Assessment of the probability of bankruptcy of construction companies

Next step is to validate the model through the ROC analysis.( See table 17.)

Table 17

Picture 2. ROC curve and AUC area

It is clearly seen as ROC is 0.9135 that suggests that prediction accuracy of the model is considered to be very good. However, at the standard 0.5 threshold of probability level the share of correctly predicted bankrupts firms is only 62.22% whereas non-bankrupt firms 93.75%, on the whole sample - 87.84%.

However the predictive accuracy of the model can be improved by changing a threshold level (G. Khaidarshina 2008). According to the analysis, the most appropriate threshold is equal to 0.4. At this level, the number of correctly specified bankrupts is equal to 80% whereas the number of correctly classified healthy firms = 88.14% (see table 18)

Table 18. True predicted bankrupt and non-bankrupt firms

The model without regulatory coefficients:

Logit model:

Next, I estimate the model without regulatory coefficients as mentioned in the methodology section (see table 19)

Table 19. The model without regulatory coefficients

On the next step I estimate a quality of the model by means of ROC analysis (see table 20).

Table 20. ROC analysis

As the ROC-analysis shows, the AUC Area is equal to 92.74% that suggests a very good quality of the model.

For the prognosis purpose I have set a threshold equal to 0.42. In this case the number of correctly specified bankrupts = 75.56%, the number of correctly classified non-bankrupts = 89.83% (See table 21)

Table 21. Number of correctly specified bankrupts and non-bankrupts

Generally, the model without regulatory coefficients is a bit better in terms of McFadden R² and Akaike criterion though the difference is not significant.

3.4 Interpretation of the results

The results show that two out of six variables appeared to be statistically significant in the model with regulatory variables: Absolute liquidity (CASHCL) and leverage ratios (TLTA).Both coefficients represent two groups of financial ratios: liquidity and solvency correspondingly and the coefficients conform to the economic sense discussed in the methodology section.

Leverage (EQTD, TLTA) is considered one of the most reliable predictors in the previous studies as it showed statistical significance in a variety of studies (e.g Altman 1968, Booth (2003), Ohlson (1980), Zmijewsky(1984), G. Khaidarshina (2008) and others.). Firms that have a large share of debt in liabilities side of the balance-sheet are more likely to go for bankruptcy procedure next year, that proves the hypothesis №1

Surprisingly, CASHCL has the most significant influence on firm's failure. This implies that the building firm that has a large amount of cash relative to current liabilities is more likely to pay its debts and remain solvent. This actually proves the idea suggested by Timofeev and Fedorova (2014) who investigated a significant role of absolute liquidity ratio to the health of building companies as “Liquidity on the market of construction services is a basis for the business potential of the company. Consistent estimation of the asset's liquidity is a key determinant of the company's market value. High market value of the building firm in its turn can attract new investors and make business more sustainable. Thus, hypothesis № 2 is confirmed. Nevertheless, another regulatory coefficient (current ratio) that was more frequently used in the previous studies does not show statistical significance. This may be a result of different structure of current assets. Firms that have difficulties with operating performance are likely to have more inventories than amaount of cash. Inventories for building companies are principally represented by construction materials that are not liquid enough as cash reserves. And in the case of sudden creditor's claim the company might be unable to pay the debts. On top of everything else the firm may have large amount of account receivable that can be potentially an overdue any time soon and this frequently leads to “break” in the cash flow and therefore damage to firm's health. Therefore hypothesis №3 is rejected. The same logic works for The Working capital to total assets ratio which is also insignificant. That means that the size of net liquid assets relative to Total assets is similar for viable companies and unhealthy ones.

On the preliminary data analysis, it is clear that on average both bankrupt and non-bankrupt firms have negative net income per annum. And all the profitability ratios (earnings to sales ratio, return on assets) in both models seem to be insignificant that is astonishingly different from many studies before. Probably it can be explained by the fact that managers can deliberately underreport financial statements to avoid heavy taxation.

