Evaluation of pollution containment policies in the US and the role of machine learning algorithms

The aim of this study is to analyse policy actions and institutional changes in local governance structures as determinants of air pollutant reductions in US urban areas. First, we construct a dataset on traffic-related air pollution and socio-economic characteristics across urbanized areas of the US. Some of these data are available through Google Earth engine, others are instead provided by institutional sources. In general, raw data come from application of machine learning techniques to either satellite images or monitoring station records and are available at different temporal and spatial resolutions. Then we adopt regression discontinuity design techniques for the evaluations of pollution reduction policies, exploiting the designation of US Transport Management Areas as a quasi-experimental framework.