% 4.8% 5 days 17.4% 2.6% 13.4% 4.7% 6 days 5.5% 0.3% 4.6% 1.5% 7 days 23.0% 1.0% 25.1% 3.5% Minutes of activity per day on days with at least 10 consecutive minutes of activity 0 to 9 minutes 18.0% 80.7% 19.1% 58.8% 10 to 29 minutes 19.4% 2.8% 16.6% 6.6% 30 to 59 minutes 32.8% 6.3% 29.7% 14.2% 1 hour or more 29.9% 10.3% 34.6% 20.4% Source: Krizek et al. (2007). Table 16-119 Pilot Program Five-Community NMT Activity with National BRFSS Physical Activity Comparison From Table 16-119 it may be inferred, accepting various simplifying assumptions, that probably some 30 percent of the five-community adult population was fully meeting the HHS minimum physical activity recommendation by either walking or bicycling. In addition, the tabulation suggests that for more than five out of six of the surveyed population, walking or cycling made at least some contribu- tion toward the activity recommendation. It may be similarly inferred from Table 16-119 that the aver- age qualifying walk/bike activity contribution for the five-community adult population slightly exceeded 1/2 of the 150-minute weekly minimum physical activity recommendation. One rough esti- mate of the relative contribution of walking versus bicycling suggests that walking may make up as much as 96 to 98 percent of the active transportation contribution to qualifying physical exercise.78 78 The Nonmotorized Transportation Pilot Program Evaluation Study researchers make no assertion that the five-community data represents the universe of travel activity in the United States. The four Pilot Program communities were selected by act of the U.S. Congress, which in itself suggests at least some proactivity in the communities’ approach to walking and bicycling. The control community, Spokane, was selected to be as representative as possible of the other four Pilot Program communities but without a notably proactive program of NMT facility improvements or programs (Krizek et al., 2007, Federal Highway Administration, 2007). National sources tend to show substantially higher percentages of persons not walking in the preced- ing week (35 percent in the 2001 NHTS) or month. (See the “Extent of Walking and Bicycling” subsection also under “Related Information and Impacts.”) Note that the analytical assessments in this paragraph are by and fully the responsibility of the Handbook authors.

More narrowly focused research, carried out by the CDC, has determined that 29 percent of public tran- sit users achieve the recommended 30 minutes or more of physical activity a day simply by walking to and from their transit service. This research employed a rigorous examination of the previously described 2001 NHTS travel data for all trip purposes, focusing on adults. On the other hand, the 2001 NHTS also indicates that only 3 percent of all U.S. adults undertake travel via the walk/transit mode on any given day (Besser and Dannenberg, 2005). This suggests that just under 1 percent of the U.S. population presently meets or exceeds the recommended physical activity levels solely by virtue of the walking connected with public transit use. A related question of interest is the proportion of all health-benefit-qualifying exercise that comes from active transportation. The one dataset found that comes close to addressing this question is from five Pacific Northwest cities of various sizes (see Table 16-124). In this dataset, compared to a total composed of active transportation and sports exercise, the transportation component ranged from 37 to 72 percent, averaging 59 percent (Socialdata, 2008). When including other forms of qualifying exercise in the total, such as certain activities of gardening, the transportation proportion would be somewhat reduced. A “typical” urban area range of 50 to 70 percent for the active transportation exercise component of total exercise seems reasonable, and appears to conform with data presented above in Table 16-119. Health Benefits for Adults of Enhanced NMT Systems and Policies Physical activity has been clearly shown to have a causal and positive relationship with good adult pub- lic health. The Surgeon General’s report of 1996 reported “an inverse association between physical activity and several diseases that is ‘moderate in magnitude, consistent across studies that differed sub- stantially in methods and populations, and biologically plausible’ ” (Committee on Physical Activity, Health, Transportation, and Land Use, 2005). The HSS document “2008 Physical Activity Guidelines for Americans” lists the health benefits now shown to be associated with physical activity, with evi- dence ratings of “strong,” “moderate to strong,” and “moderate.” A Physical Activity Guidelines Advisory Committee developed the list and ratings taking into account “dose response” (benefit per given amount of exercise) and evidence of causality (Department of Health and Human Services, 2008, Physical Activity Guidelines Advisory Committee, 2008). Table 16-120 consolidates this listing and includes the HHS findings for children and adolescents. Health benefits for children are further exam- ined following this discussion of adult benefits. Confirmatory research also demonstrates that endurance-type physical activity, with walking and bicy- cling as key examples, “reduces the risk of developing obesity, osteoporosis, and depression” and “may improve psychological well-being and quality of life.” A regimen of brisk walking has explicitly been associated—in public health research—with lower risk of both all-cause mortality and cardiovascular disease, especially in women, as well as lesser Type 2 diabetes, increased cardiovascular fitness, and other health benefits (Committee on Physical Activity, Health, Transportation, and Land Use, 2005, Physical Activity Guidelines Advisory Committee, 2008). 16-360

Physical inactivity and excessive caloric intake are the key contributors to the energy imbalance asso- ciated with obesity, a critical public health problem (Committee on Physical Activity, Health, Transportation, and Land Use, 2005). Almost 1/3 of U.S. adult males were obese in 2007–2008 (32.2 per- cent), and over 1/3 of adult females were (35.5 percent).79 In contrast, 1976–1980 statistics show 1/8 of males (12.5 percent) and 1/6 of females (16.4 percent) as being obese. That represents almost a tripling in adult obesity for males, and over a doubling for females. The only bright spot, if one can call it that, is that obesity increases for women between 1999–2000 and 2007–2008 were not statistically significant, suggesting stabilization. Although obesity definitely continued increasing for men, it may have stabi- lized over the latter three years (Flegal et al., 2010).80 16-361 Adults and Older Adults Children and Adolescents Lower risk of early death (S) Lower risk of coronary heart disease (S), stroke (S), high blood pressure (S), and adverse blood lipid profile (S) Improved cardiovascular biomarkers (S) Lower risk of Type 2 diabetes (S) and metabolic syndrome (S), and reduced abdominal obesity (M/S) Improved metabolic health biomarkers (S) Lower risk of colon cancer (S), breast cancer (S), lung cancer (M), and endometrial (uterine) cancer (M) Improved cardiorespiratory fitness (S) Improved cardiorespiratory and muscular fitness (S) Improved muscular fitness (S) Prevention of weight gain (S), weight loss, especially when combined with fewer calories (S), and weight maintenance after weight loss (M) Favorable body composition (S) Reduced depression (S), improved sleep quality (M), and — for older adults — better cognitive function (S) Reduced symptoms of depression (M) Increased bone density (M) and lower risk of hip fracture (M) Improved bone health (S) Prevention of falls (S) and — for older adults — better functional health (M/S) Note: (S) = strong evidence, (M/S) = moderate to strong evidence, (M) = moderate evidence. Source: Department of Health and Human Services (2008). Table 16-120 Health Benefits Found to be Associated with Regular Physical Activity 79 Body mass index (BMI) is used to define obesity and overweight in large-scale public health studies and epi- demiological analyses. BMI is a simplified proxy for percentage of body fat based solely on weight and height. In Metric units, BMI is equal to weight in kilograms divided by the square of height in meters. (In customary units, the weight in pounds must be multiplied by 703 before dividing by the square of height in inches.) A BMI of between 18.5 and 25 is judged to approximate normal weight; 25 to 30, overweight; and above 30, obesity (World Health Organization, 2011). 80 The emphasis here is deliberately on obesity rather than overweight and obesity. Studies agree on the adverse effect of obesity on life expectancy and health, though the mortality effect has decreased perhaps 20 percent in recent years, presumably in response to improved medical care. A major work drawing on three rounds of National Health and Nutrition Examination Surveys using normal weight as the reference category has found no excess deaths associated with non-obese overweight. Indeed, the estimated excess deaths for the overweight category were negative. It should be noted that there may be health (morbidity) and quality of life disadvantages of overweight that are not reflected in mortality (Flegal et al., 2005). In any case, the predominant weight-related public health problem is obesity, thus the focus on that category.

Adult Physical Health and Activity Relationships. Three studies highlighting the inverse rela- tionship between active transportation and physical disease are summarized in Table 16-121. Although none of these aggregate-data analyses were structured to prove causality, the presumed relationships stand out clearly. The first-listed study related nation-level active transportation mode shares (including transit) with obesity rates. It found the lowest active transportation shares and the highest obesity rates to be in the United States, and estimated inverse relationships across countries that apparently explained more than 3/4 of all variation. Measured body mass index (BMI) ranged from 34 percent obesity in the United States (12 percent active transportation) to 11 percent obesity in the Netherlands (52 percent active transportation) (Bassett et al., 2008). A subsequent reworking of this same international data, but focusing on walking and cycling alone, found the same general relationships, although the proportions of variation explained were lower. Pearson correlations of r=−0.54 and r=−0.20 were obtained for self-reported and measured BMI, respec- tively (Pucher et al., 2010). One may speculate that the loss in explanatory power is to some degree related to not having—in the revised analysis—the effect of transit use, with its associated walking for access. The 2nd study in Table 16-121, as already discussed, repeated the international analysis of the 1st study without inclusion of transit mode shares. It then extended the analysis to U.S. states and cities, on the basis of commute trips instead of all trips, with the results shown in the table. The city-level results were less robust than the state-level results, possibly in part because the health and travel data were collected using different area definitions, All results were, however, statisti- cally significant except when using measured BMI data at the international level. (The interna- tional measured-BMI dataset, in contrast to the self-reported-BMI dataset, was quite limited in size.) The study demonstrated that whether measured on the basis of countries, the 50 states, or the 47 largest U.S. cities, exercise and health are related to walking and bicycling prevalence in the expected manner: the more walking and bicycling, the more physical activity, and the less disease (Pucher et al., 2010). 16-362

The 3rd study, a 2010 “Benchmarking Project” covering U.S. bicycling and walking, related levels of walking and cycling to work with rates of high blood pressure and diabetes at the state level. It, too, used 2007 ACS and BRFSS data and found negative disease relationships. It confirmed a strong positive correlation across states between walk/bike-to-work rates and percent of adults with 30 or more minutes of daily physical activity, and a negative correlation with percent of adults who were obese. The correlations for the exercise and obesity relationships were r=0.72 and r=0.45, respectively (Alliance for Biking & Walking, 2010). As can be seen, the state-level analyses of the 2nd- and 3rd-listed studies support each other, but with somewhat different reportings of statisti- cal strength of the relationships. Adult Mental Health and Activity Relationships. The mental health component of public health was for a long time less well studied in terms of how it is affected by exercise, although HHS guide- lines now give the relationship a “strong evidence” ranking for adults (Table 16-120). A number of studies do tend to support a relationship similar to that for other types of illnesses (Heath et al., 2006, Department of Health and Human Services, 2008). The first two study examples presented in Table 16-122 link reduction in risk of dementia to greater amounts of exercise. The 1st-listed study, from Finland, used long-term follow-up data to confirm earlier findings that had been derived using relatively short follow-up times: findings that physical activity seems to promote 16-363 Study (Date) Process (Limitations) Key Findings 1. Bassett et al. (2008) Aggregate, national-level, cross- sectional analysis of obesity rates as they relate to active transportation mode shares (walk, bike, transit) in Australia, No. America, and Europe. Shares cited here are for all purposes of travel. (Raw and published data sources, some work- only shares, no exogenous-factor controls.) Active transportation shares ranged upward from 9% walk, 1% bike, and 2% transit in the U.S. (34% measured national obesity) to 30% walk, 5% bike, and 32% transit in Latvia (14% obesity). The inverse active transportation vs. obesity relationship had a Pearson correlation of r=-0.86 for self-reported and r=-0.76 for measured BMI. 2. Pucher et al. (2010) Aggregate state- and city-level, cross-sectional analysis of exercise sufficiency, obesity, and diabetes rates as they relate to percentages of commuters walking and bicycling, utilizing 2007 ACS and BRFSS data. (No controls, city-level ACS/BRFSS area coverage mismatches.) Positive relationships were established between walk/bike share and activity sufficiency, paired with negative relationships for obesity and diabetes. Pearson correlations, state and city, respectively, of r=-0.59 and r=-0.14 for activity, r=-0.31 and r=-0.28 for obesity, and r=-0.55 and r=-0.22 for diabetes. 3. Alliance for Biking & Walking (2010) U.S. “Benchmarking Project” state- level, cross-sectional analysis of hypertension and diabetes as they relate to walk- and bicycle-to-work rates. Hypertension correlation r=-0.54; diabetes correlation r=-0.66. (No controls. None of these three studies demonstrate causality.) Inverse disease vs. active transporta- tion relationship running from a 9% walk/bike commute share in Alaska, with 6% of residents ever told they had diabetes, and 25% ever told they had high blood pressure; to a 1% walk/bike commute share, 10% diabetes rate, and 33% hypertension rate in Alabama. Note: In the 3rd entry, the walk/bike commute shares shown are the highest and lowest, but the full range across states for diabetes is 5% to 12%, and for hypertension is 20% to 34%. Sources: As indicated in the first column. Table 16-121 Illustrative Examples of Aggregate Studies Relating Active Transportation to Physical Health at the National and State Level

lower risk of Alzheimer’s disease and other dementias (Rovio et al., 2005). The 2nd study exam- ined the linkage between exercise, specifically walking, and memory capability. It utilized a con- trolled experimental environment. Normal age-related decline in the anterior hippocampus of the brain, linked to spatial memory, was reversed for monitored walkers but not for study controls (Erickson et al., 2011). The 3rd study in Table 16-122 brings the built environment into consideration in the context of sup- porting mental health. It specifically links neighborhood walkability with reduced rates of depres- sion in the case of elderly males in the Seattle area (Berke et al., 2007a). The 4th study listed serves as a reminder that the mental health relationships with urban form remain less well established than for physical health. The original researchers found the lack of a demonstrable association between urban sprawl and depression, anxiety, and psychological well-being to be particularly surprising in view of the physical health relationships demonstrated and the frequent link between physical and mental problems. It is hypothesized that a smaller geography than the metropolitan- level used would offer a more refined analysis (Design, Community & Environment et al., 2006). Adult Physical Health Effects of Non-Motorized Transportation Features. Essentially all studies obtained from the travel behavior literature on adult physical activity effects of enhanced NMT systems and policies, and a fair number from the physical activity literature, express their findings in some form of travel behavior metrics. Such studies have been examined under individual pedes- trian and bicycle facility or program “Response by Type of NMT Strategy” subtopics. Some stud- ies obtained from the physical activity literature describe their findings, however, solely or primarily in terms of physical exertion measures or health metrics. These are presented here in this subsection, in most cases only here, and are listed below in Table 16-123. They are also taken into account in the “Adult and Child Public Health Relationships Summary” immediately following the discussion pertaining to children. 16-364

16-365 Study (Date) Process (Limitations) Key Findings 1. Rovio et al. (2005) Prospective analysis, with a 21-year- average follow-up, relating leisure- time physical activity to dementia and Alzheimer’s disease (AD) at age 65-79 years, controlling for socio- economic, health, and health-habit factors. (Physical activity definition not specifically related to NMT.) Mid-life physical activity was found negatively related to risk of dementia and AD. Finnish subjects engaging in such activity at least twice weekly had 50% lower odds of dementia and 60% lower odds of AD than more sedentary individuals. Relationships were significant for both men and women. 2. Erickson et al. (2011) Randomized, controlled, single- blind study of healthy but sedentary subjects, ages 55 to 80. Half were brought up toward optimal heart- rate moderate walking for 40 min., 3 days per week. Half were controls assigned to stretching and muscle- toning. (Memory improvement differences, per se, not significant.) After 1 year, hippocampal brain volume (HBV) declined by 1.4% for controls (normal aging) but increased by 2% for walkers. Walkers had 7 times the improvement of controls in maximum oxygen consumption. Tests correlated both participant HBV and aerobic fitness, before the trial and after, with spatial memory acuity. 3. Berke et al. (2007a) Cross-sectional analysis of fine- grained walkability scores for King County, WA, and Adult Changes in Thought (ACT) data, including measures of depression, obtained in a prospective, longitudinal cohort study. (Walkability might be a proxy for confounding variables.) Found, after controlling for various socio-economic and health status variables, a significant negative association between neighborhood walkability and symptoms of depression in older males (odds ratio approx. 0.32). (Negative relationship for females not statistically significant.) 4. Strum and Cohen – 2004, as summarized in Design, Community & Environment et al. (2006) Aggregate, cross-sectional analysis of metropolitan-level urban sprawl index calculations relative to 16 chronic physical health conditions plus depression and anxiety. (Large geographic scale to attempt using for measurement of mental health.) Higher sprawl (+1 standard deviation relative to -1) associated with 96 more chronic physical health problems per 1,000 population, but “no statistically significant or robust associations” with sprawl for mental health conditions after adjustment for other factors. Sources: As indicated in the first column. Table 16-122 Examples of Studies Relating Physical Activity, Walkability, and Sprawl to Mental Health

