Garbage In…Garbage Out? How Much Do Inaccurate Inputs Impact Transit Demand Forecasts?
Corresponding Author: David Schmitt, CTG
Presented By: David Schmitt, CTG
Abstract
Accurate forecasts of benefits and costs from planned transportation investments are required for good decision making. FTA, Flyvbjerg, Schmitt and others have analyzed the demand forecasts of large-scale transit projects and noted large inaccuracies in the demand forecasts. In 2013-2016, utilizing his personally developed Transit Forecasting Accuracy Database, Schmitt found that accurate demand forecasts are not being provided to decision-makers or transit operators. One reason for the inaccuracy of the demand forecasts was significant optimism bias he found in the project assumptions and exogenous forecasts.
In this research effort, the author expands on this finding to see if the errors in the forecasting inputs (e.g., demographics, project frequency, project end-to-end travel time, etc.) fully resolve the errors in the project demand forecasts. In other words, these tests help determine whether project demand inaccuracies are due mostly, entirely, or modestly to the inputs that forecasters are given.
This analysis is similar to one performed in the late 1980s in the Pickrell Report (Section 2.3.4 Overall Assessment of Input Errors), which used elasticities to adjust the demand forecasts. This method is the most appropriate given the sparseness of recorded information on the forecast inputs. The analysis will assess the impact of up to 10 project assumptions, including project service levels, population and employment estimates, economic conditions and auto fuel prices, for 61 large-scale transit projects.
The analysis was conducted and has the following observations:
• Correcting the project assumptions and exogenous forecasts improves the accuracy of the demand forecasts by 16%;
• Inaccurate project assumptions and exogenous forecasts are responsible for 25-30% of the demand inaccuracy;
• Several projects that originally had accurate demand forecasts, produce inaccurate forecasts when their assumptions are corrected;
• Inaccurate population and employment estimates had the largest impacts on demand accuracy; and
• There are reasons beyond inaccurate inputs for inaccurate demand forecasts.