Several studies have observed an optimistic bias in cost and ridership forecasts for rail transit projects around the globe, which has led to billions of dollars of public investment in projects that have not performed as promised. This bias has been a major cause of concern for project stakeholders, including the Federal Transit Administration (FTA), which has spent an average of over $3 billion each year over the past two decades on new rail transit projects in the United States through its Capital Investment Grants program, commonly known as New Starts. Partly in response to credibility concerns raised by forecast bias, the FTA has made changes to the New Starts program over the years. However, there has been no research to date that has examined how these changes in the New Starts program have influenced forecast accuracy for rail transit projects that receive funding.

This study addresses that gap in the literature through a mixed-methods approach involving semi-structured interviews with thirteen transit planning and forecasting professionals and a quantitative analysis of 67 completed transit projects to determine whether and to what extent forecast accuracy has changed over time and what changes in federal policy and transit planning practice might explain these changes.

I find that there have been steady improvements over time in the accuracy of ridership forecasts and cost estimates for New Starts projects. The improvement in ridership forecast accuracy can be explained in part by shorter project construction durations and a shift over time in the perceived purpose of forecasting from (1) project promotion to (2) fairness of competition to (3) use in local decision-making. Some of the improvement in cost estimate accuracy can be explained in by changes in project characteristics, particularly a tendency towards more modest projects representing incremental changes to the transit network.

This analysis of forecast bias in transit planning gives us reasons for optimism regarding the future of optimism bias in cost and ridership forecast accuracy, since forecasts appear to be on a long-term trajectory toward more accuracy and less bias.