How would employees commute today if they had the same characteristics as employees in 1980? – Using entropy balancing to decompose changes in observed commuting mode choice over time in repeated cross-sections

• The novel method decomposes travel behavior changes over time. • 11.9 percentage points increase in commuting car use from 1980 to 2016 (“raw gap”). • Between 3/4 and 7/8 of the “raw gap” explained by contextual factors (e.g., age). • Between 1/8 and 1/4 of the “raw gap” may have been affected by transport policy. In recent decades, societal, economic, and technological developments have altered commuting behavior in many European and North American regions. Understanding the reasons for observed changes in commuting mode choice is crucial to implementing effective transport policy measures. Changes in commuting mode choice stem from either changing observable characteristics (structure of the population) or variables that are often unobserved for the researcher (such as prices or attitudes). Our study contributes to the literature by applying Entropy Balancing for the first time to decompose changes in commuting mode choice over time. We are the first to analyze data from the German Microcensus as repeated cross-sections covering every four years from 1980 to 2020 in transport research. We estimate counterfactual scenarios by comparing the population of 1980 to the balanced populations of consecutive survey waves. The “raw gap” describes changes in commuting mode choice over time in the unbalanced data. For example, the proportion of employees commuting by car increased from 58.0% in 1980 to 69.9% in 2016, resulting in a raw gap of 11.9 percentage points. Controlling for all observable characteristics (e.g., population structure) in the data explains between 3/4 and 7/8 of the raw gap. We find the increasing female labor force participation to have the highest dampening and longer distances between the residential location and the workplace to have the strongest increasing impact on the rise in commuting car share. The unexplained part (between 1/8 and 1/4) of the raw gap is due to unobservable variables (e.g., changes in supply, attitudes or (relative) costs). All these variables can (at least partly) be influenced by transport policy. Therefore, the unexplained part of the raw gap is an indicator of the scope for actions by transport policy. Sensitivity analyses support the validity of the results.