Asset turnover ratio (SATA) is insignificant in both models. This suggests that managers of well operated firms and bankrupts generally have equal abilities in dealing with competitive market conditions. And this can be explained by the fact that all the sample is formed from the companies of micro and small sizes which operate in a perfectly competitive market that makes difficult to bring new technologies and take advantage over the rivals.

Conclusion

In the present-day world with the increasing number of insolvent firms assessment of credit risk is an essential part of doing business. Early prediction of signs for the default of building firms may prevent counterparties and cooperative housing investors from debt losses.

This paper presents two models based on logit regression which is one of the most frequently used tools in the scientific area. The output of the model in the form of percentage, simplicity of calculations and no requirement for the normality of the data set are the key reasons for the choice in favor of this method in comparison with Neural Networks MDA and techniques.

The literature review cleared up that there are no studies dedicated to bankruptcy prediction for Russian building companies occupied with site preparation and buildings construction. One more novelties of the research is the application of regulatory ratios suggested by the government law and testing them for predictive efficiency.

The analysis of the most fundamental studies in the research area revealed that there are no particular link between the technique and the employed set of financial coefficients. Besides, many ratios are consistent with a time of the researches of many authors from different countries. For example: asset turnover ratio, leverage, working capital to total assets ratio and others. Such variables have been being effective predictors over a half of century and therefore they were included in the research. A set of such variables is in the whole comprehensively describes a firm's health from profitability, turnover, liquidity and solvency perspectives. As a result, the ratio of cash to total assets, cash to total assets and two ratios of leverage appeared to be statistically significant. All the coefficients have signs corresponding to economic sense. The value of a regulatory coefficient of current liability ratio recommended by the government methodology does not suit to the obtained sample of bankrupts, although this ratio has the most significant impact on the bankruptcy probability. This is somewhat surprising because there are no findings of application of this variable to the research in previous studies. However it does make sense as asset's liquidity is considered to be one of the most fundamental factors for firm's market value and therefore investor's confidence.

To some extent coefficients employed in the analysis are somewhat similar ( - cash/ current liabilities and both leverages: equity/total debt - total liabilities/ total assets) and have pretty high correlation coefficients. Therefore they are divided between two models. However, only six variables are included in every model, though some authors admit that is lower than optimal. This can be considered as a limitation of the research and including new variables could describe company's health from a different perspective and in this way improve predictive ability of the models.

One more limitation of the research is a lack of data available for bankrupt firms. Generally, “Spark” database rarely provides annual reports of financial statement for every year. That makes search of data tougher and does not guarantee a homogeneity of a sample since data is collected from 2009 to 2015. For this period of time different macroeconomic and market conditions might “skew” the results. One possible solution is to broaden a range of data sources and concentrate the search on more recent cases of bankruptcy. In addition it is possible to include variables describing economic situation in the country (e.g. refinancing rate or rate of inflation).

All the limitations described above can be a prospect for another study. Still a ROC analysis depicts a high quality of the model and at specified thresholds both regressions allow predicting company's failure with probability higher than 81%. Improvement of the model can facilitate to its possible application for further prognosis. Possible application of the model can be an assessment of company's health for investors who have free access to the annual financial statements as well as counterparties in building sector may assess solvency for their business partners.

References

1. Altman, E.I., (1968). Financial ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. Journal of Finance, No. 22, 589-610.

2. Altman, E.I., Haldeman, R.C., Narayanan, P., (1977). Zeta Analysis: A New Model to Identify Bankruptcy Risk of Corporations. Journal of Banking and Finance, 29-54.

3. Altman, E.I., (2000). Predicting financial distress of companies: revisiting the Z-score and ZETA models. New York University, New York.

4. Aziz, M., Humayon, A.D., (2006). Predicting Corporate Financial Distress: Whither do We Stand? Department of Economics, Loughborough University, UK.

5. Beaver, W., (1966). Financial ratios predictors of failure, Journal of Accounting Research, No.4, 71-111.

6. Booth, D., (2011). Predicting Bankruptcy with Robust Logistic Regression Journal of Data Science, No.9, 565-584.

7. Back, B., Laitinen, T., Sere, K., Van Wezel, M., (1996). Choosing Bankruptcy Predictors Using Discriminant Analysis, Logit Analysis, and Genetic Algorithms. Technical Report, No.40.

