Effect of Smart Growth Policies on Travel Demand (2013)

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.

9C h A P T e R 2 Key Decision Points for Smart Growth in the Planning Process The Highway Capacity Planning Process State DOT highway capacity planning processes involve a series of decision points at which smart growth might be con- sidered. Figures 2.1 and 2.2 present these process maps for state DOTs and MPOs, respectively, and identify the areas where smart growth levers are used. In some cases, there are only a few agencies using these levers, but in most cases, there are many agencies incorporating smart growth levers into their processes. This map also correlates the phases from the TCAPP online tool, where the smart growth products from this study will reside. In general, there are four dimensions of the capacity plan- ning process in which smart growth considerations may be applied: • Policy (Statewide Transportation Plan and Metropolitan Transportation Plan); • Planning (planning studies); • Programming [State Transportation Improvement Plan (TIP)/Capital Program and MPO TIP]; and • Implementation, including National Environmental Policy Act of 1969 (NEPA) and project development. Consideration of smart growth issues in the highway capacity planning process in each of these dimensions varies substantially across the country and is also changing rapidly, as more agencies find that consideration of smart growth strategies is useful and necessary to achieve reductions in congestion and emissions. And while there is significant research on the topic of evaluating smart growth strategies to evaluate transportation impacts, there are few applications documented that clearly guide a planning agency in the pro- cess or consider the challenges in this type of analysis. The current state and MPO highway capacity planning process shows feedback from the project evaluation back to long- range planning based on performance measures but does not reflect feedback from project evaluations to land use planning activities. When capacity thresholds are exceeded, the response could be to adjust transportation plans or land use plans, thus providing feedback to both aspects of long-range planning. The feedback to the land use plans can identify areas suitable for new or expanded development. TCAPP (http://www.transportationforcommunities.com/) is a decision-making framework software designed to encour- age collaboration in the transportation planning process. The SHRP 2 program also has a related online resource called Transportation Visioning for Communities (http://shrp2 visionguide.camsys.com/) or T-VIZ. According to the T-VIZ website, “The information available on this site is intended to assist transportation agency practitioners in assessing the possibilities of visioning, in identifying practical steps when engaging in visioning, and in establishing links between vision outcomes and transportation planning and project development processes.” Examples of smart growth considerations in different dimen- sions of the planning process are presented in Table 2.1. These examples are planning topics that state and regional planning agencies are engaged in to consider smart growth strategies in the planning process. While this list is not intended to be com- prehensive, it does highlight the range of smart growth consid- erations that can be considered at different decision steps in the process. One important fact is that most land use planning and regulatory authority remains in local government hands. As a result, most state and MPO efforts toward considering smart growth are geared toward enhancing communication, cooperation and collaboration. In order for smart growth strategies to be effective, goals among the land use planning and transportation planning agencies could align or be com- plementary, and agencies could cooperate on the means to achieve these goals. Background Research

10 Most current smart growth strategies are developed for urban areas, and there is much less understanding of smart growth strategies in rural areas or small towns. There may often be different goals for rural areas, such as economic devel- opment, where urban areas would be more focused on mobil- ity, the environment and growth management. State DOTs are challenged to evaluate smart growth strategies in rural areas. Interviews with Planning Officials RSG conducted eight interviews on how smart growth is inte- grated and/or considered in the planning process with a small number of state DOTs, MPOs, and federal agencies: • The Capital District Transportation Committee • The Maryland DOT • The Oregon DOT Capital District Transportation Committee • Metropolitan Washington Council of Governments • Thurston Regional Planning Council (TRPC) • Sacramento Area Council of Governments (SACOG) • Federal Highway Administration • U.S. Environmental Protection Agency The candidates for the interviews were selected to reflect a variety of geographies, population sizes, and viewpoints. The list of questions varied for each type of agency, but was designed to understand the specifics of how smart growth strategies were included in the transportation planning process. The list of questions for each agency is provided in Table 2.2. The interviews are summarized along several key dimen- sions to frame the discussions of smart growth: • Legislative actions; • Goals and objectives; • Strategies; and • Performance metrics and tools. These interviews were designed to articulate the key informa- tion gaps and questions associated with them. Figure 2.1. State DOT highway capacity planning process map for smart growth strategies. ROW  right-of-way. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

11 Figure 2.2. MPO highway capacity planning process map for smart growth strategies. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx. Legislative Actions Several states identified laws mandating growth manage- ment (Maryland, Oregon, and Washington State) and one state (New York) has recently passed smart growth legisla- tion that requires state agencies to evaluate public infra- structure projects they fund against smart growth criteria. The 10 smart growth criteria include topics such as the following: • The use or improvement of existing infrastructure; • Development in areas that are already developed or in areas that are designated for concentrated infill develop- ment in local land use plans; • Mixed land uses and compact development; • Preservation of open space; • Improved public transport and reduced automobile dependency; and • Collaboration among state agencies and localities to pro- mote intermunicipal and regional planning. In addition, several states have set greenhouse gas (GHG) reduction targets (Washington State, Oregon, and New York), which will lead to the integration of land use and transportation planning. California has also mandated incorporation of land use with transportation analysis and adoption of GHG reduc- tion targets through SB 375 legislation, which encourages smart growth. The Sustainable Communities Strategy (SCS) provides land use and transportation connections to help meet these GHG reduction targets for MPOs in California. These sustain- able community strategies must be included in the periodic update and revision of regional transportation plans (RTPs). Also, there is an outlet that allows communities that are unable to meet GHG reduction targets through smart growth pursue TDM strategies, such as parking restraints or road pricing—or what is called alternative planning strategies (APS). There are, of course, likely synergies from pursuing SCS and APS in com- bination; however, this is an area in which empirical knowledge lags and for which forecasting and scenario testing models probably fail to account for synergistic benefits.

12 Table 2.1. Examples of Smart Growth Considerations in Planning Processes Decision Step Dimensions of Planning Process Examples of Smart Growth Considerations Definition of Corridor Corridor Planning • Recognition of impacts beyond the corridor Problem Statement/ Purpose and Need Corridor Planning Permitting/NEPA • Land use patterns and growth forecast are critical • Consistency with vision/community plans • Accessibility, economic, congestion, and mobility measures Goals Long-Range Planning Corridor Planning • Mobility • Growth management • Economic development • Environmental • Quality of life Scope of Analysis and Review Corridor Planning • Induced development? Induced travel? • Integrated corridor planning? Evaluation Criteria and Performance Measures Long-Range Planning • Built environment metrics • Modal balance, accessibility, and demand metrics • Congestion and impact metrics • System performance and safety • Economic, social justice and social equity • Environmental sustainability Identify Transportation Needs Long-Range Planning Programming Permitting/NEPA • System performance and safety • Modal balance • Federal and state funding criteria, such as “livability,” impact avoidance • Social equity • Effects of smart growth on travel demand, congestion, conformity • Triple bottom line: economic, environmental, societal return on investment Financial Assumptions Long-Range Planning • Federal and state funding criteria, such as “livability,” impact avoidance Identify Potential Strategies Long-Range Planning • Land use, transportation, and policy considerations Create Alternatives Long-Range Planning Corridor Planning Permitting/NEPA • Integrated land use and transportation “blueprint” alternatives • Trade-off and balance between transportation and land use criteria Analyze Alternatives Long-Range Planning Corridor Planning Permitting/NEPA • Integrated land use and transportation modeling • Postprocess travel model results to account for smart growth (sketch planning approach) • Interactive, quick-response tools (for local factors, site-specific evaluation) • Validate/adjust models as needed to account for smart growth and create consistency between local and regional analysis • Consider induced demand Select Preferred Alternative Long-Range Planning Corridor Planning Permitting/NEPA • Triple bottom line: economic, environmental, societal return on investment Conformity Determination Long-Range Planning Permitting/NEPA • Effects of smart growth on travel demand and congestion Project Prioritization Programming • Does the project encourage smart growth patterns? • Does the smart growth alternative reduce congestion? • Does the smart growth alternative meet other criteria above? Sequencing/Phasing Plan Corridor Planning • Consider growth inducement, primary/indirect impacts by phase

13 Goals and Objectives All of the interviewees cited goals and objectives that were for- mally adopted, although, to be fair, this short list of agencies was chosen because of their advances in this area. Goals were cited in statewide and regional transportation plans, climate action plans, and freight plans. Some goals were aimed at coordinating land use and transportation planning better; some goals were aimed at communicating and cooperating to achieve mutually beneficial land use and transportation objec- tives; and some goals were aimed at reducing transportation impacts through land use policy. The Albany, New York, MPO cited a transportation land use linkage program as an important tool for achieving these goals. The Sacramento MPO adopted a “Blueprint” in 2004, which was a bold vision for growth that promoted compact, mixed-use development and more transit choices as an alternative to low-density development. The Olympia MPO (Thurston County, Washington State) stated that congestion reduction is no longer a goal, since this improves the system for auto users and they are striving to improve the system for all users (not just auto users). Focus- ing on congestion reduction may be counterproductive, because smart growth includes compact development, which may result in more congestion for auto users, but also more options or more mobility for non-auto users. This is an Table 2.2. List of Questions for Each Agency Questions for MPO Staff 1. Does your region or state have any laws mandating integration of land use and transportation planning? 2. Has your agency formally adopted any objectives related to smart growth (e.g., jobs-housing balance or land preservation), or goals which smart growth can significantly help achieve (e.g., carbon emissions targets)? 3. Does your agency have any specific strategies to encourage smart growth? 4. Does your agency do (integrated) scenario-based planning? 5. Does your agency consider smart growth with its technical methods? a. Do you utilize visioning and scenario-comparison tools in your planning process (e.g., MetroQuest, INDEX, CommunityViz, Envision Tomorrow, iPLACE3S)? b. Do you utilize specific smart-growth-related performance measures to help make transportation decisions? • Balanced accessibility by variety of travel modes • Benefits of location-efficient placement of transportation and land use to reduce travel demand • Triple-bottom-line performance evaluation of the transportation system: economic, environmental, and livability metrics • Social impact and equity metrics such as health and safety • System speed suitability to adjoining land use and activity c. Are your models reliably sensitive to urban form variables (such as land use mix and walkability), and to TDM measures (both incentive based and cost based)? d. Do you try to estimate induced travel? What about induced growth? 6. What strategies work best to accomplish the following goals: reduce congestion and reduce emissions? Questions for DOT Staff 1. Does your region or state have any laws mandating integration of land use and transportation planning? 2. Has your agency formally adopted any objectives related to smart growth (e.g., jobs-housing balance or land preservation), or goals which smart growth can significantly help achieve (e.g., carbon emissions targets)? 3. Does your agency have any specific strategies to encourage smart growth? 4. Is your agency involved in funding any smart growth–related research or studies? 5. Does your agency consider smart growth within corridor/environmental studies? a. Do you utilize visioning and scenario-comparison tools in your planning process (e.g., MetroQuest, INDEX, CommunityViz, Envision Tomorrow, iPLACE3S)? b. Do you utilize specific smart-growth-related performance measures to help make transportation decisions? • Balanced accessibility by variety of travel modes • Benefits of location-efficient placement of transportation and land use to reduce travel demand • Triple-bottom-line performance evaluation of the transportation system: economic, environmental, and livability metrics • Social impact and equity metrics such as health and safety • System speed suitability to adjoining land use and activity c. Are your models reliably sensitive to urban form variables (such as land use mix and walkability) and to TDM measures (both incentive-based and cost based)? d. Do you try to estimate induced travel? What about induced growth? Questions for National Agency Staff 1. How does your agency encourage consideration of smart growth in the transportation planning process? The project develop- ment (e.g., EIS) process? 2. Is your agency involved in funding any smart growth related research or studies? 3. What are some noteworthy examples of incorporating smart growth into transportation planning and project development efforts?

14 example in which the goals and performance measures to achieve that goal would need to be aligned with each other and with the overall purpose of any smart growth strategies. Strategies There were many land use and transportation policy strate- gies cited as examples in the interviews and many of these were cited by more than one agency. Some of the strategies were specifically aimed at coordination between land use and transportation. A selection of strategies cited in these inter- views is provided in Table 2.3. These strategies have some common features around coordination (among policies, modes, centers, streets), growth management [urban growth boundaries, transit-oriented development (TOD), centers], and non-auto alternatives (transit, bike, and pedestrian modes). FHWA mentioned that it is providing scenario plan- ning workshops to provide more focus on smart growth strat- egies, and scenario planning was also mentioned by several agencies as a potential strategy. Performance Metrics and Tools The interviews were designed to ask specific questions about a series of tools and performance metrics. • Visioning and scenario planning tools. The University of Maryland has a Scenarios Project being used by the Mary- land DOT. The Oregon DOT has a scenario planning tool for greenhouse gas reduction called GreenSTEP (Green- house Gas Statewide Transportation Emissions Planning), which is also being enhanced by FHWA for general use by other planning agencies; Thurston County will begin to use a scenario planning tool called CommunityViz as part of a regional sustainability grant. Some agencies did not use any such tools. EPA supports CommunityViz in vari- ous locations and the Utah Envision Tomorrow Plus effort. Sacramento Area Council of Governments (SACOG) uses the Internet-based land use modeling tool Planning for Community Energy, Economic, and Environmental Sus- tainability (I-PLACE3S) to evaluate urban and rural land use changes and has engaged in keypad polling to identify values and games to help develop inputs to I-PLACE3S. • Smart growth related performance measures. Most agen- cies responded that they do include smart growth related performance measures in making transportation deci- sions. These measures include community quality of life, urban equity or environmental justice (EJ), economic, environmental measures, livability, safety, health, sustain- ability, and energy supply. EPA has been supporting the development of smart growth–related planning tools, such as the INDEX and Smart Growth Index, and has funded the creation of a map of the 4 Ds at the Census block group level. The concept of the 4 Ds is discussed in the next section. • Tools sensitive to urban form or TDM strategies. The general consensus to this question was mostly “no,” with some cur- rent work described that will provide some of these capa- bilities, such as expanding zones to represent mixed-use centers better, modeling nonmotorized modes directly, and modeling dynamic traffic to test the effects of staggered start times and improved parking access. One agency men- tioned interest in a development tool to identify changes in trip making and VMT reduction in a planning area. SACOG was an exception here, since their travel demand model is an activity-based model with parcels and can address some of the urban form and travel demand management strategies. Table 2.3. Example Land Use Policy, Transportation Policy, and Coordinated Strategies Land Use Policy Strategies Transportation Policy Strategies Coordinated Strategies • Set urban growth boundaries. • Provide transit-oriented development and mixed land use. • Support regional activity centers, urban reinvestment, and concentrated development patterns. • Set aside agricultural and natural resource lands. • Break down barriers for better land use and mixed use by working with private sector through public– private partnerships (PPPs). • Exempt urban development from concurrency regulations. • Down-zone rural areas. • Establish connected streets policies (e.g., complete streets). • Provide transportation demand management such as telework partnerships and guaranteed-ride-home programs. • Establish arterial management program to promote properly located and spaced driveways and signalized intersections, use of raised medians. • Set design details for sidewalks and bike lanes in street standards and provide impact fees to pay for these improvements. • Coordinate signal priority for transit and other operational improvements for traffic and incident management. • Develop a partnership for safe walk routes to school and education on why you should not drive your kids to school. • Provide alternatives to driving in the regional core and into regional activity centers. • Build bicycle and pedestrian improvements. • Price transportation corridors, areas, or facilities. • Coordinate policies between MPOs and cities and counties. • Provide funding for cities and towns to prepare community plans that coordinate land use and transportation. • Conduct scenario planning. • Conduct public outreach/ education.