16-366 Study (Date) Process (Limitations) Key Findings 1. Eleven studies focused on exer- cise opportunities, and/or quality of urban environ- ment, summa- rized per SR 282 and not covered elsewhere in this subsection U.S., European, Australian, and Canadian studies reported on in 1989-2001, mostly cross-sectional, relating level of exercise to physical environment and/or availability of exercise facilities close at hand. (Practically all availability measures and neighborhood environmental quality measures used in these particular studies were based on self-reported perceptions). Physical activity was positively related to neighborhood opportunities for such activity in 3 studies, 1 of which found satisfaction with the neighborhood environment insignificant. In 1 study walking was positively related to neighborhood environment but not exercise facilities. Unmet desire to exercise was correlated with lacking or inadequate facilities in 1 study. No notable vigorous exercise relationships to neighborhood opportunities or characteristics were found in 2 studies that did find availability of home equipment to be significant, while 1 study found pay facilities important. Vigorous exercise or degree-of-exercise effects were not found significant in 3 studies, though 1 found more walk- ing in Chicago than in rural areas. 2. Booth et al. – 2000, Brownson et al. – 2001, and Powell et al. – 2003, each as summarized per SR 282 Australian national, U.S. national, and Georgia state survey analyses. National studies controlled for socio-economic factors. (Perceived facility availability, self-reported exercise, no controls in GA study.) Meeting recommendations for physical activity was significantly and positively related to conveniently located safe places to walk or exercise ranging from local sidewalks (or streets in GA) to parks, paths, treadmills, and gyms. 3. De Bourdeaud- huij et al. – 2003, Eyler et al. – 2003, and Wilcox et al. – 2000, each as summarized per SR 282 Ghent, Belgium, and two U.S. national women’s survey analyses, focusing on effects of sidewalks, related factors (e.g., lighting), and environment/safety. Controlled for socio-economic factors. (Self- reported facilities and exercise.) Sidewalks significantly related with more walking for some but not all gender/ethnic stratifications, with no significance at all in Wilcox study. A few other urban environment factors were found sometimes significant: no loose dogs, lighting, land use mix, etc. 4. Giles-Corti and Donovan (2003), and SR 282 (see “Sidewalks and Along-Street Walking” under “Response by Type of NMT Strategy” for Cross-sectional analysis in Perth, with controls, relating walking suffi- ciency to individual characteristics (e.g., behavioral control), social environment (e.g., exercise partners), and objectively measured physical environment (e.g., shops, sidewalks, attractiveness). (Environment measures limited to resident’s street; fairly well-off Only 17% met 12-walks, 360-minutes 2- week minimum standard. Those who added other exercise were found more likely to get a sufficient amount (78% of multiple exercisers). Individual, social- environment, and physical- environment factors had roughly equal effects on walking. Walking odds 47% higher with high vs. low access to open space, 25% higher with sidewalk and/ related studies) population.) or shops, 49% higher if area attractive. 5. Giles-Corti et al. (2003) Parallel study to Perth research described above, but focusing on BMI. Assessed prevalence of both overweight and obesity vs. normal weight. (Self-reported height and weight, and perceived acceptable walk to store, and walk or auto access to paths.) Living on highway strongly associated with overweight but not obesity. Odds of overweight/obesity were 32%/57% higher living on street with sidewalk on one side only and 40%/69% higher with no sidewalk as compared to dual sidewalks. Adequate shop/path access associated with normal weight. 6. Huston et al. (2003) Cross-sectional analysis of effects on exercise of availability of places for activity including sidewalks in Presence of trails and perceived general access to places for physical activity were the only factors found positively Table 16-123 Summary of Findings on Direct Relationships between the Non-Motorized Travel Environment and Measures of Adult Exercise and Health

16-367 Study (Date) Process (Limitations) Key Findings 6 North Carolina counties. (Limited list of places, all data self-reported.) and significantly associated with achieving recommended activity levels. 7. Hennepin County (2005) (see also Tables 16- 63 and 16-106 for seasonal usage and purpose of use distributions) The Hennepin County summer of 2005 trail user survey obtained 3,127 user responses with an approach designed to obtain data only once from each intercepted user of 3 rail trails in/around Minneapolis. (No non-user or before-trail data or comparisons, relationship of mod- erate vs. vigorous users unclear.) Trail users self-reported 4.8 days average of moderate physical activity, 3.0 of the days on the trail system; and 3.7 days of vigorous activity, 2.5 of the days on-trail. Among users, 61% met recommended exercise minimums through total moderate activity and 70% through total vigorous activity. When intercepted, 84% were bicycling. 8. Gordon, Zizzi, and Pauline (2004) Analysis of a trail access intercept survey of users of two new rail trails in Morgantown, WV, totaling 12 mi. of paved surface. Of randomly approached users, 98% participated. Ten types of exercise were probed, with “regularity” defined as at least 3 times weekly for 20 min. (Study design did not provide assessment of impact on overall community.) Of trail users, 22% were new exercisers and 78% were habitual. Increases in exercise following trail opening were found for 98% of new and 52% of habitual exercisers. Median increases overall were roughly 80% for new and 20% for habitual exercisers, with 25% overall becoming regular exercisers because of the trails, the only physical activity venue for 31% of new users. 9. Evenson, Herring, and Huston (2005) Longitudinal analysis of a rail-trail extension in Durham, NC, using before and after (follow-up) surveys of residents living within 2 miles. (Before/after survey seasonal mismatches, no proximity analysis.) Retrospective survey questions indi- cated an increase in physical activity but the primary longitudinal analysis did not support a finding of overall activity increase in this area already above average in sidewalks/trails. 10. Troped et al. (2001) (see text for cross- references) Health, environment, and trail use survey of Arlington, MA, residents with cross-sectional analysis. (Did not report on direct relationships be- tween trail proximity and exercise.) Trail use declined with distance of residence from trail. Trail users exer- cised 60% more days/week than non- users (3.7 vs. 2.3 days), for nonsignifi- cantly longer times (46 vs. 44 minutes). 11. Ewing et al. (2003), and Ewing, Brownson, and Berrigan (2006) Nationwide cross-sectional analysis, relating U.S. BRFSS data to a sprawl index based on population density and block-size averages/distribu- tions. In 2006, focused on young adult/adolescent BMIs, and added longitudinal analyses. (County- level aggregate-index application, longitudinal results lacked significance.) In 2003 found, controlling for individ- ual socio-demographics and behavior, small associations between sprawl and leisure walking (negative), and BMI, hypertension, diabetes, and heart disease (positive), statistically signifi- cant except for the last two. In 2006 obtained similar BMI results for young adults and adolescents. 12. Frank, Andresen, and Schmid (2004) Atlanta region cross-sectional study of 10,878-participant travel survey with health questions. Used logistic regression with socio-demographic indicators, built environment vari- ables, and calculated daily driving time and walking distance all in same formulation. (Little variety of land use forms in most of area, self- reported height and weight.) Land use mix associated with a 12% obesity likelihood reduction. Lesser reductions for residential density and intersection density, mainly for whites. Inclusion of the physical activity vari- ables moderately dampened the built environment effect estimates. Each km./day walked was associated with a 5% obesity likelihood reduction, vs. a 6% increase per daily hour of driving. 13. Rutt and Coleman (2005) El Paso border community cross- sectional analysis of health survey Higher BMIs were associated with lesser reported amounts of moderate results and objective environmental physical activity, higher socio-economic Table 16-123 (Continued) (continued on next page)

16-368 Study (Date) Process (Limitations) Key Findings data. Sample was predominantly semi-skilled, 79% Hispanic, moderately acculturated, and 71% female. (Unfamiliar modeled relationships encountered.) status, poorer health (which caused barriers to physical activity), and greater land use mix. No significant relationship found with density or sidewalk availability. 14. Saelens et al. (2003), and SR 282 Cross-sectional analysis using survey and accelerometer results, with socio-economic controls, related 8-factor walkability scores to moderate and vigorous activity in two San Diego neighborhoods on the basis of exercise time per 7 days. (Small sample; 107 adults.) Found significant moderate and total activity relationships, with high vs. low walkability moderate exercise of 195 vs. 131 min. respectively (210 vs. 140 min. for total exercise). Other measures with significance included walking for errands (30 vs. 15 min. unadjusted) and BMI (35% vs. 60% overweight). 15. King, et al. (2003) Similar approach as Saelens et al. (above) but relating a neighborhood convenience score based on 20-min.- walk destination accessibility (14 possible destination types) to walking levels and physical activity of older women in southwest Good utilitarian trip accessibility and perception of walkability linked to more physical activity. Greatest total activity differences (> +20%) were for accessibility to department, discount, or hardware store (+54%), food store (+37%), library (+26%), and walk/bike Pennsylvania. (Self-judged-and- reported walk times to destinations.) trail (+22%). An area’s 2 most walkable destinations produced most of effect. 16. Berke, Koepsell, Moudon, Hoskins, and Larson (2007b) (see also “Ped… …cycle Friendly Neighborhoods”) Cross-sectional analysis of fine- grained walkability scores for King County, WA, and Adult Changes in Thought (ACT) cohort study data, including BMI and frequency of walking. (Extent of walking self- reported.) Found, after controlling for various socio-economic and health status variables, a significant positive association between neighborhood walkability and walking, and a mostly negative albeit not significant asso– ciation between walkability and BMI. 17. Handy, Cao, and Mokhtarian (2007) Northern California cross-sectional and quasi-longitudinal analysis of days with moderate or vigorous physical exercise, in the neighbor- hood, in the last 7 days, as related to built environment features, control- ling for pro-bike/walk attitudes, neighborhood preferences, and socio-demographics. Land use mix, distance to nearest health club, and a number of other measures were objectively measured; other neigh- borhood characteristics were as perceived by respondents. (Impre- cise, self-reported activity measure.) Pro-bike/walk attitudes were signifi- cantly and positively associated with physical activity in the neighborhood, but neighborhood preferences (stand- ing in for “self-selection”) were not. Socio-demographics and neighborhood characteristics were of roughly equal importance to each other and more important than the attitudes. The features significantly and positively related to exercise included land use mix; distance away from nearest health club; and perceived attractiveness, socializing, and activity options and stores within walking distance. 18. Lawrence Frank & Co., SACOG, and Mark Bradley Associates (2008) Cross-sectional analysis of BMI and accelerometer-based activity, meas- ured in NIH Neighborhood Quality of Life Study (NQLS), in relation to plat-level land use within a 1-km. buffer; trip, person, and household sociodemographic data; network- based transportation system meas- ures; and traffic analysis zone (TAZ) regional auto/transit accessibilities. (NMT facilities not accounted for.) Greater physical activity (PA) related to more children, more workers, fewer cars/adult. Lower BMI and greater PA both related to higher residential density, higher intersection density (system connectivity), and presence of a park within 1 km. Lower BMI also related to better transit accessibility. Denser retail associated with more PA, but food outlets within walking dis- tance linked to slightly higher BMI. Table 16-123 (Continued)

16-369 Study (Date) Process (Limitations) Key Findings 19. Besser and Dannenberg (2005) (see also “User Factors” under “Underlying Traveler Response Factors”) Descriptive statistics were calculated from the 2001 National Household Travel Survey covering the walking activity involved in accessing U.S. public transit. Predictors were estimated for achieving in this way the recommended 30 min. or more of daily physical activity. (Trips with 2nd access mode, 5%, excluded.) During their reported-on travel day, 3.1% of respondents walked to/from transit, averaging 19 minutes total walk time, reaching recommended activity levels for 29% of transit walkers. The highest odds for being transit walkers were found among lower income, less educated, and non-white populations, and in denser urban areas. 20. CDC – 1999 as summarized per SR 282 Related perceptions of neighbor- hood safety from crime (4-point scale) to reported physical activity. (Self-reported measures.) Proportions of persons found active on the basis of walking, moderate activity, and vigorous activity were positively related to perceived safety from crime. 21. Matthews, Jurj, and Shu – 2007, Andersen et al. – 2000, and Evaluer- ing af Odense, Danmarks Nationale Cykelby – 2010, each as summarized per Pucher et al. (2010) and also (1st two only) de Nazelle et al. (2011) Prospective longitudinal health studies in China and Denmark followed Shanghai women (5.7 years average) who exercised or cycled for transportation and Danish men and women who cycled to work along with other physical activity. Odense, Denmark, reported health outcomes over time of a multi- faceted bicycling demonstration project. (Few details, no reporting on system changes in 1st studies.) two The Shanghai women who exercised or cycled for transportation had a 25% to 35% lower all-cause mortality risk. Cycling to work reduced premature mortality risk by 1/3 to 2/5 in Den- mark. In Odense, a 20% 1996 to 2002 increase in cycling levels paralleled a 5- month life expectancy increase for males. Odense has worldwide recog- nition for proactive bicycling policies and programs, 500 km. of cycling routes for a population of 186,000, and a 25% bike mode share for utilitarian trips. a 22. Socialdata (2008), Horst and Brög (2010) (see also the case study, “Variations on Individualized Marketing in the North West United States”) “Before” and “after” surveys of an individualized-marketing target population in Bellingham, WA, provided the opportunity to calculate person minutes and hours of active transportation, inclusive of transit access, from trip diary data. The large samples reflected 76% and 78% survey response rates. The “before” survey included a “sports hours/year” question. (No “after”- survey sports activity investigation.) Bellingham residents of all ages were in 2007 obtaining 175 hours/year of phys- ical activity on average (relative to the 130 hours/year annual equivalent of the HHS baseline exercise recommen- dation), 119 hours (68%) via active transportation and 56 hours (32%) from sports. In the individualized-market- ing target area, 122 hours/year average of active transportation before the 2008 marketing project increased to 153 hours in 2009, up 25%, going from 94% to 118% of the HHS physical activity standard. Notes: Where substantial additional information on individual studies is provided in text and tables or figures, this is noted — and the location within the chapter is given — in the first column. a The Odense bicycling policy, program, and mode share information is from the “Cycle City Odense” webpages (Cycle City Odense, 2011). Sources: As indicated in the first column. The notation “SR 282” is shorthand for Committee on Physical Activity, Health, Transportation, and Land Use (2005) together with Handy (2004). Table 16-123 (Continued)

Those of the studies in Table 16-123 that went beyond walking sufficiency to seek relationships with overall exercise sufficiency, and particularly with body weight and health, faced two layers of analytical burden. Not only were they subject to confounding exogenous transportation and physical environment factors, they were also exposed to additional socio-economic and cultural factors. These range from attitudes toward exercise to dietary implications of ethnicity. All such considerations reduce the likelihood of establishing statistically significant relationships from even the most seemingly obvious patterns. The 1st entry in Table 16-123 looks beyond the basic scope of this chapter to examine 11 studies focused primarily on activity impacts of access to facilities for exercise, a traditional area of pub- lic health research. Roughly 1/2 of these studies found significant relationships between level of exercise and perceived availability of neighborhood facilities, setting what might be regarded as a baseline standard for degree of success in establishing associations. The next four table entries cover eight studies that included assessing effects of having sidewalks. Six out of seven of these research efforts found positive effects on walking sufficiency or exercise to be associated with presence of sidewalks, although not necessarily for all social/ethnic groups. The 8th study (5th Table 16-123 entry) managed to quantify a substantial relationship between objectively-measured lack of sidewalks and both overweight and obesity, finding two sidewalks to be better than one, and either to be preferable to none. Four of these same eight studies also lend some support to the usefulness for promoting exercise of having walking or bicycling paths available within reasonable access. In addition, of the next five table entries (6th through 10th), one provides data implying a strong positive relationship between shared use trail presence and physical activity, one finds roughly 2/3 of trail users meet- ing physical exercise recommendations (likely above the norm even for the Minnesota location), two provide explicit evidence of the contribution of paths to increases in exercise, and one fails to find a relationship to exercise levels. The 3rd of these studies (8th Table 16-123 entry) provides evi- dence that paths may be particularly useful for attracting to exercise persons who were previously largely sedentary. New exercisers were found to choose the trail-use modes with least apparent risk, primarily walking, and nearly 1/3 were totally dependent on trail use for all of their exercise (Gordon, Zizzi, and Pauline, 2004). The 4th of this group, the one that found no evidence of trail impact on exercise (9th Table 16-123 entry), employs the “gold standard” of a prospective study, but mainly serves to illustrate how dif- ficult this can be when dealing with quasi-experimental designs. For example, trail construction was delayed but study deadlines were not, so the “after” survey was only 2 months after trail open- ing instead of the planned 1 year (Evenson, Herring, and Huston, 2005). The 5th study of this group (10th table entry) is perhaps out of place, as its outcome measure was trail use, but it does present implicit evidence of apparently strong positive trail impact on exercise levels. More detail on this Arlington, Massachusetts, study and its findings on trail access effects was presented in the “Response by Type of NMT Strategy” section under “Street Crossings” and also “Shared Use, Off- road Paths and Trails”—“Preferences, Route Choice, and Walk/Bikesheds.” The following eight Table 16-123 entries (11th through 18th) focus primarily on exercise effects of neighborhood land use characteristics. The first of these nine entries (11th table entry) examines urban sprawl, finding it to be negatively related to walking activity and positively related to sev- eral major diseases. The next six of these eight studies (12th through 17th) all find some measure of land use mix to be significant, although the study of largely Hispanic El Paso border community residents obtains the 16-370

notable result of finding land use mix to be positively related to higher BMIs. Each of the other five studies, as well as two studies earlier in the Table 16-123 listing, find some measure of mix to be negatively related to obesity or BMI, or positively related to physical activity. Three of these stud- ies explicitly measured land use mix objectively, including the El Paso border community study; three (with some overlap) used objectively determined or perceived acceptable walking accessi- bility to stores; and four (again with some overlap) used a measure that did not address mix explic- itly but effectively included it in a walkability, accessibility, or attractiveness score. The study based on Atlanta region data (12th table 16-123 entry) also identified significant nega- tive relationships between obesity and residential density, and also intersection density, a connec- tivity measure. These findings consistently held only for white respondents, however, with weaker or mixed associations in the case of African Americans (Frank, Andresen, and Schmid, 2004). Density often stands in as a surrogate for related effects, as discussed at length in Chapter 15, “Land Use and Site Design.” In this and similar cases, it may in part be standing in for good pub- lic transit service and concomitant higher use of transit, which is shown in the 19th table entry to be associated with substantial walking. Only the last of this group of six studies (the 17th study in Table 16-123), of all the studies summa- rized in the table, offers a strong claim to demonstration of causality. It controlled for pro- walk/bike attitudes and neighborhood preferences, and then found objectively measured land use mix and distance to health clubs to be significantly and positively related to walking. (The further away that health clubs were, the more walking seemed to be induced, presumably as an exercise substitute.) Significance was also found for positive relationships between walking and reported perceived attractiveness of the neighborhood, presence of socializing, convenience of activity options, and stores within walking distance. BMI was similarly related negatively to most of these factors, but the association did not reach significance (Handy, Cao, and Mokhtarian, 2007). None of the other studies summarized in Table 16-123, while they may have shown strong and/or obvi- ous relationships, found themselves in a position to claim demonstration of causality. That said, the 18th entry in Table 16-123 offers illustrative ties to the other studies, and has advan- tages of detailed, quantitatively-measured (not perceived) travel, exercise, and physical environ- ment parameters. (BMI assessment did use self-reported height and weight.) Although the King County, Washington, I-PLACE3S modeling involved did not include walking for transit access in the transportation dependent variables, its effect was implicitly included in the BMI and accelerometer-based activity measurement and modeling. The work adds another study to the tally of those finding residential density, intersection density, and availability of a nearby park or recreation opportunity to be positively related to activity and negatively related to BMI. A nega- tive BMI relationship was also found for regional transit accessibility (Lawrence Frank & Co., SACOG, and Mark Bradley Associates, 2008). This finding meshes with the determination recorded in the 19th table entry, discussed previously and below, that transit riding contributes significantly to meeting recommended daily levels of walking for health maintenance (Besser and Dannenberg, 2005). The King County exercise/BMI modeling did not find statistical significance in closeness to a bus stop, perhaps because the greater transit use associated with having a close-by transit stop (iden- tified in the transit use model) is counterbalanced by the lesser exercise obtained when the stop is close at hand. Finally, the King County study did not find its land use mix variable to be statisti- cally significant for exercise or BMI, but density of retail—measured as retail floor area ratio—was positively related to exercise. The numbers of fast food and other retail/food establishments within 1 km. were positively related to BMI, perhaps as an indicator of opportunity to obtain food, and lending weight to the researchers’ notation that, “Health outcomes are distal outcomes—more 16-371

