Click for next page ( 13


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 12
12 Table 1.4. Selected TOD auto trip rates (total trips in and out). AM Peak Hour PM Peak Hour Study Location Apartments Office Apartments Office (trips per (trips per (trips per (trips per dwelling unit) 1,000 sq. ft.) dwelling unit) 1,000 sq. ft.) Pleasant Hill BART 0.33 1.20 0.41 1.10 San Mateo 0.44 NA 0.49 0.92 Portland TODs 0.29 NA 0.38 NA Pleasanton Apartments 0.43 NA 0.47 NA ITE Apartments (use 220) 0.51 ITE Mid-Rise Apartments (use 223) 0.30 Source: Lee, 2004 Table 1.5 shows trip rates for trips leaving The Merrick Transit System and TOD in Portland, compiled by Dill (five bicycle trips were Land Use Influences recorded and that mode is not shown). These numbers were recorded via travel diaries (not tube counters) and thus will 1. What levels of transit connectivity to desired origins and be slightly lower than reality, as they do not include trips by destinations are required to promote transit ridership at visitors and The Merrick employees. However, these are likely TODs? to be a small number of trips. 2. What TOD land-use and design features (e.g., mixed land- Assuming every resident who leaves The Merrick returns, use, traffic calming, bus bulbs, short blocks, street furniture) the numbers can be doubled to approximate total trips to and have had an effect on travel patterns, transit ridership, or from The Merrick. Thus, the daily trip generation rate is ap- the decision to locate in a TOD? proximately 5.4 total trips per apartment, and 2.8 auto trips. This is lower than the rate the MPO uses from the ITE Trip Key Conclusions Generation book (about 6.6 total trips per apartment). Like Lapham, Dill speculates this is probably due to smaller house- Research shows that system extensiveness is positively hold sizes. The average number of people per apartment at correlated with transit ridership. The Merrick was 1.3, with 73% of the households having only Extensive transit networks, worse traffic congestion (i.e., one person. In contrast, in the 2001 National Household slow auto trip times), and higher parking costs work Travel Survey (NHTS), the average household size for people together to increase TOD transit ridership. living in apartments was just over 1.9 persons per household, General consensus is that transit service headways of with 26% only having one person. In addition, about 40% 10 minutes are ideal to support a transit lifestyle. have three or more people. Since the ITE rates are based on There is no single, definitive threshold for connectivity, and an average from trip counts taken at apartments all across measures such as "track miles" and "number of transit sta- the United States, it is likely that the average household size tions" on their own are not the best predictors of ridership. for the apartments measured by ITE is larger than at The What matters is transit travel times relative to auto travel Merrick. Given this likely difference in household size, the times. For example, an extensive but very slow transit system lower total trip rate seems reasonable, and highlights the fact likely will attract few riders if highway congestion is not se- that current ITE trip generation rates may differ significantly vere. Conversely, a single fast rail corridor adjacent to a highly from actual TOD trip rates. congested auto corridor likely will attract high ridership. Table 1.5. Trip rates by mode at The Merrick TOD. Trips From Merrick Trips From Merrick Per Person Per Apartment Unit Per Week Per Day Per Week Per Day Total Trips 16.72 2.39 18.81 2.69 Private Vehicle 8.81 1.26 9.91 1.42 Walk 4.82 0.69 5.42 0.77 Bus 1.10 0.16 1.23 0.18 Light Rail 1.93 0.28 2.17 0.31 Transit (Bus + LRT) 3.03 0.43 3.41 0.49 Source: Dill, 2005

OCR for page 12
13 The systems that will generate the highest commute rider- of whether TOD households use cars or transit (Tables 1.6 ship will have a high percentage of regional jobs accessible and 1.7). by fast transit. That said, less is known about specific accessibility thresholds For work trips, proximity to rail stations is a stronger in- (e.g., number of accessible jobs, households) to support a given fluence on transit use than land use mix or quality of walk- TOD. In TCRP Project H27, the research team noted that the ing environment. The most effective strategy to increase highest recorded rail capture rates are in the Washington, TOD ridership is to increase development densities in close D.C. area, and surmised this likely is related to the fact that proximity to transit. Metrorail has the most extensive network of any recent- Employment densities at trip ends have more influence on generation system in the country. Lund, Cervero, and Willson ridership than population densities at trip origins. It is crit- (2004) partly attribute higher transit mode shares for TOD ical to locate jobs near transit in order to attract households residents in the Bay Area (e.g., Pleasant Hill, Alameda City) to TODs. to a more extensive and mature rail system than other TOD Relative travel time (transit versus auto) is still more im- places [e.g., Long Beach (LA), Mission Valley (San Diego)]. portant than any land use factor (density, diversity of uses, In that research, the authors found a significant relationship