Financial distress prediction of Turkish manufacturing companies using generalized linear models

The study set out to develop financial distress prediction models based on various combinations of three types of predictor variables namely: Accounting variables (financial ratios), macroeconomic indicators and market variables. Models solely based on accounting variables and on market variables also were developed. In actual fact, the study considered five types of models with one year and two years prior to the financial distress (t-1 and t-2) and also crossover design models were implemented. The study employed one type of Generalized Linear Models known as logistic regression since it is appropriate for binary response variables such as financial status (distressed or non-distressed). Apart from macroeconomic indicators of the Turkish economy which are standard, data used was obtained from companies in the manufacturing industry listed on Istanbul Stock Exchange (Borsa Istanbul). The study covered data spanning from 2009 to 2013. Main findings of the study are as follows: Accounting variables yielded the best prediction model at t-1, suggesting that they are sufficient to predict financial distress at this time lag. Nonetheless, their explanatory power decreases at t-2 implying that they need to be combined with other variables (especially market variables) to get reliable predictions. Macroeconomic variables have relatively good explanatory power, but because of the existence of multicollinearity (considering pairwise and partial correlations) in the sample data, most of these variables could not enter the models. In fact, only one macroeconomic variable (GDP) could enter one of the models (i.e. combined with market variables) at t-1. However, it was found that unlike market variables, GDP is positively related with the probability of financial distress. This finding was explained by the possible effects of macroeconomic shocks and monetary policy in the sector. Finally, crossover designs could not yield better models. Key words: Financial distress, accounting variables (financial ratios), macroeconomic indicators, market variables, Generalized Linear Models (logistic regression).