Assessment of the probability of bankruptcy of construction companies

Размещено на http://www.allbest.ru/

Размещено на http://www.allbest.ru/

ПЕРМСКИЙ ФИЛИАЛ ФЕДЕРАЛЬНОГО ГОСУДАРСТВЕННОГО

АВТОНОМНОГО ОБРАЗОВАТЕЛЬНОГО УЧРЕЖДЕНИЯ
ВЫСШЕГО ОБРАЗОВАНИЯ

«НАЦИОНАЛЬНЫЙ ИССЛЕДОВАТЕЛЬСКИЙ УНИВЕРСИТЕТ

«ВЫСШАЯ ШКОЛА ЭКОНОМИКИ»

ФАКУЛЬТЕТ ЭКОНОМИКИ, МЕНЕДЖМЕНТА И БИЗНЕС-ИНФОРМАТИКИ

COURSE WORK

ASSESSMENT OF THE PROBABILITY OF BANKRUPTCY OF CONSTRUCTION COMPANIES

Рецензент

Итурриага Феликс Хавьер Лопес

Руководитель

А.М. Емельянов

Пермь

2016



Abstract

This study is dedicated to the problem of bankruptcy prediction. Nowadays the evaluation of firm's creditworthiness has become a vital part of any business since there are an increasing number of insolvent firms. The aim of the research is to investigate what financial ratios can be treated as effective predictors of failure for Russian firms in building sector. The data consists of nine coefficients derived from the financial statement for 45 firms that faced bankruptcy procedure from 2009 to 2015 and 177 companies which operated well during this period. As a result two logit models are constructed with predictive ability over 81%. One model includes government regulatory coefficients and shows that absolute liquidity ratio has the most significant impact on firm's failure. Another model proves the fact that coefficients which were widely employed by Western and Russian authors can be reliable predictors for company's bankruptcy.

Аннотация

Данное исследование посвящено проблеме прогнозирования банкротства. В настоящее время оценка кредитоспособности фирмы стала неотъемлемой частью любого бизнеса, так как в России наблюдается все большее число банкротств. Цель исследования - проанализировать, какие финансовые показатели могут рассматриваться как эффективные предсказатели банкротств российских строительных фирм. Данные представляют собой 9 коэффициентов, построенных на основе финансовой отчетности 45 фирм, в отношении которых открывались процедуры банкротства с 2009 по 2015 г. и 177 компаний, которые успешно функционировали в течение этого времени. В результате были построены две логит модели с предсказательной способностью более 81%. Одна из моделей включает в себя государственные нормативные коэффициенты и показывает, что коэффициент абсолютной ликвидности имеет наиболее существенное влияние на вероятность банкротства предприятия. Другая модель содержит финансовые коэффициенты, которые широко использовались западными и российскими авторами. Данные показатели также имеют высокую предсказательную силу.



Contents

Introduction

1. Theoretical background

1.1 Bankruptcy

1.2 The earliest studies

1.3 Multiple Discriminant analysis

1.4 Logit modeling in bankruptcy prediction

1.5 Validation of the logit model

1.6 Neural networks

1.7 Sources for data

2. Research question

2.1 Comparison of MDA logit model and neural network

2.2 Limitations for financial ratios of building firms

2.3 Financial ratios employed to the research

3. Methodology

3.1 Data collection

3.2 Economic purpose of the financial ratios

3.3 Descriptive statistics

3.4 Interpretation of the results

Conclusion



Introduction

Prediction for the company's bankruptcy has always been an issue of much concern and tends to be more essential nowadays due to deteriorating economic situation and increasing number of insolvent building firms, especially in Russia. For instance, according to statistics, 2700 Russian building companies went for bankruptcy for 2015 year which is 5 times more in comparison with the previous year (Internet resource №1). Probably, the only positive side of this is that researchers have got a new data to test their predictive models.

Investigations on this field of study could be valuable for wide audience. First of all, managing directors have to be informed about financial health either of their own company, or the firms they deal with, in order to make sure if the current business policy is successful and the counterparty risk is minimized. It is especially important for managers to estimate risks of firm's solvency before offering commercial credit or assigning new contracts. Historically the traditional approach of the counterparty risk evaluation is to rely on rating issued by credit rating agencies; however, many specialists claim that these ratings are more likely to be reactive rather than predictive. Secondly, possible users of information on bankruptcy prediction may be investors who need to decide whether to participate in the funding of construction or it is too risky. Thirdly, banks have been using different models based on ratio analysis to evaluate client's creditworthiness. Hence, the study can facilitate to more reliable lending practice and correct assessment of interest rates that reflect credit risk.

Large variety of practical purposes mentioned above cause much interest in study of the bankruptcy prediction area. The aim of my investigation is to evaluate how particular financial indicators can serve as predictors for the company's bankruptcy in Russian building sector. For this purpose several tasks should be accomplished. First of all, it is essential to conduct theoretical justification for the most fundamental studies of western and Russian authors in the scientific area. Next, relying on the literature review I am going to define what factors can comprehensively describe a financial condition of the firms. Thirdly, I am intended to select financial ratios for the companies from building sector. Then, I will collect data set for these financial indicators prior one year to actual company's bankruptcy procedure including government regulators. Thirdly, I am intended to define, what indicators are the strongest at increasing the predictive power of the model. And finally I am going to test predictive accuracy of the model by means of ROC curve and per cent correctly predicted value. The methodology used in the research is a logit model which is one of the most common tools in the scientific area.