or design). between transit ridership and an accessibility measure that Mixed uses in TODs allow the transit service to be used for divides jobs reachable by transit in 30 minutes by jobs reach- a variety of trip purposes throughout the day and week, but able by auto in 30 minutes. As one would expect, the more as a travel benefit, this is not a primary consideration for accessible a trip origin is to jobs by transit (relative to auto), prospective TOD residents. Employment access is a pri- the more likely the trip is to be made by transit. While regional mary consideration. travel models cannot predict the number of jobs or house- Mixed uses (e.g., local restaurants) and urban design treat- holds needed to support a particular TOD, they can predict ments (e.g., pedestrian pathways) are important for their reasonably well the ridership that will result from a TOD amenity and design value in attracting residents and visitors/ based on regional accessibility measures. customers. TOD residents highly value good neighborhood Transit travel times have a strong bearing on relative acces- design in addition to transit access to work. Urban design sibilities (by transit versus auto) and the decision to use transit. and the local land use mix may influence which TOD Cervero (2003) found that for non-transit users, auto travel prospective residents choose to live in. Good design also was on average 42 minutes faster than transit (for all trip pur- may make a TOD a more desirable location to travel to. poses), but for transit users, auto travel was only 23 minutes faster. This is consistent with many other studies that find Findings There is no absolute dividing line or tipping point for tran- Table 1.6. 2000 transit shares sit connectivity that translates into high transit ridership. for work trips. From a transit perspective, connectivity can relate to the Transit Metro number of origins and destinations that can be accessed, the Area Zones Area speed of transit service, and/or the frequency of service con- Chicago 25% 11% necting origins and destinations. Mode choice studies of Washington DC 30% 9% TOD residents and office workers typically show that transit Memphis 6% 2% travel times and their comparison to private car travel times is Cleveland 13% 4% Denver 12% 5% the strongest predictor of transit ridership. In other words, Charlotte 4% 1% travel time differentials are a critical factor, and these differ- Los Angeles 16% 5% entials can vary greatly depending on local circumstances. Source: CTOD, 2004 Census research by Reconnecting America's Center for Transit-Oriented Development (CTOD, 2004) provides a Table 1.7. 2000 percent auto commuters macro-level view of this dynamic. CTOD looked at 3,341 by transit system size. fixed guideway transit stations in 27 metropolitan regions. Transit zones were defined as the half-mile radius around the Transit % Auto stations, and the 27 transit systems were categorized as small, System Size Commuters Small 72% medium, large, and extensive. Like Renne, CTOD found that Medium 77% commuters in transit zones were much more likely to use Large 65% transit, and concludes that the size (i.e., extensiveness) and rel- Extensive 49% ative speed of the rail transit system is a significant determinant Source: CTOD, 2004

OCR for page 12
14 that slow transit travel times retard ridership growth. Riders 10% improvement in off-peak service levels increases rider- also care a lot about service reliability. Riders have been shown ship by 7% on average). to be more sensitive to unpredictable delay than predictable In Portland, for instance, TriMet has pursued a strategy of waiting times (Pratt, 2000, Chapter 9). TODs should be fo- improving off-peak bus service in its most dense and mixed cused toward transit facilities that offer clear travel speed and use (i.e., TOD-like) corridors to expand its nonwork trip mar- reliability advantages (e.g., rail lines or bus corridors with pri- ket. From FY 99 to FY 03, TriMet improved service on 10 lines ority design treatments). to "frequent service" (15 minutes or less all day, every day). On Numerous studies under the broader topic of transit oper- the improved lines, TriMet experienced a 9% increase in ations have been completed to understand how improved overall ridership, whereas ridership generally remained level transit service (i.e., faster speeds, improved frequency, differ- for routes with only nominal increases in frequency. For the ent configurations) affects transit ridership. These studies frequent lines, weekday ridership increased 8%, Saturday rid- have typically been undertaken to increase transit ridership in ership increased 14%, and Sunday ridership increased 21%. general, although the findings are directly applicable to im- Frequent bus service now accounts for 45% of weekly bus proving TOD-focused transit service and/or locating new hours and 57% of weekly bus rides. TODs. These studies have not been exhaustively reviewed for A generally accepted service level threshold for TODs is this literature review. Rather, only some general findings are headways of 15 minutes or less during most of the day presented here. (Dittmar and Ohland, 2004). It makes little sense to build TOD As would be expected, improved transit service levels makes in places