Binary Response and Logistic Regression in Recent Accounting Research Publications : A Methodological Note

In this research note, we review 35 recent articles in leading accounting journals that performed a logistic regression or a standard linear regression analysis based on OLS for a binary dependent variable. Our review shows that many of these articles have ambiguities and even outright errors in the presentation of the binary regression model. We discuss why various presentations of the model are incorrect. We also remind readers of the weaknesses of using ordinary linear regression for a binary response variable. We explain that incorrect presentations of the model, even in conjunction with a correct analysis, may lead to a serious misinterpretation of research findings by authors and readers alike. Two articles are critiqued to demonstrate the reporting problems. We close our note with comments on some alternative approaches for the analysis of binary response data. Our note is a call for an improvement in model presentation and related statistical analysis in the accounting field