A Comparison of Beta Regression and Copula Regression for Partial Lapse Rate Estimate

In actuarial analysis, it is very useful to analyze the behavior of an interval-bounded random variable, as a percentage, a proportion, or a fraction, conditioned to other explanatory variables. For this kind of variables, considering the presence of bounds, in general in (0,1), the estimate of the conditional mean and/or conditional quantiles is more trivial than other continuous or discrete variables. This work aims to show the application of a copula-based regression model on the study of the percentage lapsed in partially lapsed life policies, demonstrating how this approach can be an effective and powerful tool compared to an alternative standard regression models like Beta Regression.