that receive only hourly bus service, as service is not transit more convenient to use and improves transit ridership. frequent enough to make transit use convenient. Table 1.9 Services may be so frequent that riders don't need schedules, describes in more detail generally recommended transit ser- and frequent service provides more flexibility regarding de- vice levels for different types of TODs. parture and arrival times. For TODs it is important to have Other studies have focused more on the geographic aspects good service levels all day. Because TODs typically have a di- of transit service (e.g., system configuration) to see how rid- verse range of land uses, they require good service frequency ership is impacted. Ewing (1995) and others have found that during both the peak and off-peak periods, to serve both work accessibility to regional activities has much more effect on and nonwork trips. Table 1.8 gives a rough indication of rid- household travel patterns than density or land use mix in the ership impacts due to different transit service changes, and immediate area. Whereas accessibility to shopping or work- shows that off-peak frequency improvements can improve places alone is relatively less important, good access to shop- ridership more than other strategies (the data indicate that a ping, services, schools, work, and other households has a strong influence on travel patterns. While Ewing's research focused on vehicular hours of travel, the findings for TOD are Table 1.8. Typical ridership clear. Even if TODs show a propensity to generate higher than response to one percent change in listed factor. average transit ridership, they should not be built in remote locations with reduced accessibility (by all modes) to a wide Percent range of activities. Factor Change Recent research on the relative performance of alternative Peak Fare 0.20% transit configurations reveals that network orientation greatly Peak Frequency 0.20% Off-Peak Fare 0.58% affects the performance of rail and bus service. Based on data Off-Peak Frequency 0.70% from the National Transit Database, Thompson and Matoff Out-of-Pocket Auto Costs 0.70% (2000) conclude that: Source: ECONorthwest, 1991. APTA, 1991. Note: Influencing factors are: preexisting The best performing systems tend to be express bus-based service levels, geographic and demographic environment, and period of day or week. systems oriented to strong central business districts (CBDs) The response is greatest when prior service in rapidly growing regions, and multi-destinational, coor- is less than three vehicles per hour, when upper and middle income groups are dinated bus/light rail systems in growing regions. In multi- served, when a high number of short trips destinational networks, a rail line is a feeder to suburban can be served, and the local economy is strong. In some suburban places, off-peak buses, just as buses are feeders to the rail line. Multidesti- frequencies have achieved elasticities national networks typically appear in two configurations: near 1% when the service expansion was comprehensive and carefully planned. as a grid in high-density areas where frequent service on all (Pratt, 2000, Chapter 9) routes can be supported and as a timed transfer network in

OCR for page 12
15 Table 1.9. TOD types with land use and transit characteristics. TOD Type Land Use Mix Minimum Regional Frequencies Housing Density Connectivity Urban Office Center > 60 units per acre High <10 minutes Downtown Urban Entertainment Hub of regional Multiple Family system Retail Urban Residential > 20 units per acre Medium access to 10 minutes peak Neighborhood Retail downtown 20 minutes Class B Commercial Sub regional hub off peak Suburban Office Center > 50 units per acre High access to 10 minutes peak Center Urban Entertainment downtown 20 minutes Multiple Family Sub regional hub off peak Retail Suburban Residential > 12 units per acre Medium access to 20 minutes peak Neighborhood Neighborhood retail suburban center 30 minutes Local Office Access to off peak downtown Neighborhood Residential > 7 units per acre Low 25-30 minutes Neighborhood retail Demand responsive Source: Dittmar and Ohland, 2004 lower-density places where frequent service on all routes In its guidance for air quality conformance testing, FHWA can't be justified. notes that accessibility (i.e., the number of jobs accessible Whereas express bus systems are more oriented to peak pe- within a certain distance or time by mode) has a much stronger riod commuters traveling to CBD's, multi-destinational influence on travel than the 3 Ds, and unless density is above rail/bus networks are oriented to a broader mix of passengers 7-10 dwelling units per acre, it is unlikely that the other Ds and destinations. will have any effect, even in combination. (See www.fhwa.dot/ In comparison, traditional CBD-oriented bus transit sys- gov/environment/conformity/benefits/benefitsd/htm.) tems in rapidly growing regions are in decline. In this case, Density, or high shares of development within a 5-minute individual routes, or collections of unrelated routes, cannot walk of a station, has generally been shown to be the strongest compete in a dispersed trip market as each route only serves determinant of transit riding and walking among the land