Transit’s New Normal Needs New Tools

Promising Variables to Enhance Ridership Models

Transit’s New Normal Needs New Tools

Promising Variables to Enhance Ridership Models

Published: May 06, 2024

Overview

  • Ridership was falling in many agencies pre-pandemic and few systems have recovered fully.
  • Factors previously associated with ridership growth like work and telework have shifted markedly.
  • Our research identifies crucial variables like hybrid schedules and factors related to transit hesitancy, necessitating adjustments for enhanced accuracy in ridership models.
The factors behind when and why people use transit to get around have changed dramatically since the spring of 2020. The trends in annual boardings for all the transit agencies that report to the Federal Transit Administration’s National Transit Database are clear: there has been a steep drop in ridership that has not yet returned to pre-pandemic conditions. So which fundamentals have changed?

Source: National Transit Database, Federal Transit Administration, 2023

    Across 2021 and 2022, there was far too much volatility related to how people made travel decisions to even begin to narrow in on what factors might persist into the future. In November 2023, APTA reports that ridership has recovered to above 77% of pre-pandemic levels, and trends are pointing upwards. As we begin 2024, there are new travel surveys, improved travel pattern information from big data sources, and more stable transit ridership data that can provide better insight about the probable factors that will drive transit ridership trends in the near to mid-term. (The future will always have considerable uncertainty – see our TrendLab+ tool for more information about how uncertainty affects travel).

    With this backdrop in mind, our Transit Discipline Group undertook a deep dive into the variables that are included in traditional, pre-pandemic transit ridership tools. Our work included applying a 2022 model by Erhardt et. al., that was designed to evaluate changes in ridership across the country between 2012-2018. This model was selected because it performed well under pre-pandemic conditions and included many variables typical of transit ridership models. However, to understand how things have changed post-pandemic, we also applied the Erhardt model to 2022 conditions (the most recent full year of ridership data from the USDOT).  In the 2022 analysis, we looked particularly for the variables and relationships in those models that may no longer be meaningful today. Below are some key takeaways from our research:

      • When we apply the legacy transit ridership models with today’s data, they do not estimate current transit ridership levels well.
      • Historically, employment was one of the most highly correlated variables with transit ridership, but this correlation is now insignificant.
      • Changes in work from home patters in 2022-2023 only explain about one-fifth to one-third of the decrease in transit ridership, leaving much of the drop in transit ridership unexplained with traditional transit ridership model input data.

    Nearly 70% of the 2018 to 2022 ridership decline on the nation’s bus systems are unexplained by traditional modeling factors:

    Over 30% of the 2018 to 2022 ridership decline on the nation’s rail systems are unexplained by traditional modeling factors:

    Digging a bit deeper, our experts have identified key data and variables that are likely to improve pre-pandemic transit ridership models. In partnership with clients like BART and King County Metro we are testing and applying new variables. The two most promising data enhancements include:

    Localized Remote Access Data

    Current travel survey data about working from home or telecommuting often lacks the granularity that is necessary to greatly improve ridership models. Similar shifts in e-commerce and travel for retail goods and services are also too complex and nuanced for legacy travel surveys. More detail about the number of days or hours people work from home in a given week or month would strengthen the next generation transit ridership models. Fortunately, new data sources and improved travel survey data collection techniques are emerging in this space along with more data about people obtain retail and service goods (traveling themselves, on-demand delivery, and traditional parcel delivery for example). Specific to King County Metro, we worked to refine peak-period transit ridership and park-and-ride forecasts by using regional big data sources to identify the number of downtown commuters are on-site on a typical mid-week day. This reduced the model error by more than 50 percent.

    Transit Hesitancy

    A term we have identified as “transit hesitancy” is related to a shift in customer preferences for transit that appears to be persistent. In both qualitative and quantitative efforts, we have identified factors such as perceived asocial behavior on transit, safety or security concerns at transit stops (with important gender differences), and increased unpredictability of service caused by labor shortages and budget shortfalls as potential factors contributing to transit hesitancy. We applied a transit hesitancy factor to improve BART ridership forecasts at urban locations to account for a 12 percent drop in ridership that could not be accounted for with more traditional forecasting variables.

    What’s Next?

    Knowing that current ridership tools cannot reliably forecast future conditions, we have recalibrated our ridership variables to deemphasize the importance of employment and developed new model inputs that can better account for hybrid work patterns and transit hesitancy.

    As more data emerges on the specific factors that are behind transit hesitancy, we will continue to refine our variables to adapt to a highly fluid market for transit ridership. With more accurate transit forecasts in place, communities can better plan for future transit services, ensure transit is provided to those who rely on it most, and identify strategies to fund this critical service.

     

    Questions? Would you like to provide some insight into our work?
    Please reach out to our Transit Discipline Group members who led this research.

    Contributors

    Jennifer Ziebarth

    Jennifer Ziebarth

    Statistical Expert

    PhD

    Jeremiah LaRose

    Jeremiah LaRose

    Transit Discipline Leader