Causality vs. Correlation: Rethinking Research Design in the Case of Pedestrian Environments and Walking

Abstract

This paper investigates the causal effect of pedestrian environments on walking behavior and focuses on the issue of research design. The paper differentiates between two types of research designs: treatment-based and traveler-based. The first approach emphasizes the variation of the treatment (pedestrian environments), and generally compares distinct neighborhoods, such as urban vs. suburban or transit-oriented vs. auto-dependent. The second approach emphasizes the homogeneity of subject (pedestrians), and aims at the same pedestrian under different environments normally due to home relocation, or the improvement of pedestrian environments. The first approach can easily identify a correlation between the pedestrian environment and walking, but proving it causal is a challenge. The second approach may not even find a correlation, but if it does, such a correlation is more likely to be causal. Which approach is better depends on whether the first approach can effectively control for the unobservable personal heterogeneity, and whether the second approach can find sufficient variation in the pedestrian environments experienced, arguably, by the same person.

Most studies used the first approach but produced inconsistent results in terms of whether self-selection exists and if it does, whether it nullifies a causal relationship. This paper supports the second approach but argues that the few existing studies failed to capture sufficient variation of pedestrian environments in their research design. The paper then follows a traveler-based research design, and proposes a new method based on pedestrians’ path choice. By comparing the preference from the same pedestrian towards multiple walking paths with different pedestrian environments, this research is able to control the personal heterogeneity while still retain a sufficient variation in the pedestrian environments, thus represents a quasi-experimental design. It is able to do so because the path-based measure is sensitive enough to capture even minor differences in the pedestrian environment. More importantly, path choice is less likely to correlate with job and housing location choices, and therefore largely avoids the self-selection problem.

In the empirical analysis, the paper targets subway commuters’ egress path choice from a station to their workplaces in downtown Boston. The results confirm a causal relationship and suggest that the pedestrian environment can significantly affect a person’s walking experience. The perceived utility change of walking, due to the sidewalk amenities, averages between 21 and 31 percent. In several street segments in downtown Boston, walking is actually perceived to have a positive utility instead of as a derived demand. Methodological issues regarding the method and generalizability of the findings are also discussed.

Keywords: causality, research design, path choice, pedestrian environment, walking, Boston