use origins and destinations on that single line. variables. Cervero (2005) estimated the following density elas- ticities for transit ridership during the course of developing The implications for TOD are that ridership is likely to be local ridership models for BART, Charlotte, North Carolina, maximized when TOD is located in express bus corridors and St. Louis, Missouri: linked to a healthy CBD, or located near rail corridors with robust connecting bus service. Charlotte Transitway TOD Scenarios: 0.192 (for persons Land use variables that affect travel are frequently described per gross acre within a half mile of a station). as pertaining to density, diversity (i.e., mixed uses), and design - BART Extension: 0.233 (for population and employment the 3 Ds. Cervero and Kockelman (1997) found that the within a half mile of a station). elasticities between various measures of the 3 Ds and travel St. Louis MetroLink South Extension: 0.145 (for dwelling demand are generally in the 0.06 to 0.18 range, expressed in units per gross acre within half mile of a station). absolute terms. They conclude that the elasticities between the land use factors and travel demand are modest to moder- While other studies have estimated much higher ridership ate, and higher densities, diverse land uses, and pedestrian- impacts attributable to development density, these studies friendly designs must co-exist if ridership benefits are to accrue. typically did not use control variables to hold the extraneous

OCR for page 12
16 factors of transit service levels, household demographics, and zone" for longer-distance work trips. That is, once work com- parking constant (e.g., prices). As a result, these factors may muters are within one-quarter mile of a rail station, factors like have influenced the results. The TCRP H-1 study, for instance, mixed land uses, traffic calming, pedestrian amenities, and estimated a high population density elasticity of 0.59, but failed even density seem to matter little. This is a consistent finding to include a measure of transit service levels. After accounting from studies on the ridership impacts of TOD, including the for transit service levels and other factors, Cervero re-estimated previously-cited research by Lund, Cervero, and Willson the density elasticity to be 0.192 (and the elasticity for the (2004). Availability, price, and convenience of parking strongly number of morning inbound trains was 0.59). determine whether or not those working in TODs take transit. Employment densities at destinations are more important Lund, Cervero, and Willson found that the only than population densities at trip origins. Having an office or neighborhood-design variable that explained commuting workplace near a transit stop is a strong motivator for many transit ridership among TOD residents was street connectiv- Americans to reside near transit and motivates people to buy ity at the trip destination. Once controlling for the influences into high transit-accessible neighborhoods. The end result is of factors like travel time and transit accessibility, no attributes that having both ends of the trip within a convenient walk to of walking quality or land-use composition in the neighbor- and from a transit stop is key to high ridership levels. hoods of TOD residents had a significant impact on transit Several studies have shown that good job accessibility via mode choice. Some of the correlations with transit ridership transit is among the strongest predictors of whether station- found in that study are: area residents will take transit to work. The 1994 Cervero study of commute choice among TOD residents of Bay Area TODs Pedestrian connectivity at trip destination: 0.37; found that having a workplace near a rail station strongly Sidewalks along shortest walk route: 0.16; encouraged rail commuting. Commuting to a job in BART- Street trees: 0.079; served downtown San Francisco or Oakland, for example, Street lights: 0.178; and increased the likelihood of taking transit by 35% to 60% Street furniture (benches, bus shelters): 0.137 among residents of suburban East Bay TODs. In another study of California TODs, Cervero (1994) found that four The researchers also found that urban design variables exert variablesemployment density, employment proximity to a stronger influence for station area workers than for station transit, commute behavior at the worker's previous job, and area residents. Furthermore, within each TOD, some will value occupationexplained 92% of the mode split variation. Orig- pedestrian treatments highly, while others will not be deterred inal research conducted by the team under TCRP H-27 for by their absence if transit is nearby. Thus, resident attitudes the Rosslyn-Ballston corridor of Arlington County, Virginia, matter considerably. That said, good urban design treatments showed that nodes of concentrated development along transit probably make living at higher densities more attractive. corridors translates into higher transit commute shares. In Ewing and Cervero (2001) note that individual urban de- Arlington County, every 100,000 square feet of office and retail sign features seldom prove significant. Where an individual floorspace added from 1985 to 2002 increased average daily feature appears to be significant, as