steps removed from the urban form variables . . . [with] other factors . . . [e.g.] diet . . . play[ing] a much larger role . . .” (Lawrence Frank & Co., SACOG, and Mark Bradley Associates, 2008). Greater opportunity to obtain food could possibly underlie the finding in an El Paso border com- munity of a positive relationship between land use mix and BMI. The 19th entry into Table 16-123 demonstrates that substantial walking is inherently built into tran- sit use (Besser and Dannenberg, 2005), showing that the practice of some researchers of treating public transit as “active transportation” along with regular walking and bicycling has a strong basis in fact. This demonstration also at least partially explains why most studies that have exam- ined perceived or measured transit accessibility or use, primarily overseas (and not identified in Table 16-123), have found it to be related to more walking or exercise. The final three Table 16-123 entries (20th through 22nd) cover five studies addressing a mix of envi- ronmental, policy and program, and promotion and information situations and outcomes. The 20th entry identifies perceived neighborhood safety from crime as being a significant contributor to higher activity levels. The 21st table entry covers three international studies. The first two are not linked to any reported change in pedestrian and bicycle facilities, but on the basis of prospective lon- gitudinal health studies in Shanghai and Denmark, provide further evidence of linkage between reg- ular walking/bicycling and better health as reflected in longer life expectancy. The third study reports a life expectancy increase in parallel with increased bicycling, presumed to be in response to a demon- stration project furthering the proactive bicycle programs and policies of Odense, Denmark (Pucher et al., 2010). The last study, the 22nd and final Table 16-123 entry, provides information on both the total exer- cise pattern of residents of the Pacific Northwest city of Bellingham, Washington, and the effect on active-transportation physical activity of an individualized environmentally friendly transporta- tion marketing program. Bellingham is obviously somewhat of an outlier in the amount of physi- cal activity its residents obtain. The analysis provides a comparison with four Oregon cities, presented in Table 16-124. Among the five cities in an area of the United States known for outdoor activity, Bellingham had the highest average hours/year in both active transportation and in the total physical activity of transportation and sports combined. Individualized marketing achieved shifts to walking, bicycling, and transit use, in this already active environment, that increased indi- vidual hours of active transportation 25 percent (Horst and Brög, 2010, Socialdata, 2008), as set forth in Table 16-123. Quantification of individualized marketing effects on physical activity in other cities is found in the “Response by Type of NMT Strategy” section under “Walking/Bicycling Promotion and Information”—“Individualized Marketing”—“Home/Community-Based Program Effects on Physical Activity.” There it indicates that while Bellingham achieved a 31-hour annual increase in physical activ- ity per individual, 11 to 13 hours per year per person is more typical. 16-372

The considerably expanded pace of empirical investigation into relationships between the built environment and walking in particular provides a continuing flow of additional information. Perspective is provided by a 2005 and early-2006 update of not only the review in TRB Special Report 282 (Committee on Physical Activity, Health, Transportation, and Land Use, 2005) but also other reviews. The update authors conclude that, “Many of the conclusions from prior reviews are supported by this more recent evidence, particularly in the consistent associations found between walking for transportation purposes and density, land use mix, and proximity of non-residential destinations” (Saelens and Handy, 2008). The 2005–06 update authors report, however, that some of the associations with the urban envi- ronment do not strongly pertain to recreational walking. The observation is also made that side- walk infrastructure appears to be of differing importance, depending on travel category, with recreational walking and walking to school noted as types more influenced by quality of sidewalk infrastructure (Saelens and Handy, 2008). Selected individual-study findings from the update’s study summaries are presented in the applicable pedestrian and bicycle facility or program “Response by Type of NMT Strategy” subsections. Finally, a newer-still infusion of synthesis information is provided by an international review of research through 2010 prepared by 39 authors—many represented elsewhere in this chapter’s refer- ences listing—from research agencies, educational institutions, public health departments, and consul- tancies in Belgium, Canada, Chile, Denmark, Finland, France, Greece, Ireland, the Netherlands, New Zealand, Spain, Switzerland, United Kingdom, and the United States. This international review finds many areas of concern still poorly understood, stresses “the complexity of interactions among people, places, and the natural environment,” and highlights need to consider possible unintended consequences of actions and policies. Firm conclusions are nevertheless offered, as follows (de Nazelle et al., 2011): • Strong evidence links walkability factors involving transportation infrastructure and land use “with more active transportation and less driving.” 16-373 Measure Bellingham , Washington Bend, Oregon Eugene, Oregon Portland, Oregon Salem, Oregon Active Transportation Hrs./Yr. 119 63 80 91 74 Sports Hours/Year 56 106 70 36 38 Total Active Transp. and Sports 175 169 150 127 112 Transportation as Pct. of Total 68% 37% 53% 72% 66% Transp. as Pct. of HHS Minimum 92% 48% 62% 70% 57% Total as Pct. of HHS Minimum 135% 130% 115% 98% 86% Notes: Represents all individuals in each study area. Bellingham data is from a city-wide 2007 sample, pre-full-scale individualized marketing. Data for the Oregon cities was obtained circa 2005-2007. Source: Socialdata (2008). Table 16-124 Average Annual Hours per Resident of Active Transportation and Sports Physical Activity in Five Pacific Northwest Cities

• Active travel policies offer the potential for large public health benefits through physical activ- ity increases, combined with smaller benefits accruing from transportation pollution reduction. Information provided on crash and pollution risks in this international review is contained within the upcoming “Tradeoffs Between Health Benefits and Crash/Pollution Disbenefits” discussion. Additional, strategy-specific perspectives from the overall review are inserted in the “Adult and Child Public Health Relationships Summary” that concludes this “Public Health Issues and Relationships” subsection. On the whole, previously synthesized outcomes of providing new NMT facilities and pur- suing NMT-supportive policies and programs conform well with the new information. Health Benefits for Children of Enhanced NMT Systems and Policies The relationships of child and adolescent health to physical activity have not been as well developed as for adults, in part because key adverse outcomes—notably premature mortality—do not much evi- dence themselves prior to adulthood. There is not much doubt, however, but that the benefit of pre- adult physical activity is every bit as important. Table 16-120, in the preceding adult-oriented discussion, contains a column listing HHS determinations as to benefits for children and adolescents of regular physical activity. By examining this listing, juxtaposed with adult benefits, it can be seen that many childhood conditions ameliorated by physical activity are precursors to chronic health con- ditions of adulthood. In addition, low levels of physical activity among children have been linked to more immediate adverse effects including low physical fitness, low bone density, and higher risk of obesity (Davison and Lawson, 2006, Department of Health and Human Services, 2008). Obesity in children has—over the course of three decades—more than tripled for 6 to 11 year-old children, and more than doubled for other child and adolescent age groups. Although obesity in childhood may not be immediately linked with the most serious clinical symptoms, the adult obe- sity which most often follows is. Moreover, the social and emotional effects of childhood obesity, including negative stereotyping, stigmatization, and discrimination by their peers, are immediate. Some physical disorders are also immediate, including high blood pressure, sleep disturbances, menstrual abnormalities, orthopedic problems, impaired balance, insulin resistance, and even Type II diabetes, to name a few. As a result of the childhood obesity epidemic, diet and physical inactivity seem destined to supersede smoking as the leading cause of death (Committee on Prevention of Obesity in Children and Youth, 2005). The National Association for Sport and Physical Education has recommended that elementary school children partake in at least 30 to 60 minutes of appropriate physical activity on all or most days (Committee on Physical Activity, Health, Transportation, and Land Use, 2005). The HHS now seeks higher child and adolescent involvement in physical activity, 60 minutes or more each day (Department of Health and Human Services, 2008). Childhood Health, Development, and Activity Relationships. In the context of childhood health and obesity, the expenditure of calories is both a positive indicator and desirable outcome. Studies conducted in the London suburb of Hertfordshire in March and May of 2002 and 2003 sought to determine the energy expenditure of children in all activities of a week during the school year. Boys and girls in grades 6 (ages 10–11) and 8 (ages 12–13) were asked to wear tri-axial accelerom- eters and keep an activity/event diary for 2 weekdays and 2 weekend days. A total of 195 children successfully completed the assignment, representing 98 percent of the sample. Results were expanded to a 7-day week. Energy expenditure was calculated in activity calories per unit of time. Activity calories are those calories consumed by the body in physical activity, as contrasted to base- line bodily functions (Mackett et al., 2005a). 16-374

One subdivision of daytime events/activities examined employed 19 different event types. (Swimming, which might be considered a 20th type, had to be omitted.) Overall energy expenditure was 0.9 activity calories per minute, ranging from 3.1 when at school in PE class or a “games lesson” down to 0.6 when in other classes or at the child’s own home. The top six activities for energy expen- diture, in addition to PE/games lessons, were unstructured ball games at 2.8, structured ball games at 2.4, walking at 2.3, unstructured events not including ball games or play at 2.1, and school break/recess at 1.9 activity calories per minute. Next after that came bicycling at 1.7 activity calories per minute.81 A measure of the importance of travel mobility to exercise is the finding that average energy expen- diture outside the home is 1.1 activity calories per minute, twice the at-home rate if correction is made for the children’s weights. The number of cycling trips in the sample was too small for detailed analy- sis, but walking—mostly to and from school—was shown to be a major contributor to physical activ- ity. In contrast to PE or games lessons, which involved an average of 70 minutes per week, walking averaged 153 minutes per week. Overall, this makes the walking a roughly equal contributor to total exercise when compared with school PE and games lessons for younger children, and much more important in the case of older children, especially boys (Mackett et al., 2005a). In addition to the physical and mental health benefits, compelling evidence is accumulating that phys- ical locomotion adds to the quality of childhood development, most specifically to enhancement of spa- tial perception skills (Yan, Thomas, and Downing, 1998). Findings are mixed as to whether or not the benefit is greater for those school-age children allowed to walk and bicycle unaccompanied by adults. A pair of Italian researchers has postulated that differences in study outcomes relate to study approach. Measurable declarative knowledge of landmarks in the environment may be enhanced by walking with adults and learning place identifications from them. Measurable cognitive mapping of routes and points of interest—spatial or survey knowledge—may best be acquired by independent active travel. The researchers’ own study of schoolchildren in a suburb of Rome found 9- to 11-year-old elementary students who walked to school on their own more accurately mapped their route and locations of points of interest than children who walked accompanied by parents or who were driven (Rissotto and Tonucci, 2002).82 Roughly similar research covering elementary students in the vicinity of London, England, likewise found a positive relationship between level of independence and accuracy and detail 16-375 81 No mention was made of inability to fully record bicycling energy expenditure, but at least one other accelerometer-based study has noted difficulties. 82 The 46 children in this intriguing study participated in three special classroom exercises. They located their home and drew their route to school first on a sheet of paper that was blank except for the location of the school and two boundary features, a railroad and an arterial road. Later they did the same on a map of streets and buildings that was unlabeled except for the school, railroad, and arterial. In between they located pre- specified landmarks, along with any more they could think of, on the streets and buildings map. In both route drawing tests, students who walked to school on their own more accurately mapped their route (based on an average of four accuracy measures) than students who walked accompanied by parents. The accom- panied walkers in turn more accurately mapped their route than children who were driven. Two-thirds of the differences in mapping quality were statistically significant. Results from the landmark-identification test were less clear. Students who walked unaccompanied by parents did best, always significantly so com- pared to those walking with parents. Students driven to school did somewhat better, however, than students who walked with parents. A confounding factor may have been that although auto-driven children were less likely to be allowed out to play without adults than children who walked to school unaccompanied, they were more likely be allowed to play unchaperoned than children who walked to school guided by parents (Rissotto and Tonucci, 2002).

of maps drawn. Children driven to school were, for example, less accurate in recalling locations of local area landmarks (Mackett et al., 2007a). Fitting children with GPS and activity monitors, the London area study noted immediately above found that 8- to 11-year-olds walk from origin to destination at speeds about 2 to 3 times faster when accompanied by an adult than when left to their own devices. In the process, their intensity of energy expenditure as measured in 10−2 activity calories per kilogram of body weight per minute is 6.7 for boys and girls accompanied by adults versus 7.4 for unaccompanied boys and 3.7 for unaccompanied girls. The higher caloric energy expenditure by unaccompanied boys, despite slower origin-to-destination speeds, resulted from numerous detours and vigorous play enroute. Girls did much less of this (Mackett et al., 2007a). Nevertheless, given that more time was taken enroute when walking unaccompanied, the total caloric energy expenditure per unit of direct route distance would still be higher for unaccompanied than accompanied girls, and much higher for unaccompanied versus accompanied boys. Childhood Health Effects of Non-Motorized Transportation Features. Studies addressing effects on child and adolescent physical activity of enhanced NMT systems and policies are primarily from the physical activity literature and are smaller in number than for adults. Moreover, most express their findings in terms of physical activity measures. Practically all of the relevant studies encountered have been covered in a review of the literature by researchers at the University at Albany, New York (Davison and Lawson, 2006). Those of the studies reviewed that offer findings in terms of travel behav- ior metrics have been examined individually under the applicable pedestrian and bicycle facility or program “Response by Type of NMT Strategy” subtopics. Table 16-125 below presents those study findings expressed in physical activity metrics, and also summarizes the findings of the few child- focused studies covered in “Response by Type of NMT Strategy” subtopics. 16-376

16-377 Studies Process (Limitations) Key Findings 1. Boarnet et al. – 2005a, Ewing et al. – 2004 (both U.S.) Boarnet studied change in walking and biking to school in response to CA Safe Routes to School program (SRTS) improvements and Ewing modeled the effect on walk/bike school access of a variety of factors. (SRTS school access mode changes were as perceived by parents.) Boarnet found increases in walking/cycling relative to study controls for children who passed via SRTS sidewalk and road crossing improvements, and Ewing identified a signifi- cant relationship between main road sidewalk availability and higher student walking rates, while failing to find any relationship between bicycle lanes and walking/cycling to school. 2. Carver et al. – 2005 (Australia), Mota et al. – 2004 (Portugal) Carver conducted cross-sectional analysis with parent and child perceptions of various facilities and environmental conditions and Mota similarly studied various factors based on adolescent reports. (Both studies used self-reported physical activity measures as well as perceived environment measures) Carver found adolescents to walk/bike more where roads were perceived safe, there were more sports facilities, and — oddly — where cycling was less easy and convenience stores farther from home. Mota found higher activity adolescents to report greater access to stores and transit stops, more local recreational opportunities, and better neighborhood aesthetics. No significance was found for other “friendly neighborhood” measures. 3. Jago et al. – 2005 (U.S.) Jago analyzed neighborhood and ped/bike system characteristics and accelerometer-measured physical activity. (Accelerometers did not pick up cycling well.) Desirable sidewalk characteristics such as distance from curb and trees serving as a buffer showed a positive relationship with light-intensity physical activity. Cycling provisions showed no discernible effect. 4. Timperio et al. – 2004, Timperio et al. – 2006 (Australia) Timperio conducted cross-sectional analysis with parental or adolescent perceptions (2004) and objective measures (2006) of various area or school access conditions. (Parent reporting of walking and cycling.) Lesser walking/cycling was, for most age/sex combinations, associated with poor public transportation, heavy traffic, and multiple road crossings. Lesser walking/cycling to school was also associated with distances over 800 meters and (for ages 5-6) steep grades. 5. Braza et al. – 2004, Norman et al. – 2006 (both U.S.) Both studies employed cross- sectional analysis of objective measurements of school area (Braza) or neighborhood (Norman) characteristics along with objective measures of walking/biking rates or physical activity to explore explanatory relationships. Braza found higher surveyed walking and biking rates to school to be associated with greater population and intersection densities. Norman found accelerometer-measured activity to be significantly related to measures of walkability including retail accessibility (boys), intersection density (girls), and recreation opportunity accessibility. Note: Drawn from summaries of 33 studies, omitting those not directly relevant. For additional information see the “Sidewalks and Along-Street Walking,” “Street Crossings,” “Bicycle Lanes and Routes,” and “Pedestrian/Bicycle Friendly Neighborhoods” subsections of the “Response by Type of NMT Strategy” section. Source: Davison and Lawson (2006). Table 16-125 Summary of Findings on Transportation Infrastructure and Land Use Effects on Children’s Travel and Physical Activity

