The objective of this project is to quantify the impact of COVID-19 pandemic on tourist visits (pedestrians + bikes) in regional parks in cities in North America (San Francisco, Seattle, New York, Indianapolis, Charlotte, and Long Beach).
The availability of pedestrian and bicycle count data allows us to estimate the causal impact of COVID-19 on non-motorized travel patterns. To quantify the causal effects of COVID-19, a Bayesian structural time series (BSTS) model is proposed, with the “treatment” date defined as the date on which the national emergency was declared. The model is intended to (1) account for variations in local trends, seasonality and exogeneous covariates before the treatment, (2) make predictions about the counterfactual trends after the treatment, (3) infer the causal effects between observed series and counterfactual series, and (4) evaluate the uncertainty about the causal inference.
The BSTS model is applied to quantify the drops or increases in non-motorized activities based on data collected from 12 pedestrian-bicycle trails in 11 cities in the United States. The model validation demonstrates the reliability of the prediction of counterfactual variables. According to the estimation results, COVID-19 led to losses in non-motorized activities in densely populated cities, but walking and bicycle activities in less densely populated cities increased. In two cities studied, trends in non-motorized activities reversed about 10-20 days after the first confirmed case of COVID-19.
Finally, datasets for this research can be found online. We are currently asking for research-use permission from the client. After the permission, we can upload several estimation results related to it.