A comparison of time series methods for post-COVID transit ridership forecasting

Transit agencies conduct system-level ridership forecasting for planning, budgeting, and other administrative purposes. However, the COVID-19 pandemic introduced substantial changes in transit ridership levels and seasonal patterns, which has impacted the performance of ridership forecasting. Although time series methods are commonly used for forecasting transportation demand, they have received limited use in practice for public transit ridership forecasting. This study compares the performance of seven time series forecasting methods for predicting system-wide, monthly transit ridership for heavy rail agencies in the continental United States. The forecasting methods are: ETS, ARIMA, STL with ETS, STL with ARIMA, TBATS, a neural network, and a hybrid model. Ridership was forecasted for pre- and post-COVID periods (pre- and post- March 2020), as well as for the full series (January 2002 to December 2023). The MAPE and MASE were used to compare forecast performance. Using the pre-COVID period, 43% of the models produced a MAPE below 5% and 82% produced a MAPE below 10%. Using the full-series and post-COVID periods, only about 10% of the models produced a MAPE below 5% and half produced a MAPE below 10%. The classical and hybrid methods outperformed the other models using the full series period, and the TBATS, neural network, and hybrid methods outperformed the other methods using the post-COVID period. The findings suggest that even a few years into the post-COVID era, patterns that were typical of heavy rail ridership before the pandemic have not returned at most agencies in the United States, posing challenges to forecasting post-COVID ridership.