In considering the findings in Table 16-125, it is useful to establish context by quickly examining the role of non-transportation physical environment effects on childhood physical activity. It is also important to note that relationships between the physical environment and the physical activity of children differ from those of adults. Children are in different circumstances: they are not able to drive, spend long hours at school, have extensive time for recreation, gain substantial physical activity through play, and are under restrictions judged wise by adults (Davison and Lawson, 2006). Importantly, there is some evidence from studies of male children, adolescents, and children in general, that those who do not walk or bicycle to school, or who encounter restricted physical activity at school, are less active—not more—during out-of-school hours (Boarnet et al., 2005a). Focusing on non-transportation physical environment effects, no association was found in four of six studies between home exercise equipment and childhood physical activity. Significant positive association was found in 10 of 14 studies for availability of recreation areas or proximity to the home of recreation areas or parks and playgrounds. No discernible relationship was found in seven of nine studies between perceived safety and children’s physical activity, but three out of three studies found a significant negative relationship between physical activity and objective mea- sures of crime or area deprivation (Davison and Lawson, 2006). Review of Table 16-125 suggests that children’s physical activity may be related to a shorter list of trans- portation and physical environment characteristics than for adult physical activity, though final judg- ment should be withheld until there is a larger body of research on the activities of children. Of the nine studies in the table, three specifically identify higher incidence of walking where sidewalks are pres- ent, of higher quality, or improved over prior conditions. Similarly, three of the studies found walking and bicycling to be more prevalent in the presence of street crossing improvements, fewer roads to cross, and less traffic. Relationships such as these presumably have a strong association with the safety perceptions of parents and guardians, though explicit exploration of this aspect was not encountered. (For perspective on the role of adult supervision on childhood travel choices, see “Underlying Traveler Response Factors”—“Behavioral Paradigms”—“The Travel Choice Making of and for Children.”) Two studies attempted to find effects on children’s physical activity associated with presence of bicycle lanes or other cycling provisions and found none. Two studies found accessibility to stores to be indicators of more walking/cycling, while one did not. Measures of transit service adequacy were positively related to physical activity in two studies, which is logical, since not only is pub- lic transit a form of active transportation, it is also the only option children have for independent travel over longer distances. Two studies found a positive relationship between intersection den- sity and walking and biking activity, while one found no impact for most walkability measures. One study in Table 16-125, the first-listed within the 5th and last table entry, found a positive asso- ciation between walking and biking rates to school and higher population densities (Davison and Lawson, 2006), logical since higher density is presumably associated with shorter distances to school. Along the same vein, one of the studies in Table 16-123 (Ewing, Brownson, and Berrigan, 2006) examined adolescent weights along with those of adults, and found adolescent obesity to be positively related to urban sprawl. Tradeoffs Between Health Benefits and Crash/Pollution Disbenefits Concerns related to the exercise and health benefits of walking and bicycling are the negative effects on well-being of crashes, associated fatal and non-fatal injuries, and also exposure to pollutant emis- sions (de Nazelle et al., 2011). The concern about crashes, as they affect cyclists, was addressed in depth in a 1992 study for the British Medical Association. That research concluded that the benefits in terms 16-378

of life years gained from the increased physical activity of bicycling far outweigh any possible negative effects in life-years lost from injuries or fatalities. It was estimated that the aerobic exercise provided by bicycling compensated for crash risk by a factor of 20 to 1 in terms of average life expectancy (Hillman, 1992, Zegeer et al., 1994, Reynolds et al., 2010). Since the crash rate for bicycling exceeds that for walk- ing, it follows that walking benefits likely also strongly favor the activity over crash concerns. Additional support is provided by studies showing positive association between engaging in active transportation and reduction in all-cause mortality risk (Reynolds et al., 2010), a risk that would include crash-related deaths, and also air pollution effects. Modeling the effects of a twofold increase in walking and an eightfold increase in bicycling in London produced an estimate that the exercise would decrease premature mortality by 528 per mil- lion people, while the increased crash exposure would result in 11 fatalities per million. The corre- sponding years-of-life impacts were an exercise-induced increase of 5,496 life-years per million people versus 418 life-years per million lost through crashes (Reynolds et al., 2010). For pollutant risks to active transportation participants, however, unknowns presently impede making any such estimate. Some studies indicate that pedestrians and bicyclists are exposed to pollutants at lower con- centrations than persons using motorized transportation. Indeed, walkers and cyclists often can and do choose routes away from heavy vehicular traffic flows. Other research suggests that higher breath- ing rates and longer trip times for active transportation participants result in higher inhalation expo- sure. Evidence of in-travel active-transportation pollutant-exposure effects has been developed in controlled experiments, but degree of effect on health under everyday conditions has not been estab- lished. Drawing of firm conclusions regarding air pollution risks for pedestrians and bicyclists will require additional research (Reynolds et al., 2010, de Nazelle et al., 2011). Adverse health effects of air pollution are of elevated concern when considering the benefits and dis- benefits of compact living environments. Higher densities and street interconnectivity at the place of residence have been shown to reduce key air pollutants carbon monoxide (CO), volatile organic com- pounds (VOCs), and nitrogen oxides (NOx) overall, but may produce higher NMT exposures to CO and VOCs along with particulates in street environments within compact developments themselves. Measurements in Vancouver, British Columbia, Canada, have found higher concentration of primary traffic pollutants, but not secondary pollutants such as ozone, in more walkable as compared to less walkable neighborhoods. Not only the unknowns noted above, but also the lack of research simulta- neously addressing the multiple relationships between the built environment and public health, impede understanding of risk/benefit tradeoffs (Frank and Engelke, 2005, de Nazelle et al., 2011). The appropriateness of considering fatal and non-fatal injury and pollutant exposure risks as a partial trade-off against the healthful exercise benefits of walking and cycling may pertain more to the indi- vidual perspective than to society as a whole. Studies reported on in the “Safety Information and Comparisons” subsection (see “Other Traffic Safety Issues and Findings”—“Safety in Numbers”) find that where there are more walkers or bicyclists, crash rates tend to be markedly lower. It is argued on the basis of these relationships that achieving growth in walking and cycling will similarly result in reduced crash rates. Likewise, even if certain individual pedestrians and bicyclists are exposed to more pollutants by engaging in active transportation, the overall effect of more walking and cycling must be some degree of areawide pollutant emissions reduction, however modest (de Nazelle et al., 2011). From a societal perspective, then, the health benefits of more walking and cycling should not have crash injury or pollutant exposure costs deducted from them. There should perhaps even be a credit for pedestrian and cyclist injury reductions and lessening of overall pollution-related disease. It is reasonable to have some reservations about “safety in numbers” conclusions, even though practically all available evidence worldwide supports the relationship. There is the possibility of exogenous influences, and one must infer cause-and-effect relationships from analyses that are 16-379

largely cross-sectional (de Nazelle et al., 2011, Bhatia and Wier, 2011). At the least, however, it would seem appropriate to neither deduct nor credit fatal and non-fatal injury costs—in calcula- tions of societal health benefits accruing from increased walking and bicycling activity—until such time as crash-reduction effects can be more firmly established. Similarly, any attempt to adjust for pollutant exposure risks would seem premature, given the present state of the art. Credence is lent to the approach of not penalizing for safety and pollutant risks by two international comparative risk assessments recently reported on. These risk assessments, while addressing the uncertainties, conclude that physical activity benefits of active travel dominate other benefits and amply com- pensate for increased risks of injuries and pollutant inhalation (de Nazelle et al., 2011). Adult and Child Public Health Relationships Summary Research by both the public health and transportation planning professions makes it clear that there is no one “silver bullet” for achieving more walking and bicycling in the interests of either exercise or motorized transportation impact reduction. Individual outcomes appear to be largely incremental, but with significant synergism possibilities. For multi-pronged thrusts involving pol- icy shifts and comprehensive programs, results may be combinative to a substantial degree. They tend, however, to come gradually as program elements are put in place. Achieving fundamental shifts toward more healthy and environmentally sustainable levels of active transportation will take long-range commitment and comprehensive effort, hopefully informed by information such as that provided in this chapter on both traveler response and recreational/exercise response to NMT facilities, policies, and actions. The following adult and child public health summary is in effect an extension of the “Traveler Response Summary” within the “Overview and Summary” section at the beginning of the chapter. It looks at each pedestrian and bicycle strategy from a public health perspective and is based on mate- rial presented in both this “Public Health Issues and Relationships” subsection and, secondarily, the “Response by Type of NMT Strategy” section. Some additional summary observations are drawn from the 39-author international review “Improving health through policies that promote active travel: A review of evidence to support integrated health impact assessment” (de Nazelle et al., 2011). These additions, identified as being from “the de Nazelle et al. international review” or simply “the international review,” serve to enhance the coverage and currency of the summary. Physical and environmental factors such as pedestrian and bicycle infrastructure and programs are, of course, only one component of the influences on choice to walk or bicycle or otherwise exer- cise. Individual factors and socio-demographic circumstances play important roles as well, as has been highlighted where such considerations most notably pertain. For strategy categories where public health research is largely lacking, effects on public health must be inferred from reported changes in walking and bicycling activity, volumes, and travel mode shifts. This requires the fairly logical assumption that if active transportation is made more prevalent, then exercise and public health benefits will naturally follow. The inference-making is done from an overview perspective only—the primary travel demand impacts summary remains concentrated in the “Traveler Response Summary.” Sidewalks and Along-Street Walking. The preponderance of public health research on effects of sidewalk availability has found significant and positive relationships with walking sufficiency, exercise, or normal body weight, although not necessarily for all demographic groups. The research overall found relationships as strong as or possibly stronger than a group of studies on availability effects of a wide variety of mostly non-transportation activity opportunities including 16-380

parks and various forms of exercise facilities. On the other hand, it has been found that persons who engage in both active transportation and other exercise forms have the highest likelihood of meeting minimum activity recommendations. Transportation planning studies, although often less statistically rigorous, tend to support the findings of significant and positive effects on walking of sidewalk availability. Commercial area sidewalk ade- quacy and adjacency to stores appears to be a critical sidewalk system component for inducing utilitar- ian walking. Sidewalk continuity seems to be positively related to neighborhood walking for exercise and pleasure, and interesting streetscapes and activities to look at along the way may help. Isolated pub- lic health and transportation research efforts suggest that narrow and pleasant low-volume streets may offer suitable compensation for lack of continuous sidewalks, at least for able-bodied adults. Limited research has found that strategically located sidewalk improvements are associated with increased walking to school. A positive relationship between presence of main-road sidewalks and walking to school has also, in another study, been established. Overall, the most critical sidewalk system elements for support of exercise-inducing active trans- portation appear to be sidewalks providing school access; commercial area sidewalks; sidewalks along busy streets vital for linkage with public transit, shopping, and other centers of activity; and sidewalk continuity in support of recreational walking. Research on trade-offs between exercise benefits of walking and associated disbenefits of crash risk and pollution exposure suggests that health outcomes strongly favor walking over non-active alternatives. Risks of breathing emissions while engaging in active transportation need much more study, but conclusions presented in the de Nazelle et al. international review make it clear that even with poorly understood exposure risk for individual walkers and cyclists, the societal benefits of more active transportation and corre- spondingly lowered pollution overall are definitive. Walking has been shown to be positively asso- ciated with not only physical health but also mental health. Street Crossings. Scattered and diverse evidence, primarily from transportation planning studies but also from public health investigations, identifies the need to cross multiple, busy, or major arterials— particularly at locations without traffic signals—as a barrier to choice of walking and even bicycling. There is weak evidence that painted crosswalks encourage a slight increase in pedestrian activity. Unfortunately, except on two-lane streets with low to moderate speeds, painted crosswalks without traffic signals or pedestrian-activated lights or beacons may increase serious crash incidence. In con- trast, a redistribution and increase in traffic signals along a central London neighborhood boundary saw an increase in pedestrian volumes that was more than mere route-shifting. Small-scale studies indi- cate that painting crosswalks on routes to school has little effect on schoolchild pedestrian volumes, but full traffic signal installation of key crossings can lead to schoolchild crossing volume increases. Pedestrian grade separations, in addition to being expensive, will not be much used if they impose sig- nificantly more crossing time than crossing at grade. Pedestrian Zones, Malls, and Skywalks. Creation of special pedestrian places and ways in central business districts (CBDs) has not much attracted the attention of active living specialists. Traditional CBD pedestrian streets (malls) have been greatly affected by secular business activity trends and many were converted back to streets as retail activity fled U.S. downtowns. Some exam- ples, especially the variations known as transit malls because of shared transit and pedestrian use, have been moderately to outstandingly successful and are believed to have helped preserve the pedestrian-friendly downtowns they serve. As such, they have lessened sprawl with its associated health disbenefits. Sufficient studies were done of Boston’s Downtown Crossing pedestrian zone and the Nicollet transit mall in Minneapolis to demonstrate that their implementation brought increased pedestrian activity. The much-newer Broadway mixed-design mall in Manhattan’s Midtown saw short-term pedestrian volume increases equivalent to twice the preceding annual 16-381

long-term growth. Count-based studies of the Minneapolis and St. Paul weather-protected Skyway systems seem to show that they work with parallel sidewalks and crosswalks to maintain fairly constant downtown lunchtime pedestrian activity throughout the year rather than enduring a dip during northern winters. Bicycle Lanes and Routes. On-street bicycle facilities are another NMT improvement approach without much coverage by original research from within the public health sector. Bicycle lanes have been found to reduce both perceived and actual conflicts with traffic and to attract cyclists from nearby parallel roads, as well as potentially tapping latent demand. A small number of both Census-based corridor studies and city-level aggregate studies have linked either new bicycle lanes or more extensive networks of lanes with additional commuter bicycling. Representative of major successful installations is the Minneapolis-St. Paul experience of a 64 percent average increase in commute travel bicycle share, representing a 1.38 percentage point increase in corridor work-trip bicycle share. No research specific to non-commute bicycle travel share or physical activ- ity increases linked to bicycle lanes has been encountered. A number of installations have averaged roughly a 50 percent increase in total bicycle traffic on the treated streets but with major proportions representing bicycle trips diverted from other streets rather than more cycling. The user makeup of bicycle lanes, relative to other facility types, may possibly be tilted toward use by adults commuting to work. In that context, however, it is relevant that prospective studies in Shanghai and Denmark have shown bicycling to work can reduce pre- mature mortality risk by roughly 1/3. No conclusive evidence has been encountered that bicycle lanes are attractive to children. Bicyclist route-tracking research provides evidence that they are more attractive to experienced male cyclists than either inexperienced/infrequent cyclists or female cyclists. Infrequent cyclists and females biking the street system appear more attracted by quiet streets (some of which can be logically designated bicycle routes) and in particular by spe- cially treated bicycle boulevards. The response to new bicycle facilities may be gradual. Peak usage of the St. Kilda Road bike lane in Melbourne, Australia, did not quite double in the first year after installation, but had increased by a factor of 12 after 10 years. The bicyclist injury rate gradually declined, after an initial spike, to 1/5 that in the “before” condition. Cycle tracks and other forms of traffic-separated but on-road bicycle facilities appear to attract more usage than standard bicycle lanes, based primarily on over- seas experience, and were found in Montreal to attract well over twice the bicycle volume of par- allel undifferentiated streets. Shared Use, Off-Road Paths and Trails. The combined public health and transportation planning research on urban/suburban off-road shared use paths has, for the most part, isolated significant, positive contribution of path proximity to active transportation and exercise levels. Transportation researchers were for many years not able to establish as strong a relationship with commuter bicy- cling levels for paths as for bike lanes, which may be because a number of studied path systems featured indirect parkland routings and/or lacked hard surfaces, but newer U.S. research has placed them on at least equal footing for commuter attractiveness when well designed, reasonably direct, and well integrated. Shared use paths serve a broader clientele of bicyclers of all skill levels along with walkers, jog- gers, in-line skaters, and groups/families seeking recreation and exercise. Indiana studies that avoided common methodological problems found users of five out of six trails to be roughly equally split between wheeled users and users on foot, with the sixth trail—like other long partially–rural trails—bicycle-dominant. Although path use seems to be most common for adults neither particularly young nor old, scattered but generally strong evidence indicates that shared 16-382

use paths are attractive for learning cyclists, inexperienced cyclists, and new exercisers, predomi- nantly walkers. Two flat trails developed on old railroad roadbeds (rail trails) in a city of steep grades and limited sidewalks, Morgantown, West Virginia, attracted new exercisers (22 percent of users), increased exercise rates for a majority of users, and established regularity of exercise. Paths well aligned with business destinations serve utilitarian users who may, limited findings suggest, be deliberately combining need to make a trip with achievement of exercise. Survey-based studies show that among summer users of the Hennepin County (Minneapolis area) trail system, 60 to 70 percent meet exercise sufficiency guidelines. Of the total qualifying exercise of the trail users, 62 to 68 percent is obtained on the trails themselves. Pedestrian/Bicycle Systems and Interconnections. Systems approaches have produced the more notable shifts to use of active transportation. In Portland, Oregon, a 215 percent increase in bikeway system extent and critical improvements to four central area bridges were accompanied, over a 13 year period, by an estimated 210 percent increase in bicycle trips. Downtown-destined walk-to- work shares in Brisbane, Australia, tripled over two decades and bike-to-work shares increased six fold in parallel with development of a three-corridor path system, a major pedestrian and bicycle bridge, and more downtown housing. Individual bridges and closings of missing links in paths have seen anywhere from modest to substantial usage with tributary facility volume growth increases ranging from 1/3 to tripling (including redistribution effects) depending on context. On a new bridge with pathway over the harbor in Charleston, South Carolina, 2/3 of all walkers and cyclists reported increased physical activity. A majority of commuters on both this bridge and the Brisbane pedestrian and bicycle bridge were found to be deliberately combining exercise with their commute. Various analytical approaches have shown the importance of good connections. Research is begin- ning to show that route directness is a walking inducement, as are higher-than-average ratios of neighborhood pedestrian connectivity relative to vehicular connectivity. Effects on exercise per se, and health, have not been quantified but presumably are proportional to positive outcomes of the types identified for sidewalk, bikeway, and path improvements. Pedestrian/Bicycle Linkages with Transit. Almost one in six of all U.S. walk trips in 2001 and 2009, and roughly one in 10 of U.S. bicycle trips, involved accessing public transit stops and stations. An estimated 29 percent of transit users achieve the recommended 30 minutes or more of physical activity a day solely by walking to and from transit. There has been little definitive research on active transportation increases achievable through transit access improvements other than transit system expansion effects. Positive association has been noted in some cross-sectional studies between good transit access and incidence of walking by adults and by children. Stated preference experiments and modeling based on cross-sectional travel survey data have been used to explore effects of access improvements, including provision of bicycle parking. Parking for cyclists has been estimated to be nearly as important, to much more important (depending on cyclist experience levels) relative to other access improvements. Walking and bicycling transit access shifts in the range of 2 to 7 percent increases have been estimated with cross-sectional mod- els for suburban Chicago commuter rail stations. Bike-on-bus and bike-on-rail programs, a rela- tively new development, expand the reach of transit service and typically serve on the order of 1 percent of riders on systems well equipped to offer the service. The exercise benefits of accessing transit (including users of bike-on-bus programs) reach a disproportionately lower-income popu- lation relative to most active transportation enhancement programs. Point-of-Destination Facilities. Secure and weather-protected bicycle parking is an obvious example of point-of-destination facilities that can be provided to make it more feasible or easier to use non- motorized transportation. Workplace showers and changing facilities, and nearby convenience services 16-383