15 • Induced demand. Most agencies said that they had dis- cussed induced demand, but not formally estimated induced growth or traffic. The Albany MPO said it con- sidered induced growth in the context of scenario plan- ning rather than land use modeling. The Washington, D.C., MPO considers induced growth by using Delphi methods. SACOG considers induced growth using quali- tative analysis because its current modeling tools are not able to estimate induced demand reliably. SACOG also has a policy not to fund capacity expansion at the urban fringes. From these series of interviews, it was determined that there is room for improvement in the use of tools and perfor- mance measures to evaluate smart growth policies. Key Practitioner Information Needs The review of planning processes with a focus on smart growth and the interviews conducted with planning offi- cials on this same topic revealed two primary areas that planning agencies are engaged in that may be useful and supportive of engaging smart growth in planning pro- cesses. The first area is that most agencies are either engaged in or interested in scenario planning as a strategy for evalu- ating smart growth. Scenario planning offers many oppor- tunities, but to date has not been developed into a tool for this purpose that could be shared or adapted for use by planning agencies. The second area is that many agencies reflected on the need for coordination, cooperation, and communication with local governments on land use pol- icy, since land use regulations are primarily governed by local governments. This interaction between land use and transportation planners has provided opportunities to engage in discussions about integration, interaction, and common goals. The review also highlighted several topics in which plan- ning agencies feel additional guidance or tools would be worthwhile: • Metrics and tools for induced demand, TDM, and urban form. • Understanding which strategies work best, that is, what outcomes can be expected? • Tools to evaluate impacts of smart growth on project selection. • Goals for congestion reduction may be counterproductive to smart growth. These topics were considered during the development of the software tools to ensure that the needs of the planning agen- cies were met, if possible. The Built environment’s Impacts on Peak Auto Demand Considerable Evidence on the Effects of Smart Growth on Daily VMT Ewing and Cervero (2010) conducted a meta-analysis that focused on aggregate vehicle trip and VMT results rather than specifically on peak-hour trips. After more than 200 built environment studies were reviewed, it was found that VMT is most strongly correlated to measures of accessibility to desti- nations and secondarily to street network design variables. The Ewing and Cervero meta-analysis provides elasticities tied to D built environment variables. These include • Density gauges how many people, workers, or built struc- tures occupy specified land area, such as gross hectares or residentially zoned land. This is defined as the population and employment per square mile. • Diversity reflects the mix of land uses and the degree to which they area spatially balanced (e.g., jobs-housing bal- ance) as well as the variety of housing types and mobility options (e.g., bikeways and motorways). This is defined as the ratio of jobs to population. • Design captures elements, like street network characteris- tics, that influence the likelihood of walking or biking (e.g., pedestrian- and bike-friendliness). Street networks vary from dense urban grids of highly interconnected, straight streets to sparse suburban networks of curving streets forming loops and lollipops. • Destinations accessibility measures ease of access to trip destinations, such as the number of jobs or other attrac- tions reachable within 30 minutes travel time. • Distance to transit measures the distance to the nearest transit stop. The first four of these built environment variables are often referred to as the 4 Ds and when the fifth variable (distance to transit) was added, the term was adjusted to reflect 5 Ds. These are not separate dimensions and indeed are often code- pendent. Having high-rise housing and office towers will yield few mobility benefits if the two activities are far from each other. A diversity of uses and improved accessibility to destinations from home or work are needed if denser devel- opment is to translate into more pedestrian and transit trips. The densest parts of most cities, which are downtowns, also tend to be the most land use diverse and most walkable (e.g., small city blocks, complete sidewalk networks, and fine- grain grid street patterns). For each variable, weighted-average elasticities of VMT are provided. The body of work reviewed in the study, as well as the resulting elasticities, focuses almost exclusively on VMT or vehicle hours traveled (VHT) per house- hold rather than on peak auto demand. The meta-analysis

16 builds off work previously conducted by Cervero and Kockelman (1997). Studies Focusing on Peak Auto Demand There are a few studies that have focused on connecting built environment characteristics specifically to peak auto demand. Generally, the built environment factors that have been high- lighted to give some reduction to peak auto demand include the overall characteristics of a TOD, the mix of uses at the employ- ment site, and the jobs-housing balance of an area. Historically, studies on peak auto demand have focused on commute trips. The National Household Travel Survey (NHTS) briefs show that nonwork vehicle trips are an increasing percentage of peak-period trips and thus highlight a need to study the built environment relationships to all type of vehicle trips. While a considerable literature has evolved for measuring the impacts of smart growth on travel, broadly defined [e.g., average daily traffic (ADT), VMT, modal splits], work on peak-period impacts, and by implication the effects on-road congestion, has been far more limited. This could reflect the numerous objectives that propel smart growth initiatives, which might include traffic congestion relief but more often than not stress other factors like reducing energy consumption and GHG emissions, expanding housing choices, encouraging increased physical activity, and reducing fiscal outlays for infra- structure and services relative to sprawl. For gauging energy consumption and tailpipe emissions, VMT might be a pre- ferred performance metric. For the study of how mixed-use development and sidewalk investments might promote phy- sical activity, the output metric of interest is apt to be modal splits (e.g., percentage of trips by walking and cycling). Add to this the fact that little VMT data are broken down by the peak period and that the sample sizes of household travel surveys are sometimes too small to partition trips by time of day for small geographic areas, a scarcity of data points has signifi- cantly constrained the ability to conduct research on how built environments influence peak auto travel. One might be inclined to examine the effects of built environments on work trips under the premise that journeys-to-work are concen- trated in the peak. According to the NHTS Brief: National Household Travel Survey, in 2001, however, more than half of all trips during the 6:00 to 9:00 a.m. period were for nonwork purposes and during the p.m. peak, the share exceeded 70%. On Fridays, four out of five vehicle trips during the afternoon peak were for purposes other than commuting. There are no easy alternatives to gauging the impacts of built environ- ments on traffic congestion other than to study relationships during the peak period itself. The NHTS briefs highlight that a significant number of nonwork vehicle trips are being made during peak periods (FHWA 2007a). On an average weekday, nonwork travel con- stitutes 56% of trips during the a.m. peak and 69% of trips during the p.m. peak. The trends show that the amount of travel for nonwork purposes is growing faster than work travel. Growth in these kinds of trips is expected to outpace growth in commuting in the coming decades. After trips to work, and giving someone a ride, the next largest single rea- son for travel during the peak period is to shop. Just since 1995, 25% more commuters stop for incidental trips during their commutes to or from work, and stopping along the way is especially prevalent among workers with the longest com- mutes. While e-commerce and Internet shopping have reduced the need for some physical travel to retail outlets, evidence sug- gests that such shopping can also have a stimulating effect by promoting consumerism and expanding knowledge networks, prompting some individuals to comparison-shop more often (Ferrell 2005). Two older Cervero studies (Cervero 1988; Cervero 1989a) provide some evidence on how to reduce peak auto demand specifically for suburban environments. The 1988 study looked at the effects of current land use mixes on the commuting choices of suburban workers based on an empirical analysis of some of the largest suburban employment centers in the United States. Overall, the findings show that single-use office settings seem to induce solo commuting, whereas work envi- ronments that are more varied generally encourage more ridesharing, walking, and cycling. While the synchronization of job and housing growth around suburban centers could be expected to encourage more foot and bicycle travel, at the same time, ridesharing and vehicle occupancy levels could be expected to fall off some. The 1989a study found similar results showing that single-occupant vehicle commuters decrease as a suburban employment center becomes denser and it features a wider variety of land uses. The availability of retail activities appears to induce a number of suburban workers to carpool and vanpool to work because in these set- tings they can get to banks, shops, restaurants, and the like without a motor vehicle. This section divides the literature on the impacts of built environments on peak auto demand into two groups: case- based analyses (A. on Figure 1.2) and variable-based analyses (B. on Figure 1.2). This division partly reflects how the body of research appears in published literature. Some studies compare neighborhoods with versus without TOD or other smart growth forms, ideally matching the cases on other fac- tors that influence travel, such as household income and lev- els of regional accessibility. Matched-pair analyses, sometimes also referred to as quasi-experimental studies, can provide real-world, grounded insights and contrasts into the travel impacts of land use interventions. With the availability of rich GIS data, far more studies—particularly those over the

17 last decade—have been based on statistical relationships between variables using various model structures, what is being called variable-based analyses. To the degree that pre- dictive models of population density on VMT are well speci- fied, controlling for other explanatory variables, variable-based models are generally preferred. This is partly because results can be expressed in metrics, like elasticities, that provide order- of-magnitude estimates of impact and partly because they are considered more internally valid, reducing the chance of con- founding influences or spurious results. That said, cases often resonate with politicians and the general public. Politicians often rely on case examples to drive home points. They also may be more inclined to listen to cases, in part because their constituents do. A study of urban poverty in Boulder, Colo- rado, showed that case-based analyses were more effective at influencing political outcomes than variable-based analyses derived from statistical techniques (Brunner et al. 1987). Together, case-based and variable-based findings provide a rich and often complementary perspective on the subject at hand: built environments and peak-period travel. Case-Based Analyses From a case-based perspective, research on built environ- ments and travel occurs at multiple scales. They are (a) micro: project and neighborhood scales; (b) meso: community, corridor, and subregional scales; and (c) macro: regional scales. Examples of microscale smart growth initiatives include traditional neighborhood development (or design) (TND), new urbanism, and TOD. At the mesoscale, smart growth might take the form of a mixed-use suburban activity center (versus a single-use office park) or a transit-oriented corridor (TOC) (versus an auto-oriented corridor). Regional- scale initiatives might include jobs-housing balance and urban containment programs such as urban growth boundaries (UGBs). Table 2.4 provides a summary of geographic scales and the settings and place types typically associated with each. Throughout this report, these scales will be mentioned, particularly with regard to tool applicability and geographic extent of case-based analysis. It was hoped that empirical evidence of smart growth’s influ- ences on peak auto travel would be available at multiple scales. After an extensive canvassing of the literature, by using various bibliographic search platforms such as TRIS Online, Google Scholar, TRANweb, Melvyl, and ISI Web of Knowledge data- bases, case materials on smart growth and peak travel fell into a more limited grouping, notably, two scale and two forms: micro–TOD; and macro–jobs-housing balance. Transit-Oriented Development The congestion-relieving potential of TOD has long been debated. Downs (2004a) argued that TOD will not reduce car traffic unless three conditions are met: (1) a critical mass of TODs in a region, (2) relatively high residential and/or employment densities within each TOD, and (3) a high per- centage of employed-residents and workers of the TOD who transit commute. Both residences and destinations, such as job sites and shopping venues, need to be concentrated around transit stations to assure both trip origins and desti- nations are linearly aligned along a rail- or BRT-served cor- ridor (Cervero 2007a). Even then, not everyone believes that Table 2.4. Smart Growth Typologies Geographic Scale Setting/Place Type Urban Centers Close-in Compact Communities Suburban Rural/Exurban Macro/Regional Adaptive Reuse/Infill/ Redevelopment Mixed-Use Development/ Activity Center Adaptive Reuse/Infill/ Redevelopment Jobs-Housing Balance Mixed-Use Development/ Activity Center Adaptive Reuse/Infill/ Redevelopment Jobs-Housing Balance Telecommunities Mixed-Use Development/ Activity Center or Traditional rural township Meso: subregional/ corridor Jobs-Housing Balance Transit-Oriented Corridor Transit-Oriented Corridor Jobs-Housing Balance Transit-Oriented Corridor Jobs-Housing Balance Mixed-Use Development/ Activity Center Telecommunities Mixed-Use Development/ Activity Center or Traditional rural township Micro: neighborhood/ community Transit-Oriented Development Transit-Oriented Development Traditional Neighborhood Design/New Urbanism (residential focus) Transit-Oriented Development Traditional Neighborhood Design/New Urbanism (residential focus) Telecommunities

18 TODs will delivery mobility benefits in car-dependent societies such as the United States. In an interview for Com- mon Ground, a trade journal of the National Association of Realtors, Wendell Cox expresses this view: “TOD increases congestion. The overwhelming majority of travel to proposed transit-oriented developments will be by automobile. This will strain road space, slowing traffic and increasing pollution as a consequence” (Still 2002). While concentrated develop- ment might lead to more spot congestion at intersections near rail stations, incidents of increased congestion needs to be weighed against research that shows smart growth in gen- eral and TOD specifically tend to be associated with fewer VMT per resident and per worker than does conventional, more auto-oriented growth (Ewing and Cervero 2001; Ewing and Cervero 2010; Cervero 2007b). Several studies provide hints of how TOD might influence peak-period travel. The first study, by Zhang (2010), simu- lated the peak-hour benefits of TOD at a regional scale while the second, by Arrington and Cervero (2008), empirically compared peak-period trip generation rates of TOD versus conventional rates for non-TODs for specific projects. Zhang Macro Scale Study Zhang (2010) applied conventional four-step travel demand models to simulate traffic outcomes across three scenarios with varying levels of TOD for Austin, Texas: do-nothing; a rail-based TOD scenario with a limited number of TODs; and an aggressive express-bus TOD scenario with numerous TODs spread across the region. It should be noted that such an analysis is fairly coarse and may exaggerate or dampen relatively small changes in effects. As a result, results should be interpreted with caution. Densities for the rail-based TOD scenario ranged from 20 to 75 dwelling units per acre. For an express-bus scenario, densities were assumed to be 1.5 times higher than 2030 density levels under the do- nothing alternative. In the four-step modeling process, modal split estimates were adjusted to account for the rider- ship premium of TOD. In addition to TOD scenarios reducing estimates of VMT and personal miles traveled (PMT), 2030 projections showed that TOD could also significantly reduce peak-period conges- tion. Under the base case 2030 scenario, 3,729 lane miles (20.3%) of roadways in the study area are predicted to be congested in the morning peak. The rail-based TOD plan was projected to reduce congested roadways by 433 lane miles versus the base case, representing 18% of the region’s lane miles. The most aggressive (All-Systems-Go) TOD scenario was expected to reduce congestion by an additional 341 lane miles, or to 16.1% of the regional total. According to Zhang’s analysis, the mid-level rail-based TOD can be expected to reduce traffic congestion by 11.7% relative to the base case. The All-Systems-Go TOD option would likely reduce it by an additional 9%, or a total of 20.7%, relative to the base case. A more aggressive postprocessing of the model results, reflecting for example evidence on the influences of density on ridership from direct-ridership models (Cervero 2006), might have yielded more sizable drops in peak-period congestion levels. Zhang concluded that most of TOD’s role as a congestion relief strategy lies in concentrated development that shortens trip lengths and thus lowers VMT and PMT relative to low-density sprawl. Specifically, “as a land use strategy, TOD reduces congestion by bringing closer trip origins and destinations and hence reducing average trip length, although shifting travel from cars to transit is ultimately desirable” (Zhang 2010, p. 154). Because TODs were estimated to reduce VMT and PMT relatively more than peak-period traffic congestion, Zhang’s study found that most of the congestion-relieving benefits were outside TOD neighborhoods. Within TOD, congestion could worsen due to the concentration of people and jobs. Promoting walking and biking to minimize local driving, he concluded, will be critical for TOD success in Austin. Arrington and Cervero Micro Scale Study The Arrington and Cervero (2008) study of TOD and peak travel occurred at a much finer grain of analysis: individual projects. This TCRP-funded study surveyed travel at 17 multi- family housing units of varying sizes near rail transit stations in four parts of the country: Philadelphia, Pennsylvania/ northeast New Jersey; Portland, Oregon; metropolitan Wash- ington, D.C.; and the East Bay of the San Francisco Bay Area, California. Pneumatic-tube recorders were placed on all curb cuts and driveways to the surveyed projects and recorded daily and peak-period trip generation rates were compared to those for the same residential land use categories in the Institute of Transportation Engineers (ITE) Trip Generation Handbook (ITE 2001). Figure 2.3 shows results for the 17 surveyed TOD-housing projects. These averaged 44% fewer vehicle trips than that estimated by the ITE handbook (3.754 versus 6.715). The weighted-average trip rate differentials were even larger dur- ing peak periods: 49% and 48% for the a.m. peak and p.m. peak, respectively. In general, denser, more urban TOD-housing had the greatest peak-hour trip rate differentials. For example, the p.m. trip rates for Portland’s Collins Circle and Alexandria, Virginia’s, Meridian projects were 84.3% and 91.7% below ITE predictions, respectively. Statistically, a relationship was established that every 10 additional dwelling units per acre for a development located within one-half mile of a rail station would be associated with a lowering of the p.m. peak trip gen- eration rate of TOD projects relative to the ITE rate of 26%

