Challenges in Predicting Financial Distress in Emerging Economies: The Case of Croatia

Abstract

Using an extensive dataset of over 44,000 firms for a period that includes an economic upswing and a recession, this article examines how time-varying, firm-specific variables and changes in the macroeconomic environment affect the probability of firms’ financial distress. Traditional single-period approaches used to predict financial distress are based on unrealistically restrictive assumptions and cannot dynamically account for changes in financial indicators and macroeconomic conditions. Therefore discrete-time hazard models (using logit and cloglog) are applied; these indicate that the probability of distress is strongly influenced by both firm-specific and macroeconomic variables. Although firm-specific variables play an essential role, the results show that macroeconomic variables are important for understanding the fluctuations in the probability of distress over time. Furthermore, the findings provide evidence that both the legal criteria and a firm’s financial health should be considered when identifying firms in emerging economies that are in distress because existing bankruptcy laws are not applied.

Keywords:

• financial distress
• firm-specific variables
• macroeconomic shocks
• multiperiod logit model

JEL Classification:

• C35
• C52
• G33

ACKNOWLEDGMENTS

The author is grateful to Mira Dimitrić, Saša Žiković, Ervin Čeperić, Kristina Kaštelan and the anonymous referees for valuable comments and suggestions.

FUNDING

This work is supported by the Croatian Science Foundation under Project 6558, Business and Personal Insolvency—the Ways to Overcome Excessive Indebtedness.

Notes

1. The most commonly applied models in survival analysis rely on continuous-time data. This means that the analyzed event can occur at any point in time: a random process happens in continuous time. However, it often happens that the observations occur at discrete time intervals, for example, at the end of each year (Jenkins 2005).

2. Although the time of distress can be viewed as a continuous variable, data concerning a firm’s distress are available on a discrete-time basis, usually annually.

3. Cloglog is a discrete-time representation of a continuous time proportional hazard model that can be applied when available data are interval censored-grouped or banded into intervals (for details, see Jenkins 2005).

4. The Orbis database by Bureau van Dijk contains basic data, financial ratios and items from financial statements of firms around the world. Access to the database enabled downloading data for all firms (successful and unsuccessful).

5. Based on previous research, Platt and Platt (2009) summarized the different events that classify firms into the distress group: evidence of layoffs, restructurings, missed dividend payments, a low interest coverage ratio, cash flow less than current maturities of long-term debt, changes in equity prices, or a negative EBIT, the negative net income before special items. They formulated their own classification of firms in distress based on whether they meet the following three criteria for two consecutive years (Platt and Platt 2006): negative EBITDA interest coverage, negative EBIT and negative net income before special items.

6. To overcome the widespread illiquidity, the Croatian government introduced the Financial Operations and Pre-Bankruptcy Settlement Act (the Act) in October 2012.

7. Negative equity corresponds to the accounting view of the firm distress. A firm with negative equity does not have enough funds to cover its liabilities. However, negative equity, as such, does not mean that a firm will eventually fail and/or declare bankruptcy. The book values of assets and liabilities do not necessarily represent their fair value. The balance sheet does not reveal whether the firm has sufficient liquid assets to cover its liabilities, but it gives a strong warning (Hazak and Männasoo 2010).

8. This criterion is justified because, even though the Orbis database is one of the best databases for emerging markets, the accounting information for micro-firms is rather scarce.

9. Firm size was determined in accordance with the Accounting Act (Official Gazette 109/2007, Article 3), where: (1) Small firms are those that do not exceed two of the following conditions: (a) total assets of EUR 4.33 million, (b) income of EUR 8.67 million, (c) average number of employees during the financial year 50; (2) Medium-size firms are those that do not exceed two of the following conditions: (a) total assets of EUR 17.33 million, (b) revenues of EUR 34.67 million, (c) average number of employees during the financial year 250; (3) Large firms are those that exceed two of the three conditions for medium-size firms.

10. A complete list of annual report variables containing ratios from four categories (liquidity, leverage, profitability and cash flow) is available upon request.

11. In addition to the stepwise procedure, univariate analysis between dependent and each individual independent variable are examined and the variables showing the strongest correlation with distress risk are considered.

12. Current liabilities to total assets (CL/TA) also confirms a strong association with distress risk but is excluded from further analysis because it is highly correlated with total debt to total asset ratio (correlation coefficient = 0.74). Results on short-term debt financing are available upon request.

13. Some of the statistically significant variables have only a marginal economic effect: although they are statistically significant, they might not bear a significant economic impact on the probability of financial distress in real-life situations.

14. The variables relating to the inflation rate and unemployment rate did not prove statistically significant in predicting financial distress when standing alone in the model. The results are available upon request.

15. According to descriptive statistics, distressed firms finance over 90 percent of all assets through debt.

16. The LR test evaluates the difference between nested models: one model is considered nested in another if the first model can be generated by imposing restrictions on the parameters of the second.

17. The cut-off points for models estimated in Table 5 range from 0.055 to 0.057 for in-sample testing and 0.04 to 0.08 for out-of-sample testing. Likewise, in Table 7, in-sample testing uses cut-off points from 0.054 to 0.055, while cut-off points from 0.04 to 0.08 are used in out-of-sample testing.

Additional information

Funding

This work is supported by the Croatian Science Foundation under Project 6558, Business and Personal Insolvency—the Ways to Overcome Excessive Indebtedness.

Notes on contributors

Ivana Tomas Žiković

Ivana Tomas Žiković is an assistant professor in the Faculty of Economics at the University of Rijeka, Rijeka, Croatia.