Theoretical background includes literature review for the most famous western and Russian studies dedicated to bankruptcy prediction modeling. In this section I am going to pick out the most fundamental techniques used in the scientific sphere and stress out financial ratios which serve as good predictors for firm's insolvency. On the basis of the theoretical background I will pose a research question in the next part. Then, I will describe the model and collected data on the research in the methodology section. There I also ground for the selected variables and make hypothesis. In the results section I am going to define the most effective ratios and conclude with the predictive accuracy of the models



1. Theoretical background

1.1 Bankruptcy

Before making an analysis it is important to clear up what does bankruptcy mean according to Russian legal system. A Federal Law "On Insolvency (Bankruptcy)" suggests that bankrupt firm is a debtor's inability to meet the claims of creditors on monetary obligations and (or) to meet the obligation for making compulsory payments. A sign for company's bankruptcy is formulated in the following way: A legal entity is considered to be unable to meet creditors' claims for monetary obligations and (or) to meet a requirement for making compulsory payments, if these obligations and (or) the obligation is not met by this within three months from the date when they should have been done. According to Russian law, a legal entity is recognized as bankrupt only after the Court of Arbitration states signs of insolvency. In the case of recognition of the debtor as bankrupt, the court adjudges one of the following bankruptcy procedures:

Bankruptcy proceedings;

Introduction of financial recovery

Introduction of the external control

Introduction of observation

Amicable agreement

The date of actual company's failure is considered as a date of Arbitrage court's adjudication of one of these stages.

The date when a creditor applies to Arbitrage court to adjudge his counterparty's insolvency is considered to be a date of the firm's default. As a rule, firm's default takes priority of the bankruptcy. Typical bankruptcy case is represented as follows:

1. Arbitrage court considers a complaint.

2. After creditor's meeting the Arbitrage court adjudges one of the bankruptcy procedures.

3.A potential bankrupt is, then either going for
turnaround or liquidation.

Generally, the researchers are willing to predict exactly the date of default but not the date of an actual company's bankruptcy as it takes rather long time (about a year). And practioners are interested to predict a date when his counterparty will potentially fail to meet its obligations, but not the date when the court adjudges a firm to be bankrupt.

1.2 The earliest studies

The history of the bankruptcy prediction takes roots in 1930s. P. Fitz in his study collected data on financial ratios for 19 failed and nonfailed firms and maintained strong differences in ratios at least three years prior to failure. The Winakor and Smith (1936) calculated mean ratios of failed companies for 10 years prior to failure and investigated the deterioration in ratios as failure approached. Almost the same technique was implemented by Merwin in 1942, who compared the mean ratios of continued firms with those of discontinued during the period from 1926 to 1936 and proved the increase in differences between ratios as the firm became more close to the bankruptcy. Although the aforementioned studies are devoid of any modeling, they gave a rise to further researches in the problem area.

One of the most fundamental studies on this issue was done by Beaver in 1966. As well as Merwin's study, the paper includes empirical verification for 30 financial ratios and comparison of its means between failed and non-failed companies. But the innovative idea was the application of dichotomous classification test to make prediction for the bankruptcy, considering separately the impact of the particular ratio. As a result, Beaver derived set of ratios, which let him define a health of businesses and estimate probability for the company's failure after a certain period of time. (See table 1.)

Table 1. Coefficients in Beaver's study (1966)

Ratio	Formula	Value of ratio
		Prosperous company	5 years prior to failure	1 year prior to failure
Beaver's coefficient		0,4-0,45	0,17	-0.15
Return on assets		6-8	4	-22
Leverage		? 0,37	? 0,5	? 0,8
Asset coverage ratio		0,4	? 0,3	0,06
Coverage ratio		? 3,7	? 2	? 1

As a result, Cash flow and Total asset ratios appeared to be the most effective as they predict more than 90% of the insolvent firms before 1 year to actual bankruptcy. creditworthiness firm building bankruptcy

Regardless of its simple procedure, obtained results have certain limitations. Especially, the method reveals just a value of the ratio corresponding to different firm's financial status and it does not account for a degree to which the importance of certain indicator could be evaluated for bankruptcy prediction in the whole.

1.3 Multiple Discriminant analysis

E. Altman in 1968 managed to cope with the problem stated by W. Beaver (1966) by developing innovate predictive Z-score model. The model combined five most important ratios and considered its particular impact on the prediction of the bankruptcy on the whole. The paper is very famous as it first introduced a ground-breaking discovery of the multiple discriminative analysis (MDA). Applying this technique, E. Altman managed to calculate the weights of the specific ratio, thereby define overall Z-value as a prediction for firm's failure.

Xl =

To maintain the critical points of Z-score for failed and non-failed companies, the F-test is used which implies the calculation of ratio of the sums-of-squares between-groups to the within-groups sums-of-squares. As a result, Z-score boundaries for the bankrupt and non-bankrupt companies one year prior to the probable insolvency were maintained.( See table 2).

Table 2. E. Altman Z-score model's output

Z-score value	Probability of bankruptcy within 1 year
< 1,8	Very high
1,81-2,7	High
2,71- 2,9	Possible
>2,9	Low

The 5 financial coefficients generally proved to be successful in predicting bankruptcy one year before as they have an accuracy of 95%.

The model is widely used by companies due to its simplicity of the calculations and high accuracy as well. However, this model does not account for firm's national belongings and therefore market conditions therefore weights of the factors and overall Z-score index can be misleading. Meanwhile, the model is denied with profitability ratios which can be good indicators for firm's health.

Later, Altman, using the same technique, managed to build the model with 2 factors:

Where:

X1 - Current liquidity ratio;

X1 - Total debt/ Equity

If Z-score > 0, then a company is considered to be close to insolvency and vice a versa if Z-score < 0, the company operates well.

