Click for next page ( 104


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



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

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

OCR for page 103
Prospects" (Cervero et al., 2004) presents examples and case studies from a variety of TODs around the United States. Although the focus of the report is not travel behavior, there are certain travel- behavior-related findings reported. Similarly, the final report, "Statewide Transit-Oriented Development Study: Factors for Success in California" (Parker et al., 2002) touches on travel behavior related information within a larger, comprehensive treatment of TOD. Recent general interest books on TOD that touch on impacts include The New Transit Town: Best Practices in Transit-Oriented Development (Dittmar and Ohland, 2004) and the Urban Land Institute's Developing Around Transit: Strategies and Solutions That Work (Dunphy, 2005). Of particular use from a TOD travel data perspective is "Travel Characteristics of Transit-Oriented Development in California" (Lund, Cervero, and Willson, 2004a), identified throughout this chap- ter as the "2003 California TOD travel characteristics study." The researchers surveyed several hundred residents and workers in station-area developments in a variety of contexts in California to facilitate some disaggregate analyses and to perform selected comparisons with background travel behavior. Substantial region-specific compilations of station-area travel shares and related information include the Washington Metropolitan Area Transit Authority's 2005 Development- Related Ridership Survey (WMATA, 2006a) and the Metropolitan Transportation Commission's Characteristics of Rail and Ferry Station Area Residents in the San Francisco Bay Area: Evidence from the 2000 Bay Area Travel Survey. The latter report uses a regional data set obtained from nearly 35,000 residents to examine demographic profiles and travel characteristics of individuals residing within various sidewalk walking distances of rail stations and stops and ferry terminals (Metropolitan Transportation Commission, 2006). The National Center for Transit Research report, Impacts of Transit Oriented Development on Public Transportation Ridership, already provides a well organized synthesis of relevant literature and studied recognition of research difficulties and needs, even though original research findings must await the project's Phase II (Hendricks, 2005). Two TCRP projects will likely produce future project reports of interest to those concerned with the ridership impacts of TOD: TCRP Project H-31, "Understanding How Individuals Make Travel and Location Decisions: Implications for Public Transportation" and TCRP Project H-27A, "Ensuring Full Potential Ridership from Transit- Oriented Development." In addition, the "Benchmarking TOD" project for the Federal Transit Administration by the University of California Transportation Center is developing performance measures for TOD with case examples from five U.S. cities. CASE STUDIES Portland, Oregon, Metro Region TOD Travel Effects Investigation Situation. A modest original research investigation was undertaken of transit oriented development's (TOD's) effects on non-work travel in Portland, Oregon, as part of TCRP Project B-12B, the TOD component of the "Traveler Response to Transportation System Changes" Handbook. The objective was to see whether or not an advanced travel demand model, in this case the Oregon Model Steering Committee's research model of mode choice 17-103

OCR for page 103
for home-based non-work purpose trips,23 would fully account for the particular characteristics and effects of the TOD urban form. The model in question had been developed to test different approaches for including an urban design descriptor to both improve model performance and provide forecast sensitivity to alternative urban forms. Advantages of using Portland for this investigation have included the large number of TODs in the urban area and the availability of a sophisticated travel demand model set and database, backed up by extensive applied research and experience on how best to represent urban form in travel models. A disadvantage has been that the most recent regional household travel survey, taken in 1994/1995, dates from a time when much of the TOD along the west light rail transit (LRT) line was in preliminary stages of development. However, TOD was in place along the east LRT line opened in 1986, as well as along major trunk bus lines where it is accompanied by streetcar-era tra- ditional development equivalent to TOD. Actions. A two-part model-based investigation was undertaken. A set of TOD indicator variables was added to the Oregon Model Steering Committee's final research model to see if the variables would prove significant, exhibit logical parameters, and provide improved estimation of modal shares for non-work travel to and from TODs. A comparison was then made of walk, bike, transit, and auto mode shares for TODs and non-TOD areas within and outside of the central area. Compared were the actual non-work mode shares observed in the travel survey; the shares estimated with the Steering Committee's initial, base-case model lacking an urban design variable; the Committee's final model containing a composite urban design variable; and the final test model with the set of TOD variables added into it. The Steering Committee's base-case model already contained variables describing auto, transit, walk, and bike travel times and costs as well as selected demographic variables. Thus the inherent proxim- ity of TODs to good transit service was already represented in their initial model, along with effects of auto ownership and availability. Their final research model adds a quantitative continuous urban land use and design variable describing the intensity of the mix of retail businesses, households, and local intersections, the latter being a measure of street and pedestrian facility continuity.24 The TOD variables which were then superimposed indicate whether the conventional and urban land use and design variables are sufficient to express TOD effects on non-work travel mode choice or whether they leave unexplained some degree of walk, bike, or transit mode attractiveness peculiar to non-home destinations in travel analysis zones (TAZs) judged to exhibit TOD characteristics. 23 The Oregon Model Steering Committee's model, characterized here as a "research model" to distinguish it from models adopted for official use, is not a limited-scale research formulation such as might be prepared to support a one-time investigation of a specific phenomenon. It is a full-scale, network-based regional travel demand model. The mode choice component estimates home-based, non-work travel shares for the auto, transit, walk, and bike travel modes. 24 The land use and design variable selected by the Oregon Model Steering Committee is the normalized harmonic mean of three measures: the number of retail businesses within 1/2 mile, the number of house- holds within 1/2 mile, and the number of local intersections within 1/2 mile. The harmonic mean formula- tion increases rapidly when all of its terms increase. Places with a moderate mix of retail businesses, households, and local intersections score higher than a place with very high amounts of one measure and little of the others. 17-104

