Puget Sound Transportation Panel Research Articles

Analysis of the Role of Traveler Attitudes and Perceptions in Explaining Mode-Choice Behavior

An analysis was carried out using the 1991 wave of the Puget Sound Transportation Panel data set to determine the role played by attitudinal and preference variables in explaining commuter mode-choice behavior. Different modal market segments were compared to determine the extent to which attitudes and preferences differ across mode choices. A factor analysis was performed on the sample to identify a few distinct factors that would summarize the multitude of attitudinal variables present in the data set. Multinomial logit models of mode choice were estimated using different utility specifications. Three types of models were estimated: one that included only demographic variables, another that included only attitudinal factors, and another that included both demographic and attitudinal variables. Likelihood ratio tests were applied to assess the significance of the contribution of different types of variables in explaining mode-choice behavior. Results show that demographic variables and attitudinal variables are extremely important in explaining mode-choice behavior. More noteworthy, however, is the finding that the contribution of attitudinal factors is greater than that of demographic variables, thus emphasizing the need for greater consideration of attitudinal and preference variables in travel-demand-modeling applications.

Application of Poisson Regression Models to Activity Frequency Analysis and Prediction

The Poisson regression model and its variants are used to estimate individuals’ daily activity frequencies by activity type using the first 4 years of the Puget Sound Transportation Panel. The estimated model coefficients are applied to the 5th-year observed data to predict daily activity frequencies by activity type. Forecasting accuracy is measured with indicators of deviation between observed and predicted values. Examples of model estimates and their forecasting performance are provided. Comparisons between the observed and predicted 5th-year data show the predictions to be fairly accurate. Subsistence activity is the most accurate among all types, followed by trip chains, out-of-home leisure, and maintenance activities. The analysis also indicates that different theoretical distributions should be used for different dependent variables.

Travel trends using the puget sound panel survey: a generalized estimating equations approach

This paper examines longitudinal mode use trends using four waves of the Puget Sound Transportation Panel. The analysis is conducted using generalized estimating equations for model estimation. In addition to examining mode use frequencies over time, we also consider mode use trends conditioning on household income and lifecycle stage. As expected, results indicate an overall increase in the number of worktrips made between 1989 and 1993 and these trips were marked by increasing use of single occupancy vehicles. The full parameters of the model were also to estimate the rate of increase in terms of percentage increase and their confidence intervals. Results indicate that the mean number of worktrips made by driving alone significantly increased from wave 1 to wave 4; with a 95% C.I. the rate of percent increase was estimated between 8.2 and 24.5%. The ranges for rates of change in high occupancy vehicle modes and non-motorized worktrip frequencies overlap with the range for single occupancy vehicle rate of change, and thus, it cannot be said that rate of change for the high occupancy modes was significantly different from the single occupancy rates of change. The rate of change in the mean frequency for the high occupancy–transit mode is not only below the range for single occupancy vehicle trips but also suggests, with 95% confidence, the rate of percent decrease was between 2.88 and 44.0%.

Systematic self-selection and sample weight creation in panel surveys: The Puget Sound Transportation Panel case

The Puget Sound Transportation Panel (PSTP) is a survey that contains sampling stages which lessen the sample's representativeness of the Puget Sound region. This paper presents an analysis of survey participation and a method for sample weight creation when systematic self-selection in survey participation can be detected and accounted for. The method uses probability of participation models for selected stages of the survey to first identify determinants of participation and then create sample weights. A variety of weight schemes are thus derived and used to correct for systematic self-selection in panel participation. The performance of these sample weights is tested using as a benchmark sample the Public Use Microdata Sample (PUMS) from the census of 1990 in the U.S.

A dynamic analysis of person and household activity and travel patterns using data from the first two waves in the Puget Sound Transportation Panel

Daily activity participation and travel patterns are examined using data from the Puget Sound Transportation Panel (PSTP), which contains two-pairs of daily travel diary information (wave 1 in 1989 and wave 2 in 1990). Summary data of the travel diaries at the person and household levels are obtained using cluster analysis. At the person level, four clusters are found reasonable for both activity and travel. The four-cluster solutions indicate substantial day-to-day variation in activity participation and similarity in travel behavior over time. Temporal changes are analyzed using contingency table methods and log-linear models. The analysis reveals that activity participation and mobility present many regularities over time. Transitions within each wave show strong dependence between two days for both activity and travel, with higher dependency for the travel patterns than for the activity patterns. These are consistent when the analysis focuses on a longer period such as a year. Temporal dependencies appear to be stronger in the household-based than the person-based analysis. A hierarchical structure is also found in the relationship between activity and travel clusters. The link between activity and travel is much stronger within a day, weaker from one day to another, and the least strong from one year to the next. Important irregularities, however, are found and may be due to scheduling time-frames adopted by the respondents that are only partially captured in the data used.

