Spatio-temporal Event Studies for Air Quality Assessment under Cross-sectional Dependence

Abstract: Event Studies (ES) are statistical tools that assess whether a particular event of interest has caused changes in the level of one or more relevant time series. We are interested in ES applied to multivariate time series characterized by high spatial (cross-sectional) and temporal dependence. We pursue two goals. First, we propose to extend the existing taxonomy on ES, mainly deriving from the financial field, by generalizing the underlying statistical concepts and then adapting them to the time series analysis of airborne pollutant concentrations. Second, we address the spatial cross-sectional dependence by adopting a twofold adjustment. Initially, we use a linear mixed spatio-temporal regression model (HDGM) to estimate the relationship between the response variable and a set of exogenous factors, while accounting for the spatio-temporal dynamics of the observations. Later, we apply a set of sixteen ES test statistics, both parametric and nonparametric, some of which directly adjusted for cross-sectional dependence. We apply ES to evaluate the impact on NO2 concentrations generated by the lockdown restrictions adopted in the Lombardy region (Italy) during the COVID-19 pandemic in 2020. The HDGM model distinctly reveals the level shift caused by the event of interest, while reducing the volatility and isolating the spatial dependence of the data. Moreover, all the test statistics unanimously suggest that the lockdown restrictions generated significant reductions in the average NO2 concentrations.