to decrease need for an automobile, are other examples. These are key strategies that remove barriers to walking and bicycling for utilitarian purposes, but have not been studied in isolation to assign exer- cise encouragement or health effects to them. A study in the United Kingdom based on stated prefer- ence research estimated a 22 percent increase in commuter cycling would be associated with provision of secure indoor bicycle parking and showers. A Los Angeles area study found walk- and bike-to-work shares about 1/3 higher (starting at 2 to 4 percent) with an assortment of workplace amenities includ- ing a high aesthetic appeal and perception of safety. Although early results from a Minneapolis bike- sharing application show a majority of users treating it as an alternative to other forms of active transportation, 1/3 reported their alternative as traveling by auto or not making the trip at all. Pedestrian/Bicycle Friendly Neighborhoods. A major thrust of transportation planning and pub- lic health active transportation research, including a number of multi-disciplinary studies, has been to examine the effects of urban design. While individual design elements have been found unim- portant in one or a number of studies, and some logical relationships have not achieved formal sta- tistical significance, the overall direction seemingly established is that pedestrian and bicycle friendliness of neighborhoods matters—especially in the case of walking. Bicycling evidently is an individual choice only moderately associated with the local land use and design environment. Urban sprawl has been found negatively related to walking and positively related to several major health problems. One study is of particular interest in that it offers a claim to demonstration of causality, having con- trolled for pro-walk/bike attitudes, neighborhood preferences, and socio-demographics. It found positive relationships for walking with objectively measured land use mix, perceived attractive- ness of the neighborhood, stores within walking distance, and convenience of activity options. BMI was related negatively to most of these factors, albeit not with statistical significance. The study took place in Northern California. The same factors show up consistently with positive relationships to walking and health measures in the vast majority of studies, joined by higher land use density (which brings activities closer together), proximity of jobs, street intersection density (a surrogate for pedestrian system connec- tivity), and better public transit accessibility. In the case of children, broadly-defined land use mix may not be important, although retail proximity is, and accessibility from the home to recreation areas or parks and playgrounds should be added to the list of indicators of greater physical activ- ity. Distance to school is critical, with multiple studies consistently finding distance to school to be inversely related to choice of active transportation for school access. Objectively measured crime or area deprivation shows, in a number of studies, a negative relationship with activity of children. Further support for the importance of land use density and mix; proximity to home of shopping, services, and transit stops; more and better-quality sidewalks; system connectivity; and adequate and safe bicycle facilities as built-environment features associated with higher probability of walk- ing, bicycling, and using transit is provided—on the basis of research through 2010—by the de Nazelle et al. international review. The strong association with density of many supportive built- environment attributes continues to make difficult the isolation of density effects and attributes usually but not always found in combination. Two recent studies in Belgium and the United States found residents of neighborhoods classified as having good walkability to be spending 35 to 49 minutes more per week engaged in physical activity than persons in low-walkability neighbor- hoods. A Minneapolis study, however, found neighborhood type to be associated not with the amount of physical activity but with its nature, such as walking for transportation versus for recre- ation versus visiting a gym. This exercise trade-off possibility plus the question of whether added physical activity accrues from sedentary people starting to exercise or active people exercising more are questions not yet resolved. 16-384

NMT Policies and Programs. Similar to the situation with pedestrian and bicycle system expansions and interconnections, exercise and health effects have not been empirically derived for instances of translating policy into substantial city-wide non-motorized transportation programs, although promis- ing forecasting has been done in connection with policy planning in U.S. urban areas such as Seattle. In terms of travel effects, the exemplary programs in Portland and Brisbane were covered above. An additional 4 years of Portland experience has shown an accelerated (pre- “great recession”) growth, giving a 1991 through 2008 bicycling exponential growth rate of 9.6 percent per year, despite a slow- ing of system expansion in the later years. Possible reasons for the continued and accelerated growth include a lag effect in the response to earlier pre-2004 bikeway system expansions, a doubling of gaso- line prices from 2004 to 2008, Portland’s ongoing individualized marketing program, and growing vis- ibility and general acceptance of cycling. Davis, California, still an ideal place for the ordinary citizen to bicycle, offers a cautionary tale. The outstanding bicycle-to-work-trip travel mode share of 14 percent in 2000 and the University of California Davis student bicycle-to-campus share of 48 percent in 2007 actually represent major declines attributed in part to loss of citizen involvement and municipal expertise along with weak- ened university parking policy. Boulder, Colorado, has focused not on a single active transportation mode but instead has been pursuing a goal of concurrent enhancement of pedestrian, bicycle, and transit facilities and services. Active transportation mode shares between 1990 and 2006 grew by 26 percent for all trips by all residents, by 16 percent for the commute trips of employees working in the city, and by 30 percent for worker midday-trips. Northern European programs provide the exam- ple of major non-motorized transportation turnarounds starting in the 1970s and leading to circa 1995 combined walk and bike mode shares 5 to 6 times higher than 1995 U.S. NMT travel shares. Of particular interest in the context of childhood exercise are the Safe Routes to Schools (SRTS) pro- grams in the United States and elsewhere. California elementary school SRTS studies showed a 46 percent average increase in schoolchild walking in response to sidewalk improvements and one- half that in response to crosswalk signalization, but inconclusive results for other intersection improvements. Among forms of encouragement programs, multifaceted approaches and “walk- ing school bus” programs achieved walking increases—or walking and bicycling increases— within the 6 to over 60 percent range, omitting outliers. Basic coordination and encouragement programs in England produced negligible results, but encouragement with daily tracking of stu- dent modes of access combined with student recognition and sometimes other program actions has proved effective in English, Canadian, and U.S. applications. In a 19-category breakout of child activities, ranging from PE classes to relaxing at home, walking and bicycling were in the top seven for energy expenditure. Walking to school has been shown to improve spatial information retention. Some but not all studies suggest that walking indepen- dently improves spatial cognition, such as is reflected in mapping ability. The de Nazelle et al. international review observes that assessing policies designed to effect behav- ior change requires consideration of “bundles” of programs and strategies, and that both interac- tions and opportunities for “co-benefits” (multiple benefits) require attention. The desirability of considering auto disincentives among policy options is noted: In the example of London’s central area congestion pricing, the congestion charge and cycling infrastructure investment saw a dou- bling of bicycling levels. The international review also speaks of a “cultural shift” that may occur when walking and cycling reach a level that signals “that these are safe and enjoyable and perhaps even fashionable activities.” This sort of “virtuous circle” or “critical mass” was noted above as a possible reason for continued and accelerated cycling growth in Portland, Oregon. Walking/Bicycling Promotion and Information. Results of most mass-market walking, cycling, and transit use promotion programs tend to be inconclusive and not encouraging. Exceptions are 16-385

seen in some instances where individuals have been induced to try an active transportation mode they have had little direct experience with, as in a bike-to-work-day event. As marketing becomes more focused, short-term results become more positive, though long-term impacts have been lit- tle studied and dissipation of beneficial effects is a major concern. A higher order of targeted mar- keting intensity is provided by the promotional and informational protocol known as individualized marketing. In environments typical of Australia and the United States, benchmark applications have produced walk-share gains among trips for all travel purposes of 1 to 4 percent- age points and bicycle-share and transit gains of 1 to 2 percentage points. Surveys on three continents have demonstrated substantial durability of mode shifts after 1 to 4 years. The mode shifts translate into added physical activity. Three U.S. estimates of physical activity gain linked to individualized- marketing-induced active transportation lie in the range of 11 to 13 hours per person per year averaged across the contacted target area population. In a fourth estimate, a 2008 large-scale program in Bellingham, Washington, averaged a walking and cycling time increase of 31 hours per person per year including transit and parking access mode changes. Public health individual or household interventions have also been tried in an effort to increase walking. A major synthesis effort focusing on outcomes of such interventions, primarily under- taken at a research scale in the United States and Australia, concluded that more walking clearly can be encouraged when the interventions are tailored to individual needs. In a typical example, three interventions involving 12 to 16 counseling sessions over 12 to 24 weeks, communicated to sedentary adults via telephone or Internet, resulted in net increases in self-reported walking of 32 to 62 minutes per week as measured after 3 to 6 months. Evidence was found less convincing in the case of measures taken at the institutional level, whether workplace, school, or community. A major question, which only five of the 27 reviewed studies examined, is sustainability of interven- tion results over time. A majority of the five studies determined that walking increases recorded at 4 to 16 weeks were not sustained as measured at 24 weeks or 1 year. A Pittsburgh intervention, however, that started with 8 weeks of twice-weekly walking training for post-menopausal women, followed with various encouragements and even home visits, produced sustained walking increases that stood at 7.3 miles per week in a 10-year follow-up. The de Nazelle et al. international review, as discussed above with respect to policies and pro- grams, identifies promotional strategies as a partner in bundles of strategies that may even bring active transportation to a “certain ‘critical mass’ ” of greater public acceptance and normality. On the other hand, the same international review notes that practitioners are putting more emphasis on changes in the urban environment to engender more physical activity. These conclusions are not necessarily in conflict—strategy synergism appears to offer enhanced outcomes. Traffic, Energy, and Environmental Relationships Walking and cycling trips may be broadly characterized according to purpose as being either recreation/exercise NMT trips or transportation/utilitarian NMT trips. Both are important from the perspectives of public health and quality of life. Only utilitarian NMT trips, however, can nor- mally be viewed as possible substitutes for auto use. Therefore, it is only walking and cycling trips made for utilitarian purposes that in theory have the potential to affect congestion, energy use, and pollution (Krizek et al., 2007). Driving Avoidance Estimates The Nonmotorized Transportation Pilot Program Evaluation Study developed initial estimates, based on the 2006 baseline survey alone, of the amount of driving currently avoided in five U.S. urban 16-386

areas thanks to present-day walking and cycling choices. Refined estimates along with energy and emissions savings calculations are expected as part of the final pilot program documentation (Krizek et al., 2007). The baseline survey was briefly described above under “Public Health Issues and Relationships” (see “Baseline Walking and Bicycling Activity,” including Footnote 78). Only utilitarian travel was considered in the estimates of NMT substitution for driving. Survey constraints required limiting the analysis to adult travel, thus chauffeuring of children was not addressed. Work commute and other utilitarian trip distances and daily walk and bike trips per adult were estimated from the five-area pilot survey. A low estimate was prepared on the basis of calculated “reference trip” distances and a high estimate was drawn from the daily walk and bike travel time totals reported. These steps were followed by survey-based estimation of degree of walk or bike substitution for auto driving relative to other travel options. Commuter driving substitution was computed using the ratio of walk or bike commuters listing driving as their alternative mode to the total of walk or bike commuters reporting any alternative mode. Across the five communities, 32 percent of bicycle commute trips and 36 percent of walk commute trips were estimated to represent driving substitution. Non-commute utilitarian trip driving substitution was estimated based on alternative modes reported for applicable reference trips. In this manner, 93 percent of non-work utilitarian bicycle trips and 95 percent of such walk trips were estimated to be replacements for driving. The overall five-community driving avoid- ance estimate is summarized in Table 16-126 (Krizek et al., 2007). 16-387 Community Low Estimate (miles/day) High Estimate (miles/day) Adult Average (miles/day) Daily Driving per Adult (mi.) Percentage Reduction Columbia, MO 0.40 0.50 0.45 15.1 3.0% Marin Co., CA 0.56 0.78 0.67 23.6 2.8% Minneapolis 0.69 0.94 0.82 20.7 3.9% Sheboygan 0.16 0.35 0.26 22.3 1.2% Spokane 0.22 0.40 0.31 25.9 1.2% Total 0.40 0.59 0.49 n/a n/a Source: Krizek et al. (2007). Table 16-126 Reduction in Auto Driving Estimated for 2006 Levels of Walking and Cycling The walk and bike modes of travel together were estimated to replace approximately 1/4 to 3/4 miles per day of driving per adult resident, depending on urban area characteristics. Present day use of NMT modes, in the context of 15 to 25 miles per day auto travel in the communities studied, thus appears to reduce driving by 1 to 4 percent. Roughly 70 percent of this avoided driving was attributed to walking, and the rest to cycling. Although bicycle trips are longer than walk trips, trip length dif- ferences are overshadowed by the much larger number of people who make utilitarian walk trips on any given day. The researchers note the many factors that render difficult the estimation of NMT effects on driving, and report an earlier analysis by Handy and Clifton in 2001 that estimated walk trip substitution of 2.1 miles of driving per month (Krizek et al., 2007). That would be 1/5 the walk component of the average overall interim driving substitution estimate prepared for the pilot project, but within range of the low estimate for Sheboygan County, Wisconsin.

Facility and Project Impacts A different perspective is provided by estimates of the traffic and emissions reductions attainable from individual new pedestrian and bicycle facilities and programs. An evaluation and assessment of Congestion Mitigation and Air Quality Improvement Program (CMAQ) projects funded between 2002 and 2007 takes such a perspective, noting that pedestrian and bicycle projects generally serve multiple goals ranging from improving mobility and access for non-drivers to improving NMT safety. That said, such projects were determined to have modest effects on vehicle trip and emissions reductions. The sum of estimated VOC, CO, and NOx reductions for individual projects examined were 4.6 kg./day for the 8.3 mile Swansea bikeway facility in Massachusetts, 3.6 kg./day for a 4.3 mile bike path to Pinhook Park in Indiana, 8.5 kg./day for a Transit Bike Depot in Colorado, and 42.8 kg./day for New York City’s CyclistNET marketing program. Estimated reductions in vehicle trips per day ranged from 83 to 902, with the higher reduction pertaining to the New York City program. Effects on congestion were not esti- mated given the modest numbers of vehicle trips removed relative to total travel (Grant et al., 2008). It will be noted that, outside of the New York City program, the largest CMAQ pedestrian and bicycle program emissions reduction—among the four projects examined—was for the enhanced bicycle-parking-at-transit in Colorado (Grant et al., 2008). Looking at a different mix of four alter- natives, Chicago-area analyses reported in FHWA’s National Walking and Bicycling Study of the early 1990s found secure bicycle parking at transit stations (for bike-and-ride) to be the most cost- effective approach for reducing hydrocarbon emissions incurred in accessing transit service. Not among the alternatives studied were enhanced paths and walkways (Replogle and Parcells, 1992). Transit access trips are of particular interest for emissions reductions. Such trips, when made with conventionally powered vehicles, have higher-than-average pollutant emissions per mile because they usually begin with a cold-start for the engine and—when logical candidates for walk and bike substitution—are normally short enough that there is not much opportunity for engine warm-up. A conventional automotive engine running cold emits over 4 times the CO and about twice the VOCs per mile as when running hot. Indeed, an estimate prepared for the Chicago area’s Metra commuter rail system found that Metra pas- sengers who drove to the station were producing between 50 and 90 percent of the pollution they would if they drove all the way into the downtown “Loop” district (Wilbur Smith and Associates et al., 1996c). A related consideration is that although most bicycling in an area such as Chicago occurs in the 7 months from April to October, that is also the most critical time of year for atmospheric pollution in the form of ozone, for which VOCs (light hydrocarbons) are an essential ingredient (Pinsof, 1982). In 1980, the Chicago Area Transportation Study (CATS) and the Illinois Department of Transportation (IDOT) undertook a rare quantitative evaluation of bike-and-ride effects on emissions. IDOT had installed bicycle racks at nine Edens Expressway corridor commuter rail stations near Chicago, in July 1979, with a capacity of 457 bicycles. Additional bicycles parked at the stations in the new racks totaled 222 by August. The associated vehicle travel reduction was estimated at 1,739 VMT per day. (The emissions estimates are not reported here because of the many significant automo- tive emissions-control improvements made in the three-plus decades since.) National Walking and Bicycling Study estimates suggest that, for the circa 1990 vehicle mix, 150 gallons of gasoline per year are saved for each park-and-ride commuter attracted to bike-and-ride. In the case of com- muters previously using an automobile for the entire trip, the corresponding savings is an average of 400 gallons per commuter diverted to bike-and-ride (Replogle and Parcells, 1992).83 16-388 83 It does not appear that these gasoline savings estimates have any adjustment for auto-access or auto-commute carpooling, or for reduced bicycling in cold or inclement weather. Both would reduce the annual fuel savings.

Program Impact Model Findings Bicycle program estimates of energy savings are available from a “Conserve by Bicycle Program Study” conducted by the Florida Department of Transportation (FDOT). A key goal of the FDOT study was to determine the energy savings that could be realized if more and safer bicycle facili- ties were built. Under the study research plan, a corridor-level multinomial logit mode choice model pertaining to utilitarian bicycle trips was developed. The travel data source was intercept survey results from 17 corridors with various types of bicycle facilities: shared use lanes, bicycle lanes, shared use paths adjacent to the roadway, and independent shared use path alignments. With the model, the energy savings (in terms of fuel costs saved) could be estimated based on esti- mated mode shift from the motor vehicle to the bicycle mode (Petritsch et al., 2008). The mode choice model, calibrated from a data set of 1,554 motorists, 55 transit riders, 11 bicyclists, and 21 pedestrians, exhibited an R-Square of 0.91. The model included variables representing high- way congestion, transit quality of service, trip length, network friendliness for bicyclists and pedes- trians, quality of bicycle and pedestrian facilities, and population and employment density. It was not possible to include a household income variable and an additional bicycle friendliness mea- sure because of limited variability in the relevant variable values within the data set and the lim- ited number of non-motorists making utilitarian trips. The Nebraska Avenue Corridor in Tampa was selected to illustrate the calculations of energy savings resulting from modeling different types of bicycle facilities with the mode choice model. Given key assumptions concerning selected parameters such as average utilitarian trip length, miles per gallon of fuel, and fuel price per gallon, the study predicted fuel costs that would be saved relative to the no- bicycle-facilities condition by providing bicycle lanes, shared use paths adjacent to the roadway, and independent shared use path alignments. The estimated annual transportation corridor fuel savings were $3,452, $113,858, and $387,596, respectively (Petritsch et al., 2008). A Washington State Department of Transportation (WSDOT) study of urban form alternatives and pedestrian and transit improvements as greenhouse gas (GHG) reduction strategies is the first the study authors knew of to relate sidewalk availability with VMT and GHG emissions (CO2 in this case). Information on 2,699 King County households and their associated 39,297 trips from the Puget Sound Regional Council (PSRC) 2006 Household Travel Survey formed the travel data for the analysis. Seattle and six suburbs provided GIS sidewalk data indicating lack or presence of sidewalks on one or both sides of all streets. A “Sidewalk-to-Street-Ratio” variable was used that described the ratio of total side- walk length within a 1 km. network buffer compared to total length of street right of way within the buffer. The maximum theoretical value was 2.00, which would indicate sidewalks on both sides of all streets. In addition to the key sidewalk data, other variables studied in the analysis fell into the cate- gories of urban form, transit service, travel costs, and socio-demographic and household characteris- tics (Frank et al., 2011). The multivariate regression equations derived from the study exhibited the expected direction of sidewalk effects on VMT and CO2 reduction and the sidewalk variables were thus retained in the model. The sidewalk variable was not found to reach statistical significance in explaining VMT, but was marginally significant in explaining CO2 emissions.84 The lack of variation in the data 16-389 84 This is a logical outcome to the extent that vehicle trips diverted to walking would tend to be short and not very consequential from a trip mileage perspective, but important to emissions reduction because of the vehicle start-up and turn-off cycles eliminated.