Transit and Carpool Commuting and Household Vehicle Trip Making: Panel Data Analysis

Most of the evaluation of commute trip reduction programs centers on the effect on reducing driveway trips at the work site. Little has been done to explore the effects of such programs on overall trip making. This work is meant to help understand this relationship by determining the association between transit and carpool commuting and overall household vehicle trip making. Household activity models have indicated that an individual's travel decisions influence (and are influenced by) the decisions of other household members. It is thus postulated that one household member's choice of transit or carpool for the morning trip to work will affect the mode choices of the other household members. In addition, the choice of mode to work will affect the individual's mode choices for trips other than the morning trip to work. The analysis uses the first four waves of the Puget Sound Transportation Panel survey to construct linear econometric longitudinal trip generation models (random effects models). The results of the modeling indicate that one household member's mode choice to work does affect overall household vehicle travel. Compared with driving alone to work, if one household member chooses transit to work, overall household vehicle trips are reduced (on average). Carpooling by one household member is associated with little change in overall household travel, with average daily household vehicle trips ranging between one trip fewer and one trip greater than households where each driver in the household drives to work in a separate, single-occupant vehicle. The analysis strengthens the associations to imply causal links. The model results suggest that an improved understanding of interactions among household members is necessary to define strategies in support of commute trip reduction policies to better meet air quality and congestion-management goals.

Sample Weights for Puget Sound Transportation Panel Using Stratification Anchors in Public Use Microdata Sample and Probabilistic Models for Self-Selection

The repeated measurement of the same units over time (panel survey) contains many sampling stages such as initial response, sample stratification used as a tool to decrease costs, and panel participation. In all stages, selective (systematic) survey participation needs to be accounted for if panel data are to be used as a tracking device of the population under study. The Puget Sound Transportation Panel (PSTP) offers a unique opportunity to assess the performance of weighting methods because benchmark data, such as the Public Use MICRO DATA Sample data set, are also available. The PSTP weighting scheme contains sequential weighting steps for exogenous and endogenous sampling stratification, missing data, prewave self-selection, and multiwave panel attrition. Statistical comparisons show that weights based on probabilistic models of self-selection succeed in eliminating part of the bias introduced by selective participation.

Sample Weights for Puget Sound Transportation Panel Using Stratification Anchors in Public Use Microdata Sample and Probabilistic Models for Self-Selection

The repeated measurement of the same units over time (panel survey) contains many sampling stages such as initial response, sample stratification used as a tool to decrease costs, and panel participation. In all stages, selective (systematic) survey participation needs to be accounted for if panel data are to be used as a tracking device of the population under study. The Puget Sound Transportation Panel (PSTP) offers a unique opportunity to assess the performance of weighting methods because benchmark data, such as the Public Use MICRO DATA Sample data set, are also available. The PSTP weighting scheme contains sequential weighting steps for exogenous and endogenous sampling stratification, missing data, prewave self-selection, and multiwave panel attrition. Statistical comparisons show that weights based on probabilistic models of self-selection succeed in eliminating part of the bias introduced by selective participation.

Multivariate Marginal Frequency Analysis of Activity and Travel Patterns in First Four Waves of Puget Sound Transportation Panel

An analysis of activity and travel patterns that explicitly accounts for transportation level of service via composite accessibility measures, land use type, and density around residence and workplace of survey participants in the Seattle region is presented. Using the Puget Sound Transportation Panel (PSTP) data, individuals' activity and travel patterns are first grouped into a few relatively homogeneous behavioral groups using cluster analysis. This is done to reduce the great diversity in individuals' behavior into a few representative patterns of behavior using activity and travel indicators simultaneously. Taking the contextual analysis approach, four-level multilevel models that include temporal, spatial, household, and person effects are constructed. The derived accessibility measures of individuals' residence and workplace are then included in the four-level multilevel pattern selection models to study the relationship between individuals' characteristics and their activity and travel pattern choices. The analysis is done by separating employed from unemployed persons and provides evidence that level of service and land use are strong determinants of activity participation and trip making. Their effect, however, depends on the person's employment status.

Multivariate Marginal Frequency Analysis of Activity and Travel Patterns in First Four Waves of Puget Sound Transportation Panel

An analysis of activity and travel patterns that explicitly accounts for transportation level of service via composite accessibility measures, land use type, and density around residence and workplace of survey participants in the Seattle region is presented. Using the Puget Sound Transportation Panel (PSTP) data, individuals' activity and travel patterns are first grouped into a few relatively homogeneous behavioral groups using cluster analysis. This is done to reduce the great diversity in individuals' behavior into a few representative patterns of behavior using activity and travel indicators simultaneously. Taking the contextual analysis approach, four-level multilevel models that include temporal, spatial, household, and person effects are constructed. The derived accessibility measures of individuals' residence and workplace are then included in the four-level multilevel pattern selection models to study the relationship between individuals' characteristics and their activity and travel pattern choices. The analysis is done by separating employed from unemployed persons and provides evidence that level of service and land use are strong determinants of activity participation and trip making. Their effect, however, depends on the person's employment status.

The puget sound transportation panel after two waves

Begun in 1989, the Puget Sound Transportation Panel is the first general-purpose travel panel survey in an urban area in the United States. The overall survey sample was stratified by county of residence and by usual mode of travel to work. This paper reports descriptive results from the first two waves of survey data. Panel attrition between these waves was about 19 percent of all households. There was some demographic bias in panel attrition. The paper documents changes among the retained households in their demographics, their residence and work locations, their trip making, their work trip lengths, and their travel mode to work. Residential moves were numerous among young adults and young families, more outward from the city core than inward, and correlated with changes in work trip length. Work locations changed for 20 percent of continuing workers, with a tendency to increase trips and work trip length, and to change travel mode after a relocation. For all workers, there were clear shifts from transit to drive alone, from carpool to drive alone, and from drive alone to carpool as well. The PSTP continues, with a modified third wave in 1991, and a fourth wave scheduled in 1992.