19 (Cervero and Arrington 2008). The importance of density and proximity to the core in reducing p.m. peak-period trip generation rates is further revealed by Figure 2.4. Based on model results, the figure shows that a transit-oriented apart- ment 20 miles from the central business district (CBD) in a neighborhood with 10 units per residential acre can be expected to have a p.m. trip rate that is 55% of (or 45% below) the ITE p.m. rate. If the same apartment in the same density setting were 5 miles from the CBD, the p.m. trip rate would be just 38% of the ITE rate. A follow-up survey focused on parking demands at TODs, including some surveyed by Arrington and Cervero (2008), shed further light on TOD’s transportation impacts (Cervero et al. 2010). In the case of Portland’s transit-oriented housing projects, parking demand was 11% less than that estimated by the ITE Parking Generation Manual, which is based on p.m. peak trip rates for peak parking periods (typically in the early morning). On average, the supply of parking exceeded peak demand by 30% at Portland’s TOD projects. Jobs-Housing Balance Balancing the locations of jobs and housing confers mobility benefits by shortening trips, promoting alternatives to single-occupant car travel, and rationalizing commute sheds (e.g., less criss-cross, and lateral-moving traffic) (Cervero 1989b; Cervero 1996; Ewing 1996). To date, no research has been conducted specifically on the influences of jobs- housing balance on peak-period auto travel; however, most studies have looked at influences on commute trips, many of which occur in peak periods. On the one hand, evidence that balanced regional growth can reduce work-trip VMT has been unearthed in studies of the San Francisco Bay Area (Cervero 1989a); Puget Sound, Washington State (Frank and Pivo 1994); and metropolitan Portland (Kasturi et al. 1998). Studies in Toronto, Ontario, Canada (Miller and Ibrahim 1998), and greater Los Angeles, California (Giuliano and Small 1993), on the other hand, found little or no evidence that balanced growth can drive down commute VMT or durations. Indirect evidence of the influences of balanced growth on travel performance, notably speeds, comes from empirical work by Cervero and Duncan (2006) of the San Francisco Bay Area. This study measured the number of jobs within four highway network miles that were in an employed-residents occupation, adding an important qualitative dimension to typical metrics of accessibility and jobs-housing balance. Occupational matching allowed the accessibility to jobs that Source: Cervero and Arrington (2008). Figure 2.3. Comparison of weighted-average vehicle trip rates: TOD housing and ITE estimates. PM P ea k Tr ip R at es a s a pr op or ti on of IT E Tr ip R at es (% ) Source: Cervero and Arrington (2008). Distance to CBD (miles) Figure 2.4. Influences of residential densities and distance to CBD on transit-oriented-housing p.m. trip rates as a proportion of ITE rates.

20 individuals qualify for to be gauged. The research found that a doubling of occupationally matched jobs within 4 network miles of workers’ residences was associated with a 32.9% reduction in commute VMT and a 33.8% reduction in com- mute VHT. The slightly larger elasticity of work-trip VHT as a function of job accessibility suggests that, on average, improved job access translates into slightly faster commute speeds. Cervero and Duncan (2006) conjectured that this could be due to the rationalization of commute patterns, with subregional balances in jobs and housing marked by less cross-town, lateral, and zigzag patterns of commuting from one quadrant of a region to another. The research also showed that larger commute-trip VMT and VHT reductions occurred as a function of job accessibility than did shop-trip reduc- tions as a function of retail access. While balancing where people live and shop matters in driving down VMT and VHT, balancing where they live and work matters even more. Variable-Based Analysis The Ewing-Cervero 2010 meta-analysis (Ewing and Cervero, 2010) computed elasticities for individual studies and pooled them to produce weighted averages. However, their work focused exclusively on daily auto demand: VHT and VMT. The mixed-use development tool (Ewing et al. 2011), based on 239 mixed-use sites from six U.S. regions, provides daily, a.m. peak-hour, and p.m. peak-hour external vehicle trips at both the meso and micro scales. Hierarchical linear models are used to calculate the probability that trip making will occur externally or internally from a mixed-use site, resulting in peak-hour auto demand estimates. Mobility by Mode and Purpose Two meta-analyses along with other recent studies provide connections between mode choice, particularly transit usage and walking, to built environment factors. The VMT and VHT results from these same studies were described in the prior section. Mixed-use developments with good transit access tend to generate a significant share of walk and transit trips. Walking trips are most strongly correlated to jobs- housing balance, mix of uses, intersection density, and prox- imity of destinations. Transit trips are correlated strongly with transit access of a development, transit supply, job acces- sibility via transit, intersection density, street connectivity, and population centrality. Ewing and Cervero (2010) found that walking and transit trips have strong correlations to various characteristics of the built environment. The meta-analysis shows that mode share and likelihood of walk trips are most strongly associated with the design and diversity dimensions of built environments. Intersection density, jobs-housing balance, and distance to stores have the greatest elasticities. The mode share and likeli- hood of transit trips are strongly associated with transit access. Next in importance are road network variables, such as high intersection density and street connectivity, and then, measures of land use mix. The meta-analysis did find that jobs-housing balance is a stronger predictor of walk mode choice than land use mix measures. Linking where people live and work allows more to commute by foot, and this appears to shape mode choice more than does sprinkling multiple land uses around a neighborhood. The 2009 TRB meta-analysis, Driving and the Built Envi- ronment (National Research Council 2009), linked transit mode share to built environment characteristics. Population centrality and transit supply have a nonnegligible effect on the share of commuting by rail, bus, and nonmotorized modes (i.e., walking and bicycling). After controlling for self- selection, job accessibility via transit remains statistically sig- nificant. TOD studies conclude that the location of a TOD in a region—its accessibility to desired locations—and the quality of connecting transit service are more important in influenc- ing travel patterns than are the characteristics of the TOD itself (e.g., mixed uses, walkability). For work trips, proximity to transit and employment densities at trip ends exert a stronger influence on transit use than do land use mix, population den- sity at trip origins, or quality of the walking environment. Transit Modal Shares and TOD A number of research studies have demonstrated that hous- ing in close proximity to rail transit stations averages high transit modal splits for commute trips and that improved walking connections to rail stops increases this modal share even more (Cervero 1994; JHK and Associates 1987, 1989; Stringham 1982). Similar relationships hold for employees who work near rail stops (JHK and Associates 1987; Cervero 1994; Lund et al. 2004) and shoppers heading to retail outlets near rail (Bragado 1999; Cervero 1993; Lund et al. 2006). In the case of transit-oriented housing, some analysts (Cao et al. 2009; Chatman 2009) show that ridership premiums are partly due to self-selection (i.e., a lifestyle proposition to live in a neighborhood with good transit services); however, even for pro-transit types, living in a well-designed TOD can induce even more transit travel (Cervero 2007b). Transit modal splits are also thought to increase when TODs take the form of a transit-oriented corridor, akin to a string of pearls. Perhaps the best U.S. example of this is the Rosslyn– Ballston corridor in Arlington, Virginia. Surveys show that 39% of residents living within a quarter mile of a rail stop along the corridor take Metrorail to work compared to just 17% of residents who reside farther away but also within

21 Arlington County (Cervero et al. 2004; National Research Council 2009). Walk/Bike and Traditional Neighborhood Development Many early studies of built environments and travel focused on modal split impacts using cross-neighborhood comparisons. Typically, neighborhoods would be matched on the basis of household income and other sociodemographic controls, but would fundamentally differ in terms of built environments [e.g., auto-oriented versus pedestrian- or transit-oriented (Ewing et al. 1994; Cervero and Gorham 1995)]. While such cases provide order-of-magnitude insights and receive high marks for understandability, the fact that such cases generally rely on statistical means when representing travel character- istics raises suspicions about possible aggregation biases. This led to the use of predictive models that included dummy and interactive variables to distinguish relationships between places with contrasting built forms (e.g., Cervero and Radisch 1996; Holtzclaw et al. 2002; Lund et al. 2006). Several case-based matched-pair studies that specified regression models to study relationships reveal that tradi- tional neighborhood development (TND) significantly pro- motes walking and cycling over automobile trips, particularly for retail shopping and neighborhood-scale activities. A com- parison of two East Bay neighborhoods with similar house- hold incomes, regional access, and transportation services showed that residents of the TND setting averaged 1.07 walk trips per day for nonwork purposes compared to 0.33 daily walk trips for those living in a conventional auto-oriented sub- urb ( Cervero and Radisch 1996). For nonwork trips less than a mile in distance, 28% of residents in the TND walked compared to just 6% in the conventional suburb. Matched-pair compari- sons of TND versus conventional neighborhoods in Los Angeles County (Cervero and Gorham 1995), the San Francisco Bay Area (Handy 1992; Cervero and Gorham 1995), Palm Beach County, Florida (Ewing et al. 1994), and Austin, Texas (Handy 1996) reached similar conclusions: compact, mixed-use, tra- ditionally designed neighborhoods encourage internal walk- ing trips that substitute for out-of-neighborhood shop trips. A six-regional analysis of mixed-use developments found that jobs-housing balance most strongly predicted whether trips made by residents to nonwork destinations (i.e., home- based other trips) were internal to the project (Ewing et al. 2011). Balanced job and housing growth was also strongly associated with walking and shorter car trips for external trips made by residents. The research concluded that “for traffic impact, greenhouse gas, and energy analyses, the VMT gener- ated by a mixed-use site depends . . . on the site’s placement within the region, specifically, on the share of jobs located within a 20- or 30-minute drive of the site” (Ewing et al. 2011). Activity and Health In a comparison of new urbanist and conventional suburban communities in central North Carolina with similar income and sociodemographic characteristics, Rodriquez et al. (2006) found little difference in the amount of leisure time involving physical activity among residents of both communities. Over- all, however, new urbanist residents logged 40 to 55 minutes more walking and cycling each week than did their counter- parts in the conventional suburban neighborhoods. Utilitar- ian travel, such as to work or shopping, accounted for the difference. This finding concurs with that of Saelens et al. (2003) that neighborhood design is not related to leisure- time physical activity when one controls for individual- and household-level characteristics. Also, the North Carolina study found that increased numbers of walking trips came at the expense of automobile trips, consistent with prior evi- dence (Cervero and Radisch 1996). Emissions A case-based study of office workers who relocated from rail- served downtown San Francisco to a low-density, single-use, campus-style office park in the East Bay served by freeway estimated that commute VMT increased by a factor of three following this relocation (Cervero and Landis 1992). The largest contributor to the VMT gain was modal shifts from transit to solo commuting. The study concluded that since tailpipe emissions are directly related to VMT, air quality impacts attributable to this workforce’s commuting increased by a similar order of magnitude. Greenhouse Gas Emissions Most studies on built environments and GHG emissions focus on VMT per household as an intermediate explainer. For the cases of metropolitan Los Angeles, Chicago, and San Francisco, Holtzclaw et al. (2002) found that higher residen- tial densities were significantly associated with fewer VMT per household in all three cities, with the relationship follow- ing an exponential decay function, thus implying that the largest VMT reductions accrue when going from very low to moderate densities. Some observers claim that lifestyle prefer- ences explain much of the lower levels of VMT in denser, more walking-friendly neighborhoods, and that failure to account for self-selection could bias results. In a study of neighborhoods in the Puget Sound area, Krizek (2003) removed possible self- selection biases by longitudinally examining changes in travel when households relocated. He found that moving to a neighborhood with denser, mixed-use, well-connected street patterns was associated with lower VMT and PMT reductions (Figure 2.5).

22 Induced Traffic and Induced Growth Few contemporary issues in the urban transportation field have provoked such strong reactions and polarized interest groups as have claims of induced travel demand. Experience shows that supply-side solutions to traffic congestion provide mobility benefits that are mostly short-lived. Within a few years, newly expanded road capacity is sometimes fully absorbed, with traffic conditions largely the same as prior to the investment. The contention that “you can’t build yourself out of traffic congestion” has become a rallying cry of many environmental advocacy groups aiming to halt new road con- struction altogether. Figure 2.6 diagrams the flow of events attributed to the demand-inducing impacts of an expanded road. In the near term, increased capacity unleashes behavioral adjustments— for example, trips previously suppressed are now made because of improved flows (i.e., latent demand); motorists switch routes, modes, or time-of-travel to take advantage of a new facility; motorists travel to destinations that are further away because of speedier flows (Downs 1962, 1992, 2004b; Cervero 2002b; Noland and Lem 2002). New trips, longer trips, and modal shifts contribute to increased VMT, the strongest cor- relate to overall resource consumption and tailpipe emissions in the transport sector. Other adjustments, such as route and temporal shifts, do not noticeably increase VMT and thus are largely redistributive in nature. Time-of-day shifts from the off-peak to the rush hour underscore the limited congestion- relieving impacts of road expansion. A meta-analysis found a mean short-term elasticity (between lane-km capacity and VKT) of several dozen roadway invest- ments in the United States of 0.40 [i.e., all else equal, a doubling of road capacity was associated with a 40% increase in VKT within 1 to 3 years of the investment (Cervero 2002b)]. Over the long term, added road capacity led to more deeply rooted structural shifts, such as increased car-ownership rates and more auto-oriented land-development patterns, or what is sometimes referred to as induced growth. Adding structural impacts to accumulated short-term ones markedly increases long-term elasticities—on average, to 0.73 in the United States (Cervero 2002b). Other studies have estimated even higher long-term elasticities (Heanue 1997; Fulton et al. 2000; Metz 2008). Overall, experiences reveal that travel adjusts to form a new supply-demand equilibrium of traffic congestion fol- lowing road improvements. This traffic-inducing and thus benefit-offsetting impact is incompletely accounted for by most economic appraisals of transport-facility investments (Downs 1992; Salomon and Mokhtarian 1997; Pells 1989; Cervero 2002b; Cervero and Hansen 2002; Ory et al. 2004). The economic benefit for additional users is typically accounted for in these appraisals. Figure 2.5. Daily VMT by neighborhood type and preference. Source: Krizek (2003).

