AbstractPrevious studies targeting accuracy improvement of default models mainly focused on the choice of the explanatory variables and the statistical approach. We alter the focus to the choice of the dependent variable. We particularly explore whether the common practice (in the literature) of using proxies for default events (bankruptcy or delisting) to increase sample size indeed improves accuracy. We examine four definitions of financial distress and show that each definition carries considerably different characteristics. We discover that rating agencies effort to measure correctly the timing of default is valuable. Our main conclusion is that one cannot improve default prediction by making use of other distress events.