A number of capital structure models presented even in prominent papers are estimated with incorrect estimators

The capital structure is most often measured using a debt ratio, which usually takes values in the interval (0; 1). This makes the linear regression model a linear probability model. A basic shortcoming of such a model is that fitted values for some observations can be less than zero, which is inconsistent with the definition of the debt ratio. In the vast majority of articles, the authors did not pay attention to these defects and if they noticed them, they proposed solutions that cannot be considered fully satisfactory. A simple solution, known for many decades, is the application of the logit transformation of the dependent variable, which ensures that the fitted values of the debt ratio are in the interval (0; 1). The aim of the study is to draw researchers’ attention to the inappropriateness of using linear probability models and to show the advantages of the logit models of debt ratios. The considerations were illustrated by models of debt ratios estimated for the companies listed on the WSE in 1998–2019.