16-390 set, skewed as it was towards the more urban and walkable parts of King County, may have contributed to this lack of strong statistical significance as well as to insignificance of variables such as residential density and intersection density—known to usually relate importantly to VMT and CO2. Such limitations notwithstanding, the study results still provide early evidence of the potential effectiveness of sidewalk availability for reducing VMT and CO2. For example, increasing sidewalk coverage on both sides from 30 percent to 70 percent of all streets was estimated to result in 3.4 and 4.9 percent decreases in VMT and CO2, respectively. By comparison, using the same model, an increase in hourly parking cost from $0.28 to $1.10 resulted in 11.6 and 9.9 percent estimated reduc- tions in VMT and CO2, respectively (Frank et al., 2011). Economic and Equity Impacts Active transportation has often been largely overlooked and therefore undervalued in conventional transportation studies. The full extent of walking in particular is difficult to survey, and once sur- veyed, may be partially “defined out” of trip data by classifying multimodal trips according to a hier- archy that never affords NMT “primary mode” status when combined with motorized travel. This particular analytical problem, among others, is introduced within the “Analytical Considerations” subsection of this chapter’s “Overview and Summary.” Ironically, the oversight diminishes under- standing of one of walking’s particularly important functions—the critical system interconnectivity it provides. Walking links homes and businesses to public transit, connects parking facilities with shop- ping and workplaces, and knits transportation services together. There are other ways that conventional survey and analysis procedures have contributed to under- statement of NMT valuation. One is the heavy emphasis on commute trips, not the province of highest mode shares for NMT. Another is the frequent omission of travel by children, heavy users of the walk and bicycle modes as an alternative to parental chauffeuring. Still another is the poor representation of trips short enough to “disappear” within the confines of traditional traffic analy- sis zones (TAZs), the finest level of geographic disaggregation typical until recently. Short NMT trips link businesses with each other, offices with restaurants and stores, and homes with neigh- borhood activities. Only with analysis of such surveys as the 2001 NHTS (and its successor the 2009 NHTS) is the full scope of NMT activity and contributions becoming clearer (Victoria Transport Policy Institute, 2007). It is now understood, still not counting those walk trips linked to auto park- ing, that some 12 to 13 percent of all trips in the United States involve walking or bicycling either as the only mode used or as part of transit travel (Kuzmyak et al., 2011). The array of evaluation factors historically employed in motorized-transportation-based planning presents still other set of issues for NMT economic evaluation and equity analysis. Conventional transportation planning focuses heavily on travel time saved as a benefit, and NMT travel gener- ally does not save time. The pedestrian or cyclist, if he or she has a choice at all, trades off accep- tance of greater travel time for other benefits. One of these benefits may be cost savings, which is covered by conventional analysis, but typically only to the extent of out-of-pocket costs. Other user or public benefits typically go unquantified. If the NMT trip is purely for recreation or exercise, conventional transportation planning offers no metric at all for measuring benefit (Victoria Transport Policy Institute, 2007). In summary, economic and equity analysis for NMT facilities and support is not well developed. The discussion which follows briefly offers a few glimpses of ben- efits unique to NMT and some indications of the full scope of NMT economic and equity valuations and concerns.

16-391 Societal Economic Impacts On the one hand, as observed in NCHRP Report 552: Guidelines for Analysis of Investments in Bicycle Facilities, “the majority of past [NMT benefit-cost] work has a clear advocacy bent; it is not always known how and where much of the data are derived” (Krizek et al., 2006).85 On the other hand, practi- cally all available NMT benefit analyses—particularly those done with demonstrable rigor—focus solely on one area of concern and thus omit major components of benefit. The very small number with a broader focus have generally been limited in their ability to include the full range, although the Australian example to follow does make the attempt for walking only. Pedestrian and bicycle facilities and programs provide benefits in many forms. While analyses of a sin- gle category of benefit can be instructive, no single benefit category can properly illustrate the overall societal economic impacts. This situation is very important to keep in mind when reviewing the several examples presented here of single-benefit analyses. An unimpressive or even unfavorable benefit-cost ratio for a single benefit may become strongly favorable when the full range of benefits is considered. Indeed, some NMT investment outcomes—such as improvements in quality of life—will require effec- tiveness (goals attainment) analysis in lieu of benefit-cost analysis to receive due consideration. Breadth of Benefits. An indication of the breadth of factors worthy of including in NMT benefit and cost analyses is provided by an Australian computation of monetized benefits to society of shifting 1,000 km. of travel from driving to walking. The total value estimated, in 2001 Australian Dollars (AUD), was AUD 181 in the current year and AUD 2,339 over 30 years (Victoria Transport Policy Institute, 2007). Of arguably greater interest, however, are simply the categories of benefit and cost employed and the sign and relative magnitude of the benefits/costs. Table 16-127 lists the benefits and disbenefits considered and the corresponding monetized net-benefit estimates. Benefit/ Disbenefit Category Current Year 10 Years 30 Years Vehicle operating cost savings AUD 113 AUD 819 AUD 1,446 Improved health 84 607 1,071 Crash risk (from increased walking) b -95 -687 -1,212 Crash risk (from reduced driving) 34 246 435 Reduced air pollution 20 145 256 Reduced greenhouse gas emissions 20 145 256 Reduced traffic noise 3 22 38 Reduced water pollution 2 11 19 Total Benefits AUD 181 AUD 1,318 AUD 2,339 Notes: a Estimated in 2001 Australian Dollars, using a 7% annual discount rate. b This computation apparently does not reflect the “safety in numbers” benefit explored above under “Safety Information and Comparisons” — “Other Traffic Safety Issues and Findings” — “Safety in Numbers.” Also, it is important to note that the finding of NMT crash risk costs in excess of monetary benefits of improved health conflicts with contrary evidence reported above under “Public Health Issues and Relationships” — “Tradeoffs Between Health Benefits and Crash/Pollution Disbenefits.” Source: Ker – 2001, as reported in Victoria Transport Policy Institute (2007c). Table 16-127 Estimated Valuea of Shifting 1,000 km. of Travel from Driving to Walking 85 The users of this subsection should be aware that this critique undoubtedly applies to some of the benefit- cost values reported herein on the basis of the available literature.

16-392 Another benefits listing, this one from the perspective of bicycling, is provided by NCHRP Report 552. “Mobility” is first listed, focusing on the ability of cyclists to reach their destinations faster, more safely, and via more attractive routings when provided with bicycling improvements. “Health” addresses the inducement of more children and adults to shift from inactivity to meeting recom- mended basic physical activity guidelines. “Safety” covers the benefits of crash prevention, although the researchers reported finding little agreement or conclusive evidence to support safety benefit calculation, except perhaps from the perspective of “safety in numbers” as might be com- puted for an entire metropolitan area. “Reduced auto use” is set forth as a benefit measure to cover societal benefits of congestion reduction, air quality improvement, and transportation energy con- servation. “Livability” is presented as a benefit that can be measured on the basis of housing pre- miums paid by purchasers of homes with good accessibility to bicycle facilities. “Fiscal” is couched in terms of future cost savings through preservation of linear rights-of-way that may facilitate future transportation uses. “Recreation” is the final benefit introduced. NCHRP Report 552 offers suggested benefit value computations for mobility, health, reduced auto use, and recreation (Krizek et al., 2006). Even the benefits listings in Table 16-127 and in NCHRP Report 552 taken together must be viewed as illustrative and not fully complete. Examples of more benefit categories introduced below include certain revenue benefits, ADA and schoolchild transportation cost savings, and commercial sales increases. Goals attainment categories beyond those covered in the benefit analysis examples include enhanced mobility for the transportation disadvantaged, more travel options for the gen- eral population, and support for sustainability. Some benefits are not explicit in the Table 16-127 and NCHRP Report 552 category headings, but are covered, such as land value enhancements (pro- posed in NCHRP Report 552 to quantify “Livability”) and expanded opportunities for social inter- action (an aspect of “Livability”). At the same time, “the other side of the coin” from benefit omission is benefit overlap or duplication, impermissible in benefit-cost assessments. Given the typ- ical approach to estimating recreational benefits, described below, recreational benefit and health benefit valuations may be duplicative to the extent that exercise and training is an objective of the seeker of recreation. Health Benefits. The “Improved health” benefit of Table 16-127 is a good example of a benefit not covered in conventional transportation evaluations. Even NMT benefit-cost analyses focused exclusively on public health benefits apparently have their problems. A published review of 16 benefit-cost studies covering health effects of transportation policies with data on walking and bicycling illustrates several issues. The 19 analyses lacked “transparent and standardized methodologies,” only three “were considered to be of high quality,” and only one was in the United States, the rest having been sited in Europe. The benefit-cost ratios reported for health benefits varied widely. Irrespective of these limitations, the median outcome was a substantial 5 to 1 ratio. One study reported a ratio smaller than one, while the other 15 were all positive in outcome. The one U.S. study covered five trails in Nebraska, for which a benefit-cost ratio of 2.95 was found for health benefits relative to “costs associated with trail construction and use” (Gotschi, 2011).86 86 Benefit-cost ratios are reported in three different formats. A 5 to 1 ratio indicates there is $5.00 of benefit for each $1.00 of cost. This result may also be expressed as a 5:1 ratio, or alternatively, as 5.0, the result of divid- ing 5 (the benefit) by 1 (the cost). Examples of results indicating lack of cost effectiveness would be a ratio of 0.5:1 or an 0.5 benefit-cost ratio.

16-393 CDC studies and other research provide background perspective on health benefits. The CDC has estimated that direct medical expenses attributable to sedentary behavior totaled over 76 billion dol- lars nationwide in 1987, expressed in year 2000 dollars, for the 88 million inactive Americans 15 years of age or more and without physical limitations. Not included in this figure are indirect costs, such as the lost productivity resulting from the physical and mental disabilities to which physical inactiv- ity contributes. These same estimates of direct medical expenses related to physical inactivity, when adjusted to 1993 dollars and compared to smoking-cost studies, placed inactivity just 9 to 15 percent below smoking-related direct medical excess costs. Adding a rough estimate of indirect costs, the 1987 cost in 2000 dollars of inactivity likely exceeded $150 billion dollars. A 1980s study by RAND estimated that the total costs imposed on society by sedentary lifestyles may actually be larger than those imposed by smokers. None of these particular estimates directly addresses the question of what could be saved through enhanced walking and bicycling programs and facilities, but the excess societal-cost pool is enor- mous enough that any draw-down would be quite significant. In a 1990s study specific to active transportation, it was estimated that if 10 percent of U.S. adults were to take up walking on a reg- ular basis, the savings in heart disease costs alone would be an estimated 5.6 billion dollars (Committee on Physical Activity, Health, Transportation, and Land Use, 2005, Committee on Prevention of Obesity in Children and Youth, 2005, Pratt, Macera, and Wang, 2000). An evaluation of the benefits and the costs of City of Portland, Oregon, bicycling investments made starting in 1991 goes further, focusing primarily but not exclusively on health benefits. Benefits of the increased physical activity engendered were drawn from some of the same stud- ies as those enumerated above plus others. An average per capita estimate of annual health care costs attributable to physical inactivity of $544, inflated to 2008 dollars, was developed. Value of added life, derived using the Health Economic Assessment Tool (HEAT) for bicycling provided by the World Health Organization, was considered in a parallel set of calculations. The statistical value of life employed was $5.8 million, a figure suggested by the U.S. Department of Transportation (Gotschi, 2011). The replacement cost for Portland’s 274-mile network of bikeway improvements, in place as of 2008, was estimated to be $57 million. The cumulative cost of Portland’s Smart Trips individual- ized marketing program and associated promotion and education, from 2003 through 2012, was calculated at $7.2 million. Growth in bicycling, from which reduction in inactivity was derived, was estimated using basically the same bicycle count and mode share data for 1991 through 2008 as presented earlier in the “Response by Type of NMT Strategy” section (see “NMT Policies and Programs”—“New World Program Examples”—“Portland, Oregon”), including the exponential growth rate of 9.6 percent per year in central area Willamette River bridge bicycle volumes from 1991 through 2008. Bridge volumes were translated into bicycle miles of travel using the Portland Metro regional travel model. Decrease in inactivity, corresponding health care cost savings, and also energy savings, were estimated on the basis of bicycle miles over the 1991 baseline. Cyclists were estimated to have accumulated 109 million miles in excess of baseline cycling by 2008. This translated into $42 million in health care costs saved plus $16 million in saved energy costs. Forward projections were made in line with past experience and Portland’s 2030 bicycle master plan to put 80 percent of residents within 1/4 mile of a “low stress” bikeway. It was estimated that cumu- lative benefits since 1991 would begin to exceed cumulative costs in 2015, on the basis of health care and energy savings alone, without including the statistical value of life saved. The evaluation went further, estimating costs and benefits through 2040, representing a 50-year time span. A benefit-cost ratio of 2.3 to 1 was derived for the “80 percent” plan. Two other options produced benefit-cost ratios of 3.8 and 1.3, and benefit-cost ratios were of the next order of magnitude higher when statistical value

of life saved was included. The calculations explicitly excluded opportunity costs, and utilized a dis- count rate of 3 percent (Gotschi, 2011).87 Energy Saving and Emissions Reduction Benefits. The preceding subsection, “Traffic, Energy, and Environmental Relationships,” presented generalized energy savings estimates for transit access mode shifts to bike-and-ride from park-and-ride commuting and from auto-only commut- ing, and also for estimated shifts to bicycling given hypothetical corridor improvements in Tampa, Florida. Neither of these analyses was carried forward to the point of making benefit-cost or rate- of-return calculations. NCHRP Report 552, apparently focusing on shifts to bicycle-only travel, cautions that “these [auto substitution] benefits are relatively small” and “of only minor significance” (Krizek et al., 2006). This might not be the case for shifts to bike-and-ride from park-and-ride at transit stations, where in the local context the benefits of reduced parking demand may be important, especially where space is constrained and spot emissions of automotive pollutants are critical. As the last example in the “Traffic, Energy, and Environmental Relationships” subsection, modeled estimates were presented for relative VMT and GHG emissions reductions in response to sidewalk coverage expansion in Seattle and eight of its inner suburbs. These results were carried forward to the point of drawing conclusions, on the basis of elasticity-based sensitivity analyses, about the cost effectiveness “in terms of VMT and CO2 outcomes” of expanding the proportion of streets with side- walks. The analysis indicated that there are diminishing returns as full coverage is approached. For the sample under study, adding sidewalks was deemed cost effective up to but not beyond the point of having 1.42 miles of sidewalk per mile of street (as compared to the mean ratio under existing con- ditions of 1.16). A ratio of 1.42 is equivalent to having sidewalks on both sides of 71 percent of the street mileage (Frank et al., 2011). Obviously, consideration of a broader range of benefits would extend the cost-effectiveness break-even point beyond this degree of coverage. People with Disabilities Mobility Benefits. The “Pedestrian/Bicycle Linkages with Transit” dis- cussion, within the “Response by Type of NMT Strategy” section, notes that construction of suit- able bus stop provisions combined with critical links of sidewalk have been shown in specific cases to be quite cost effective in the service of providing mobility to people with disabilities. The Maryland Mass Transit Administration (MTA) has done calculations of cost savings and capital recovery for constructing improvements that allow a wheelchair-bound patron to use accessible 16-394 87 The researcher, in excluding opportunity costs and utilizing a 3 percent discount rate, chose not to follow tra- ditional engineering economic analysis protocols and U.S. Federal guidelines stating that benefit-cost out- comes should be “determined using a real [inflation-free] discount rate of 7 percent” (Office of Management and Budget, 1992). However, FHWA guidance observing that state governments mostly use 3 to 5 percent for discounting highway investments, based on “best practice” real discount rate calculations, was not violated (Federal Highway Administration, 2011). Indeed, some authors hold that benefits of life and health should not be discounted at all, although in other circles such deviations from uniform discount rate applications are held to be inadvisable. New federal guidance is anticipated. Meanwhile, a recent reassessment has recom- mended the social opportunity cost of capital approach and used it to determine real discount rates—for benefit-cost analysis—in the range of 6 to 8 percent (Burgess and Zerbe, 2011). Recomputation at a 7 percent discount rate would substantially lower the estimated Portland bicycle program benefit-cost ratios (Federal Highway Administration, 2011) and might well lead to one or more with a ratio less than one. On the other hand, inclusion of more benefits than the two considered, or of benefits of shared-use facilities to walkers, would tend toward counterbalancing the effect of a higher discount rate.

fixed-route bus service instead of having to receive and rely on federally-mandated ADA paratran- sit door-to-door service. MTA estimates its fully-loaded cost of ADA paratransit at $76.64 per one- way trip. They find the capital cost of simple bus stop “landing” improvements, inclusive of minor sidewalk improvements, to average $7,000 per stop. More extensive improvements can come to $58,000 per stop on average. Assuming these improvements allow an ADA paratransit patron who uses the service 5 days a week (10 trips) to switch to fixed-route bus service, MTA estimates the lesser of these capital costs will be recovered in 10 weeks from ADA paratransit cost savings. The more extensive improvements take an estimated 18 months for capital cost recovery (Goodwill and Carapella, 2008).88 There are also ben- efits to the user, who is no longer constrained to ADA paratransit trip prearrangement and reliabil- ity issues. This one example of benefit to the disability community and savings to providers of services to people with disabilities is likely representative of a number of other savings that could accrue from ADA-compliant sidewalk provisions and improvements in general. Transportation Cost Savings Benefits. A transportation cost saving benefit example, in addition to the one used above as a stand-in for people with disabilities mobility benefits, comes from school transportation operations. The Auburn School District in Washington State initiated an early SRTS program both to address childhood obesity and the high cost of running school buses. A late 1990s pilot project, with $121,770 in state funding, focused on infrastructure improvements paired with education plus student walk-to-school tracking and recognition. A $185,000 federal SRTS infra- structure grant followed in 2007, but results presented here pertain to a time prior to full federal project completion. With more students walking, reaching a milestone of 20 percent walk mode share, it has been possible to scale back school bus service. Transportation cost has been reduced by $220,000 annually. This savings equates to 180 percent each year of the one-time pilot grant, or 72 percent of the pilot grant plus the full amount of the partially implemented 2007 federal grant (Pedestrian and Bicycle Information Center, 2011). The overall Auburn SRTS program was intro- duced in the “NMT Policies and Programs” section (see “Schoolchild-Focused Programs”— “Infrastructure and Traffic Engineering Improvements”). Transportation Revenue Benefits. Benefits of increased transportation revenue have only been examined in the case of individualized marketing (see “Response by Type of NMT Strategy”— “Walking/Bicycling Promotion and Information”—“Individualized Marketing”). Benefits of shifts to active transportation accruing from this informational and promotional strategy should logi- cally be as broad as the list in Table 16-127. Nevertheless, most individualized marketing rate-of- return computations provide examples of focusing on only one clearly tangible benefit if that alone is sufficient to demonstrate cost-effectiveness. Individualized marketing generally produces shifts to public transportation along with walking and cycling, and the tangible benefit typically selected for cost-effectiveness demonstration is the incremental increase in those transit farebox revenues attributed to the individualized marketing program under study. It could be argued that this is not a walking or bicycling benefit, but it must be remembered that much access and egress walking takes place as part of transit riding, and in any case the added revenue accrual is an outcome of the individualized marketing expenditure. 16-395 88 The cited source gives the longer cost recovery period as “eighteen weeks” (sic) but other information pro- vided, along with data in an FHWA/FTA Transportation Planning Capacity Building Program Peer Workshop Report (including a presentation by C. Scott Windley of the United States Access Board at the May 7, 2007, meeting in Nashville, TN), clearly indicates “eighteen months” was meant.

