Modeling executives' remunerations

In a company's life, the executive director's remuneration package is an important matter for the board. We investigate the topic of executives' compensation from two angles. First, we analyse the differences in pay level in Europe and how the governance structure of the company affects those differences. This part can be seen as a extension of the work conducted by Fernandes et. al [2012], who focused on the comparison between the United States and the rest of the world. Second, as there exists a whole industry for remuneration consulting services, we examine the possible algorithms that could be used and the steps of progress necessary to create a robot-advisor to replace the manual process of compensation advice. For the analysis of European CEO's remuneration, we have constructed different regressions fitted using the Ordinary Least Squares, initially on a data set comparing the companies based in the United Kingdom and quoted in 2016, and then by changing the considered country or year. Finally we have used more complex methods, the Lasso and the Elastic net, and compared these results with the initial ones obtained with the Ordinary Least Squares. Our analysis shows that, regardless of the country, the year or the method, the country matters in the total remuneration given to an executive director. Moreover, the Ordinary Least Squares is a sufficiently sophisticated method for the problem at hand. Regarding the development of a robot-advisor, we have compared a decision tree regressor and a k nearest neighbours algorithms. The first method enables to retrieve the features importance, which is important in order to understand the prediction given by the model. The second approach is very similar to the manual process of the consultants. We have found that the k nearest neighbours gives the best predictions. Therefore, we have discussed possible improvements that could be made on this algorithm.