23 Figure 2.6 shows near-term (i.e., first-order) and long term (i.e., second-order) impacts of expanded capacity. Initially, a road investment increases travel speeds and reduces travel times (and sometimes yields other benefits such as less stressful driving conditions, on-time arrival); increased utility, or a low- ering of “generalized cost,” in turn stimulates travel, made up of multiple components, including new motorized trips (e.g., latent demand, previously suppressed), redistributions (modal, route, and time-of-day shifts), and over the longer term, more deeply rooted structural shifts such as land use adjustments and increased vehicle ownership rates (which in turn increase trip lengths and VKT). Some of the added trips are new, or induced, and some are diverted. Relevant to discussions on the potential traffic impacts of smart growth is the flip side of the induced-demand choice, what is sometimes called reduced demand or suppressed demand. Studies have gauged the effects of transportation programs that often accompany smart growth initiatives, like the cre- ation of pedestrian-only districts, rededication of traffic lanes to buses only, and other measures that reduce, instead of expand, road capacity. In a study of more than 100 cases of road-capacity reductions in Europe, North America, Japan, and Australia, Goodwin et al. (1998) found an average overall reduction of 25%, even after controlling for possible increased travel on parallel routes. This “evaporated” traffic was assumed to represent a combination of people forsaking low value-added (discretionary) trips and opting for alternative modes, including transit, walking and cycling. In the United States, perhaps the most dramatic example of promoting the objectives of smart growth and livability over automobility has been the tearing down of elevated freeways replaced by surface boulevards and transit improvements. The experiences with a freeway-to-boulevard conversion in San Francisco hints at the traffic inducement impacts of this early form of what might be called “complete streets” (Cervero et al. 2009). The closure of the middle section of San Francisco’s Central Freeway in 1996 prompted officials to predict a traffic nightmare, with “bumper-to-bumper traffic for 45 miles east across the Bay Bridge and south into the San Francisco peninsula” (Cervero et al. 2009, p. 47). A survey mailed to 8,000 drivers whose license plates had been recorded on the freeway prior to the closure revealed that 66% of respondents had shifted to another freeway, 11% had used city streets for their entire trips, 2.2% had switched to public transit, and 2.8% said they no longer made the trip previously made on the free- way (Figure 2.7) (Systan, Inc. 1997). The survey also found that 19.8% of survey respondents stated they had made fewer trips since the freeway closure. Most were discretionary trips, such as for recreation. Some 6 months after the September 2005 opening of Octa- via Boulevard, the former level of 93,100 vehicles recorded on the Central Freeway in 1995 had dropped by 52%, or to Figure 2.6. Tracking induced travel demand.

24 44,900 vehicles. While this suggests substantial reduced demand, there likely was some rebound effect that had eroded the traffic-reducing impacts over time, and certainly traffic conditions did not radically change along the corridor. Today, Octavia Boulevard and the network of streets that link to it operate at capacity during peak hours (Cervero et al. 2009). As a result, some motorists have opted to continue using street detours that were planned more than a decade ago for the first Central Freeway demolition (San Francisco Depart- ment of Parking and Traffic 2006). While VMT or traffic con- ditions might not have been altered over the long run, this does not mean the project did not deliver net social benefits: more walking and cycling trips are now being made along the corridor, which is a positive public health outcome, and based on the higher land values and rents in the surrounding neigh- borhood, residents and merchants clearly have placed a higher premium on living near a well-landscaped boulevard than near an elevated freeway (Cervero et al. 2009). Little is known about the induced traffic and induced growth impacts of smart growth initiatives, as reflected by changes in attributes of the built environment, such as higher residential densities, increased mixed land uses, or improve- ments in the pedestrian environment. On the basis of a litera- ture review, it does not appear that any empirical studies of this specific question have been conducted to date. Concep- tually, however, the same dynamics should be unleashed by land use initiatives such as TOD or new urbanism designs that reduce or suppress travel demand. The near-term impact of most smart growth measures will be less car traffic matched by more transit usage, walking, and cycling, perhaps over shorter distances. This normally translates into less VMT, both in peak periods and the off-peak. The question becomes, however, will the vacated slots on nearby roads and smoother flowing traffic induce intermediate and long-term responses? That is, will the short-term mobility benefits soon erode as people take advan- tage of better traffic conditions and react to the lowering of transportation costs? Over the long term, might some of the attractive elements of smart growth that draw households and firms to locate in these communities diminish as traffic read- justs and perhaps congestion levels creep upward? Similar questions could be posed about the intermediate to longer- term impacts of TDM strategies, such as improved parking management and dynamic ridesharing, as well. Most attention about the possible induced demand, or rebound effects, of smart growth have centered on one compo- nent: mixed land uses. In the case of neighborhoods with a mix of housing, retail shops, and other commercial outlets, home- based trips that would otherwise be made to destinations out- side of a neighborhood by car might now be made within the neighborhood by walking or cycling. This is what transporta- tion engineers refer to as “internal capture.” However, shorter trips and driving less reduce the cost of travel, which over the long term could prompt residents to make more trips. That is, the travel-reducing benefits of mixed-use development could erode over time and perhaps totally evaporate. Crane (1996) first raised the possibility that smart growth strategies might Source: Systan, Inc. (1997). Figure 2.7. Source of traffic shifts following removal of San Francisco’s Central Freeway.

25 have unintended consequences of inducing travel. Crane examined the potential impacts of three elements of neotradi- tional neighborhoods (grid street networks, traffic calming, and mixed land uses) on three measures of travel demand (number of car trips, VMT, and modal splits). Only traffic calming was found to contribute to an overall reduction in automobile travel. The other elements, Crane conjectured, could actually increase motorized trips and VMT. Crane and Crepeau (1998) later empirically tested this idea of induced travel spawned by smart growth, finding that grid street net- works in San Diego, California, had no significant effect on the amount of automobile or pedestrian travel. The 1998 Crane study was based on a San Diego Association of Governments data set from 1986 and was not entirely conclusive regarding the built environment–travel demand relationship. Induced travel can also take the form of more non-auto travel, which does not necessarily increase VMT but nonethe- less represents a second-order rebound effect. In a survey of residents in six neighborhoods of Austin, Texas, Handy (1996) uncovered evidence of induced travel among residents making shopping trips. From a survey of residents who had walked to a local store, about one in eight stated they would have stayed home instead of driving if there had been no nearby store within walking distance. This implied that the opportunity to walk to a store likely induced some extra pedestrian trips. Since these were not motorized trips, the presence of induced trip making does likely mean no change in VMT or an erosion of the traffic-reducing impacts of smart growth strategies. If any- thing, such inducements are positive outcomes: more physical activity and perhaps social interaction. A recent analysis of mixed-use development in Plano, Texas, provides further insight into the possible induced travel impacts of smart growth strategies over time (Sperry et al. 2010). Intercept surveys were used to ask those entering a destination of a mixed-use employment center on the edge of Plano: Would you be making this trip if you had to travel outside of <study site name>? A “no” answer implied the trip was induced because the marginal cost to travel off-site was perceived to be higher than the respondent valued the trip. Around one-quarter of internal trips, the researchers esti- mated, were induced, meaning that one out of four internal trips were additional trips and not replacements for trips that would have been off-site, on the external street system. Many of these internal trips were by foot; however, a number were also by private car. Among internal car trips, 17.2% were esti- mated to be induced. While these trips contributed to the mixed-use project’s VMT, because they were internal to the site, they did not appear to contribute to increased traffic congestion on the external road network. The analysis con- cluded: “It is evident that some of the internal trips at mixed- use developments are not ‘captured’ from the external street network, but represent additional trips, induced by the characteristics of the mixed-use environment that reduces overall travel costs” (Sperry et al. 2010, p. 22). Perhaps the element of induced travel with the strongest implications for peak travel and thus infrastructure capacity is time-of-day shifts. To the degree that congestion prompts some travelers to switch to the shoulders-of-the-peak, any measures— be they road expansions or smart growth initiatives—that improve rush-hour conditions will have the opposite effects, encouraging some to switch from the off-peak to the peak. Pells’ (1989) literature review of induced travel suggested most redis- tribution via time-of-day shifts. These shifts, however, can be considered discretionary reactions to lower travel impedance that produce greater mobility, accessibility, and possibly other social and economic benefits without creating a need to expand roadway network capacity. Recent research indicates that the nature of growth pattern changes is materially dependent on the context of the highway investment (Funderburg et al. 2010). Funderburg et al.’s research in three diverse California counties pointed to strong linkages between growth patterns and the type of highway improvement (new extensions and expanded capacity, for example) and locational characteristics (rapidly growing urban area or a more rural context). A highway expansion may pro- vide new benefits through enhanced access in one location, while a similar expansion could impose costs on a small town bypassed by new investment. Travel inducement is not necessarily all bad. While the inducement of car trips can erode the benefits of both supply- side expansions and smart growth initiatives, there are pre- sumably benefits to travelers from the ability to make extra trips that were previously suppressed. Quite likely, however, these are low value-added trips (e.g., less essential, discretion- ary ones) since they were not worth making when the per- ceived marginal costs of making them were too high. The questions of whether new roads or smart growth are, on bal- ance, beneficial to society cannot be informed by studies of induced demand; such important questions require a full accounting of social benefits and costs. Relationship Between Smart Growth and Congestion The top 100 metropolitan areas in the United States cover just 12% of the nation’s land area, but hold 65% of its population and are responsible for 76% of its gross domestic product (GDP) (Sarzynski et al. 2008). The success of urban regions is critical for the success of the nation, but the land use patterns and transportation system characteristics in most of these areas greatly impede their travel efficiency, economic produc- tivity, and quality of life. With much of the functional por- tions of these areas built after World War II, following the popular theme of outward expansion, lower densities, and

26 separation of land uses, travel in these areas substantially relies on highways and motor vehicles making trips over rela- tively long travel distances, equating to high rates of VMT per household and per individual traveler. Between 1976 and 2001 (dates of the FHWA’s National Household Travel Survey), population grew at a rate of 0.45% per year, while the VMT generated by households grew at a rate of 2.02% per year: a ratio of 4.5 to 1. It has been virtually impossible to match this disparity in growth of demand with new highway investment, resulting in ever-growing congestion and delay. These patterns have also greatly affected rates of freight and commercial vehicle traffic, as addressed in the next chapter. There is a growing consensus that how community and activity centers are designed and built has a considerable impact on how efficiently they can support both personal and economic travel needs. Transit most likely needs more compact development forms and higher densities in order to perform efficiently. Walking and bicycling often become viable travel options when urban design comingles activities and brings them closer together. Transit is more likely to be used if it can be reached by walking (or bicycle) at both ends of the trip. The earlier sections in this chapter provide but a small portion of the evidence from both empirical and statistical modeling research that areas with reasonable densities, a balanced mix of uses, effective design that ties the uses together in a way that allows them to be accessed by pedestrians, cyclists and transit users, and high regional accessibility via transit result in fewer vehicles owned by households, fewer trips made by private vehicle, overall shorter trip lengths, and rates of VMT produc- tion that are only one-half to one-third of those seen in con- ventional suburban/Euclidean-zoned settings. Litman (2011) refers to a Surface Transportation Policy Project look at the Travel Time Index (McCann 2001) to explain how sprawling areas tend to have better levels of service on each mile of roadway or at various intersections, but higher per capita delays. He also cites 2002 Urban Mobility Report rankings for Portland, Oregon, versus Atlanta, Georgia, in terms of Travel Time Index values and congestion delays (where Portland ranks high/poorly) versus overall hours of delay per capita (where Portland ranks much lower/better than Atlanta). Litman presents Cox’s (2003a) simple (bivariate) plot of overall/regional densities versus commute times, which shows how job-access/work-travel times tend to rise in larger, denser regions (though other travel times may well fall, along with emissions and heart disease, for example). Cox (2003b) also estimates VMT per square mile versus population densi- ties, showing an expected upward trend—but one that is highly concave (once both axes are linearized), suggesting significant travel economies in the presence of added density. Reduced VMT and greater shares of nonmotorized travel are expected to reduce petroleum dependence and GHG emissions, but congestion can dramatically reduce vehicle fuel economies. Figure 2.8 shows that fuel economy of vehicles more recent than the 1997 models is typically maximized around steady-state speeds of about 30 mph on local streets or highway speeds of 50 to 60 mph (Rakha and Ding 2003). Reduced fuel economy is associated with higher emissions of GHGs, NOx, VOC, PM, toxics, and other pollutants, as well as delays to personal travel and goods shipments. Lower speeds also reduce the attractive- ness of vehicle travel, thus reducing emissions directly via fore- gone trips. A critical consideration in determining the effects of highway capacity expansion on congestion-related impacts is Source: Rakha and Ding (2003). Figure 2.8. Fuel economy–constant speed relations.

27 the degree by which reduced travel speed increases emissions and energy use relative to the degree to which it reduces travel volumes. Goodwin (1996) estimated an elasticity of travel demand with respect to travel time of -0.27 in the short run and -0.57 in the long run on urban facilities. If one considers slow- ing traffic from 60 to 30 mph, this will result in a doubling of travel time (adding 1 minute per mile traveled), and one can expect VMT to fall by 27% to 57%. If this slowed speed results in 3 fewer miles to the gallon, Figure 2.8 suggests roughly an 8% increase in fuel consumption and CO2 emissions, which would be more than fully offset by a 27% short-run reduction in VMT. However, this would assume that the 30 mph speed would be a relatively uniform, or steady-state, condition rather than stop- and-go travel, a scenario that might only be achieved through advanced in-vehicle and out-of-vehicle ITS technology. Another way to look at the trade-off would be to note that fuel economy would need to decline by about 27% (from 35 mpg at steady- state 65 mph to an average of 25 mpg at a slower more con- gested speed) to offset the short-run VMT reduction that would result from travelers’ avoidance of congestion. To offset the long-run effects, fuel economy would need to decline by 57% (to 15 mpg). Thus slowing traffic down may reduce energy consumption and carbon emissions overall for personal travel. While a considerable body of research has successfully iso- lated and begun to qualify the effects of smart growth land use design on trip making, there has been a noted lack of research on the subsequent link between smart growth development and traffic congestion. The principal findings on the first- order effects strongly suggest that when communities incor- porate higher levels of the Ds in their design, households that reside in those communities own fewer cars, make fewer trips by vehicle, and generate lower rates of VMT than do house- holds of comparable demographic composition living in more conventional single-use settings. Similar results occur in employment and commercial activity centers. When these destination areas combine uses in more compact walkable settings, commuters, shoppers, and visitors are found to be much more likely to travel to these locations by modes other than driving, and once there, to conduct a higher percentage of their work-related or non- home-based trips locally by walking or by transit. Other than Cervero’s early work on suburban activity centers (1991), these relationships have not been nearly as well studied as the effects of built environment on the residential end of the trip—largely because that is where the travel behavior data (obtained from household travel surveys) are richest and most plentiful. Renaissance Planning Group and Fehr & Peers are currently performing research under a Lincoln Land Institute grant that is examining these destination-end rela- tionships in greater detail in the Los Angeles region. Where the connection between the built environment and travel has been least studied, however, is in the link between travel behavior in response to these land use designs and the traffic that is actually occurring on the street and highway system. Skeptics of smart growth approaches suggest that, even if higher-intensity land use designs reduce auto dependency for their residents, the fact that the designs still amount to putting more activity in a given land area space likely implies that traffic levels will increase in these places or along the facilities that serve them. The following section presents summary findings from two research studies performed by members of the study— from Phoenix, Arizona (the Arizona DOT), and suburban Washington, D.C. (Prince George’s County, Maryland)—that are relatively unique in addressing this link between smart growth land use and traffic congestion. Arizona DOT Land Use and Congestion Study In 2007, the Arizona DOT’s Transportation Research Center (ATRC) commissioned a study of the impact of higher den- sity development on traffic congestion (Kuzmyak et al. 2012). The study was in response to growing questions as to why the state was not more actively considering smart growth land use practices to manage sprawl and to reduce congestion and demand for new highway capacity. The Ari- zona DOT sought to improve its understanding of how land use affects travel behavior and how it affects traffic condi- tions on adjacent roads. A two-part approach was devised to address these issues, both focused on the Phoenix metropolitan area. The first part used travel survey data from the Maricopa Association of Governments’ (MAG) 2001 regional household travel survey combined with detailed GIS and transportation system data to create models of travel behavior in relation to land use. The second part used case study analysis to examine the relation- ship between development patterns and on-road traffic con- ditions in four different locations where traffic congestion was perceived to be the result of local development patterns. To address the question of whether Phoenix residents did, in fact, exhibit differences in travel in relation to development conditions, a set of regression models were estimated to explain household vehicle ownership, total daily household VMT, and daily household work and nonwork VMT. The models accounted for household size, composition, and income; regional transit accessibility to all jobs and retail jobs only; and local land use as measured through the variables of household density, land use mix (entropy) and walk oppor- tunities. The models showed vehicle ownership to be nega- tively correlated with the 4 Ds variables of household density, land use mix, and walk opportunities (but not transit acces- sibility); total daily VMT negatively correlated with auto ownership, transit accessibility to all jobs and retail jobs, and land use mix; home-based work-trip VMT negatively