In 1977, Altman, Haldeman and Narayanan made a dramatic contribution to the research by developing a new ZETA-model which has several improvements in comparison with the original Z-score approach. Especially, it is concentrated not only on the manufacturing companies, but on the retailed ones. In general, 7 chosen factors were selected from 27 ones and subjected to the conscious scrutiny. (See table 3).

Table 3. ZETA model coefficients

Variable	Formula
X1	- Return on assets	Earnings before Interest and Taxes/Total Assets
X2	- Stability of earnings	Normalized measure of the standard error of estimate around a five to ten-year trend in ROA
X3	- Debt service	Earnings before Interest and Taxes/Total Interest Payments
X4	- Cumulative profitability	Retained Earnings/Total Assets
X5	- Liquidity	Working Capital/Total Assets
X6	- Capitalization	Common Equity/Total Capital
X7	- Size	Logarithm of Total Assets value

The model also incorporates enhances in the MDA which allow ZETA-model to outperform Z-score one in terms of prediction accuracy. (See table 4)



Table 4. Comparison of Z-score and ZETA models

Period of time prior to	ZETA Model	Z-score, 1968
bankruptcy	Bankrupt, percent	Non-Bankrupt, percent	Bankrupt, percent	Non-Bankrupt, percent
1	96,2	89,7	93,9	97,0
2	84,9	93,1	71,9	93,9
3	74,5	91,4	48,3	n.a.
4	68,1	89,5	28,6	n.a.
5	69,8	82,1	36,0	n.a.

In Russia one of the most famous studies belongs to Zakharova (2014). The author used MDA methods and obtained a following model:

Where:

K1 is the ratio of its own working capital to total assets;

K2 - the ratio of net profit to equity;

K3 - asset turnover ratio;

K4 - the rate of return: the share of net income per unit of cost

Company bankruptcy probability is estimated using the values of the R-score, a ranking on which is presented in table 5

Table 5. Value of R-score corresponding to bankruptcy probability

Bankruptcy probability	Very high	High	Average	Low	Very low
Probability	90-100%	60-80%	35-50%	15-20%	Менее 10%
Value for R	<0	0?R<0,18	0,18?R<0,32	0,32?R<0,42	R>0,42

1.4 Logit modeling in bankruptcy prediction

Predictive accuracy of the model is 81%, however due to some reasons it is not widely accepted by the practioners.

Although ZETA and Z-score models gained a wide range of followers, the MDA was subjected to criticism. In particular, Ohlson (1980) admits that the output of an MDA technique is simply a score which can be hardly interpreted, since it is basically a typical discriminatory tool. Ohlson advocated a radically new approach that was founded upon the logit-model theory. Alexeeva (2011) admits that logit modeling has certain advantages e.g. it can incorporate any type of regressors including discrete variables. Deakin (1972) also maintains that normality of distribution for the sample of bankrupt and non-bankrupt companies has been no longer a requirement for obtaining unbiased estimators. In addition, the model makes possible to consider interrelation between independent variables. But the most significant advantage is that, this technique allows using the ratios as “real predictors but not as indicators for matching purposes” in comparison with MDA technique (Ohlson, 1980). It means that the output of the logit model is a certain percentage that explains a probability of becoming bankrupt. Traditionally the researchers establish the cutoff point at P=0.5 that means, companies which have higher values than P=0.5 are more likely to go bankrupt rather than companies whose probability is less. If there is a special importance of the prediction for the costs of the first and the second kind of error, then it is recommended to find a threshold based on the minimization of the average forecast error, weighted by the importance of the error. In this case, the optimal threshold depends on the content of the problem (electronic resource №2). Setting thresholds is rather subjective and it may depend on the results of percent correctly predicted procedure. For example, G. Khaidarshina (2008) defines a threshold at P=0.44 reasoning that exactly this cutoff point allows to distinguish bankrupt and non-bankrupt companies in the most accurate way. The author is renowned for its logit model for Russian production companies. The logit regression included 61 bankrupt firms and involved the same financial ratios as western researchers. Following ratios showed statistical significance at 1% level: Account Payable/ account receivable ratio, the natural logarithm of the ratio of assets to GDP deflator, Total debt/ Total assets, Total debt/ equity.

Estropov M. (2008) built a model which included data on 100 companies of agricultural sector and 150 industrial enterprises. Along with the “standard” variables the author considered such indicators as: age of the company, a rate of growth of assets and liabilities and a refinancing rate of the Russian central bank.

Another methodology that is widely used among the researchers is a multinominal logit regression. Instead of the classical dichotomy failed/ non-failed this methods uses several financial states that describe a severity of the firm's financial distress. For example, Lau (1986) classified a dependent variable into 5 states. These states are: “state 0: financial stability; state 1: omitting or reducing dividend payments; state 2: technical default on loan payments, state 3: protection under Chapter X or XI of the Bankruptcy Act; and state 4: bankruptcy and liquidation.”

1.5 Validation of the logit model

Along with the output, prediction accuracy of logit model is an issue that deserves much attention. For this purpose, researchers use percent correctly predicted companies, which is used as a goodness-of-fit measure. The more per cent correctly predicted better accuracy model has. However, this indicator works well if a sample contains enough bankrupt in comparison with healthy ones. Lenard, Alam and Madey (1995) stress that percent correctly predicted can be misleading as the number of bankrupt firms in normal economic periods is very small and they can be treated as outliers. Booth D.(2011) sticks to the same opinion: “Occurrence of bankruptcy is usually very low. Thus, in the case of corporate bankruptcy, it is possible to get rather high percentages correctly predicted even when the model's prediction of bankruptcy is extremely poor.”