Many examples of this benefit-cost analysis approach predate the inclusion of the walk-only and bicycle-only travel modes as options promoted by individualized marketing. In those early (1990s) applications, primarily in Germany and other European countries, the benefit-cost ratio was typ- ically in the 3:1 to 4:1 range with full coverage of costs in the first year (UITP and Socialdata, 1998). Economic analyses based on this transit revenue measure alone will, however, probably show cost effectiveness only in those metropolitan areas with substantial transit usage. An example of a first-year rate of return calculation comes from the city of Linz, Austria. Based only on cost recovery from increased transit revenues, the first year rate of return for individualized marketing costs was in the range of 1.1 to 1.6 (more than 100 percent cost recovery in the first year). Similar calculations for conventional direct marketing in Linz obtained only an 0.5 first year rate of return, not cost effective (Ashton-Graham and John, 2006). Individualized marketing benefit-cost analysis in South Perth, Australia, a situation more compa- rable to those seen in North American cities, examined a broader range of benefits of which tran- sit net fare revenue gain contributed the next to largest component. It accounted for 1/3 of all monetized benefit after deduction for additional bus capital costs. The largest benefit, at just over 1/3 of the total, was avoided road construction costs. The remaining benefits computed, summing to not quite 1/3 of the total and listed in decreasing order of importance, were public health sav- ings from reduced air pollution, avoided traffic control costs, and public health savings from improved health and fitness. Benefit-cost estimates were prepared in 2002 covering rather long time periods for a marketing strategy, arriving at benefit-cost ratios of 44:1 for 10 years and 77:1 for 25 years (Parker et al., 2007). Contemporary follow-up surveys following the South Perth full- scale application showed durability of shifts to walking and bicycling for at least 4 years, but with a decay in the shifts to bus transit after 18 to 24 months, as illustrated in Table 16-61 within the “Individualized Marketing” discussion cross-referenced above (Australian Government, 2005). A consulting study done for the U.K. Department for Transport reported that the early individualized marketing pilot projects undertaken in England had been found to exhibit an average benefit-cost ratio of 31:1. The consultants advised that benefit-cost analyses of individualized marketing taking a broad range of impacts into account typically report positive ratios on the order of 30:1, and that individual- ized marketing cost effectiveness appears to improve “as the scale of implementation is increased” (Parker et al., 2007). The proportion of benefits attributable to transit net fare revenue gains was not specified in the published reporting of U.K. experience or advice. Recreational Benefits Net economic impacts of NMT facilities can legitimately include user benefits not covered from trans- portation or public health perspectives. For example, quality of life improvements—such as availabil- ity of a preferred recreational venue—offer tangible value to users. Since use of NMT facilities is usually free, such value represents consumer surplus: benefit received that does not incur user cost through pricing. Valuation procedures have been developed to address consumer surplus in such instances. Although these approaches likely produce user benefit valuations that include certain effects already captured in transportation- and public health-based benefit-cost analysis, the overlap is only incidental, and it is instructive to look at typical recreational benefit results. Shared use paths are the facility type most obviously possessed of characteristics that would pro- duce recreational and associated benefits, although it could well be argued that a sidewalk system or bicycle boulevard in a pleasant neighborhood should produce quality-of-life value. A case example involving an off-road rail-trail is provided by economic benefit valuations prepared for 16-396

the Washington and Old Dominion (W&OD) facility in Northern Virginia. The W&OD Trail extends some 45 miles west from Alexandria, Virginia, through urban, suburban, exurban, and rural communities. The analysis technique applied relies on relating the number of trips taken to costs incurred in traveling to the facility. Empirical data for the W&OD model were obtained from surveys taken in 2003–04 using a nonprobability quota sampling approach. Two functional forms of regression model were used, with variables including annual W&OD trips, round trip access distance and time, perceived availability of a viable substitute recreational venue, annual house- hold income, and group size (number of W&OD users in the household). Travel cost for access was assumed to be $0.131/mile, with no cost for time. Only users whose trail-use purpose was recre- ation, and who did not live directly on the trail, were included (Bowker et al., 2004). The two different mathematical functions employed, truncated negative binomial and truncated stratified Poisson, produced consumer surplus estimates per trip of $9.08 to $13.63 in 2003 dollars. Consistent with other studies, W&OD per-user trip making activity was found to be negatively related to group size. Income and perceived availability of suitable recreational alternatives did not prove to have statistical significance, although the variables were retained in the model for log- ical consistency. Survey respondent reporting of benefits important to them would suggest that perceived health benefits were the primary area of potential overlap with transportation- and pub- lic health-based benefit-cost analysis. W&OD Trail usage was estimated at 1,707,000 visits per year. Setting aside persons estimated to be commuting or “not on a primary purpose visit to the trail,” the $9.08 to $13.63 per trip benefit estimates translated into approximately $14.4 to $21.6 million in annual recreation value. By comparison, com- parable estimates for the similar but shorter (7.6 mile) Lafayette/Moraga Trail in the San Francisco East Bay Area, converted to 2003 dollars, were $5.82 to $20.22 per visit depending on the statistical model, or $2.3 million annually. Aggregate annual recreational value estimates for two U.S. trails more rural in character were, in 2003 dollars, $5 million and $10.6 million annually (Bowker et al., 2004). An esti- mation for the Monon Trail in Indianapolis, detailed further below in connection with off-road path added land value and benefit-cost estimation, gave an annual recreational benefit of $3 million (Lindsey et al., 2004). Land Value and Commerce Impacts Financial benefits in terms of increased land value and added commerce do not fit neatly within existing categorizations of societal economic impacts. Together with recreational benefits, they make up an almost “parallel universe” of approaches to valuing NMT facilities, or in some cases, approaches to validating that reallocation of street space to NMT does not detract from the con- duct of commerce. No global attempt to rationalize these approaches amongst themselves or to integrate them with the types of societal economic impact analyses discussed above has been found. Lacking an overall NMT economic impact paradigm, the following presentation of land value and commerce impacts simply starts with walkability and path/trail effects on property sal- ability and value, then moves into commerce impacts of trails, and concludes with several perspec- tives on downtown commerce impacts. Neighborhood Walkability Effects. Home buyers, despite desire for a larger home and highway access, were found in a National Association of Home Builders and National Association of Realtors survey to be concerned about neighborhood walkability. Top-ranked out of 18 listed community amenities were: highway access (44 percent), jogging/bike paths (36 percent), sidewalks (28 percent), parks (26 percent), playgrounds (21 percent), and shops within walking distance (19 percent) (Victoria Transport Policy Institute, 2007). 16-397

Land Values and Off-Road Paths. The economic benefits of shared use, off-road paths and trails have been fairly extensively reported upon, although many of the available studies have focused on rural rather than urban-area facilities. No one study has been encountered that seems to cover all the different types of benefits isolated in individual studies. Paths join other NMT facilities in providing transportation and health benefits of the types already discussed, along with user ben- efits examined under “Recreational Benefits.” There are also land value and commerce impacts. Many of the available impact assessments, such as provided below for the Pinellas Trail, border on the anecdotal but nevertheless provide useful insights. The implementation and early years of the Pinellas Trail in Florida were accompanied by success- ful downtown revitalizations in Dunedin, Largo, and Tarpon Springs. These positive develop- ments have been largely attributed—at least by some—to presence of the trail and its users. The Pinellas Trail is a rail-trail. It has been theorized that since now-gone rail lines often dictated the original pattern of community location, the old alignments frequently provide downtown to downtown connectivity, such that trail conversion reestablishes historic linkages—with accompa- nying economic benefits (Guttenplan and Patten, 1995). The Indiana Trails Study of six urban, suburban, and rural trails throughout the state sought quan- titative data. Trail neighbor surveys found, among the six trails, that 60 to 88 percent felt their trail had improved neighborhood quality. Some 86 to 95 percent perceived that their trail had either increased or had no effect on their property value, and 81 to 93 percent felt that the trail would make it easier to sell their property or would have no effect on ease of selling. Realtors, however, in focus group settings, reported not seeing either major increases in property value or ease of selling (Indiana University, 2001). A subsequent study in Indianapolis alone sought to capitalize possibly elevated home sale prices along greenways and trails. The study was designed to explore relationships between property values and public choices about public investments having significance to the housing market. After quantifying the effect of property characteristics such as housing age and number of bath- rooms, it was established that neighborhood characteristics ranging from property taxes to school quality (expressed as test scores) have significant effects on housing values. Then, with the study’s final three hedonic property value cross-sectional models, it was determined that greenways (some with trails) have important and mainly-positive effects on prices. These effects are almost all sig- nificant, although not as strong as for property and neighborhood characteristics. Adjusted R2 val- ues were 0.79 for all three models. Model 1 differentiated between location within 1/2-mile of the central feature of a greenway and location outside that band, but did not differentiate between greenways with and without trails. It estimated that location adjacent to a greenway corridor was worth $3,700 in the price of a home. Model 2 made the trail/no-trail differentiation. It estimated the average added value of adjacency to a greenway without trail at $5,300 versus $4,400 for a greenway with trail. Finally, Model 3 further differentiated between the “flagship” trail of Indianapolis, the Monon rail-trail, and other greenways with trails. It found a sales premium for location within 1/2-mile of the Monon Trail of slightly over $13,000, nearly $4.4 million dollars total of added value for the 334 sales along the Monon in 1999. In Model 3, adjacency to a greenway without trail was worth $2,200 in home value, but the effect of being next to a greenway with a trail other than the Monon was not significant and slightly negative (Lindsey et al., 2003). The Monon Trail results alone were expanded to estimate the added value accruing to all 8,862 house- holds located near the trail, rather than just those sold in 1999. Applying the average trail vicinity sales premium to all the homes within 1/2-mile, an added home value estimate of $116 million was 16-398

obtained. It was noted that such estimates represent an approach to valuing “amenity or ecologi- cal values” of greenways that accrue with or without active use (Lindsey et al., 2004). In conjunction with this estimate, recreational benefits of the Monon Trail were also calculated, using a variation of the opportunity cost estimation procedure described above in connection with W&OD Trail recreational benefits derivation. The primary differences were the inclusion of all trips, with round trips counted as one trip; inclusion of cost of time to access the trail, with time value taken to be 1/2 the prevailing Indianapolis wage rate; and the use of four distance-zones of access. Automotive driving costs were not applied to users in the closest zone, who were assumed to use walk or bike access, and were applied to only one-half the users in the next zone out. In this manner, applying a late 1990s-based usage estimate of 373,581 visits, the Monon Trail recreational benefits were estimated as a total consumer surplus of approximately $3 million. The Monon Trail economic analysis researchers posited that land value and recreational benefits estimates are largely complementary, with only limited aspects having risk of overlap. Adding the two benefits estimates, they developed Monon Trail benefit-cost ratios on the basis of trail con- struction and maintenance costs, a 10-year time frame, and a discount rate of 6 percent. The result was an estimated benefit-cost ratio of 5.7. In a sensitivity analysis, the researchers recalculated the estimate excluding the value of travel time in the recreational benefits recomputation, and arrived at a lower-bound benefit-cost ratio of 2.9 (Lindsey et al., 2004), still a substantial return. A cross-sectional research model for estimation of the value of bicycle facilities as capitalized into home sale prices was also developed for the Minneapolis-St. Paul Twin Cities area, and similarly obtained non-uniform land value results. The Twin Cities area at the time of the study had 1,692 miles of shared use off-street trails and a number of on-street bicycle facilities as well. Physical housing attributes, city versus suburban location, distance to open space, school district and population mea- sures, and distance to downtowns and highways, were all included as model variables. Three types of bicycle facilities were assessed: on street bicycle lanes, shared-use off-street trails in a non-roadside position, and shared-use off-street roadside trails more-or-less in a “sidewalk” position. Substantial dif- ferences were found between city and suburban price relationships, similar to earlier Twin Cities research on value of open space proximity and size. Bicycle facility amenities, like open space amenities, proved to be more valued by city dwellers than suburbanites. Overall, the effects on home prices of bicycle facility proximity were estimated to be limited given all the other modeled considerations involved in home valuation. Measures of facility extent were not significant. Proximity of shared-use off-street trails in a non-roadside posi- tion was valued positively in city locations, but proximity of roadside trails was valued negatively, and presence of on-street bicycle lanes was not found to be significant. In the case of suburban loca- tions, proximity was estimated to be a negative factor for all three facility types. The suburban rela- tionship being different than the city relationship for shared-use off-street trails in a non-roadside position, the researchers offered several possible explanations. These included possible suburban- ite lack of interest and overriding desire for seclusion, the wintertime use of exurban trails by snowmobiles, and legacy effects in the case of the many rail trails, wherein lingering depression of property values owing to railroad proximity may still pertain (Krizek, 2006). Commerce Impacts of Off-Road Paths. The commerce impacts of paths and major NMT bridges have one aspect largely unique to these particular types of facility. This aspect might, for easy iden- tification, be called the “tourist dollars” contribution of the facility to the economy. More formally, what is of interest in this regard is the local economic impact of non-local trips attracted to the path or other facility in question. Economic impact evaluation protocol does not allow inclusion of vis- itor spending by local residents or of non-local visitor spending outside the local area. The W&OD 16-399

Trail economic studies introduced in the “Recreational Benefits” discussion did, nevertheless, develop an estimate of total spending as a matter of general interest. The trail, very much oriented to use by local residents, was found to generate about $5.3 million annually in local economy expenditures by local resident users. The non-local visitor contribution, examined further below, equated to some $1.4 million annually, bringing the total local expenditure to on the order of $7 million annually. The grand total recreational spending associated with the W&OD Trail, when non-local spending was included, was determined to be an estimated $12 million annually (Bowker et al., 2004). The W&OD non-local trail user spending contribution to the local economy, the spending deemed legitimate for economic impact analysis, was found in surveying to consist of lodging (25 percent), restaurants and bars (42 percent), groceries and carry-out (6 percent), gas and oil (22 percent), use fees (4 percent), and miscellaneous (1 percent). The total per individual visitor was $15.40. The W&OD Trail’s percentage of non-local users was determined from survey responses to be just 5.24 percent. The $1.4 million annual direct effects contribution to the economy derives from 5.24 percent of the 1,707,000 estimated annual visits making the average expenditure, within 25 miles of the facility, of $15.40. Using a National Park Service “Money Generation Model” (MGM2), indirect and induced effects were added in. In this manner, the total boost to the local economy flowing from non-local vis- itors to the W&OD Trail was estimated to be $1.8 million of economic output, 34 jobs (full-time equiv- alents), and $642,000 in personal income (Bowker et al., 2004). The W&OD Trail is a long-established radial facility, opened in stages from 1974 to 1988. It also lies adjacent to the Nation’s Capital and extends a lengthy 45 miles, so it is important to consider that the commerce impacts may be less impressive for other urban-area paths not as ideally situ- ated relative to a well developed path visitor market. For example, the Indiana Trails study set out to do a similar analysis on the six trails it examined. So few of the intercepted trail users were vis- itors, as compared to trail neighbors, that the sample of expenditure data was deemed too small for reliable reporting (Indiana University, 2001). Analysis of a 27-mile section of the Little Miami Scenic Trail, at the time of the study a 60 mile subur- ban, rural, and small-town facility just east of the Dayton, Ohio, urban area, found a $13.54 per per- son per visit non-durable-goods expenditure. The largest expenditures on average were for restaurants and auto-related costs. Trail visits for the 1996–97 study year were estimated at 150,000 to 175,000 annually. The study area was within Warren County, and of the total dollar value, 77 percent was expended in the county. Virtually all the remaining expenditures were in other Ohio counties (Ohio-Kentucky-Indiana Regional Council of Governments, 1999). The result of multiplying the Little Miami Scenic Trail average non-durable-goods expenditure of $13.54 by the 150,000 to 175,000 annual visitors is $2.0 to $2.4 million, or $1.6 to $1.8 million if restricted to Warren County expenditures. This range of figures should roughly compare with the $7 million grand total annual local economy expen- ditures seen for Virginia’s W&OD Trail. Downtown Pedestrianization Effects. Scattered economic data is available for introduction of pedestrian zones and malls, primarily for those deemed successful. Boston’s Downtown Crossing pedestrian zone was created in 1978 and beautified in 1979. Despite increasing competition from other areas such as Faneuil Hall Marketplace, store visits from 10:00 AM to 4:00 PM weekdays were up 6 percent in 1980, compared to 1978 before the closing of streets to private auto traffic. Individual purchases were up 26 percent. High-end purchases declined somewhat, however, such that the total dollar value of all purchases increased at the same rate as upkeep-goods-and-apparel price inflation. Most of the pedestrian and associated retail activity increase was attributable to mid- day visits by nearby office workers, and their typical midday purchase price was modest (Weisbrod and Loudon, 1982). 16-400