28 correlated with vehicle ownership, transit accessibility for all jobs and land use mix; and nonwork VMT negatively corre- lated with vehicle ownership, transit accessibility to retail jobs, and household density. The region was then separated into 17 different areas (jurisdictions) of different character, and the comparison demonstrated some fairly substantial differences in the rates of vehicle ownership and VMT associated with differences in density, mix, design, and transit accessibility. Older, more urban and walkable areas such as East and West Phoenix and South Scottsdale had rates of daily per capita VMT that were more than 30% less from newer but less compact communi- ties like Mesa and Gilbert, and more than 70% less than the newest and most outlying places such as Glendale, Peoria, and Chandler. The differences in VMT rates were comparable for both work and nonwork travel, in contrast to similar stud- ies in Baltimore, Maryland, that showed much bigger differ- entials among nonwork VMT rates. Again, this second part of the analysis assumed a case study format. Four areas were identified in the Phoenix region that featured different land use patterns, with each cited by local stakeholders as probably having traffic issues related to local development. Three of the sites were located in the most densely devel- oped portions of the region: Scottsdale Road near Old Town Scottsdale, North Central Avenue just north of the CBD, and the Mill Avenue/Apache Boulevard corridor through the most built-out portions of Tempe. A fourth corridor, West Bell Road, served as something of a control site, being located in a medium-density (but intensely developed) typical sub- urban setting on the region’s northwest edge. Each site sur- rounded one or more major arterial highways and each was no closer than two miles from the nearest expressway. A key finding was that despite the considerably higher den- sities in the three urban examples, measured traffic condi- tions on key roadways were found to be considerably better than those in the much lower density Bell Road corridor. Lacking information on intersection level of service (queuing and delay), the researchers focused on traffic level of service on key links in each study area, measuring volume-to- capacity (V/C) ratios in both the mid-day and p.m. peak time periods. These results, summarized in Table 2.5, revealed sur- prisingly reasonable traffic flow on most of the critical links in the Scottsdale and Central Avenue corridors, with both mid-day and p.m. peak V/C readings below 1.0. Tempe does not show as well, with measurably higher V/C readings, par- ticularly on Mill Avenue, which is the area’s commercial strip. However, traffic conditions on Bell Road were easily the worst of the group, with V/C ratios in the 1.3 to 1.6 range, reflecting heavy traffic congestion. An important consideration in examining local traffic levels is accounting for the proportion of traffic that is simply passing through, having neither origin nor destination in the study area. This is always a key factor in evaluating the efficiency of a land use design, since travel which is totally unrelated to the development activity is part of the total volume contributing to demand on the facilities, and counting in any traffic test—in effect, being used as part of the test to determine the perfor- mance of local land use. The previous chapter dealt with the related issue of induced demand, whereby efficiency improve- ments attributable to good design (more trips made internally or by transit) free up capacity on adjacent roadways, which then attracts trips that previously would not have been made or would have been made on other facilities. Select link procedures were used to estimate the through traffic percentage on each of the sample roadway links in the Table 2.5. Volume-to-Capacity Ratios on Select Links (Adjusted to Counts) Study Area Location Mid-Day PM Peak North/East South/West North/East South/West Scottsdale Scottsdale Road, North of Indian School 0.59 0.57 0.66 0.61 Indian School, West of Scottsdale Road 1.05 0.85 1.11 0.99 Bell Road Bell Road, between El Mirage and 115th 1.88 1.63 1.68 1.91 Central Avenue Central Avenue, North of Osborne 0.41 0.64 0.59 0.49 Thomas Road, West of Central Avenue 1.27 0.83 1.11 1.26 Tempe Mill Avenue, North of University Drive 1.38 1.25 1.33 1.70 Rural Road, North of University Drive 0.60 0.46 0.71 0.38 Apache Boulevard, West of McClintock 0.56 0.58 0.99 0.56 Broadway Boulevard, West of McClintock 0.71 0.74 0.96 0.87 Source: Kuzmyak et al. (2012).

29 Phoenix examples. Each of the areas’ facilities was determined to be carrying appreciable levels of through traffic, with Scottsdale being least affected (23% to 28% range, peak and off- peak), but with half or more of all peak-period traffic in the other three areas being through traffic. What this showed was that while Bell Road could attribute half of its peak-period traf- fic to through trips, both Central Avenue and Tempe were supporting similar ratios, but with much better net V/C mea- sures. Indeed, if the through travel proportion on Bell Road were reduced to the 22% to 28% moderate ratio in Scottsdale, it would still have a V/C well over 1.0. The net take away from this exercise was to find that while the three urban higher density, mixed-use sites had residential densities twice that of the suburban example, and employment densities greater by multipliers of 7 to 25, traffic conditions were in fact much better—and certainly not worse, as might have been pre- dicted based on the differences in densities. Several important differences helped account for this apparent paradox. The first difference is the presence of an articulated street grid in the three urban sites. While most of the region is served by a 1-mile super grid, Central Avenue and Scottsdale Road are embellished with a secondary street grid that features smaller capacity streets on quarter- or eighth- mile spacing. This not only makes walking and access to tran- sit more convenient, but provides more effective capacity to handle traffic, plus the ability to specialize links, signals and turns to optimize flow for particular travel segments (e.g., local versus through) or by time of day. Bell Road clearly does not possess such a network, and while there are many roads, few are designed to connect arterials, but mainly to serve internal circulation within subdivisions. In addition, the siting of commercial activity in strip centers and malls along the main arterials means that virtually all access to and between residential areas and these centers must be by driving. The other difference has to do with how the smart growth design in the three urban areas is correlated with more effi- ciency in terms of travel demand. Resident households in the Scottsdale and Central Avenue corridors own fewer vehicles (1.4 to 1.47) than those in the Bell Road corridor (1.7), while auto ownership levels in Tempe (which is generally less urban than Scottsdale and Central Avenue) are higher (1.63) and more like those of Bell Road. Daily household VMT rates are much lower in Scottsdale (19.5) and Central Avenue (17), and even appreciably lower in Tempe (24.2) than Bell Road (31.8). Reasons for this may be seen in higher rates of internal capture for work trips (18% to 21% versus 13%); nonwork trips have about the same high rate of capture (40% to 42%) in each corridor, but the Bell Road corridor likely earns this status because of its large size (17 square miles versus 3 to 5 square miles for the urban sites). Average trip lengths are much longer for all trip purposes in the Bell Road corridor than at any of the three urban sites (about half as long for work trips, between 12% and 25% as long as for nonwork trips). The three urban sites also capture decent shares of trips either from or to the area by transit (3% to 10%), com- pared to less than 1% in the Bell Road corridor (where all transit is park and ride). Prince George’s County Smart Growth Development Study The Prince George’s County, Maryland, planning department commissioned a study in 2009 to investigate alternatives to traffic level of service (LOS)–based adequate public facilities (APF) requirements for evaluating the performance of com- pact mixed-use centers and corridors (Kittelson and Kuzayak 2010). The county’s adopted 2002 General Plan emphasized strategic development around its numerous Washington Metrorail and MARC commuter rail stations, as well as in other designated centers and corridors. Unfortunately, the county’s planners found themselves stymied by local traffic violations of APF standards when they attempted to move forward with these plans, causing them to seek alternative mechanisms to measure performance and adequacy for these activity areas. Because the APF test is performed in proximity to a pro- posed development project, the use of standard trip genera- tion and impact assessment methods place the burden of meeting local traffic standards on adjacent development, regardless of (a) whether the development is inherently effi- cient in its design or (b) whether it is the primary source of traffic in the measured stream. The county planners were in search of an alternative way to determine both “adequacy” and “attribution,” thus seeking a broader and more revealing set of tests and indicators that would be more appropriate and use- ful in encouraging the right types of development in the des- ignated growth locations. Believing that research on the 4 Ds provided strong support to the premise that smart growth (compact, mixed-use, pedestrian friendly and transit-served) development reduces vehicle dependency and use, the goal was to establish protocols for defining the functional bound- aries of these areas, the desired attributes of the development, and measures to more accurately represent the performance of the planned development. A two-part methodology centered on case studies was developed for this assessment. The first part was to measure and assess traffic conditions and the composition of traffic. The second part was to look at the characteristics and design of the given study area to ascertain whether it possessed good smart growth design properties, and the degree to which its design was beneficial to transportation objectives. Six representative areas were selected as case studies, to allow for a thorough investigation of the relationships between land use patterns and traffic conditions. Each of these areas had been designated for intensified development under the 2002

30 General Plan, and they varied with respect to regional location, proximity to Metrorail and key highway facilities, density and mix of development, and overall scale. The areas ranged in size from 2.8 to 4.9 square miles, in household density from 0.3 to 3.8 households per acre, in employment density from 631 to 6,660 employees per acre, in jobs–housing ratios from 0.82 to 3.88, and in retail jobs–housing ratios from 0.09 to 1.51. All areas were on or adjacent to one or more major state or U.S. highways supporting interregional travel. Three of the areas had one or more Metrorail or MARC train stations. Those principal road segments likely to be used in an ade- quacy determination were identified, and data on their utiliza- tion and performance was recorded. Traffic levels in the current (base) year were established by comparing model-generated link volume estimates with actual counts, and concluding that the estimating accuracy was acceptable. Conditions in 2030 were then forecast by using the county’s travel model, with planned development and transportation improvements in place throughout the region. (The county’s model is based in TransCAD, includes the entire metropolitan Washington, D.C., region, and has a highly detailed road network and assignment process.) These analyses showed that most of the identified facility segments in the case study areas would be carrying 2030 traffic volumes that would exceed established LOS thresholds. Hence, the development planned for these centers would prob- ably not be permitted to go forward. A first step in assessing these traffic conditions was to determine the proportion that was attributable to develop- ment in the subject study area versus direct pass-through. This assessment was done by using the select link procedure in the travel model, and showed that the major portion of traffic on the representative links was comprised of through traffic, with no less than 50% in any of the situations, and as much as 100% in the worst case (Brandywine Road). The clear implication was that the planned growth in almost of these areas was not the reason for a likely traffic LOS failure, but rather that these areas are serving as conduits for through travel that substantially determines their performance. The first part of the analysis thus demonstrated that a local traffic congestion test to determine the worth of a smart growth center plan would probably be inappropriate in several ways: first, by making the local area responsible for traffic volumes that were unrelated to local development activity; second, by reducing the development design and likely compromising the transportation efficiency potentials; and third, by focusing solu- tions on actions to increase road capacity instead of improving efficiency (such as through provision of a street grid). The second part of the analysis was to look in depth at the trip generation characteristics of the study areas themselves. If such smart growth designs were to be given special treat- ment for their presumed efficiency on travel, their character- istics should satisfy design standards and protocols that research has found to be associated with reduced vehicle dependency and VMT. The Ds provide such a checklist, offer- ing guidelines on minimum densities, synergistic mixes of different uses, proper layout and design to support pedes- trian, bicycle and transit use, and both good regional transit service and accessibility, as well as efficient access to transit within the study area. Since the tested scenarios incorporated 2030 design assump- tions and population/employment allocations (thereby imply- ing that the county’s design plan for the area had been implemented), it was possible to test each area’s smart growth legitimacy by using the following measures of performance: • The number of trips generated by residents, by trip purpose: home-based work, home-based shopping, home-based other, and non-home-based; • The destinations to which these trips were made, allowing measurement of how effectively they design retained trips internally; • Average trip lengths; • The modal split for trips made for each of the four purposes for trips made from, to, and within the study area (and par- ticularly the number made by transit or nonmotorized modes); and • VMT generation rates for households residing in the study area versus comparable households outside of such areas. What this analysis showed was that the design of the desig- nated growth areas fell far short of smart growth ideals: Over- all densities were much lower than desired; the balance of residential, employment and retail was insufficient to retain a respectable portion of travel with the study area, and high rates of nonhome-based VMT were observed, suggesting auto-based trip chaining to accomplish basic travel needs. In terms of transit viability, aside from home-based work trips being made by Metrorail to well-served destinations in downtown Washington or Arlington, transit use for work trips by visitors to the study area or by residents to any other location were nominal, and negligible for nonwork travel purposes. A contributing factor to the low transit use rates was the location of the actual transit station in a noncentral location relative to the rest of the developed center, making access inconvenient. This analysis was very revealing to the county’s planners, making evident that what many people thought was smart growth was not reflected in the actual designs put forward. Thus, the dual message was taken that (a) smart growth proj- ects can have a major impact on vehicle trip generation and congestion, reduced need for additional road capacity, and therefore deserve special performance criteria to measure their impact and worth; but (b) there are critical elements that define a legitimate smart growth design, that clearly were not