One of the strategies to validate a predictive power of the model is to form two samples. One sample or so-called training sample is used for building a model, another one (test sample) is implemented to test its prediction accuracy. Identical results obtained from running the model on both samples can be treated as a good sign and such model is more likely to give appropriate results in practice.(Ievlev 2014).

One more way to check quality of the model is to use ROC analysis. This tool implies that all the outcomes can be divided into positive and negative ones depending on whether the company is bankrupt or not. ROC analysis is a graphical tool which allows estimating quality of the model through the indicators of specificity and sensitivity. These indicators show how many positive and negative outcomes, respectively, will be revealed by classifier (Sorokin 2014).

For the validation of the multivariate logit models the researchers commonly use MLA-generated probabilities. For example, Lau (1986) Used the MLA-generated probabilities technique for the model with 5-j states which implied that “entering each of the five possible states, the firm is "classified" into the state j with the highest predicted probability of entering. A firm is considered to be" correctly classified" if it actually entered j, otherwise the firm is "misclassified."

Probit regressions are not as frequently used as logit ones. One of the most fundamental studies dedicated to probit analysis belongs to Zmijewsky (1984) who built regression based on three factors: Net income/ Total assets, Total debt/ Total assets, Current assets/ short-term borrowings. All the coefficients appeared to be statistically significant and the model showed prediction accuracy higher than the model constructed by E. Altman in 1968 year

1.6 Neural networks

Along with the multivariate discriminant analysis and logit regressions neural networks method is also widespread in the research field. P.Coats and F.F ant (1994) claim that in the case of the usage of neural networks there is no more requirements for the linear decision making set to distinguish between healthy companies and the distressed ones. Moreover, the study indicates that neural networks model can be at least as successful as using MDA technique in terms of prediction accuracy.

Tarn and Kiang (1992) studied banks for solvency with the use of multi-layer neural network and came to the same conclusion as multi-layer neural network proved to be the best for a one-year time span, whereas logistic regression gave the best result for a two-year period. The authors considered the standard coefficients used in the E. Altman's Z-score model.

Salchenberger, Cinar and Lash (1992) made a comparison of neural networks to logistic regression. “Neural networks performed considerably better than logistic regression regarding their classification accuracy. In an 18-month time span for example, logistic regression gave 83.3-86.4% accuracy depending on the threshold used, while the neural network reached 91.7% accuracy”. The neural network model defined a following set of significant variables:

1)

2) 3

Kerling and Poddig (1994) like Coats and Fant used the database of French companies compared neural networks and discriminant analysis performance for a three-year prediction time span. Neural network gave 87% accuracy, discriminant analysis gave 85.7% accuracy. Kerling and Poddig also tried a number of interaction studies and the early stopping algorithm.

Leshno and Spector (1996) experimented with a new type of neural network

based on cross-conditions and cosinus relations. Depending on the type of network, classification accuracy for the prediction model prior two years to actual insolvency varied between 74% and 76%, as compared to the 72% accuracy of linear perceptron network.

Back (1996) implemented genetic algorithms to select the inputs of multi-layer neural networks. The method was used for one-, two- and three-year periods before bankruptcy, and significant improvement was reported compared to discriminant analysis and logistic regression in terms of model accuracy, however, the same variables showed the same statistical significance. They are:

X1: Quick Liquidity Ratio

X2: Return on Sales;

X3: Cash-flow / Total Debts

X4: Current Assets / Total Assets

X5: Accounts Receivable / Accounts Payable

Kiviluoto (1998) used almost the same predictors employing self-organising maps with different time spans, based on Finnish companies' annual reports and obtained 81-85% accuracy of the model. All the significant coefficients defined in the Back's study except for quick liquidity ratio showed statistical significance.

Regulatory analysis

In Russia, practioners, in order to assess a creditworthiness of the building firms, usually resort to estimating regulatory factors suggested by the government. Regulatory approach in prognosis for the company's bankruptcy implies a comparison of the obtained financial coefficients with the value of the regulatory one or its belongings to a certain regulatory interval. If the value of the obtained coefficient is laid beyond a particular interval, then, the company is considered to be unhealthy. According to Russian government's regulation on assessment of the financial condition from 1994, 6 financial coefficients are widely used in the analysis (See table 6)

Table 6. Government regulatory coefficients

Regulatory coefficient	Formula	Value for healthy firms
Current ratio		?2
Absolute liquidity ratio		>0.2
Liquidity ratios in fund-raising		>0.7
The ratio of debt to equity		<0.7
Own funds ratio		>0.1
Maneuverability ratio of working capital		>0.5

However these regulatory coefficients are devoid of any relation to the specific economic sector. Meanwhile Fedorova and Timofeev (2014) revealed a very weak predictive ability of the regulatory coefficients for the sample of companies selected from different sectors. (See table 7)

Table 7. Predictive ability of the government regulators

Coefficient	Formula	Predictive accuracy
Current ratio		49,3%
Absolute liquidity ratio		42,9%
Liquidity ratios in fund-raising		27,3%
The ratio of debt to equity		35,4%
Own funds ratio		64,7%
Maneuverability ratio of working capital		29,7%

To strengthen a predictive power of the regulatory coefficients, the authors implemented binary classification method and Ginni coefficient which enabled them to obtain financial ratios for the specific economic sector. Regulatory coefficients for the building companies are represented in table 8.

