Regression analysis of the relationship between COVID-19 pandemic and the change in unemployment in EU27 countries

This thesis aims at investigating the statistical relationships between the change in unemployment from 2019 to 2020 and COVID-19 severity across the EU27 countries using regression analysis. It thereby takes into account the other factors that may have influenced the change in unemployment, such as government expenditure on health, the EU-issued SURE package, and the stringency index. The theoretical part of the thesis explains in detail the concepts behind the methodology implemented in the thesis, such as linear regression analysis, least squares estimations, LASSO regression, principal component analysis, and inferential regression analysis. In the empirical part, the variables used in the regression models are analyzed, and the validity of the classical linear regression assumptions is assessed. Then, using a descriptive approach, the partial and paired linear relationships between the variables are investigated through the analysis of the outputs of the built regression models. The inferential analysis revolves around the randomized permutation of statistical correlation tests. The predictive capabilities of the regression models were tested by the implementation of various cross-validation techniques. The thesis concludes with a discussion of the results derived from the research and attempts to answer the questions raised in the section of the research hypothesis. It found that most of the statistical relationships between the response variable and the explanatory variables satisfy the initial expectations, except the main one. The results generated by the three regression models suggest that the partial linear relationship between the change in unemployment and excess mortality is negative, holding the rest of the variables fixed.