In the early stages of the COVID-19 pandemic, cities across the globe converted street space to non-automobile uses. This project studies four of these slow street programs in the U.S.: in Los Angeles, Portland, Oakland, and San Francisco. In each city, the slow streets (implemented in late spring to early fall 2020) are used as a treatment and compared to non-implemented control groups. The dependent variable is counts of dockless scooter trips passing a mid-block screenline for time periods both before and after slow street implementation. Those dockless scooter counts were obtained from historical data provided by Lime, a dockless scooter provider in each of the study cities. Two methodological approaches were used: differences-in-differences (DID) and panel regression analysis with block fixed effects. For the DID analysis, the researchers used networks of candidate slow streets that were not implemented as the control group. Such control networks were available in Los Angeles, Oakland, and San Francisco. For the panel analysis, they used slow street segments implemented later in the study period as control segments for earlier implemented slow street segments, including fixed effects for blocks and for time periods in the panel regressions. The findings show statistically significant associations between increased dockless scooter trips and slow street implementation in each study city, using both DID and panel analyses. The associations are robust to different specifications. The authors calculate the magnitude of the slow street treatment effect by dividing the estimated treatment effect by a 2019 baseline of dockless trip counts. In the DID analysis, they find that slow street implementation increased dockless scooter trip counts from 22.16% to 74.5%, relative to a 2019 (before slow streets) baseline. In the panel analysis, the increase in dockless trip counts on slow streets ranged from 10.77% to 16.75%, relative to a 2019 baseline.