Table 8. Obtained values for the regulatory coefficients

Regulatory coefficient	Formula	Value for healthy firms
Current ratio		>0.8
Absolute liquidity ratio		>0.5
Liquidity ratios in fund-raising		>0
The ratio of debt to equity		[0; 10]
Own funds ratio		[-0.25; 1]
Maneuverability ratio of working capital		[-0.25; 0.75]

Obtained values of the coefficients differ dramatically from the regulatory ones and increase predictive power of the ratio to an average up to 75%.

1.7 Sources for data

Deriving financial data is an issue that is also widely discussed among scientist. Some authors analyze the data given by the banks and international rating agencies like Moody's and S&P (e.g. Altman 1968, Beaver 1966). However, many of them rely only on the auditor's report (e.g. McGouth 1978,). The underlying assumption of this is that financial statements provided by auditors have a proof that the firm has the ability and intension to operate well in the future. Some researchers prefer auditor's researches to bank's filings at it allows to “capture” a real date of the financial distress starting but not the date of real bankruptcy procedure begins (K. Coats F. Fant 1993). More contemporary Russian studies principally acquire information from online- data bases e.g “Spark”, “Ruslana” “FIRA” and others (e.g. Ievlev 2014). There is an argument that such data can be fake one due to the fact that it is not always proved by auditor's control. However, Sokolova (2013) claims that such data is considered to be reliable as it first transmitted in the control and supervision authorities, the statistical service (FNS, Rosstat), and only then, into analytical database.

Regardless of the large variety of studies the problem of bankruptcy prediction still have not been so far investigated due to these reasons:

1. Different models lead to different results depending on the different economic sector, different time horizon and different macroeconomic conditions of the research. For example, Beagley (1996) cleared up that the misclassification of the model on the more temporary data is much higher in comparison with the studies from the past. That is why the author confirmed the hypothesis that the models generally do not show as good predictive accuracy for the present data as for old one.

2. Little advance has been achieved in Russian bankruptcy prediction sphere in comparison with the western one.

3. Models built by western scientists can be hardly implemented to the Russian bankruptcy prediction as they do not account for the specificity of the Russian economic situation as well as business organization that include different approaches to the accounting practice and taxation. These can result in differences of variables selection as well as coefficients of the regressors.

4. The majority of the studies are dedicated to the production and trading companies whereas companies from the building sector remain unheeded.

All the facts mentioned induce carrying out new researches in the bankruptcy prediction area.



2. Research question

2.1 Comparison of MDA logit model and neural network

A vast number of prediction models appeared since 1960 with over 20 different techniques employed. The majority of studies in bankruptcy prediction area are founded upon logit modeling, multiple discriminant analysis and neural networks. The most fundamental work with MDA built by Altman in 1968 gained wide range of followers and was widely accepted in practice by firms. However this method was subjected to criticism due to its obvious drawbacks. Firstly, the output of the MDA model is a simple integral Z-score which can be hardly interpreted since it is basically a discriminatory tool. Secondly, MDA requires a normality of the data set that can create certain limitations for the research. Otherwise, a dependent variable of logit regression is much simpler for understanding as it is a certain percentage that means company's probability of failure. Meanwhile logit regression does not require normality of the data set. Neural networks can effectively build a nonlinear dependence and therefore more accurately describe the data sets. Generally prediction models based on neural networks show high prediction accuracy, however, this method is not as easy for usage as statistical models and demands special skills for making analysis. Drawing on the literature review it is possible to make following conclusions:

1. The majority of the coefficients, especially coefficients introduced by Altman in 1968, are consistent in studies from 1960 to 2014 year

2. Researches including the same ratios showed different statistical significance depending on time horizon, economic sector and macroeconomic conditions

3. The majority of the models are founded upon logit modeling, multiple discriminant analysis and neural networks

4. The models based on different methods show approximately equal predictive ability

The aim of the research is to build logit model which allows estimating probability of company's bankruptcy prior one year to actual insolvency. Particularly, I am intended to investigate what financial coefficients can be treated as reliable predictors for company's bankruptcy in Russian building sector.

For this purpose several tasks should be done:

1. Selecting appropriate financial coefficients for the analysis based on the literature review that comprehensively describe financial condition of a company

2. Collecting available data set for failed and non-failed firms during 2008 to 2015 which will include information on balance sheets and Profit and Loss account.

3. Preparing financial coefficients based on the collected data and testing them for multicolinearity problem

4. Building logit regression which reflects the influence of indicators on the probability for company's bankruptcy and its statistical significance

5. Validating a model with ROC analysis.

6. Interpreting the results

2.2 Limitations for financial ratios of building firms

The majority of the building companies in Russia which faced bankruptcy procedure are micro and small-sized, i.e satisfy following conditions:

For small-sized companies:

State share of participation in its authorized capital does not exceed 25%;

It is not formed in the reorganization and privatization of state enterprises;

Profits from the production or provision of services for the departed year does not exceed 800 million rubles;

The average number of employees of the enterprise in the previous year is no more than 100 people.

For micro-sized companies:

The total share of the Russian Federation's participation, the RF subjects, municipalities, international organizations, public and religious organizations (associations), charities and other foundations, organizations that do not apply to small and medium-sized enterprises does not exceed 25% of the authorized capital;

Average number of employees should not exceed 15 persons.

The company's turnover is not more than 120 million rubles per year.

Selected companies are formed principally as limited liability companies which are not listed on the stock exchange. This fact imposes some restrictions to the research. In particular, it is no longer possible to include in the analysis financial coefficients which contain market value of a firm and dividend payments. For example,

1. Stock price trend ratio introduced by Lau (1986)

SPT=

Where:

H(t) and L(t) are respectively the high and low values of the range of stock exchange price in year `t'.

