Predicting financial distress in publicly-traded companies

Abstract

Several models for forecasting bankruptcy have been developed over the years, one of the reasons for which is the important part it plays in decision-making. However, forecasting a company's bankruptcy leaves a very short time for stakeholders to change the situation. It is in this context that this paper arises in order to develop a model for predicting financial distress, which is identified as a step prior to bankruptcy. The predictive model uses the logistic regression technique with panel data and a sample of Brazilian publicly-traded companies with shares listed on the São Paulo Stock, Commodities, and Futures Exchange between 2001 and 2014. As well as financial variables, the final model includes market expectations (macroeconomic) and sector variables. These variables are statistically tested and the hypothesis is confirmed that they improve the accuracy of the model. The research identified the existence of financial distress in 96% of the companies that went bankrupt. In addition, the relationship between the phenomena of bankruptcy and financial distress is verified, using financial and macroeconomic explanatory variables. The results demonstrate that most (83%) of the explanatory variables in the model for predicting bankruptcy are also present in the model for predicting the phenomenon of financial distress. The expected gross domestic product variables and the quick ratio, asset turnover, and net equity over total liabilities financial variables are statistically significant in predicting both phenomena. With this evidence, the study suggests the use of the concept of financial distress as a stage prior to bankruptcy and provides a model for predicting financial distress with 89% accuracy when applied to publicly-traded companies in Brazil in the period examined.

Financial, Macroeconomic and Corporative Management Indicators to Predict the Insolvency of B3 Companies

“…When it comes to the financial indicators, many of the results and expected signs corroborate with the findings by Rezende et al (2017), Scalzer et al (2015), Soares and Rebouças (2014) and Stüpp (2015). However, it is worth reflecting on the two results for which the expected signs were not confirmed: a) Concerning outstanding debt (X05) not presenting the expected sing, looking from a wider perspective, it is possible to justify the presentation of the negative sign based on some alternatives: debt level might have been well handled (debt management) by the corporations of the sample; drop on the rate of taking new loans and financings by the corporations due to the negative credit and/or difficulties imposed by the financial institutions; asset reduction to pay off debt that had already been contracted by the corporation implicating in a less valuable asset, in case there are no new investments.…”

“…Mendes (2014) was not able to validate all macroeconomic variables that he used. Rezende et al (2017) utilized the variable gross domestic product to justify introducing macroeconomic variables in models for insolvency prediction and financial difficulty, as it was statistically significant in them. As a result, the fact that not all macroeconomic variables showed statistical significance for the model might indicate that corporations design strategies to overcome economic difficulties that may appear.…”

“…Such indicators are widely known in companies' financial analysis literature as being linked to capital structure liquidity, debt, profitability, and activity levels. After the selection of variables and sample, it is common the application of statistical techniques such as logistic regression, discriminant analysis, neural networks, and so on (exemplified by Brito and Assaf Neto, 2008;Rezende et al, 2017;Scalzer, Rodrigues and Macedo, 2015;Soares and Rebouças, 2014). Bearing in mind the literature that covers financial indicators, the first research hypothesis to be followed is presented: H01: Financial indicators predict the probability of company insolvency.…”

“…The only macroeconomic variable that was confirmed as statistically meaningful for the model was the one regarding gross domestic product -X17 (embedded in the model with the natural logarithm), having a negative relationship (as expected). Rezende et al (2017) utilized this variable to justify embedding macroeconomic variables in insolvency prediction and financial difficulty models, for which it proved to be statistically meaningful. The estimated ratio was that the corporation has 3.48 times the chance to become insolvent in case the economy goes through some period(s) of recession(s) or even still, for a sluggish economy.…”

“…The current research has contributed to the debate surrounding the utilization of financial, macroeconomic and/or corporative management variable groups regarding insolvency prediction, motivated by the verification of conflicting results displayed by other studies (ex. : Brito and Assaf Neto, 2008;Castro Junior, 2003;Rezende, Montezano, Oliveira, & Lameira, 2017;Soares and Rebouças, 2014;Stüpp, 2015).…”