Kara M. Kockelman

TRAVEL DEMAND AND THE 3DS: DENSITY, DIVERSITY, AND DESIGN

Abstract The built environment is thought to influence travel demand along three principal dimensions —density, diversity, and design. This paper tests this proposition by examining how the ‘3Ds’ affect trip rates and mode choice of residents in the San Francisco Bay Area. Using 1990 travel diary data and land-use records obtained from the U.S. census, regional inventories, and field surveys, models are estimated that relate features of the built environment to variations in vehicle miles traveled per household and mode choice, mainly for non-work trips. Factor analysis is used to linearly combine variables into the density and design dimensions of the built environment. The research finds that density, land-use diversity, and pedestrian-oriented designs generally reduce trip rates and encourage non-auto travel in statistically significant ways, though their influences appear to be fairly marginal. Elasticities between variables and factors that capture the 3Ds and various measures of travel demand are generally in the 0.06 to 0.18 range, expressed in absolute terms. Compact development was found to exert the strongest influence on personal business trips. Within-neighborhood retail shops, on the other hand, were most strongly associated with mode choice for work trips. And while a factor capturing ‘walking quality’ was only moderately related to mode choice for non-work trips, those living in neighborhoods with grid-iron street designs and restricted commercial parking were nonetheless found to average significantly less vehicle miles of travel and rely less on single-occupant vehicles for non-work trips. Overall, this research shows that the elasticities between each dimension of the built environment and travel demand are modest to moderate, though certainly not inconsequential. Thus it supports the contention of new urbanists and others that creating more compact, diverse, and pedestrian-orientated neighborhoods, in combination, can meaningfully influence how Americans travel.

Driver injury severity: an application of ordered probit models

This paper describes the use of ordered probit models to examine the risk of different injury levels sustained under all crash types, two-vehicle crashes, and single-vehicle crashes. The results suggest that pickups and sport utility vehicles are less safe than passenger cars under single-vehicle crash conditions. In two-vehicle crashes, however, these vehicle types are associated with less severe injuries for their drivers and more severe injuries for occupants of their collision partners. Other conclusions also are presented; for example. the results indicate that males and younger drivers in newer vehicles at lower speeds sustain less severe injuries.

Travel Behavior as Function of Accessibility, Land Use Mixing, and Land Use Balance: Evidence from San Francisco Bay Area

The relative significance and influence of a variety of measures of urban form on household vehicle kilometers traveled, automobile ownership, and mode choice were investigated. The travel data came from the 1990 San Francisco Bay Area travel surveys, and the land use data were largely constructed from hectare-level descriptions provided by the Association of Bay Area Governments. After demographic characteristics were controlled for, the measures of accessibility, land use mixing, and land use balance—computed for trip-makers’ home neighborhoods and at trip ends—proved to be highly statistically significant and influential in their impact on all measures of travel behavior. In many cases, balance, mix, and accessibility were found to be more relevant (as measured by elasticities) than several household and traveler characteristics that often form a basis for travel behavior prediction. In contrast, under all but the vehicle ownership models, the impact of density was negligible after accessibility was controlled.

A multivariate Poisson-lognormal regression model for prediction of crash counts by severity, using Bayesian methods

Numerous efforts have been devoted to investigating crash occurrence as related to roadway design features, environmental factors and traffic conditions. However, most of the research has relied on univariate count models; that is, traffic crash counts at different levels of severity are estimated separately, which may neglect shared information in unobserved error terms, reduce efficiency in parameter estimates, and lead to potential biases in sample databases. This paper offers a multivariate Poisson-lognormal (MVPLN) specification that simultaneously models crash counts by injury severity. The MVPLN specification allows for a more general correlation structure as well as overdispersion. This approach addresses several questions that are difficult to answer when estimating crash counts separately. Thanks to recent advances in crash modeling and Bayesian statistics, parameter estimation is done within the Bayesian paradigm, using a Gibbs Sampler and the Metropolis-Hastings (M-H) algorithms for crashes on Washington State rural two-lane highways. Estimation results from the MVPLN approach show statistically significant correlations between crash counts at different levels of injury severity. The non-zero diagonal elements suggest overdispersion in crash counts at all levels of severity. The results lend themselves to several recommendations for highway safety treatments and design policies. For example, wide lanes and shoulders are key for reducing crash frequencies, as are longer vertical curves.

Bayesian Multivariate Poisson Regression for Models of Injury Count, by Severity

In practice, crash and injury counts are modeled by using a single equation or a series of independently specified equations, which may neglect shared information in unobserved error terms, reduce efficiency in parameter estimates, and lead to biases in sample databases. This paper offers a multivariate Poisson specification that simultaneously models injuries by severity. Parameter estimation is performed within the Bayesian paradigm with a Gibbs sampler for crashes on Washington State highways. Parameter estimates and goodness-of-fit measures are compared with a series of independent Poisson equations, and a cost-benefit analysis of a 10-mph speed limit change is provided as an example application.