2. Binary variable =1 if no dividend is being paid currently, 0- otherwise.

A practice also shows that some financial coefficients that are appropriate for companies which are listed on the stock exchange can be misleading for Russian limited liability companies. For example ROE coefficient which is used in many western studies (e.g. Zmiewsky 1984, Beaver (1966), Ohlson (1980)) is made up of Net income/ Equity. Equity value for companies which are listed on the stock exchange is in most cases positive as it has a market value. However, balanced equity for limited liability companies can take on negative values of Equity indicators due to negative balancing items of retained earnings. However, if Net income and equity values are less than zero simultaneously, the coefficients appear to be positive that gives false information about company's health.

One more limitation for the research is that a very few companies post its annual financial statements every year. Thus some coefficients that are widely used in the western models can be hardly implemented to the analysis. For example, it is not possible to include in the analysis following financial ratios:

Stability of earnings ratio=Normalized measure of the standard error of estimate around a five to ten year trend in ROA (Altman, Handeman and Naranyan (1978))

Binary variable CHIN in the logit regression built by J. Ohlson (1980) = (N.I_t- N.I_t-1)/(| N.I_t|+| N.I_t-1|)

Where, N.I_t- net income for the most recent period

3. Trend of Capital Expenditures (Lau (1986))

According to preliminary analysis, Russian construction firms principally do not tend to have long-run debts besides short-term borrowings are most commonly replaced by the account payable. Thus, some financial ratios which work well for companies in studies of Western authors cannot be employed for analysis of building ones in Russia. For instance, such indicators can be:

1

2.

2.3 Financial ratios employed to the research

Considering all the aforementioned results and limitations we have selected 9 ratios which satisfy four conditions:

1) It has economic purpose to be real predictors for company's solvency

2) It can be derived from the balance sheet and P&L account from annual reports of Russian building companies.

3) It is frequently mentioned in previous studies of Russian and Western authors and showed statistical significance.

4) All the coefficients in the whole represent company's health from liquidity, asset turnover, profitability, and solvency.

5) Two government regulators are included in the research.

Table 9. Ratios employed to the research

Financial ratios	Method of calculation	Study
Leverage		Beaver (1966), Ohlson (1980), Jones (2003), Zmijewsky (1984) G.Khaidarshina (2008)
The ratio of equity to Total debt		S.Jones (2003) G.Khaidarshina (2008)
Current ratio		Beaver (1966), Ohlson (1980), Regulatory coefficient
Earnings to sales ratio		Foreman R.(2003)
Return on assets		Naranyan (1977)
Asset turnover ratio		Altman (1968) Jones(2003), Booth (2011)
Working capital to Total assets ratio		Altman (1968), Ohlson (1980), Jones (2003), Booth (2011), Naranyan (1977)
Ratio Cash to Total assets		Jones (2003), Altman (1968), Booth (2011)
Absolute liquidity ratio		Regulatory coefficient



3. Methodology

The research methodology will be based on the logistic regression theory with the dichotomous choice bankrupt/non-bankrupt. The general form of logit model is presented as follows:

Where

Probability of the bankruptcy

Independent variable

= estimated coefficient

A derived value ofindicates probability of company's bankruptcy. The probability of the company is extremely high if the obtained value of is equal to 1 and it is low if strives to 0. At the chances of failure/non-failure are equal.

Generally, logit regression in the research includes four steps:

1. Construction of Z-function as a linear combination of the regressors. The independent variables will be a set of financial ratios mentioned above

2. Building the model with the regulators suggested by the government methodology

3. Building the model

The coefficients in the regressions are estimated using the maximum likelihood method. Standard errors of coefficients will be evaluated with the Newey-West correction to heteroscedasticity and autocorrelation.

3. Interpretation of results;

4. Testing the accuracy of developed models through the ROC (Receiver Operator Characteristic) analysis. Although, the aim of the research is to build explanatory model but no for the prognosis it is still interesting to see what the predictive ability the obtained model has.

ROC analysis investigates and employs the relationship between sensitivity and specificity of a binary classifier. Sensitivity or true positive rate measures the proportion of positives correctly classified observations. Specificity or true negative rate measures the proportion of misclassified (See picture 1)

Размещено на http://www.allbest.ru/

Размещено на http://www.allbest.ru/

Picture 1. ROC curve

A quantitative indicator of the quality of a model is AUC (area under curve) which is restricted by the ROC curve and specificity axis. The larger this area, the better quality of the model is. (See table 10)

Table 10. AUC area and quality of the model

Measure of AUC	Quality of the model in terms of accuracy
AUC <0.5	Bad
0.5<AUC<0.6	Dissatisfactory
0.6<AUC<0.7	Average
0.7<AUC>0.8	Good
0.8<AUC<0.9	Very good
0.9<AUC<1	Excellent

Assessment of the model can be achieved through many tests. In the research I will consider AIC test which was suggested by Akaike in 1974

The Akaike information criterion (AIC) is “a measure of the relative quality of statistical models for a given set of data. Given a collection of models for the data, AIC estimates the quality of each model, relative to the rest of the other ones. Hence, AIC provides a means for model selection. AIC is founded on information theory: it offers a relative estimate of the information lost when a given model is used to represent the process that generates the data. In doing so, it deals with the trade-off between the goodness of fit of the model and the complexity of the model.” This indicator can also inform the researchers if the model contains iirelevant variables or not. AIC is constructed as follows:

where:

RSS- residual sum of squares;

n-number of observations;

k-number of regressors

The lower indicator of AIC is associated with a better quality of the model and proves that fact that all the coefficients are relevant to the analysis.

