Nearly four decades later: Is accounting still useful in predicting business failure? Probabilistic business failure prediction on Nordic manufacturing firms

Abstract: With the aim of examining whether accounting-based failure prediction models still can be used effectively, this thesis investigates how well prediction models classify a modern sample of firms applying IFRS. In doing so, this thesis anchors on the Skogsvik (1987) study and performs a three-step plan. The first step tests the robustness of the original Skogsvik (1987) model. The second step recalibrates the coefficients in the Skogsvik model to assess how the relative importance of each financial ratio has changed. In step three an entirely new failure prediction model, the Haglund and Olufsen (2021) model, is created to see which financial ratios best predict firm failure in a modern context. The sample of firms used to generate the model consists of Nordic manufacturing firms. Probit analysis was performed from one year before failure up to five years before failure on a sample of 388 survivor firms and 52 failure firms in the time period 2005-2021. The results indicate that the original Skogsvik (1987) model works well in the shorter prediction span, while one must recalibrate or create a new model in order to get a high predictability in the longer prediction span. We conclude that accounting information still can be used effectively in predicting business failure under IFRS, as the error rates in the models generated are similar to those in Skogsvik (1987). However, the compositions of the models are to a large extent different. Therefore, we conclude that business failure prediction models must mirror the business environment of the firms they examine.