A Walk Trip Generation Model for Portland, OR Guang Tian, Reid Ewing Presented by Guang Tian Department of City & Metropolitan Planning University of Utah [email protected] Walkable street, but not within walkable distance Walkable distance, but without walkable street Research Question When assessing the benefits, costs, and priorities of proposed pedestrian improvements for the government or developers, it is frequently helpful to answer the following questions: • What kinds of built environments encourage people to choose to travel on foot? (neighborhood scale or street scale) • If a new pedestrian facility is built, how many people will use it? Trip generation in practice Trip generation in Conventional Four-Step Travel Demand Models • TAZs, block group or census tract, too big to represent the actual built environment for walk • Interzonal trips are usually ignored, including walk trips (Beimborn et al. 1997, p. 17 & p. 15) Trip Generation in Conventional Traffic Impact Studies (ITE Methodology) “Data were primarily collected at suburban localities with little or no transit service, nearby pedestrian amenities, or travel demand management (TDM) programs.” (ITE 2012, p.1) Further, ITE advises: “At specific sites, the user may want to modify the trip generation rates presented in this document to reflect the presence of public transportation service, ridesharing or other TDM measures, enhanced pedestrian tripmaking opportunities, or other special characteristics of the site or surrounding area.” (ITE 2012, vol. 1, p. 1) Trip generation in research • There are many studies on the associations between built environment and travel (Saelens and Handy, 2008; Ewing and Cervero, 2010). Specifically, positive relationships between walking for transportation and density, land use mix, distance to destinations and street connectivity (Besser and Dannenberg, 2010; Saelens et al., 2008; Sehatzadeh et al., 2011). • Challenges of walk behavior studies: sufficiently detailed data on the built environment that can be spatially matched to sufficiently detailed data on travel behavior is challenge (Handy et al., 2002); more refined spatial unit (Liu et al., 2012). Conceptual Framework “A city sidewalk by itself is nothing. It is an abstraction. It means something only in conjunction with the buildings and other uses that border it, or border other sidewalks very near it” (Jane Jacobs, the Death and Life of Great American Cities, 1961, p. 29). Study Area – Portland, OR 2011 Oregon Household Travel and Activity Survey Individual trips Households info Built environment data Population, employment Parcel land use data Street network Transit service intersections Sidewalk system Traffic control device …… Analysis Half-mile road network buffer around household location Cumulative percentage Mean Walk distance 0.31 Std.D 0.716 < 0.25 mile < 0.5 mile 63.2% 85.6% < 1.0 mile 96.4% Variables Sociodempgraphic characteristics Number of walk trips per household Neighborhood development Household size Race Number of workers Income Single-family vs. multifamily Activity density Job-housing balance Land use mix Transit stop density Employment accessibility Sidewalk quality (by PCA) Street design Intersection density % of 4-way intersection Sidewalk coverage Improved corner for pedestrian Traffic calming Traffic signal Slope Hurdle Models • The stage 1: categorizes households as either having any walk trips or not, and uses logistic regression. • The stage 2: estimates the number of walk trips generated by households with any (positive) walk trips, and uses negative binomial regression. o Stage 1 - Logistic regression • Dependent variable: households with any walk trips (1=yes, 0=no) Coefficient Household size Std. Error T-ratio P-value 0.431 0.054 7.955 < 0.001 -0.297 0.167 -1.784 0.074 0.321 0.077 4.187 < 0.001 -0.843 0.132 -6.383 <0.001 Activity density in thousand 0.035 0.008 4.400 <0.001 Job-housing balance 0.403 0.241 1.673 0.094 Sidewalk system quality 0.328 0.061 5.411 <0.001 -1.126 0.292 -3.862 <0.001 Race_dummy (1=white, 0=nonwhite) Number of workers in household Living (Single-family=1, multi-family=0) Constant Sample size: 1970 -2 log-likelihood ratio: 2404 Pseudo R2: 0.15 (Cox-Snell) o Stage 2 - Negative binomial regression • Dependent variable : the number of walk trips for the subset of households that make walk trips Coefficient Std. Error T-ratio P-value Constant 0.922 0.128 7.187 < 0.001 Household size 0.143 0.031 4.555 < 0.001 Transit stop density 0.003 0.001 2.834 0.005 Land-use entropy 0.542 0.229 2.365 0.018 Sample size: 962 -2 log-likelihood Ratio: 34 Pseudo R2: 0.01 (McFadden) Discussion and implements Socioeconomic variables are strong predictors of walk trip generation. • • • • Household size Race Number of workers Living (single-family vs. multi-family) According to the residential self-selection theory, “individuals with a preference for walking consciously choose a neighborhood that is conducive to walking” (Cao, Handy, and Mokhtarian 2006, p. 4). For the public policy makers, in order to induce more walking trips by pedestrian facilities, the people in neighborhoods might be considered. Positive influences of built environment • Higher activity density, good balance of jobs and housing supply, more mixed use neighborhoods create more opportunities that the destinations are within walkable distance. • Better transit services provides opportunities of combining different travel modes. • Interconnected street network provides places to cross the street and routing options to destinations, and generates more eventful trips. • Sidewalk coverage and improved corners provide a safe and comfortable travel way for pedestrians. Conclusion There are two distinctions between this study and earlier studies of walk trip generation. • The two-stage hurdle model - Logistic regression: the decision of a household to include walking among its mode choices - Negative binominal regression: the decision on how many walk trips to make • A full array of street design variables represented by a principal component Street design does not seem to affect the number of walk trips made by a household, but has a significant effect on the decision to walk at all. Limitations • One city (limits the external validity of the findings) • Other factors (weather condition and season changes; parking supplies and prices; residential attitudes) • Endogenous vs. exogenous (pedestrian facilities causes walk trips or walk trips causes pedestrian facilities ) • Pedestrian network buffer ( instead of street network buffer) Any questions and comments are welcome !