Identifying the determinants of soccer coach turnover in Italian League Serie A

The aim is to study the causes and consequences of head coaches turnover in the top division of the Italian Soccer League (Serie A) over the period 2010-2019. Since coaches can exit from the club because of management choice or due to their own decision, we employ competing risk models, able to estimate the effects of predictors on the risk of replacement. We gather data on coach, team and club by merging information from various internet sources. Since the set of information collected includes a large number of covariates that can influence the turnover of head coaches, we identify the best set of possible predictors, by applying a non-parametric procedure based on random survival forest, and then estimating the hazard ratios associated with the possible changes by competing risks model. Furthermore, the covariates selected by the random survival forest are compared with those chosen by a traditional variable selection, i.e. the forward selection