Integrating causal graphs and potential outcomes: theory, applications and a novel method

Covid-19 generated an unprecedented shock on the Italian economy, which severely affected firm performance. This work focuses on estimating the causal effect of implementing home-based working (HBW) after the pandemic outbreak on firms’ expected revenues. The analysis uses a unique firm-level dataset, which captures a rich set of features before and after the spread of the virus. Causal effect estimation is performed implementing an integrated approach that merges Causal Graphs and Potential Outcomes frameworks. At first, the dataset is used to learn a causal diagram that encodes theory-based assumptions and information contained in the data. An adjustment set is then selected by applying the back-door criterion on the obtained graph. Lastly, causal estimates are computed with full matching, using the chosen adjustment set to ensure unconfoundedness. The results confirm the presence of a positive effect of the implementation of HBW on expected revenues. The treatment seems to be particularly effective in providing revenue stability and mitigating of losses. The results are consistent with the fact that HBW equips firms with greater flexibility and helps contain productivity decreases in Covid times.