Aggregate level transit ridership forecasting models often are based on time series data, with potential serial autocorrelation properties that can bias parameter estimates upward and error estimates downward, and skew forecasting confidence intervals and their resulting interpretation significantly. A combined time series and cross-sectional regression model of transit ridership is developed that incorporates temporal variations as well as supply, demand, and pricingcharacteristics of the market for transit services in orange county, california, between 1973 and 1989. It was found that orange countytransit ridership exhibits significant serial and seasonal fluctuations, which were captured in the model. The temporary and lingering effects of incidents were also tested. The 1979 oil shortage was shown to have a large positive impact on transit ridership, which dwindled quite rapidly once the oil shortage ended. A work stoppage of 6 weeks' duration in 1981 had a large negative impact on transit ridership, which dwindled only slowly. A shorter work stoppage in 1986, during which limited service was provided by transit agency administrative personnel, had a much smaller negative impact than the 1981 work stoppage, which dwindled much more rapidly. Transit fare and gasoline pricing variables were found to have no significant effect on transit ridership in the preferred temporally based model. Transit fares did not increase much in real terms over the period covered, anddid not reflect variations in transit service provided, being predicated on a simple county-wide flat fare basis. Over 70% of all orange county transit riders were captive riders in 1987, having no car available to them for commuting or other travel purposes, making the price of gasoline basically irrevelant to the majority of such transit riders in the shorter term. This paper appears in transportation research record no. 1297, Public transit research: management and planning 1991.
Abstract