Intrastate Travel Behavior Differences: Implications for Statewide Travel Demand Models

INTRASTATE TRAVEL BEHAVIOR DIFFERENCES: IMPLICATIONS FOR STATEWIDE TRAVEL DEMAND MODELS* Stephen E. White INTRODUCTION. Many states are developing statewide travel demand simulation forecasting models. (J) These models require forecasts of household travel behavior that include the average number of daily automobile trips and the average length of trip for households having specified socioeconomic characteristics. A popular method of statewide transportation modeling is one patterned after urban forecasting procedures and includes: (2) 1 ) developing trip generation equations for a base year; 2)developing trip length distribution relationships for a base year; 3)assigning base year trips between traffic zones to a computer simulation network; 4)forecasting trip generations for each traffic zone; 5)calculating the distribution of future zonal trip productions from each traffic zone to all other zones; 6)assigning future trips among traffic zones to a computer simulation network; and 7)comparing forecasted travel assignments with base year assignments . Trip generation equations are often regression equations using data produced from interviews made in a sampling of households. (3) The dependent variable is average daily trips within a specific region or traffic zone, whereas the independent variables may include a variety of household socioeconomic characteristics or local land use types and * I wish to acknowledge Henry Bennett and Mohamed Taqui of the Kentucky Department of Transportation for their valuable comments and assistance in providing the travel information used here. The findings and conclusions of this study do not reflect the official views or policies of the Kentucky Department of Transportation. I also wish to thank Ms. Linda Raborn Mitchell for her cartographic assistance. Dr. White is Assistant Professor of Geography at Kansas State University, Manhattan , KS 66506. Vol. XVIII, No. 1 69 intensities. Forecasts of the independent variables are plugged into the base year equation to provide forecasts of future trips. Because identical equations are used for the base year and for the forecast year, the relationships between household composition and trip making are assumed to be constant over time—an unlikely assumption considering the variability of future gasoline supplies. After trip production forecasts are determined for each traffic zone, they are distributed to other traffic zones by the use of gravity formulations that reflect base year trip length frequency distributions (curves that indicate the percentage of trips that occur within specified distance ranges). Trip length distribution relationships are also assumed to be constant over time. After the number of trips between all paired combinations of traffic zones have been determined, they can be assigned to a computer simulated highway network by means of a minimum time path algorithm . Future highway assignments are compared with base year assignments to measure changes in route volumes so that future highway needs may be evaluated in light of social, economic, and environmental impacts associated with future construction alternatives. Travel demand studies may also be important devices for estimating potential energy consumption based upon combinations of assumptions concerning fuel availability, vehicle mix, and the elasticity of demand for gasoline. (4) PURPOSE. This study seeks to identify the risks inherent in extending urban travel behavior modeling techniques to the statewide scale. Specifically, it suggests that households having similar socioeconomic characteristics may have different trip generation rates and trip lengths if they reside in regions having different relief characteristcs or population densities. If household trip rates vary among state subregions, then a single statewide regression equation will underestimate trip rates in some regions and overestimate them in others. A summary trip length distribution used to calibrate the gravity formulation for distributing trips among traffic zones will underestimate or overestimate trip lengths in different subregions. Trip length frequency distribution data are also used in allocation models to determine the locations of future trips. For example, the Federal Highway Administration has developed a Statewide Activity 70Southeastern Geographer Allocation Model (SAAM) to predict population and other socioeconomic data for traffic zones. (5) SAAM is a descendent of other urban allocation models such as the Projective Land Use Model, designed by Rosenthal, Meredith, and Goldner, and the EMPIRIC Activity Allocation Model, by Peat, Marwick, and Mitchell and Company. (6) SAAM was developed for application to state-sized regions, however , instead of urban areas. Two of SAAM's basic allocation mechanisms...