did striped crosswalks in Metrorail boardings and alightings by nearly 50 daily board- one study, the causality almost is certainly confounded with ings and alightings. other variables. In this case, painting a few stripes across the Research shows that proximity to rail stations is a stronger road is not likely to influence travel choices, and the number determinant of transit usage for work trips than land-use mix of crosswalks must have captured other unmeasured features or quality of walking environment (Cervero, 1994). Concen- of the built environment. trating growth around rail stops often will yield high ridership Cervero (1994) concluded that for work trips, dividends almost regardless of the urban design attributes of the immediate area. Still, all transit trips involve walking Within a quarter to a half mile radius of a station, features of the built environment (ignoring issues of safety and urban to some degree, thus the provision of safe, efficient, and blight) matter little--as long as places are near a station, the comfortable-feeling walking corridors between transit stations physical characteristics of the immediate neighborhood are and surrounding communities is an essential attribute of inconsequential. successful TODs. Mixed uses like housing, offices, retail shopping, and entertainment centers are important compo- Another assessment underscores the importance of density nents of TOD since they produce all-day and all-week transit and proximity to a station, however, more value was attached trips, thus exploiting available transit capacity. to the land-use composition of a TOD: "transit use depends Studies show that the urban design features of TOD tend to primarily on local densities and secondarily on the degree of have a modest influence (relative to physical proximity) on rid- land use mixing" (Ewing and Cervero, 2001). For instance, ership patterns, and suggest the presence of an "indifference using data on more than 15,000 households from the 1985

OCR for page 12
17 American Housing Survey, Cervero (1996) found the presence (e.g., whether and how to access work or a shopping center), of retail shops within 300 feet of one's residence increased the it is more likely to affect secondary trips from an activity probability of transit commuting, on average, by 3%-- center, which can be made by car, transit, or on foot. ostensibly because transit users could pick up convenience Because of their pedestrian orientation and mix of land items when heading home after work. uses, TODs can significantly increase the number and per- Not all recent evidence diminishes the importance of cent of local trips made by walking and cycling in particular. urban design on the travel choices of TOD residents. The Table 1.10 shows how the share of walk, bike, and transit TCRP H-27 study found, for example, that the combination trips for the Portland metropolitan region are higher in of high densities and small city block patterns significantly neighborhoods with TOD characteristics. Most notably, increased the share of station-area residents in the San Francisco walk trips almost double when mixed uses are included in Bay Area who took transit to work in 2000. In addition, auto- areas with good transit service. restraint measures, like traffic calming and car-free streets, Using primary data from urban residents in the San likely have some marginal influence on ridership to the degree Francisco-Oakland-San Jose MSA and San Diego County and walking becomes safer, easier, and more enjoyable. negative binomial regressions, Chatman (2005) found that The quality of walk and bus access to and from stations access by transit to nonwork activities increased by 22.6% for should also be considered. Although parking supplies and each 1,000 retail workers within a quarter mile of residences. prices at the trip destination more strongly influence rider- This robust relationship was found for all of the nearly 1,000 ship among TOD residents than parking at the nearby rail residential households that were sampled. Adding a rail station station, the design and siting of station parking lots bears yielded a significant further bump in ridership. For residences some influence on transit demand. Peripheral parking lots within a half mile of a light-rail station in San Jose or San that do not sever pedestrian paths to nearby residential neigh- Diego, the number of nonwork activities by transit rose an borhoods, for example, may induce transit usage, although additional 6.5%. A far bigger bonus was found for high- this has not been tested empirically. performance regional rail services: for those living within a Transit travel times, which tend to be short when transit half mile of a BART heavy-rail or CalTrain commuter rail sta- enjoys high connectivity, are far stronger predictors of rail tion, the number of nonworker activities via transit rose a re- usage for TOD commuters than land-use, urban-design, and sounding 284%. Besides retail density, pedestrian connectiv- demand-management variables. Based on standardized model ity increased transit's mode share of nonwork trips. On the coefficients, the predictive power of transit travel-time vari- other hand, as walking quality increased, transit trips seemed ables tends to be two- to