The inclusion of time-varying covariates into survival analysis has led to higher predictions of the time to default in behavioural credit rating scoring models. Having said that, when these time-different covariates are endogenous, There are 2 key challenges: estimation bias on the survival product and insufficient a prediction framework https://lukasufjsu.tribunablog.com/the-ultimate-guide-to-peter-cornwell-head-43259408