Sales increased by 30 percent on Copenhagen’s Stroget, consisting of three contiguous streets in the main shopping district, after it was closed to motor vehicles in 1962. In East Anglia, England, London Street merchants saw sales increases of 5 to 20 percent. The busy State Street Mall in Madison, Wisconsin, a transit mall, was found in the mid-1980s to have average base rents for fronting com- mercial space of $9.87 per square foot, relative to a downtown average of $8.15. The on-mall vacancy rate was 3.4 percent (Robertson, 1994). The typical U.S. pedestrian mall was not found capable of stemming downtown decline on its own, however, as discussed under “Response by Type of NMT Strategy”—“Pedestrian Zones, Malls, and Skywalks”—“Pedestrian Zones and Malls.” In Minneapolis, the “percentage of metropolitan area residents who ‘shopped downtown within the past month’ ” declined from 48 percent in 1965, to 42 percent in 1967—the year the Nicollet transit mall opened, and to 33 percent in 1969. A 1973 survey indicated stabilization or perhaps even a little recovery, with post-1970 Nicollet Mall retail sales data also suggesting at least tempo- rary stabilization. In any case, declines notwithstanding, surveys of major Nicollet Mall retailers in 1977 encountered almost universal enthusiasm for the transit mall. The owner of a 20-store chain, that included suburban mall anchor stores, noted that the Nicollet Mall department store had the strongest sales. Property owners in the entire transit mall assessment district agreed to pay for a four-block extension. Of 21 merchants on the mall answering a survey question that asked them to imagine freedom to move, 18 indicated that they would stay at their present location or relocate elsewhere on the Nicollet Mall. Secondary economic indicators, including rental rates and new investment, provided more positive conclusions for the Portland and Minneapolis transit malls. In part this may be because the retail sales data “did not take into account new businesses.” The Urban Mass Transportation Administration’s Service and Methods Demonstration Program review concluded that the transit malls in Portland, Oregon, and Minneapolis (along with a now-dismantled mall in Philadelphia) “appear[ed] moder- ately positive” in their economic impact on business (Edminster and Koffman, 1979). Downtown Skywalk Impacts. A rather unique case of benefits accruing from pedestrian improve- ments is presented by downtown skywalk systems. Skywalk benefits may include pedestrian and vehicular delay reduction, pedestrian system climate control, concomitant increases in walking activ- ity, intersection traffic flow improvements, related transportation energy and pollution reductions, and land redevelopment impetus. In downtowns with enough inherent drawing power, a second level of retail-commercial and service establishments may develop without necessarily detracting from ground-level potential. A Minneapolis-St. Paul review done after 5 years of Skyways experi- ence found building managers were reporting that second-level rents had increased from “margin- ally to considerably below” street-level rents to “equal or above” street-level rents. At the same time, street-level rents reportedly did not suffer. Preliminary evidence did suggest, however, that com- mercial space at the fringes of the central business districts (CBDs) might be negatively impacted, indicative of a possible compaction of retail and service activities (Podolski and Heglund, 1976). Indeed, circa 1980, St. Paul vacancy rates for buildings on their Skyway system were found to be less than 1/4 the rate for buildings not connected. Subsequent study in St. Paul, the city cited by several observers as the skywalks city where street-level retail decline is readily evident, found that 3/4 of downtown retailing was taking place in the Skyway level by 1994. Downtown retail- ing had continued to increase, but was paired with a steady movement from ground floors to the Skyway level. The average annual lease rate had become $10.58 per square foot on the Skyway level, versus $8.90 for first floor leases (Robertson, 1988 and 1994). Building design factors spe- cifically affecting the St. Paul situation were discussed under “Response by Type of NMT Strategy”—“Pedestrian Zones, Malls, and Skywalks”—“Pedestrian Skywalks”—“Urban Planning Considerations.” 16-401

A benefit-cost estimate prepared in advance of Des Moines skywalk system implementation esti- mated a total annual benefit of $561,600 relative to an annualized cost of $375,000 (values in 1978 dol- lars). The benefits monetized in this case covered only savings in pedestrian and vehicular delay at intersections, along with related vehicle fuel consumption reductions, although an estimate of 6.5 tons in annual emissions reductions was also prepared based solely on reduced vehicle idling at intersections. The findings resulted in declaration by the FHWA that the proposed skywalk system was eligible for Federal Aid Urban funding (Heglund, 1980). Commercial Street Modification Impacts. When pedestrian or bicycle improvements along an urban street require reduction in parking, the general expectation is that merchants will lose busi- ness. This perception, right or wrong, has delayed or derailed countless NMT improvements. Investigators in Toronto addressed this viewpoint head-on with merchant, pedestrian, and park- ing surveys along Bloor Street in the Annex Neighbourhood. This neighborhood is in the older city, with the University of Toronto nearby, and has commercial development along Bloor Street. The typical (median) Bloor Street merchant was found to believe that close to 25 percent of their customers drove to the area. (Sidewalk surveys found 10 percent of interviewees to have come by car.) Despite this belief, almost 75 percent thought that installing bicycle lanes at the expense of one-half the on-street parking would either improve or not affect their business, with 25 percent of businesses feeling that they would be adversely affected. Thus past opposition of city council members to Bloor Street bike lanes was found to have apparently rested in part on misconception of merchant positions on the matter. Surveyed merchant reaction was almost the same to widen- ing sidewalks at the expense of one-half the curb parking. Interestingly, while sidewalk survey respondents overall preferred either a bike lane or sidewalk widening to not tampering with the parking supply, they also preferred a bike lane to widened sidewalks by a ratio of almost 4 to 1. This may have been because the sidewalks were already 4 meters (13 feet) in width and the survey did not indicate what additional width might be used for (Sztabinski, 2009). Overall Bloor Street merchant intuitions appeared to be backed up by investigation of spending pat- terns. The median reported monthly expenditure at stores in the area was squarely within the $100–$499 (Canadian) range for surveyed pedestrians who indicated they usually came to the area solely by walking (46 percent mode share). The median for pedestrians who usually arrived by bicy- cle (12 percent mode share) was in the same range, but with more spending less and fewer spending more. Median monthly spending by transit riders (32 percent mode share) and private vehicle users (10 percent mode share) was the range of $25–$99 (Sztabinski, 2009). Of course, an intercept survey of the type used would naturally tend toward picking up frequent visitors more than infrequent vis- itors, and the per-month expenditure findings do not speak directly to average expenditure per visit. A roughly similar study in the SoHo district of Manhattan that asked only about sidewalk widen- ing found 42 and 48 percent, respectively, of those who had shopped or dined on Prince Street in the last month thought that they would be likely to come more often with widened sidewalks and less parking. In contrast, 7 and 8 percent indicated they would come less often. The surveyed sec- tion of Prince Street, between Broadway and 6th Avenue, has quite crowded sidewalks (Schaller Consulting, 2006). Toronto’s Annex Neighbourhood and Manhattan’s SoHo both have intensive public transit service, undoubtedly a factor in the study outcomes. Equity Issues Equity for the Transportation Disadvantaged. The improved survey methodology of the National Household Travel Survey (NHTS) has allowed demonstration as never before of the importance 16-402

to the poor and minorities of good accessibility to daily activities and transit stops for persons trav- eling on foot. Walking is not only the most affordable of all transportation modes, but also it is the most important means of reaching public transit. According to 2001 NHTS results, persons in households earning less than $20,000 per year make 16 percent of their daily trips solely by walk- ing, versus 8 to 9 percent for persons in families with more income. They also make 38 percent of their daily trips by transit, a travel mode shown to involve substantive walking, compared to a range of 7 to 20 percent for income categories over $20,000. Comparisons are four or more times as dramatic for households without a private vehicle versus households with one or more vehicles (Pucher and Renne, 2003). (For more specifics, see the “Income” and “Automobile Ownership” dis- cussions along with associated tables within the “User Factors” subsection of the “Underlying Traveler Response Factors” section.) The 2007 Benchmarking Project covering U.S. bicycling and walking provided a basis for examining possible equity issues through the lens of comparison between proportion of non-white workers using NMT commute modes and proportion of all workers using NMT modes.89 Only a slight differ- ence was found in the case of bicycling, with 0.46 percent of non-white workers bicycling to work compared to 0.43 percent of all workers. A significant difference, however, was exhibited by the walk mode. The walk commute share of 3.6 percent non-white workers is 1/3 more than the 2.7 percent walk share for all workers (Thunderhead Alliance, 2007). The bicycling statistics thus provide only a small and tenuous indication of minority persons being especially captive to the bicycle mode. There is stronger indication, on the other hand, of dependence on the walk mode and of likely related inequities, particularly where pedestrian facilities may be generally inadequate. The adverse impact of inadequate pedestrian infrastructure on transportation disadvantaged pop- ulations is underscored by the already-discussed examination of neighborhood sidewalk provi- sions and associated pedestrian characteristics in 12 Seattle neighborhoods, six “suburban” and six “urban.” The neighborhoods were matched for population density, land use, and approximate income levels, and were also found to have similar per-person auto ownership rates. The sub- urban neighborhoods were deficient, however, on any number of pedestrian facility and related neighborhood design measures, while the urban neighborhoods did well on these measures. It appeared that the pedestrian facility and design deficiencies had to be a primary explanatory fac- tor for why the “suburban” neighborhoods averaged only 1/3 the amount of walking to each neighborhood’s local commercial center compared to the “urban” neighborhoods. The suburban pedestrians who were observed were disproportionately young people and persons of color, taken as an indication of persons who do not or may not have the option of driving. On aver- age, 41 percent of the suburban pedestrians were under age 18, 180 percent higher than in the neigh- borhood population, while in urban sites the proportion (16 percent) more or less matched the local population. The high proportion of young pedestrians and pedestrians representing disadvantaged minorities, combined with lack of appropriate pedestrian facilities in the suburban sites, raises sig- nificant equity issues. It also raises safety concerns, for the younger persons especially, and for the pedestrians with disabilities observed at three of the suburban sites (Hess et al., 1998, Moudon et al., 1997). Additional information on this revealing analysis is found in the “Pedestrian Activity Effects of Neighborhood Site Design—Seattle” case study below. The income-based equity case for bicycle facilities is not as strong as for walking facilities in terms of absolute numbers of current usage. Nevertheless, the fact that the 2001 NHTS found bicycling mode 16-403 89 The 2010 Benchmarking Report was not used for this assessment because of a change in method of comparison that would seem to allow differential unemployment rates to possibly cloud conclusions.

shares for zero-car households (2.4 percent) (see Table 16-78) that were three times the bicycling mode shares for households owning vehicles (0.8 percent) (Pucher and Renne, 2003) illustrates that the equity advantages of providing safe and useful bicycle facilities should be a part of bicycle planning, design, and benefit analysis. Indeed, bicycle use by persons with limited transportation options could likely be more extensive were there better bicycle facilities in poorer neighborhoods. Equity of Access. There are also facility-access equity questions, which take more than one form. One access equity manifestation is illustrated by the instance of differing categories of bicyclists vis-à-vis facility type availability. The concern in this case is equity in provision of facilities to support cycling by not just skilled bicyclists but bicyclists of all skills, ages, and degrees of real or perceived risk tol- eration. The discussion found in the “Response by Type of NMT Strategy” section under “NMT Policies and Programs”—“European Programs and Comparisons” illustrates how countries with the most intensive NMT programs, including extensive systems of separate cycling facilities, achieve gender balance in bicycling mode shares—along with bicycling shares for persons over 75 years of age that approach or equal shares for younger cohorts. In contrast, countries that rely more on “vehic- ular cycling” (bicycling in mixed traffic), such as the United States, see gender distributions that are on the order of 1/4 female and 3/4 male. They also have bicycle mode shares that decline with age— to the point of being vanishingly small over age 65 (Pucher and Buehler, 2008b, Pucher and Buehler, 2009a). These disparities are suggestive of category-of-cyclist bicycle network access inequities. The sharply differing user type distributions of off-road shared use path users compared to mixed traf- fic bike route users are highlighted in the findings on Rock Creek hiker-biker trail versus Beach Drive bike route user characteristics presented within the case study “Special Mini-Studies in Montgomery County, Maryland” (see “More—Off-Street Versus On-Street NMT User Mix,” including Table 16-129). These parallel-facility classification count results show use of the trail by a wide spectrum of cyclists including cyclists-in-training (and also joggers and walkers), while the on-road route attracts a narrow spectrum composed—by all appearances—of sports-minded adult cyclists in their prime. Facility-specific observations such as this offer confirmation for studies that use aggregate data to conclude that separate facilities are especially important for the less skilled, less fit, and less dar- ing. Provision of separate on-road cycling facilities and shared use, off-road paths thus becomes a form of social justice, facilitating engagement of a much broader spectrum of the population in cycling for pleasure, exercise, and utilitarian travel (Pucher and Buehler, 2009a) and—in the case of off-road facilities—serving walkers and joggers as well. In a sense, this situation exhibits a rough parallel to the decision already made in the United States that pedestrian access must be opened up, by means of ADA-compliant design, to persons with disabilities. Note that the role of non-separated bike lanes and bicycle boulevards in serving the various cate- gories of bicyclists is less clear given the relative lack of user-type studies on these facilities. However, information of the type presented in the “Trip Factors” subsection of the “Underlying Traveler Response Factors” section (see “Bicycle Trip Distance, Time, and Route Characteristics”— “Bicycle Route Choice”) suggests that bicycle lanes are more attuned to the proclivities of adult, frequent, often-commuter bicyclists, while bicycle boulevards and other quiet streets may serve a broader clientele, especially inclusive of female cyclists. Another access equity manifestation is that of pedestrian/bicycle facility location, and facility access-point proximity. A proximity-equity study of the trail system in Indianapolis, done on the basis of that which was in place as of 1999, examined the populations living in census tracts at least partially within 1/2-mile. It found that the completed trail segments were adjacent to a population that was poorer and with a higher proportion of African Americans than Indianapolis-Marion County as a whole, with more households lacking vehicles and high school diplomas. The study 16-404

noted, however, that as the trail system expanded it would become more focused on higher-income populations with smaller non-white components (Lindsey, Maraj, and Kuan, 2001). Limitations in this proximity-based methodology could include lack of recognition of multiple versus more lim- ited trail access options and failure to identify lack of access imposed by long segments without access points, such as might be imposed by a rail trail on high embankment or a riparian trail some- what isolated from the street system. A more sophisticated analysis of the Indianapolis trail system was subsequently undertaken, this time as of 2006 with more trail mileage in place. The research, actually focused on development of improved use-forecasting methodologies, sought to supplement or replace simple proximity measures with an accessibility measure or measures. A Hansen gravity-model-based accessibility measure was employed in logistic regression models keyed to estimation of trail use on the basis of socioeconomic descriptors and facility accessibility. Four sets of models were tested, one without distance or accessi- bility as variables, one with distance from the home but not accessibility, and two with accessibility but not raw distance. The models exhibited progressively better fit, with Pseudo-R2 values of 0.07, 0.13, 0.14, and 0.14, respectively, for the models of usage in the previous month. Importance of ethnicity variables provided a tentative indication of equity. Even without a distance or accessibility variable, neither the proportion of African Americans nor the proportion of non-Hispanic whites proved viable as model variables (Ottensmann and Lindsey, 2008). This may suggest, if the result is not clouded by other factors, that these groups were equitably served by the 2006 trail system. The variable for Hispanics was significant and positive in both the trail usage model and the use fre- quency model. It became insignificant in the trail usage model when an accessibility measure (but not the raw distance measure) was included. It remained significant in all trail use frequency mod- els, but shrank progressively in size as raw distance and then accessibility were introduced. These results may possibly indicate above-average trail accessibility for Hispanic populations. The researchers urge caution, however, noting that a less-indirect application of accessibility measures to calculate access equity (such as has been done for playgrounds and public libraries) would be more promising as an indication of relatively underserved or overserved populations. As a matter of inter- est, accessibility proved to be a strong variable in the trail use research models. Income over 300 per- cent of poverty level, college graduation, and age less than 65 years were also strong markers for trail usage and use frequency (Ottensmann and Lindsey, 2008). The income, education, and age results do not necessarily indicate inequity, as the effects may have been uniform across ethnic groups. Access may also be a skywalk use equity factor, perceived by some as reflecting a tendency of skywalk systems to segregate persons in the downtown by economic class. In this view, the lower-rent uses and the people who patronize them are left on the ground floor and outside the skywalk network. A 1988 five-city survey of skywalk pedestrians indeed found a perception that the typical skywalk user was a white-collar worker of the white race, more often female than male, and more likely to be earning a high income than a modest one.90 Duluth was somewhat of an exception, thought to be a reflection of the large presence of mining and shipping industries. Cincinnati skywalk use was perceived to be the least office-oriented and most heterogeneous, including use by large numbers of minorities. Two contributors to this more diverse use were noted. An obvious factor was the more varied makeup of downtown Cincinnati itself. A second (hypothesized) factor was provision of direct skywalk connections to sidewalks, and the skywalk system’s coordination with a major open space, Fountain Square. The other four skywalk systems were—at the time—accessed solely 16-405 90 This 8:30 AM to 6:30 PM non-random survey did not quantify socio-economic characteristics directly, but instead asked structured questions designed to obtain perceptions of fellow skywalk users.