Published in 2008
Socioeconomic forecasts are the foundation for long range travel demand modeling, projecting variables such as population, households, employment, and vehicle ownership. In Virginia, metropolitan planning organizations (MPOs) develop socioeconomic forecasts for a given horizon year at a traffic analysis zone level., and the Virginia Department of Transportation (VDOT) uses these forecasts as input to the four-step travel demand model system. This report identifies the socioeconomic forecasting practices currently used by four medium-sized Virginia MPOs, computes the accuracy of socioeconomic forecasts generated by one such MPO, and recommends practices for improving such forecasts. This research found that medium-sized Virginia MPOs are using similar techniques to forecast socioeconomic variables. These techniques are to (1) identify jurisdictional population control totals based on U.S. Census and Virginia Employment Commission data; (2) disaggregate population projections to the zonal level based on comprehensive plans, local knowledge, and historic trends; (3) apply historic ratios of households to population and autos to population to forecast households and autos; (4) use historic trends and local expertise to determine future employment; and (5) revise zone projections through coordination with local jurisdictions. Using a forecast that was developed for the Lynchburg region in 1981 with a horizon year of 2000, the study area percent error was computed as the difference between the forecasted and observed values for the entire study area. While the study area percent error for number of vehicles and employment was less than 10%, the study area percent errors for population and households were 48% and 14%, respectively. Two adjacent zones accounted for approximately 80% of the population error and 90% of the household error, and the error resulted because anticipated development therein did not materialize. The zone percent error is the average difference between forecasted and observed values for each zone. Population, households, and vehicles had similar zone percent errors of 61%, 65%, and 54% respectively, while the employment zone percent error was 154%. Four recommendations for improving forecasts are given. First, localities should provide updates to MPO or PDC staff as changes in land development occur, and such staff should perform socioeconomic forecasts more frequently than the current practice of every five years. Second, MPOs should consider providing two sets of socioeconomic variables for the travel demand model: (1) the baseline forecast (which is the MPO's best estimate) and (2) the baseline forecast modified by some percentage that accounts for the possibility of forecast error. Third, best forecasting practices should be shared among MPOs through a user's group, a workshop, or some other forum where MPO and PDC staff will be in attendance. Fourth, VDOT should communicate these recommendations to MPO staff who are responsible for completing socioeconomic forecasts.
Danielle R. McCray, Lester A. Hoel, Ph.D.
Last updated: November 23, 2023