OCR for page 103
Analysis. The TOD variables were added to the Steering Committee's final research model in the form of "dummy" (yes-no) variables, three in all, one each associated with the walk, bike, and transit non-auto modes of travel. This produced a separate TOD coefficient for each mode. TAZs identified as having TOD characteristics were given a value of 1; all others were assigned a value of 0. The TCRP Project B-12B TOD-included specification of the final research model was then calibrated and evaluated. Table 17-46 lists the variables and provides calibration results for the Steering Committee's initial base-case model, their final research model with its urban design variables, and the TOD-included modification. Table 17-46 Initial-Research, Final-Research, and TOD-Included Logit Model Calibrations (Coefficient Values and T-Statistics) Initial Model Final Model TOD-Included Variable Mode Value T-Stat Value T-Stat Value T-Stat walk ASC W -0.26 -5.2 -0.49 -9.2 -0.50 -9.3 walk time W -0.097 -36.3 -0.093 -34.0 -0.090 -32.6 walk--0 auto hh W 2.9 23.3 2.7 20.8 2.7 21.0 walk--autos < workers W 0.80 9.1 0.68 7.5 0.66 7.3 bike ASC B -3.4 -32.6 -3.4 -32.6 -3.6 -31.5 bike time B -0.14 -15.2 -0.13 -13.8 -0.12 -12.8 bike--0 auto hh B 2.5 11.8 2.4 10.8 2.4 11.0 bike--autos < workers B 0.66 3.4 0.57 2.9 0.59 3.0 transit ASC T -3.6 -27.3 -3.6 -27.3 -4.6 -26.3 transit time T -0.047 -14.5 -0.038 -7.4 -0.026 -5.0 transit--0 auto hh T 4.3 30.9 2.3 10.8 4.2 28.4 transit--autos < workers T 1.1 6.8 1.1 6.4 0.97 5.8 auto travel time A -0.13 -9.1 -0.11 -7.6 -0.092 -6.2 cost TA -0.63 -9.3 -0.53 -11.5 -0.4127 -8.6v walk--urban design W -- -- 0.00053 12.5 0.00050 10.1 bike--urban design B -- -- 0.00042 4.9 0.00045 4.8 transit--urban design T -- -- 0.00048 6.7 0.00036 4.8 walk--TOD dummy W -- -- -- -- 0.23 2.8 bike--TOD dummy B -- -- -- -- 0.014 0.1 transit--TOD dummy T -- -- -- -- 1.2 9.1 Notes: ASC stands for Alternative Specific Constants. Mode codes are A--Auto, B--Bike, T--Transit, W--Walk. "--" indicates that the variable is not included in the indicated model specification. The analysis required that each TAZ in the Portland region be characterized as having TOD characteristics or not having TOD characteristics. To keep this identification at arms-length from the model systems being examined, it was done judgmentally by two locally-based but nationally recognized experts, G. B. Arrington and Keith Lawton, one an authority on TOD development and the other on travel demand and urban design. The TOD TAZs thus identified include new development at stations of both outer and inner segments of the east LRT line, traditional urban development in the Broadway and Hawthorne Districts along frequent bus service focused on Portland's Transit Mall, and a number of areas in and around 17-105

