Are public transit investments based on accurate forecasts? An analysis of the improving trend of transit ridership forecasts in the United States

Historically, forecasts of travel demand on public transit infrastructures have been found to be optimistically biased. However, there has been a lack of data available for statistically significant analysis of factors affecting the accuracy. This paper analyzes the overall trend of transit ridership forecast accuracy in the US and contextualizes it with ridership trends based on the largest yet database of 164 large-scale transit infrastructure projects in the US. We find that transit ridership is about 24.6 % lower than forecast on average with about 70 % of the projects over-predicting ridership. Forecast accuracy varies by mode, service area characteristics, familiarity with transit, ramp up period, and time span. The accuracy has been getting better over the years, particularly after 2000 with the introduction of new analytical and evaluation tools as part of the Capital Investment Grants program. Projects that have been forecasted since 2000 have average ridership about 22 % lower than forecast, compared to about 52 % lower from pre-2000. The steadily improving accuracy, however, is offset by the unexpected decline in transit ridership since 2012. Advent of ride-hailing services and improved socio-economic trends that support auto-oriented cities have prompted this decline in ridership and have affected their forecasts as well. Despite the improving trend, we find that there remains substantial deviation in the outcomes from their forecasts. This points to the need of better scrutiny of model inputs and specifications and how they interact with the built environment to unearth the underlying reasons for inaccuracy. Planners and policymakers may make use of our results to advocate for considering the uncertainty around forecasts for any project and funding decision.