Data analysis, building the model and measurement of ROC analysis will be exercised by means of STATA.

3.1 Data collection

The data on bankrupt and non-bankrupt companies was collected from the “Spark” database and includes 49 companies which encountered bankruptcy procedure from 2009 to 2015 and 177 companies which operated well during this period. The data for viable companies has been derived with even portions for every year. As an indicator of bankruptcy I have extracted data on a certain year when company faced one of the four bankruptcy procedures:

1. Bankruptcy proceedings;

2. Introduction of financial recovery

3. Introduction of the external control

4. Introduction of monitoring procedure

5. Agreement of lawsuit

The time of bankruptcy procedure a company faces is a year of failure provided that there were no previous procedures adjudged by Arbitration court. A practice shows that normally the failed company faces a monitoring procedure the first and within 2 months after creditor meetings it encounters external control, agreement of lawsuit, financial recovery or bankruptcy proceedings. However, Russian legal system allows Arbitrage court to adjudge any of this as the first.

To avoid problem of inhomogenity, it is appropriate to choose companies for the analysis from the specific sector and of the same sizes. That is why I have selected companies from the building sector which are occupied with construction and site preparation. In more detail, it is represented by following
economic sectors described in Russian National Classifier of Economic Activities:

1. Construction site development 45.1

2. Demolition of buildings and earthworks 45.11

3. Clearance of building plots 45.11.1

4. Civil works for the construction of buildings 45.21.1

After deleting outliers, the sample consists of 24 micro-sized and 21 small-sized bankrupts whereas healthy companies are represented by 88 micro, and 89 small-sized businesses (see table 11).

Table 11. Sample

Bankrupt / non-bankrupt	Company's size	Total
	Micro	Small
Non-bankrupts	88	24	112
Bankrupts	89	21	110
Total	177	45	222

On the next stage the information on company's liquidity, profitability and solvency has been derived from balance sheets and profit and loss accounts prior one year to actual date of company's failure. Comprehensive analysis of data is essential to provide credible results. That is why the appropriate for the analysis financial indicators have been computed.

As a result, 9 financial ratios of profitability, solvency, liquidity and turnover have been prepared (see table 12).

Table 12. Financial ratios

Financial ratios	Name	Method of calculation	Type
Leverage	TLTA		Solvency
Working capital to Total assets ratio	WCTA		Solvency
The ratio of equity to Total debt	EQTD		Solvency
Ratio Cash to Total assets	CASHTA		Liquidity
Current ratio	CACL		Liquidity
Absolute liquidity ratio	CASHCL		Liquidity
Earnings to sales ratio	EASA		Profitability
Return on assets	NITA		Profitability
Asset turnover ratio	SATA		Turnover

3.2 Economic purpose of the financial ratios

Solvency:

1. Leverage ratio (TLTA):

A company with a high proportion of debt capital called financially dependent company. The company finances its activities at the expense only of equity, called financially independent company. Investors and business owners prefer a higher ratio of financial leverage, because it provides a greater rate of return. On the contrary, creditors are more willing to offer debts for companies with a lower coefficient of financial leverage as such companies are financially independent and have less risk to become bankrupts.

2. Working capital to total assets ratio (WCTA):

WCTA is a measure of firm's net liquid assets relative to Total assets. Ordinarily, a firm that experience consistent operating losses will have a lack of current assets in relation to total assets. Besides a firm, serving large current liabilities is more likely to be unable to meet its obligations.

3. The ratio of equity to Total debt (EQTD)

1) Company which has more obligations is more likely to fail to meet them.

2) Company which has little equity value is considered to be financially dependent and therefore less solvent.

Thus, I have made a hypothesis №1: EQTD negatively influences a probability of the company to fail.

Liquidity

Cash to total assets ratio (CASHTA):

CASHTA has a simple rationale that more viable company, in order to meet its obligations, is likely to have larger amount of cash relative to total assets value.

Current ratio (CACL):

Liquidity is the ability of the company to punctually meet obligations; characterized by the presence of liquid funds. Current ratio reflects the company's ability to repay the current (short-term) liabilities by means of current assets. The higher the indicator, the better a company's solvency is. Current ratio also characterizes company's ability to meet obligations in the event of an emergency. Current ratio value suggested by the government "Methodical provisions for assessment of the financial condition of enterprises and the establishment of the unsatisfactory balance structure" for healthy firms is more than 0.8. However this threshold does not correspond to reality. As there are 38 of 45 bankrupts which have CACL value prevailing 0.8 and only 148 of 177 viable firms having a current ratio more this threshold. However, I have made a hypothesis №2: CACL negatively influences a probability of the company to fail.

Absolute liquidity ratio (CASHCL)

CASHCL is rarely used in the researches but still has similar rationale to be a real predictor as current ratio: The more amount of cash reserves in combination with low debt service allows a company to successfully pay its debts and therefore be more solvent. Therefore, hypothesis №3: CASHCL negatively influences bankruptcy probability.

Profitability

Net income margin (EASA):

This ratio indicates how much actual profit company receives from each ruble of sales. So, the highest profit margin shows on the most profitable company, which better controls the size of its costs compared to its direct competitors. As a rule, this figure is expressed as a percent. For example, thirty percentage shows that the company receives a net income in the amount of 30 rubles of every 100 rubles of revenue. This rationale implies that the higher the net income margin is the more solvent company should be.

Return on assets (NITA):

ROA is a common indicator of profitability that reflects the efficiency of managing assets. Return on assets, including assets reflects the ability to generate profits. Generally it shows how much profit a firm will get investing in total assets. For example ROE=0.3 illustrates that one invested ruble will bring 30 cents to firm.