31 evident in the designs that were reflected in the scenario. This implied that county also needed additional tools and proto- cols to support better design of its smart growth centers. Smart Growth and Freight Traffic Truck and rail modes each carried 40% of the nation’s 3.34 tril- lion ton-miles of commodities moved in 2007 (U.S. DOT 2010), with average distances of 206 and 728 miles, respectively. Intermodally, truck and rail carried 5.9 percent of ton-miles captured by the Commodity Flow Survey, with a (combined) average distance of 1,007 miles (U.S. DOT 2010). FHWA (2007b) has forecasted a doubling in U.S. freight tonnage between 2002 and 2035, due to globalization and modern supply-chain management (including just-in-time manufac- ture and delivery of more higher-value goods). Congestion, crashes, pollution, noise and other issues are associated with moving goods in a world of rising population and incomes and population. Finding space for containers and vehicles, pickups and deliveries, within dynamic urban regions is a challenge. While heavy-duty-trucks generally are responsible for less than 5% of most highways’ VMT, urban truck VMT has out- paced overall freight-VMT increases (Bronzini 2008), and trucks are said to occupy 60% of road space on many “chroni- cally congested roadways” in places such as New York City (Move NY & NJ 2007). Truck’s share of U.S. ton-miles has increased over time (EPA 2006), while mode energy efficiency has fallen (Davis and Diegel 2007). Kockelman et al. (2008) suggest that this may be due to more trucks traveling empty (or “dead heading”), since heavy-duty truck (HDT) fuel economy has remained constant or increased over the same time period (FHWA 2007b; Davies et al. 2006; Bertram et al. 2008). But growing roadway congestion is another potential cause (with HDT fuel economy–speed relationships presum- ably similar to Figure 2.8 curves, though with maximum fuel economies around 6 mi/gal). Many argue for a shift of freight to rail transport (CEC 2011), where fuel use and emissions are arguably much lower (e.g., roughly 400 versus 100 ton-miles per gallon of diesel on rail versus truck), capacities are theoretically higher (e.g., roughly 200 versus 40 million ton-miles per track or lane per year, respectively), shipper costs are noticeably lower (e.g., 2.7 versus 5.0 cents per ton-mile by rail versus highway), and safety statistics are better (e.g., rail transport exhibits roughly one-third the number of injuries and fatalities per ton-mile shipped), according to Move NY & NJ’s McGregor (2006). There is hope that double-tracking of more rail corridors will dramatically improve rail’s reliability and travel times, enhancing its modal competitiveness. Rising roadway conges- tion, the introduction of road tolls, and higher gasoline taxes may incentivize shifts to rail and other freight modes. Truck presence on highways varies significantly by loca- tion. In many U.S. corridors, highways carry 30,000 or more HDTs a day, with these HDTs contributing 10% or more of the facilities’ VMT (Bronzini 2008). These U.S. corridors include major highways in the Chicago region; Atlanta’s I-285, I-75, and I-20; and Southern California’s I-710 (serv- ing the Los Angeles–Long Beach port). U.S. Interstate high- ways typically carry less than 10,000 trucks per day, but their truck traffic often contributes 20% or more of their VMT (Bronzini 2008; Wilbur Smith Associates 2003). Port areas are especially important for freight movement, with 2 billion tons of freight entering the nation at marine terminals each year. Associated population exposure to heavy vehicles, their emissions, and potentially devastating queuing are of key concern to planners, shippers, port operators, local residents, and business leaders. As Prasad (2011) put it: “Land-use decisions are critical”—to environmental justice, human health, the economy, and quality of life. Land Development and Infrastructure While mixed used and higher density land-development pat- terns are expected to reduce goods-and-services-delivery- related VMT, coordination and cooperation may be key (e.g., to fill up delivery vehicles and meet customers’ time windows). “Public logistics terminals” or multicompany distribution cen- ters have been studied and, in some instances, adopted as a method for reducing delivery burdens via capacity consoli- dation by third-party operators (see, e.g., Hassall 2005 and Taniguchi et al. 1999). Inland ports or “freight villages” exist in the United States (e.g., the Alliance, Texas, multimodal hub and North Carolina’s Global TransPark), as well as across Europe (Ballis 2006). These expertly designed transshipment points for warehousing by multiple operators facilitate inter- modal transfers and goods storage while enabling consoli- dated operations (e.g., shared pickups and deliveries within the nearby cities), often relieving competition for scarce land (and road space) in densely developed regions (e.g., Athens, Greece, and Paris) (Ballis 2006). Though many firms are more accustomed to competing, rather than coordinating their movements, there are multiple benefits to consolidation of deliveries and pickups (including reduced fuel use and fewer employees needed on site to receive added deliveries). These freight villages can have growth-inducing effects that counteract the positive reduction in truck VMT, when new exurban communities develop nearby and produce trips among new residents and workers that result in higher auto VMT, given the low densities and remote locations. This phe- nomenon may be intuitive but it is not well understood. In a recent publication on freight and land use (FHWA 2012), the positive benefits of freight villages are discussed but induced effects are not mentioned.

32 Klastorin et al. (1995) examined the decisions of six firms with distinctive logistics needs in the Seattle, Washington, region more than 15 years ago (including Safety, Avtech, and Boeing), and found that land rents drove location decisions more than transport access did (though some level of highway access is presumably fundamental to site choice, but relatively well provided within and between most U.S. regions). Four of six firms preferred denser urban form for access to customers and clients, though the move toward larger/longer vehicles (to reduce shipping costs) makes many local street designs tougher to navigate. The conclusion that site access design (e.g., provi- sion of curb loading zones, one-way alley protocols, and sign- age) “can have a big impact on urban goods movement” (Klastorin et al. 1995) was highlighted, and the use of smaller trucks (24 foot) by at least two of the six firms for intra- neighborhood operations was noted, with satellite transfer facilities for shifting goods to and from larger trucks. The proximity of freight and nonfreight activities often results in more trespassing issues and theft, more human exposure during hazardous materials incidents, and other unsafe conditions, along with complaints regarding emis- sions, noise and vibration issues, and light pollution at night- time (Strauss-Weider 2003). Relocation of freight activities requires a high degree of communication and coordination among affected parties, public and private. Urban brownfield sites present an opportunity for such land uses at reasonable cost, with thoughtful location being key for carrier access, goods consolidation, and streamlining movements (ideally across carriers and shippers). Hush-kits on airport equip- ment, alternative fuels and electrified engines, reduced idling regulations, whistle-free (or modified-whistle) zones (for rail transport), grade separation, barrier construction alongside corridors and shipyards (Figure 2.9), corridor preservation (by purchasing underused industrial parcels and rights-of- way) and other strategies are also providing valuable in U.S. applications and abroad (Strauss-Weider 2003). Designing street systems and associated infrastructure to accommodate large trucks and other forms of goods move- ment can be at odds with various smart growth strategies. For example, wider lanes, longer loading areas, and longer turn radii mean more paved surfaces and greater exposure of pedestrians and cyclists. Longer, wider, heavier vehicles can mean more damage to special street surfaces (e.g., brick or textured surfaces), close-in curbs, medians, islands, street fur- niture and roadside vegetation. Smaller vehicles address such issues, but raise labor costs (and, presumably, fuel costs and emissions) per ton-mile transported. Limited rights-of-way and freight-loading zones mean more double-parking, back- ups into and across streets, and blocking of pedestrian and bike baths, thereby worsening congestion and traveler safety. Truck-only lanes (and access ramps), truck-restricted loca- tions (enforced by size and weight, with permits for special shipments at less congested times of day), rail yard and cor- ridor investments (including staging areas for deliveries and rest areas for truck drivers satisfying work-time regulations), and congestion pricing or roadspace rationing (with travel credits for continued access and revenue-neutrality) (see, for example, Kockelman and Kalmanje [2004]) help avoid con- flicts while incentivizing socially preferred modes and routes. Freight Delivery and Pickup Pivo et al.’s (2002) interviews of truck drivers (via Seattle-area focus groups) echo such findings, along with a strong impres- sion that deliveries and pickups are now at all times of day (due to the changing nature of business) and loading zones are not often long enough (with 30 feet a desired length, per intended vehicle, ideally located at the ends of blocks [for added access]) or exclusive enough (with limousines and sales representatives with commercial license plates taking valuable space, or bus lanes precluding parking). Truck driver complaints include the clutter and congestion of alleyways Source: Strauss-Weider (2003, Figure 4). Figure 2.9. Barriers for pedestrian protection: before (top) and after (bottom). Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

33 (e.g., dumpsters, misdirected trucks, mis-parked cars, and homeless persons), and the improper design of loading docks (e.g., at the bottom of steep descents with tight turn radii). Wider alleys, turntables for delivery trucks at space-constrained loading docks, standardization of good practices in dock designs, alcoves for dumpsters, higher emergency stairwell clearances, and shorter/single-unit trucks were all desired for urban stops. All-way pedestrian phases were also cited as desirable, to minimize pedestrian exposure and risk during truck turning movements. At shopping malls and large office buildings, centralized delivery locations, with intra-mall/ intra-building delivery made onsite by specialized mall- managed vehicles or building-provided workers is also desired (to minimize parking times, freeing up limited parking space for others). Drivers reported a dislike of commercial strip devel- opment, since it is not so conducive to safe or efficient delivery practices. As congestion mounts, light-duty vehicles appear more likely to take chances around bigger trucks; business practices place more emphasis on time-sensitive pickups and deliveries while network unreliability increases, leading to a highly stressful situation for urban truck drivers. Recently, Weisbrod and Fitzroy (2008) examined the eco- nomic consequences of urban congestion in terms of freight delivery and business operations. They cited literature describ- ing the reduced customer, labor, delivery and input sheds (or catchment areas) that emerge from urban congestion, along with other potential agglomeration disbenefits, like higher input costs and shifted or narrowed delivery windows. They highlighted Vancouver, British Columbia, Canada; Chicago, Illinois; and Portland, Oregon, as examples, where business leaders were seeking to address concerns about sea, rail, truck, and airport activities being compromised by serious roadway congestion. Their interviews revealed that early morning deliveries have been rising (to avoid congested times of day) and worsening p.m. peak traffic conditions have curtained certain backhaul opportunities, affecting carriers’ bottom lines (and therefore shipper and customer costs). Just-in- time deliveries and increasingly complex supply chains are threatened by growing congestion. It was noted how air and maritime port schedules are relatively constrained (by time of day and frequency of departure to desired destinations, particularly for international shipments), putting more emphasis on truck travel, thanks to reduced uncertainties. In terms of land use relationships, Weisbrod and Fitzroy (2008) noted that warehousing and distribution centers, tra- ditionally drawn to the edges of urban regions, are finding the density of later infill development to limit their operations via congestion, vehicle-turn conflicts (on space-constrained roadways), and higher land values for any desired expansions. While the costs of such congestion is difficult to estimate, res- ervation times at port facilities, congestion-based road and runway tolling, variable pricing of capacity-constrained rail corridors, and various impact fees for existing and new land uses may ensure reliability in movement of freight and pas- sengers, raising some business costs while avoiding a host of others. The use of TREDIS software for multimodal modeling of the benefits and costs of network changes was suggested. Quak and de Koster (2009) highlight the common response of municipalities to the issues of large-truck deliveries in the urban area: delivery time windows (usually to the early morn- ing, to avoid conflicts with pedestrians and added street con- gestion and noise) and vehicle-size limitations. Apparently, restrictions of delivery timing are very common in western Europe, particularly in the larger cities, where many of the most commercially developed locations date back more than 100 years, well before the arrival of (and design for) large trucks. They model the cost and emissions impacts of differ- ent policies (by running optimal logistical patterns for vari- ous case study retailers), emphasizing the following variables at play: number of distribution centers (which proxies for the inverse of average distance to the nearest distribution center), delivery frequency, vehicle capacity, unloading time (duration of stop), and available delivery windows. Delivery windows are most restrictive (and costly) for those businesses with smaller delivery sizes (since multiple drops per journey are preferred and feasible without the schedule restrictions in place). Simi- larly, vehicle-size restrictions are most problematic (and costly) for those with large drop sizes (that can fill more than one size-constrained vehicle). Reductions in delivery fre- quencies (by aggregating shipments and reducing the num- ber of stops per journey) deliver significant cost savings for both types of businesses (but make the most sense for those with smaller drop sizes). Finally, size and timing restrictions were estimated to increase all emissions types studied (NOx, PM, and CO2), suggesting that there is an environmental trade-off in the pursuit of such policies; reductions in deliv- ery frequencies ameliorate this impact (as well as delivery cost implications). Transportation Policies for Freight Mobility Lemp and Kockelman (2009a) simulated a variety of scenarios for an Austin, Texas, comparison of traditional/aggregate and disaggregate/activity-based demand model applications. Their “centralized employment” scenario moved half of the rural- zone jobs and 30% of the suburban-zone jobs into urban and CBD zones (in proportion to these latter zones’ existing job counts). It is interesting to note that predicted levels of region- wide VMT did not rise and, instead, fell slightly under both model specifications (0.46% and 1.47%, for the aggregate and disaggregate model specifications, respectively). The strongest overall reductions in VMT were forecast on lower-level road- ways (2.14% and 4.57% reductions, respectively, on the collector/local class of coded links). Transit and walk/bike

34 mode shares rose very slightly (10% or less of their already very low values), while average speeds during peak times of day fell negligibly. The researchers had expected significant speed reductions (via congestion) to arise from moving so many jobs downtown, with no network changes (to buttress the urban and CBD roadways, for example), and so were pleasantly surprised by the results. Zhou et al.’s (2009) simulations of Austin under an urban growth boundary (UGB), like those of Kakaraparthi and Kockelman (2010) and Tirumalachetty and Kockelman (2010), resulted in significant (roughly 15%) VMT reductions, versus trend (similar to reductions stem- ming from stiff road tolls), and much higher long-term popu- lation and jobs densities (from application of land use models, in tandem with travel demand models). While Tirumalachetty and Kockelman modeled internal commercial trips directly, freight trips remain largely exogenous to modeling efforts (with external trip tables simply held constant or scaled up proportionally over time). And commercial trips remain dif- ficult to characterize and forecast accurately (PSRC 2009). Johnston (2008) reviewed more than 40 simulation exer- cises across a variety of U.S. and EU regions and concluded that many transport pricing, land use policies, and invest- ment strategies offer significant long-run reductions in VMT and emissions (relative to trend) without compromising highway levels of service or regional productivity. Increased pricing of road use, fuels and parking enhanced “the effec- tiveness of the land use and transit (provision) policies,” while highway capacity expansion often resulted in predic- tions of worse congestion. The CEC’s (2011) Destination Sustainability report men- tions the “need for more integrated land use-freight transport planning” several times, but without any details. The report offers more on the notions of enhancing recognition and inspection technologies for freight and trucks, along with better supply-chain management practices to speed up cargo checks and moderate waste in the freight industry— particularly in the context of reducing border delays (which have significant local emissions impacts, and costly time expen- ditures for cargo, vehicle, drivers, and their customers). The CEC report also mentions the benefits of maritime and rail modes over truck transport—primarily in relation to energy consumption and carbon dioxide equivalent (CO2e) emissions, but congestion also serves as a solid reason for such mode shifts in many locations. More full-cost pricing of mode choices, by all travelers, can reduce roadway delays by moderating the exces- sive use of modes and routes that carry greater social costs. More thoughtful routing and delivery timing decisions can also reduce truck VMT and associated emissions. Pitera et al. (2011) recently showed how application of an emissions min- imization algorithm for University of Washington mail ser- vices could reduce GHG emissions by 6% and costs by 9%. If service frequency were reduced to once-a-day, emissions savings estimates rise to 35%. In associated work, Wygonik and Goodchild (2011a, 2011b) examined how added density of customers (and smaller vehicles) reduces the cost and GHG emissions of delivery. Like Quak and de Koster (2009), they found that less restrictive delivery windows and/or a higher density of stops/customers enables more efficient goods movement (in terms of GHG and cost savings per delivery, within a single carrier’s routing plans). In all sce- narios evaluated, cost savings far exceed the value of saved CO2 (since carbon markets value CO2e at less than $100 per ton, now and many years into the future). While smaller vehi- cles often prove more efficient for this type of multistop, less- than-truckload (LTL) delivery system, hybrid engines offered the lowest costs and emissions. Interestingly, it was noted how higher customer densities can offset tighter delivery win- dows, better meeting customer needs (or city ordinances). Another policy for impacting freight movements is road pricing. Holguín-Veras et al. (2006) looked at carrier responses to the Port Authority of New York–New Jersey (PANY/NJ) variable-pricing policy on six bridges and tunnels. Their sur- vey results suggest that “productivity changes” (e.g., load deci- sions and vehicle sizing choices) and transfer of increased costs (to receivers) were much more common than route changes/facility-use changes, in this particular instance. While much depends on the specific context of the pricing’s implementation (e.g., price levels by time of day and avail- ability of routing alternatives)—including the carrier-receiver relationship dynamics and market competition, they con- clude that carrier responses may be much more nuanced than demand modelers expect, due in part to the many decision variables at play for carriers (as well as shippers). More than half of the respondents (54.8%) indicated that customer schedule dictated travel schedules, with congestion avoidance posting second (with 23.1%). Only 3.1% indicated that lower tolls drove their scheduling decision (presumably because the toll differentials were rather small relative to overall vehicle, driver, and fuel costs, as well as customer needs and receiv- ing costs). As expected, for-hire carriers exhibited much less trip-timing flexibility (and sensitivity to toll rates) than private carriers [who enjoy more accommodating (in-firm) receivers]. Overall, these results suggest that road (and zone- based/cordon) pricing may not have much of an impact on freight-vehicle use of congested corridors and locations, unless there are clear alternatives. Freight Trip-Making Like commercial trips, many freight trips are less-than- truckload (LTL). Holguín-Veras et al.’s (2011) recent work explains how freight-trip generation is not proportional to firm size or zone employment in most cases and across most indus- try sectors, thanks to LTL shipping, shipment indivisibility,