three-times greater than land-use to switch to travel by foot. and policy-related variables, and based on modal travel time Chatman's work strongly suggests that the quality of the differences many travel models can predict transit ridership walking environment significantly influences travel choices at TODs reasonably well. for nonwork travel. Walk/bike travel to nonwork activi- TOD land use features are more likely to affect travel behav- ties was found to increase by 7.1% for every 1,000 retail work- ior for shorter-distance, nonwork trips. To the degree that ers within a half mile radius of sampled residences. These housing, offices, shops, restaurants and other activities are results show that the combination of intensifying retail activ- intermingled, people are less likely to drive and more likely ities with good pedestrian facilities near regional rail stations to walk to nearby destinations. Similarly, while urban design can dramatically increase the use of transit for nonwork is likely to only have a marginal impact on primary trips purposes. Table 1.10. Metro travel behavior survey results, all trip purposes (Portland, Oregon). Mode Share Daily VMT per Land Use Type % Auto % Walk % Transit % Bike % Other Capita Good Transit & Mixed Use 58.1% 27.0% 11.5% 1.9% 1.5% 9.80 Good Transit Only 74.4% 15.2% 7.9% 1.4% 1.1% 13.28 Rest of Multnomah Co. 81.5% 9.7% 3.5% 1.6% 3.7% 17.34 Rest of Region 87.3% 6.1% 1.2% 0.8% 4.6% 21.79 Source: Metro 1994 Travel Behavior Survey VMT = vehicle miles traveled

OCR for page 12
18 Using 2000 data collected from more than 15,000 house- In a test model, the urban design variable was retained, and holds sampled in the San Francisco Bay Area, Gossen (2005) TAZs that contain built TOD projects were given an addi- studied travel and sociodemographic attributes for seven tional code (the TODs were identified via a qualitative assess- distance/density categories based on households' proximity to ment by local TOD experts). Table 1.11 shows how inclusion rails stations and ferry terminals. Regarding nonwork travel, of the TOD variable allows the model to more closely match Gossen found that transit made up these shares of nonwork observed mode share totals. trips for the following distance rings: 14.2% (up to 1/4 mile); Evans and Stryker's results show that in centrally located and 11.5% (1/4 to 1/2 mile); 6.1% (1/2 to 1 mile); 1.6% (> 1 mile - low- outlying TODs, walking's share of nonwork trips is more than density suburbs). Gossen also found that VMT per capita in- twice that for non-TOD areas, and that transit use is signifi- creased with distance from rail/ferry stations in the following cantly higher in central TODs (7% compared to 1%) where fashion: 19.9% (1/4 mile); 24.1% (1/4 to 1/2 mile); 29.4% (1/2 to local and connecting transit service is most robust. The results 1 mile); 45.0% (> 1 mile - low-density suburbs). also show that the standard and urban form models capture Evans and Stryker (2005) conducted research on Portland most mode choice behavior for nonwork trips. Adding a TOD TODs to see if the presence of TOD design features is de- land use variable to account for the influences of unspecified tectable using a travel demand model for nonwork trips. In factors (e.g., parking configuration, street lights) improves the other words, does designating a travel analysis zone (TAZ) as model results only modestly and most noticeably for Central including TOD add explanatory power to a base travel model TOD transit use, which increased from 5% to 7%. [The urban for non-work trips? form and TOD variables were not found to be correlated. The In the Portland travel models, an urban design variable author also cautions against using TOD dummy variables in that captures the number of retail businesses, households, travel models, because 1) it is not good practice to overuse and street intersections within a half mile of each zone is cur- dummy variables, particularly ones that may measure a contin- rently used to estimate nonwork trips. The variable is formu- uous attribute (e.g., degrees of TOD-ness) and 2) using a TOD lated so that places with a moderate mix of all three elements variable requires an analyst to arbitrarily designate TODs in score higher than places with very high amounts of only one the base year and in future years, potentially introducing bias element. into the model.] Table 1.11. Nonwork trip attractions by TOD types and travel mode (Portland, Oregon). Area Source Walk Bike Transit Auto Actual 444 16% 50 2% 198 7% 2043 75% Central TOD Standard Model 373 14% 53 2% 133 5% 2176 80% Urban Form Model 453 17% 56 2% 126 5% 2100 77% TOD-Included Model 460 17% 50 2% 184 7% 2041 75% Actual 133 17% 11 1% 12 2% 626 80% Outlying TOD Standard Model 101 13% 11 1% 14 2% 656 84% Urban Form Model 106 14% 12 1% 15 2% 649 83% TOD-Included Model 117 15% 11 1% 26 3% 628 80% Actual 1401 7% 217 1% 195 1% 19,388 91% Non-TOD Standard Model 1504 7% 214 1% 258 1% 19,225 91% Urban Form Model 1419 7% 210 1% 263 1% 19,308 91% TOD-Included Model 1401 7% 217 1% 195 1% 19,388 91% Actual 1978 8% 278 1% 405 2% 22,057 89% Overall Standard Model 1978 8% 278 1% 405 2% 22,057 89% Urban Form Model 1978 8% 278 1% 405 2% 22,057 89% TOD-Included Model 1978 8% 278 1% 405 2% 22,057 89% Source: Evans and Stryker, 2005.