Self-Selection in Home Choice: Use of Treatment Effects in Evaluating Relationship Between Built Environment and Travel Behavior

The role of self-selection in shaping travel patterns, by affecting ones home location choice, is a critical issue. Developers, planners, and policy makers regularly debate to what extent the built environment and land use patterns can alleviate roadway congestion, greenhouse gas emissions, and myriad other urban problems. A study illustrates the use of Heckmans latent index model to ascertain travel impacts of neighborhood type in Austin, Texas. Under this approach, self-selection is formulated as sample selection bias in receiving a treatment. Here, treatment is defined to be ones residence in a suburban or rural zone, rather than Austins central business district (CBD) and nearby urban zones. This approach of treatment or no-treatment is a meaningful advance in models of self-selection effects and requires estimation of three straightforward models. Depending on model specification used, results suggest that at least half (58% to 90%) of the differences in vehicle miles traveled observed between similar households living in CBD or urban versus rural or suburban neighborhoods of Austin is due to the location or treatment itself, whereas self-selection of such treatment (by households that wish to meet special travel needs or preferences) accounts for the remainder.

Accessibility Indices: Connection to Residential Land Prices and Location Choices

Specifications of accessibility indices range from simple minimum-travel-time indices, to measures of cumulative opportunities within specified distance or time thresholds, to maximum utility measures. Models are presented that relate a variety of general accessibility indices for the Dallas–Fort Worth region of Texas to property valuations for single-family dwelling units and commercial units, and to household residential location choices. Hedonic models are used to assess how important access is to property valuations, while controlling for improvement attributes and parcel size. Multinomial logit models are used to derive logsum measures of accessibility as well as to assess the effect of access on location choices, while controlling for household demographics. Three functional specifications of access measures were used. Job accessibility (a proxy for work and other opportunities) was estimated to affect residential land values positively in statistically and economically significant ways, suggesting—as hypothesized here—that land rents track property owners’ assessments of accessibility, whereas other common accessibility measures do not perform as well. After controlling for this measure, access to park space (proxy for availability of outdoor recreational activities) and access to retail jobs (proxy for shopping opportunities) were not valued in the land market. Distances to regional central business districts and household heads’ workplace locations played important roles in location predictions, often in the presence of the more general access measures. Residential location choice model results suggested which indices are better measures of accessibility. Different functional specifications appeared useful here. Cumulative opportunities access measures were most helpful in predicting residence location, but differences in predictive power were relatively small.

A model for time- and budget-constrained activity demand analysis

This paper describes and demonstrates the derivation of a system of demands for activity participation by applying microeconomic theory in a time-price setting. Both time and money constraints are incorporated explicitly and the integer nature of demand observations is accommodated via a multivariate negative binomial stochastic specification. A model based on this theory is calibrated, providing estimates of income, discretionary-time, and cross-time elasticities. The results reject a constant travel time hypothesis but do not reject a hypothesis of no income effects on total activity participation.

Use of Heteroscedastic Ordered Logit Model to Study Severity of Occupant Injury: Distinguishing Effects of Vehicle Weight and Type

A heteroscedastic ordered logit model was used to study the effects of various vehicle, environmental, roadway, and occupant characteristics on the severity of injuries sustained by vehicle occupants, conditional on the crash occurrence. As expected, the models found that heavier vehicles increased both a vehicles crashworthiness and driver aggressiveness toward others. The models also found that if all passenger vehicles were to become 1,000 lb heavier, crash injury outcomes would not change dramatically. However, if all passenger cars were to become light-duty trucks (i.e., minivans, pickups, and sport utility vehicles) of the same weight, incapacitating injuries and fatalities were predicted to rise by 26% and 64%, respectively. Beyond weight and vehicle type, many other factors were controlled for as well. For example, older occupants and female occupants were more likely to experience injury and death, particularly when navigating curved roadway sections with higher speed limits.

Safety effects of speed limit changes : Use of panel models, including speed, use, and design variables

This work estimates the total safety effects of speed limit changes on high-speed roadways by using traffic detector data and Highway Safety Information System data from 1993 to 1996. To gauge the total effects, the study applies a sequential modeling approach: average speed and speed variance models are first estimated on the basis of roadway design, use, and speed limit information. Then, crash counts (of varying severity) are estimated on the basis of the speed estimates, design, and use variables. The 4 years of data come from 63,937 homogeneous roadway segments along seven Interstates and 143 state highways in Washington State. A random effects negative binomial model was selected among several alternative panel and nonpanel models for count data. Results indicate that the average road segment in the data set can be expected to exhibit lower nonfatal crash rates up to a 55 mph (88 km/h) speed limit. In contrast, fatality rates appear unresponsive to speed limit changes. Fatal and nonfatal rates fall for design reasons, including wider shoulders and more gradual curves, which appear to be key design variables. However, fatal and nonfatal rates move differently when traffic levels rise, with nonfatal rates remaining unchanged and fatal rates falling.