35 variable truck sizes, scheduling needs, and other logistical deci- sions. In general, there is an economy of size that comes with freight shipments for larger establishments (though their data also show some peaking of trip generation rates for certain types of mid-sized-firms). Holguín-Veras et al. recommend that demand modelers turn to straightforward Economic Order Quantity equations to get a better sense of such economies in shipping decisions, along with finer-scale resolution of zones and firms, ideally to the parcel level, to replicate and forecast freight movements. One land use implication of such findings is that a mix of business types (a typical smart growth objective) may require significant consolidation and coordination of ship- ments to avoid the more-than-proportional increase in local freight movements (and their associated congestion), relative to large-firm, separated-use styles of land development. Allen and Browne (2010) point out the “deindustrialization” that has taken place in highly developed countries in recent decades (with production jobs shifting overseas), reducing the need for large industrial sites near urban areas, and their asso- ciated warehousing, while increasing the importance and activity of port locations. These trends have been accompa- nied by a “spatial centralization of stockholding,” via large regional or national distribution centers outside urban areas. Such centers or transshipment points tend to be strategically located, often at the crossroads of accessible trade/travel cor- ridors but away from congested urban sites, with their higher land values. They allow for storage and consolidation (and breakup) of shipments, preparation of items for final display and sale, and mode shifts—well away from the spatially intensive activities of the urban core. Allen and Browne (2010) describe the nature of different freight trips, from single-stop to multistop/multileg deliveries and pickups, direct versus consolidated shipments. Such deci- sions depend on the nature and size of shipment, including its time sensitivity, proximity of destinations, and travel costs. They state that land use plays less of a role in freight-related travel than in personal travel since fewer mode options exist for freight shipments (e.g., all trips must be motorized, except for final rounds of small-parcel delivery and pickup), price elastici- ties are presumably lower (though no citations are given for this), and most freight trip ends and route choices lie along arte- rial highways or urban commercial streets (rather than the more variable styles of residential and suburban development). While loading space is relevant to freight movements, parking provision (and cost) is not. Similarly, transit and sidewalk pro- vision presumably have relatively little impact on freight move- ment. In looking at 2005–2007 UK commodity-flow data, Allen and Browne (2010) estimate that the share of intra-urban goods movement rises from about 20% to 40% (of tons and ton-km moved) as region size grows (e.g., from 464,000-population Edinburgh to 7.51 million persons across the Greater London region). The average (intra-urban) haul length appears to be 20 miles (about 32 km) in the UK data, with the average carry- ing capacity of intra-urban vehicles being half that of vehicles carrying shipments to and from such regions (i.e., 10 tonnes versus 20 tonnes). Lading factors (use of vehicle weight capacity) are also much lower for intra-urban movements (generally between 30% and 40% of vehicle weight capacity) than other movements (which range from 0.51 to 0.67 in the 16-region UK data set). Freight trips departing an urban region tend to run less full than those entering (due to partial pickups). Allen and Browne’s (2010) look at commercial-space data across 16 major UK regions suggest a limited rise in retail space (just 4% over the 1998–2008 20-year period, across England and Wales, and 5% in London), only moderate intra-urban gains in warehousing (e.g., just 5% in London), and sizable office space growth (within regions and across the island— averaging 24%), as the nation de-industrializes. While ware- housing floor space across England and Wales rose 22% over the 20-year period, the number of warehouses grew just 3%. Finally, it was noted that office operations lend themselves to far less use of heavy-goods vehicles than warehouse, retail and industrial sites, per square meter of floor space. Lighter goods vehicles are also more common in urban freight movements (versus inter-urban movements) across other land use types, for reasons of maneuverability and shipment size. Truck Energy and Emissions Bronzini’s (2008) examination of Southworth et al.’s (2008) energy and truck VMT estimates across U.S. metropolitan areas indicate how controlling for regional population alone can predict 75% of the variance in commercial truck VMT. Population is less of a predictor for such freight VMT because so much freight movement entails through traffic. Truck VMT and carbon emissions (per capita) were most correlated with job and population density measures (r ≈ -0.48)—as com- pared with their correlations to the shares of metropolitan area jobs within 10 and 35 miles of the CBD, a couple of coarse jobs-housing-balance measures, and the presence of rail transit (though all were significant) (Southworth et al. 2008). Obviously, metropolitan structure is important for travel distances—with differences in origin and destination accessi- bility, roadway congestion, building sizes (per occupant) and design, parking costs, alternative mode availability, and vari- ables like climate impacting travel decisions and energy use. Southworth et al.’s (2008) examination of U.S. data sets suggest more than 2-to-1 differences in VMT per capita when compar- ing top 100 U.S. metropolitan areas such as Bakersfield, Cali- fornia, and New York, and potentially 4-to-1 ratios that emerge in simple per capita GHG calculations across such region pairs. Sarzynski et al.’s (2008) follow-on calculations suggest that freight-related GHG variations (per capita) are even more pro- nounced between low- and high-density pairings: at ratios of

36 4 to 1 or more. Location is important, and HDT travel is a part of the equation. But little research exists to quantify the distinc- tions at relatively high levels of spatial resolution. To this end, Bronzini (2008) recommended simulation studies of various land use pattern scenarios versus truck travel patterns. Done well, such simulations can anticipate a variety of travel changes, alongside system benefits and costs. Like Wygonik and Goodchild (2011a, 2011b), Allen and Browne (2010) remark on the travel and energy savings of higher density land use patterns for freight deliveries (and pre- sumably pickups as well as shipper drop-offs). They note that land use mixing has this potential as well, but the relegation of distribution hubs to exurban sites may not support such supply-chain arrangements. Finally, they recognize some value of more connected networks (e.g., grid layouts versus cul-de-sacs) for efficient multistop routing strategies, and they acknowledge an almost exclusive reliance on the truck mode for intra-urban freight movements. Unfortunately, there is no data provided to quantify such expected relationships. The trend toward electric and other clean-fuel trucks could allow freight delivery in off-peak and late-night periods because vehicles can operate quietly without disrupting the nighttime tranquility of neighborhoods as much. Thus tech- nology could enable a travel demand management response— off-peak delivery—which in turn could improve peak-period traffic conditions and reduce emissions from trucks. There has not been significant research on this topic to date. Integrating Freight and Community Goals The NCHRP 320 Synthesis (Strauss-Weider 2003), on the topic of integrating freight facilities and operations with com- munity goals, highlights the conflicts of and opportunities for mixing major freight facilities with other land uses. Best prac- tices for such colocation include replacing at-grade rail cross- ings with separated-grade facilities (to avoid traffic queue formation during train movements and stop periods) and incentivizing shippers and carriers to rely more on rail trans- port, to moderate highway congestion and safety concerns. Freight activity sites, like distribution centers, can make good sense for brownfield redevelopment projects in urban loca- tions, along with buffer zones around freight-related uses (in order to transition into residential uses) and electrification of gantry cranes (or other, alternative fuels). Such modifica- tions can improve safety and air quality (reducing particulate matter exposure from diesel engines). Strauss-Weider’s review recognizes “the growing need to balance freight transportation and community goals” (2003, p. 5) to enable commerce without compromising basic health and quality-of-life objectives. Of course, colocation of con- sumers (and workers), producers and goods is fundamental to moderating travel costs while serving final and intermediate demands. She notes how the growth in population, intensifi- cation of land development near ports and trade corridors, and shift to a largely service economy brings many conflicts to the fore amid a set of stakeholders that (mostly) do not have direct appreciation for (and understanding of) freight trans- port needs. Rising incomes and living standards reduce resi- dents’ tolerance of noise, delays, and pollution. Summary and Recommendations Strengths of Existing Work A generous body of research has been completed—literally hundreds of studies—focusing on the relationship between the built environment and trip making, on a daily basis. This work has been documented in a number of meta-analyses, which have typically provided elasticities and other analytical methodologies. With these methodologies, users have devel- oped defensible tools that allow “what-if ” estimations of potential reductions in VMT and VHT related to alternative built environment scenarios. While most of this work focuses on the project scale, there are additional tools available for meso-scale and macro-scale analysis. Case studies have pro- vided hints of how TOD might influence peak-period travel. In addition, there is some research indicating that jobs/ housing match improvements can reduce congestion. One study addressed the impact of higher density development on traffic congestion. Two recent meta-analyses, along with other recent studies, provide connections between mode choice, particularly tran- sit usage and walking, to built environment factors. Findings include strong correlations between walking and transit trips and various characteristics of the built environment. Studies have established a link between increased road capacity and increased driving; these increases are reflected in both near-term and long-term impacts. Case studies indicate that the opposite also holds: reduction in roadway capacity can lead to mode shift and elimination of some trips. Key Findings Key Decision Points for Smart Growth in the Planning Process The review of planning processes with a focus on smart growth and the interviews conducted with planning officials on this same topic revealed two primary areas that planning agencies are engaged in that are useful and supportive of engaging smart growth in planning processes. The first area is that most agencies are either engaged in or interested in scenario planning as a strategy for evaluating smart growth. Scenario planning offers many opportunities, but to date has

37 not been developed into a tool for this purpose that could be shared or adapted for use by planning agencies. The second area is that many agencies reflected on the need for coordina- tion, cooperation and communication with local govern- ments on land use policy, since land use regulations are primarily governed by local governments. This interaction between land use and transportation planners has provided opportunities to engage in discussions about integration, interaction, and common goals. The review also highlighted several topics where planning agencies feel additional guidance or tools would be worth- while: • Metrics and tools for induced demand, TDM, and urban form. • Understanding which strategies work best, that is, what outcomes can be expected? • Tools to evaluate impacts of smart growth on project selection. • Goals for congestion reduction may be counterproductive to smart growth. The Built Environment’s Impacts on Peak Auto Demand Peak-period travel remains the primary focus of demand and supply analysis, yet time-of-day travel has become increasingly complex. The simple assumption that peak-hour congestion is attributable to home-based work trips is clearly no longer valid. In 2001, for example, more than half of all trips during the 6:00 to 9:00 a.m. period were for nonwork purposes and during the p.m. peak the share exceeded 70% (FHWA 2007c). Case study analyses provide insights into smart growth and congestion relationships. Both residences and destinations, like job sites and shopping venues, need to be concentrated around transit stations to assure both trip origins and desti- nations are linearly aligned along a rail- or BRT-served cor- ridor (Cervero 2007a). Even then, not everyone believes that TODs will deliver mobility benefits in car-dependent societies such as the United States. According to one critical observer, TOD “increases congestion. The overwhelming majority of travel to proposed transit-oriented developments will be by automobile. This will strain road space, slowing traffic and increasing pollution as a consequence” (Still 2002). TOD can become another major vehicular traffic magnet or major vehicular traffic generator without a balance of residential and nonresidential uses. A 2010 study of the Austin region found that TOD sce- narios, in addition to reducing estimates of VMT (vehicle miles traveled), could also significantly reduce 2030 peak- period congestion (Kakaraparthi and Kockelman 2010). Under the base case 2030 scenario, 3,729 roadway lane miles (20.3% of the study area’s coded-network total) were pre- dicted to be congested in the morning peak. The rail-based TOD plan was projected to reduce congested roadways by 433 lane miles versus the base case, representing 18% of the region’s lane miles. The most aggressive (All-Systems-Go) TOD scenario was expected to reduce congestion on an addi- tional 341 lane miles or to 16.1% of the regional total. According to the analysis, the mid-level rail-based TOD was forecast to reduce traffic congestion by 11.7% relative to the base case. The All-Systems-Go TOD option would likely reduce it an additional 9%, or a total of 20.7%, relative to the base case. There were 17 TOD-housing projects surveyed and these averaged 44% fewer vehicle trips than that estimated by the ITE manual. The weighted-average differentials were even larger during peak periods: 49% lower rates during the a.m. peak and 48% lower rates during the p.m. peak. In general, denser, more urban TOD-housing had the greatest peak-hour trip rate differentials. A survey focused on parking demands at TODs shed further light on TOD’s transportation impacts (Cervero et al. 2010). In the case of Portland’s transit-oriented housing proj- ects, parking demand was 11% less than that estimated by the ITE Parking Generation Manual, which is based on peak parking periods (typically in the early morning). On average, the supply of parking exceeded peak demand by 30% at Port- land’s TOD projects. Other research focused on the commute trip found that a doubling of occupationally matched jobs within 4 network miles of workers’ residences was associated with a 32.9% reduction in commute VMT and a 33.8% reduction in com- mute VHT. The slightly larger elasticity of work-trip VHT as a function of job accessibility suggests that, on average, improved job access translates into slightly faster commute speeds. Cervero and Duncan (2006) conjectured that this could be due to the rationalization of commute patterns, with subregional balances in jobs and housing marked by less cross-town, lateral, and zigzag patterns of commuting from one quadrant of a region to another. The research also showed that larger commute-trip VMT and VHT reductions occurred as a function of job accessibility than did shop-trip reduc- tions as a function of retail access. While balancing where people live and shop matters in driving down VMT and VHT, balancing where they live and work matters even more. Focusing on the effects of smart growth at travel destina- tions, two studies found significant trip reduction resulting from development density, land use diversity, urban design at workplaces and other activity attractors. One study of all of Montgomery County, Maryland, found that elasticities describing the selection of non-auto travel at were twice as high for the density and diversity at destinations throughout the county as they were for residential locations within the county (Cervero 2002a). Another study in the Seattle region