8. Coats, K., Fant, L.F., (1993). Recognizing Financial Distress Patterns Using a Neural Network Tool. Financial Management, Vol.22, No.3, 142-155.

9. Deakin, E.B., (1972). A Discriminant Analysis of Predictors of Business Failure. Journal of Accounting Research, Vol.10, No.1, 167-179.

10. Jones, S., Hensher, D., (2004). Predicting Firm Financial Distress: A Mixed Logit Model. The Accounting Review, 79/4, 1011-1038.

11. Kerling, M., Poddig, T., (1994). Klassifikation von Unternehmen mittels KNN. In: Rehkugler, H. Zimmermann, H. G. (eds), Neuronale Netze in der Цkonomie. Mьnchen: Vahlen Verlag, 424-490.

12. Kiviluoto, K., (1998). Predicting Bankruptcies with the Self-organizing Map. Neurocomputing.

13. Leshno, M., Spector, Y. (1996). Neural Network Prediction Analysis: The Bankruptcy Case. Neurocomputing, 10(1), 125-147.

14. Lau, A., (1987). A Five-State Financial Distress Prediction Model. Journal of Accounting Research, 127-138.

15. Lenard, M.J., Alam, P., Madey, G.R., (1995). The Application of Neural Networks and A Qualitative Response Model to the Auditor's Going Concern Uncertainty Decision. Decision Sciences, No.26(2), 209 -227.

16. Ohlson, J., (1980). Financial ratios and the probabilistic prediction of Bankruptcy, Journal of Accounting Research, 18/1, 109-131.

17. Salchenberger, L., Cinar, E., Lash, N., ( 1992). Neural Networks: A New Tool for Predicting Thrift Failures. Decision Sciences, 23(4), 899-916.

18. Tam, K.Y., Kiang, M.Y., (1992). Managerial applications of neural networks: the case of bank failure predictions. Management Science No.38, 926- 947.

19. Winakor, A., Smith, R., (1935). Changes in the financial structure of unsuccessful industrial corporations, Bulletin, No. 51, University of Illinois, Bureau of Business Research, Urbana, IL.

20. Алексеева, Ю., (2011). Оценка финансового состояния и прогнозирование банкротства предприятия.

21. Естропов, М., (2008). Оценка возможностей прогнозирования банкротства предприятий в России. Вестник Оренбургского государственного университета, No.4, 25-32.

22. Захарова, М.И., (2014). The models of the quantity assessment of the bankruptcy probability were developed by the foreign specialists. Education.

23. Змиевский, М.Е. (1984). Methodological Issues Related to the Estimation of Financial Distress Prediction Models. Journal of Accounting Research Vol.22, No.1, 59-82.

24. Соколова, Т.В., (2013). Сравнительный анализ функциональных возможностей, преимуществ и недостатков информационно-аналитических баз данных. Управление корпоративными финансами, 280-288.

25. Сорокин, А.С., (2014). К вопросу валидации модели логистической регрессии в кредитном скоринге. Интернет-журнал Науковедение, 1-14.

26. Федорова Е.А., Тимофеев Я.В., (2014). Нормативы финансовой устойчивости предприятий: отраслевые особенности. Корпоративные финансы, No.1(33).

27. Хайдаршина, Г. Современные подходы к оценке риска банкротства предприятия.

28. [Internet-resource 1]. Information Group «Interfax», 8.04.2015, (http://www.interfax.ru).

29. [Internet-resource 2]. Паклин Н. (2012). Логистическая регрессия и ROC-анализ - математический аппарат. BaseGroup Labs - технологии анализа данных (http://basegroup.ru).

30. Федеральный закон (ФЗ) от 26.10.2002 N 127-ФЗ (ред. от 29.12.2015) "О несостоятельности (банкротстве)" (с изм. и доп., вступ. в силу с 29.03.2016)

31. Федеральному закон (ФЗ) от 8.01.1998 "О несостоятельности (банкротстве)" (гл. I, ст. 2)

Размещено на Allbest.ru