OCR for page 103
and south of downtown Portland similarly characterized by mixed-use, higher-density develop- ment featuring high quality transit service and pedestrian interconnection and amenities. Results. When the TOD variables were added no statistical problems became evident,25 all three had the expected signs (+), and the walk and transit TOD variables were statistically significant. The bike TOD variable was not. The positive signs indicate that all the non-auto modes have a higher utility when associated with trip attraction areas having TOD characteristics, other things being equal. In effect, non-auto travel to TODs for non-work trip purposes was shown to be more attractive even than estimated on the basis of transit service characteristics and the Steering Committee's composite urban design measure. The Committee's final research model "knows" which trips have good transit service and are to higher density, mixed-use, well-connected urban locations, yet still cannot quite replicate the actual degree of transit use and non-motorized travel to TOD and TOD-like areas. There are several possible explanations, one or more or all of which may pertain: TODs may provide a synergism of good transit service, easy ability to move around on foot, and placement of daily needs within easy reach whereby the overall enhancement in use of environment-friendly travel modes is greater than the sum of the parts. The holistic non-auto travel environment may attract persons who would prefer to meet daily needs without relying on auto use, an outcome referred to as self-selection in housing and des- tination choice. The existing state of modeling transportation systems, however advanced, may not yet be (and probably isn't) capable of fully reflecting the attention to pedestrian system continuity and quality of design typical of TODs. Pertinent to these possible explanations is that walking is an inherent and critical element of most transit trips. Walking environment improvements enhance transit travel at the same time as they enhance travel by non-motorized means alone. Table 17-47 illustrates how well the initial and final Steering Committee models did in replicating travel to and from TAZs identified as TODs, and how well the TOD-included modification with TOD variables added did. Combining central area and outlying auto mode estimation results for Portland TOD TAZs shows that non-work auto trip productions are overestimated by 8.6 percent with the initial research model, which already takes into account quality of transit service and traveler characteristics. Adding the urban design variable in the final research model reduces the overestimate from 8.6 percent to 1.5 percent, while inclusion of the TOD variables further halves the overestimate to 0.8 percent. Similarly, TOD non-work auto trip attractions are overestimated by 6 percent in the initial research model and by 3 percent in the final model with the urban design variable, while adding the TOD variables produces a perfect match for auto trip attractions overall. 25 The TOD dummy variables appeared to improve the predictive capability of the model without introduc- ing significant multi-colinearity effects. The urban land use and design variable remained significant and its coefficient did not change much. 17-106

OCR for page 103
Table 17-47 Comparison of 1995 Observed and Estimated Non-Work Trips and Mode Shares for Portland, Oregon, TOD and Non-TOD Area Types Area Walk or Bike Public Transit Auto Driver/Passenger Type Data Source P's A's P's A's P's A's Observed 431 33% 494 18% 103 8% 198 7% 755 59% 2,043 75% Area TOD Central Initial Model 334 26% 426 16% 79 6% 133 5% 876 68% 2,176 80% Final Model 430 33% 509 19% 95 7% 126 5% 765 59% 2,100 77% TOD-Included 430 33% 510 19% 96 7% 184 7% 763 59% 2,041 75% Observed 153 14% 144 18% 26 2% 12 2% 946 84% 626 80% Outlying Initial Model 130 12% 112 14% 24 2% 14 2% 971 86% 656 84% TOD Final Model 139 12% 118 15% 25 2% 15 2% 961 85% 649 83% TOD-Included 146 13% 128 16% 27 2% 26 3% 952 85% 628 80% Observed 1,672 8% 1,618 8% 276 1% 195 1% 20,356 91% 19,388 91% Non-TOD Overall Initial Model 1,792 8% 1,718 8% 302 1% 258 1% 20,210 91% 19,225 91% Final Model 1,688 8% 1,629 8% 285 1% 263 1% 20,331 91% 19,308 91% TOD-Included 1,682 8% 1,618 8% 281 1% 195 1% 20,341 91% 19,388 91% Notes: P's = Non-Work Trip Productions (trips observed/estimated at the home end of the trip); A's = Non-Work Trip Attractions (trips observed/estimated at a non-home end of the trip). For each observation or estimate both the absolute number of surveyed or estimated trips and the observed/estimated percentage mode share are given. The surveyed trips are not expanded to the total universe of trips. "TOD-Included" is the Oregon Model Steering Committee's final research model with TOD variables added (see text). The modeling objective is, in simplistic terms, to match the observed (surveyed) absolute number of trips and mode shares as closely as possible with the estimated trips and shares. More . . . The intent of this exercise was to test whether or not a special TOD effect is present at the TAZ level of observation. While this one test cannot serve as the last word on the subject, it does give evidence that features of TODs can give a boost to choice of non-auto travel modes that goes beyond what can be explained using an advanced set of urban design descriptors alone. Together, the urban design exhibited by TODs and the special TOD effect appear to diminish choice of the auto mode by some 6 to 9 percent compared to providing similarly good transit service in the con- text of Portland non-TOD development. This conclusion is different than claiming that the TOD-included model is either an appropriate forecasting model or better than the final research model of the Steering Committee. The use of dummy variables to approximate a continuous variable, as done here for TOD, is not usually good forecasting model practice. Furthermore, inclusion of a judgmentally applied TOD indicator requires the forecaster to identify TOD in the base year and in the future year, inherently a some- what arbitrary and prone-to-bias process. A potential line of further research is to develop and test the inclusion of a formulaic, continuous "TOD Index" variable that not only identifies the location of TODs but also measures and reflects their quality as well. 17-107