38 found significant influence of employment density on reduc- ing single-occupant-vehicle use and increasing walk and transit for work trips (Frank and Pivo 1994). A national synthesis of more than 200 research studies on travel and the built environment found consistent evidence of VMT reductions resulting from smart growth characteris- tics. Elasticities ranged from a 4% reduction in VMT per 100% increase in development density, to a 9% reduction for each 100% improvement in diversity, 12% per each 100% improvement in urban design, 22% for each doubling of des- tination accessibility, and 5% for improved transit accessibil- ity (Ewing and Cervero 2010). Mobility by Mode and Purpose Research studies have demonstrated that housing in close proximity to rail transit stations averages high transit modal splits for commute trips and that improved walking connec- tions to rail stops increases this modal share even more (Lund et al. 2006; Chen et al. 2007; Cervero 1994; JHK and Associates 1987, 1989; Stringham 1982). Others have reached similar con- clusions: compact, mixed-use, traditionally designed neigh- borhoods encourage internal walking trips that substitute for out-of-neighborhood shop trips. A six-region analysis of mixed-use development found that jobs-housing balance most strongly predicted the likeli- hood that trips made by residents to nonwork destinations would be walking trips. Overall, however, new urbanist resi- dents logged 40 to 55 minutes more walking and cycling each week than their counterparts in the conventional suburban neighborhoods. Utilitarian travel, such as to work or shopping, accounted for the difference. This finding concurs with that of Saelens et al. (2003), who found that neighborhood design is not related to leisure-time physical activity when one controls for individual- and household-level characteristics. Also, the North Carolina study found that increased numbers of walk- ing trips came at the expense of automobile trips, consistent with prior evidence (Cervero and Radisch 1996). The largest VMT reductions accrue when going from very low to moderate densities. Some observers claim that lifestyle preferences explain much of the lower levels of VMT in denser, more walking-friendly neighborhoods, and that fail- ure to account for self-selection could bias results. In a study of neighborhoods in the Puget Sound area of Washington State, Krizek (2003) removed possible self-selection biases by longitudinally examining changes in travel when households relocated. He found that moving to a neighborhood with denser, mixed-use, well-connected street patterns was associ- ated with VMT reductions. The mixed-use development tool, mentioned in Chapter 2 (Table 2.4), uses hierarchical modeling to estimate walking and transit use (for external trips) from mixed-use development (Ewing et al. 2011). The walking share of external trips is related to three types of D variables: diversity, destination accessibility, and demographics. The transit use share of external trips is related to measures of design, destination accessibility, distance to transit, and demographics. A national study of 239 mixed-use and transit-oriented development sites in Boston, Atlanta, Houston, Seattle, Port- land, and Sacramento found that statistically verifiable evidence of travel reductions of between 20% and 45% by region result- ing from trip internalization, and walking and transit use to off- site destinations. The study categorized the travel generation by trip purpose, thus allowing for the evaluation of trip reduction and trip length effects by time of day (Ewing et al. 2009). Induced Traffic and Induced Growth Research has concluded that over the long term, added road capacity led to more deeply rooted structural shifts, such as increased car-ownership rates and more auto-oriented land- development patterns, what is sometimes referred to as induced growth. Adding structural impacts to accumulated short-term ones markedly increases long-term elasticities— on average, 0.73 in the United States (Cervero 2002b). In a study of more than 100 cases of road-capacity reductions in Europe, North America, Japan, and Australia, Goodwin et al. (1998) found an average overall reduction of 25%, even after controlling for possible increased travel on parallel routes. This “evaporated” traffic was assumed to represent a combination of people forsaking low value-added (discretionary) trips and opt- ing for alternative modes, including transit, walking and cycling. A Texas study surveyed residents who had walked to a local store and found that about one in eight stated they would have stayed home instead of driving if there had been no nearby store within walking distance. This implied that the opportunity to walk to a store likely induced some extra pedestrian trips. Relationship Between Smart Growth and Congestion A number of studies cited in previous sections address travel reduction effects of smart growth either by time of day or by trip purpose and destination, allowing the deduction of peak- hour effects. These include studies performed at the macro scale (Zhang 2010), and at the meso and micro scales (Ewing et al. 2009; Cervero 2002a; Cervero 2007a; Frank and Pivo 1994). While a considerable body of research has successfully iso- lated and begun to qualify the effects of smart growth land use design on trip making, there has been a lack of research on the subsequent link between smart growth development and traffic congestion. When communities incorporate higher levels of the Ds in their design, households that reside in those communities

39 own fewer cars, make fewer trips by vehicle, and generate lower rates of VMT than household of comparable demographic composition living in more conventional single-use settings. Similar results occur in employment and commercial activ- ity centers. When these destination areas combine uses in a more compact, walkable setting, commuters, shoppers and visitors are found to be much more likely to travel to these locations by modes other than driving, and once there, to con- duct a higher percentage of their work-related or non-home- based trips locally by walking or by transit. In one of the few known studies to address these issues head-on, the Arizona DOT commissioned a study of the impact of higher density development on traffic congestion (Kuzmyak et al. 2012). Using a case study approach comparing four sites in the Phoenix area—three very urban in density and character, and one more typically suburban—the key finding was that while the three urban sites had residential densities twice that of the suburban example, and employ- ment densities greater by factors of 7 to 25, traffic conditions were actually much better in the higher density, mixed-use urban examples. Further investigation showed that this result was attributable to higher rates of internal capture of resi- dents’ trips for all trip purposes, resulting in shorter trip lengths and lower VMT rates. The urban examples also had higher rates of transit use both by residents and visitors, and featured extensive street grids that both facilitate walking and allow for better management of vehicle traffic flow. All of the areas were affected by high proportions of through traffic, though the urban examples—seemingly because of the street grid—appeared better able to absorb and dissipate the effects of this additional demand. A second example, taken from Prince George’s County, Maryland, examined the relationship between higher-intensity development in designated centers and corridors and traffic impacts on local area LOS standards (Kittelson and Kuzmyak 2010). Projected violation of traffic standards on measured facilities in the centers/corridors under 2030 build-out condi- tions imperiled adopted smart growth and TOD plans for these areas. In a detailed analysis of six centers, two key find- ings were made: (a) the majority of traffic in the areas of vio- lation could be attributed to through travel and not to the development activity of the development area, and (b) the centers/corridors could do a much better job in achieving desired travel efficiencies than their current designs enabled. Lacking tools or formal protocols for effective smart growth design, the centers were found to be deficient in terms of density, mix of uses, effective design (pedestrianization, con- nectivity, street grid), and taking best advantage of transit infrastructure. The methods developed and performance metrics used in this assessment are perhaps its key contribu- tion to the report, because they provide a mechanism for assessing this complex set of issues. Smart Growth and Freight Traffic Smart growth emphasizes accessibility, rather than mobility, though more efficient location choices and connected trans- port systems, for more “complete” neighborhoods. Like per- sonal travel, goods movement is core to the health and wealth of all communities. However, freight offers fewer mode choices, along with many challenges. Truck and rail modes dominate goods movement, each shuttling more than a tril- lion ton-miles of the U.S. commodity movement annually (CFS 2007). While rail generally is a more efficient mode of freight travel in many ways, it cannot access most buildings or penetrate most neighborhoods, thus requiring integration with trucking systems for final delivery of many goods. Inland ports or freight villages, and public logistic terminals or multi- company distributions centers facilitate such intermodal operations along with cross-company consolidation for more efficient customer service in highly urbanized environments. Simulation studies, to examine the details of design and logis- tics choices, can be essential in the definition, siting and valu- ation of such programs and policies. The research discovered the following factors linking freight traffic with land use patterns, and logistics management that might be addressed through smarter growth planning and regional and local logistics: • In recent years freight energy efficiency has fallen, possibly due to more trucks traveling empty, or dead heading. • Double-tracking of more rail corridors could dramatically improve rail’s reliability and travel times, enhancing its modal competitiveness. Rising roadway congestion, the introduction of road tolls and higher gasoline taxes may incentivize shifts to rail and other freight modes. • Port operators, local residents, and business leaders are rec- ognizing that land use decisions are critical to environmen- tal justice, human health, the economy, and quality of life. • Transshipment points for warehousing by multiple opera- tors facilitate intermodal transfers and goods storage while enabling consolidated operations, including shared pick- ups and deliveries within the nearby cities. • In terms of smart growth solutions, studies demonstrate that micro-, meso-, and macro-scale measures are needed to improve freight operations and rationalize land use and locational factors that influence them. Site access design, such as the provision of curb loading zones, one-way alley protocols, and signage can be beneficial as can use of smaller trucks for intraneighborhood operations, with satellite transfer facilities for shifting goods to and from larger trucks. • Freight operators cite the advantages of shorter/single-unit trucks for urban stops, all-way pedestrian phases to mini- mize pedestrian risk during truck turning movements. Cen- tralized delivery locations, with intra-mall/intra-building

40 delivery made onsite by specialized mall-managed vehicles at shopping malls and large office buildings. • Commercial strip development is undesirable, as it is not so conducive to safe or efficient delivery practices. Urban deliv- eries become much more difficult as congestion mounts and business practices place more emphasis on time-sensitive pickups and deliveries. Just-in-time deliveries and increas- ingly complex supply chains are threatened by growing congestion. • Reservation times at port facilities, congestion-based road and runway tolling, variable pricing of capacity-constrained rail corridors, and various impact fees for existing and new land uses may ensure reliability in movement of freight. • Metropolitan structure is important for travel distances— with differences in origin and destination accessibility, roadway congestion, building sizes (per occupant) and design, parking costs, alternative mode availability. Density of customers (and smaller vehicles) reduces the cost and emissions of deliveries. • Simulation exercises across a variety of U.S. and EU regions concluded that many transport pricing, land use policies, and investment strategies offer significant long-run reductions in VMT and emissions (relative to trend) without compromis- ing highway levels of service or regional productivity. Recommendations Key Decision Points for Smart Growth in the Planning Process Many planning agencies are evaluating smart growth policies and are looking for tools to understand the implications for induced demand, TDM, urban form, project selection, and congestion reduction as well as information on expected outcomes. The Built Environment’s Impacts on Peak Auto Demand While there has been considerable study and syntheses lead- ing to well-established relationships between smart growth and travel demand on a daily basis, the research on travel effects by trip purpose or by time of day is much more limited. This creates a challenge for the prospect of estimating the effects of smart growth development patterns and transportation man- agement on peak-period traffic conditions and congestion. Mobility by Mode and Purpose As is the case with evidence on smart growth effects on peak traffic, evidence on mode choice and mobility is much more limited under peak conditions than when expressed in term of full-day metrics. Induced Traffic and Induced Growth A moderate sampling of credible studies of induced travel and induced growth suggest that elasticities describing traffic demand growth tend to rest in the range of 0.3 to 0.4 in the short term and between 0.6 and 0.7 in the long term when expressed as functions of the amount of added traffic capac- ity. In other words, up to 70% of the added capacity would be used by induced travel. However, capacity expansion at a spe- cific location is a very crude indicator of the effect of a traffic network improvement on travel decisions ranging from route shifting, to time-of-day shifting, to mode shifting, to trip gen- eration and distribution and land investment and develop- ment. More empirical evidence is needed on the subject of induced travel measured as a function of travel time benefits afforded by a transportation improvement that captures the effects the facility’s role in the network, the effects of non- capacity operational improvements, and the degree to which land use plans represent a priori conditions rather than effects of the added transportation access. Relationship Between Smart Growth and Congestion Research is quite limited on the subject of congestion effects of smart growth. There is some evidence that the combined effects of lower trip generation per unit of development, shorter trip distances and better interconnected circulation networks that characterize smart growth reduce overall regional congestion and, in several examples, reduce congestion at the local level even in spite of the increased land use intensity. The research sample is too small, however, to develop statistically strong rela- tionships that might be transferable to other regions and situa- tions. There is a critical need for further data gathering at a macro level from sources such as Texas A&M Transportation Institute and at corridor and local levels from cities, counties, DOTs, and GPS data vendors, and for statistical analysis to ascertain the transferable relationships between smart growth characteristics such as the Ds, including network density and connectivity, and levels of traffic volume and congestion on local streets, arterials and highway. Smart Growth and Freight Traffic Smart growth lends itself to relatively narrow street systems and higher shares of nonmotorized modes (with their rela- tively vulnerable travelers), which poses issues for large-truck access and traveler safety. While density lends itself to more efficient routing of delivery vehicles, smaller businesses may generate more freight trips, per ton moved. And colocation of freight facilities and populated land uses poses safety, noise, pollution, theft, and other concerns. Ultimately, freight

41 movement must occur to sustain the enterprise of human settlement. Better design of loading docks, better vehicle and routing choices, more full-cost pricing (of fuels, scarce road and parking spaces, and vehicles), separation of various freight facilities and crossings (to protect the public and avoid bottle- neck queuing), and new systems to facilitate interfirm coop- eration and stakeholder communication all support reliable and safe goods movement within the smart growth context. Information Gaps and Limitations of Current Practices Relatively little information is available regarding the effect of smart growth on trip purpose and peak-hour congestion. Where the connection between the built environment and travel has been least studied is the link between travel behav- ior in response to land use designs and the traffic that is actu- ally occurring on the street and highway system. In addition, while there is emerging information regarding the use of alternative modes attributable to smart growth, there are no calibrated and validated trip generation rates for bicycle, walking, and transit trips tied to the built environment. Little is known about the induced traffic and induced growth impacts of smart growth initiatives themselves, as reflected by changes in attributes of the built environment, such as higher residen- tial densities, increased mixed land uses, or improvements in the pedestrian environment. No standard, widely accepted kit- bag of tools has emerged for estimating induced-demand impacts of highway or transit improvements, much less of gauging the second-order, rebound impacts of smart growth strategies. An assessment of the strengths and limitations in the cur- rent practices of assessing the effects of smart growth on transportation capacity identified the following limitations: • Most state and regional transportation agencies are either engaged in or interested in scenario planning as a strategy for evaluating smart growth but find that they lack suitable tools for this purpose. • Many agencies feel the need for coordination, cooperation and communication with local governments on land use policy, since land use regulations are primarily governed by local governments, suggesting that tools need allow the plan- ning process to operate at multiple scales, including regional (macro), corridor and community (meso) and development project such as specific plan or TOD (micro). • The underlying relationships that define the effects of smart growth on peak travel and transportation capacity needs are not well understood. While there has been considerable research and well-established relationships between smart growth and daily travel demand, research on travel effects by trip purpose or by time of day is much more limited. This creates a challenge for the prospect of estimating the effects of smart growth development patterns and trans- portation management on peak-period traffic conditions and congestion. As is the case with evidence on smart growth effects on peak traffic, evidence on mode choice and mobility is much more limited under peak conditions than when expressed in term of full-day metrics. Reliable means of efficiently predicting the effects of induced growth and travel are also lacking. Some studies sug- gest that short-run traffic growth consumes 30% to 40% of added highway capacity and that long-term traffic growth fills 60% to 70%. However, capacity expansion at a specific location is a very crude indicator of the effect of a traffic net- work improvement, as the travel responses are complex and nuanced. They include route shifting, time-of-day shifting, mode shifting, trip generation and distribution and land investment and development. There is a need for further study of induced travel when measured as a function of travel time benefits afforded by a transportation expansion in a manner that captures the facility’s role in the network, the effects of noncapacity operational improvements, and the degree to which land use plans represent a priori conditions rather than effects of the added transportation access. Research is also quite limited on the subject of congestion effects of smart growth. There is some evidence that the com- bined effects of lower trip generation per unit of development, shorter trip distances and better interconnected circulation networks that characterize smart growth reduce overall regional congestion and, in several examples, reduce conges- tion at the local level in spite of the increased land use inten- sity. However, the research sample is too small to develop statistical relationships that might be transferable among regions and situations. There is a critical need for data and statistical analysis to ascertain the transferable relationships between smart growth characteristics such as the develop- ment density and diversity and transportation network con- nectivity, and the resulting traffic congestion on local streets, arterials and highways. With regard to freight planning, there are a number of smart growth and logistical strategies that can reduce the exposure of goods movement to congestion and delay. These strategies are often interregional as well as local in scope and, as tactics, are transferable among regions. Modeling tools or resource materials should attempt to address freight logistics in public scenario planning, possibly through case studies and best practices for addressing freight issues and to test the effects of alternative regional growth patterns and transpor- tation network investments on goods movement.