A “common sense” suggests that the higher rate of profitability, the less likely a business is going to fail.

Turnover

Capital turnover ratio (SATA):

Total assets turnover ratio is the most general estimator of the business activity. Generally it illustrates assets ability to generate revenue. The decrease in sales at constant assets may be treated as a negative sign for a company's health and therefore it can increase firm's probability to fail.

3.3 Descriptive statistics

Descriptive statistics for the obtained data on the financial ratios for both failed and non-failed firms are shown in tables 3 and 4correspondingly.

Table 13. Descriptive statistics for failed firms

<...

VARIABLE	Obs	Mean	Std. Dev.	Min	Max
TLTA	45	0.901	0.337	0.169	1.755
EQTD	45	0.108	0.358	-0.634	1.535
EASA	45	-0.142	0.486	-2.224	0.120
NITA	45	-0.044	0.152	-0.535	0.316
SATA	45	1.237	1.341	0.02	6.855
WCTA	45	0.02	0.323	-0.775	0.658

Подобные документы

• Comprehensive analysis of the economic activity of the company "Turboatom"

  Directions of activity of enterprise. The organizational structure of the management. Valuation of fixed and current assets. Analysis of the structure of costs and business income. Proposals to improve the financial and economic situation of the company.
  
  курсовая работа [1,3 M], добавлен 29.10.2014
  

• Pipeline capacity and economics

  Gas pipeline construction: calculating the pipe diameter, the pressure required for the transportation of natural gas compressors. The definition of capital costs for construction and operation of the pipeline. Financial management of the project.
  
  статья [774,7 K], добавлен 05.12.2012
  

• Analysis of activity of company DTEK Zuevskaya TPP

  General characteristic of the LLC DTEK Zuevskaya TPP and its main function. The history of appearance and development of the company. Characteristics of the organizational management structure. Analysis of financial and economic performance indicators.
  
  отчет по практике [4,2 M], добавлен 22.05.2015
  

• Small business survival and development problems and countermeasures from the perspective of financing

  Analysis of the status and role of small business in the economy of China in the global financial crisis. The definition of the legal regulations on its establishment. Description of the policy of the state to reduce their reliance on the banking sector.
  
  реферат [17,5 K], добавлен 17.05.2016
  

• Electronic Construction Procurement System

  Project background and rationales. Development methodology, schedule planning. Company mission and vision. Organization of staff and company structure. Procurement system target market. Implementation of procurement system. Testing, user manual.
  
  дипломная работа [6,8 M], добавлен 28.11.2013
  

• The main principles of safety market economy

  Assessment of the rate of unemployment in capitalist (the USA, Germany, England, France, Japan) and backward countries (Russia, Turkey, Pakistan, Afghanistan). Influence of corruption, merges of business and bureaucracy on progress of market economy.
  
  реферат [15,5 K], добавлен 12.04.2012
  

• Assessment of the situation on the regional housing market in Russia

  Issues about housing prices formation process. Analytical model of housing prices. Definition a type of relationship between the set of independent variables and housing prices. The graph of real housing prices of all Russian regions during the period.
  
  курсовая работа [1,6 M], добавлен 23.09.2016
  

• Financial bubble

  Financial bubble - a phenomenon on the financial market, when the assessments of people exceed the fair price. The description of key figures of financial bubble. Methods of predicting the emergence of financial bubbles, their use in different situations.
  
  реферат [90,0 K], добавлен 14.02.2016
  

• The improvement of Russia’s investment climate: problems and perspectives

  Establishing a favorable environment for investments, removing administrative barriers. Establishing high-technology parks. Formation of financial mechanisms to attract and support investments, tax stimulation measures. Brand promotion of Russian regions.
  
  реферат [15,9 K], добавлен 04.06.2013
  

• Economy of Kazakhstan

  A variety of economy of Kazakhstan, introduction of the international technical, financial, business standards, the introduction to the WTO. The measures planned in the new Tax code. Corporation surtax. Surtax reform. Economic growth and development.
  
  реферат [27,2 K], добавлен 26.02.2012
  

• The stock market of Russia and Italy: dynamics of development and comparative characteristics

  The stock market and economic growth: theoretical and analytical questions. Analysis of the mechanism of the financial market on the efficient allocation of resources in the economy and to define the specific role of stock market prices in the process.
  
  дипломная работа [5,3 M], добавлен 07.07.2013
  

• Risk in an entrepreneurial activity

  Entrepreneurial risk: the origins and essence. The classification of business risk. Economic characteristic of entrepreneurial risks an example of joint-stock company "Kazakhtelecom". The basic ways of the risks reduction. Methods for reducing the risks.
  
  курсовая работа [374,8 K], добавлен 07.05.2013
  

• Economic Policy of the United Kingdom in the crisis of 2008-2010

  The global financial and economic crisis. Monetary and financial policy, undertaken UK during a crisis. Combination of aggressive expansionist monetary policy and decretive financial stimulus. Bank repeated capitalization. Support of domestic consumption.
  
  реферат [108,9 K], добавлен 29.06.2011
  

• Basic directions of American institutionalism

  State intervention in the economy. Assessment and the role of teaching Veblen. Economic development of the society. Process of long-term loan and the inclusion of investor-banker in industrial production. Negative aspects of American institucionalism.
  
  реферат [27,4 K], добавлен 14.11.2012
  

• Efficiency of economic safety of the coal companies and population of coal regions

  Negative consequences proceeding in real sector of economy. Social stratification in a society. Estimation of efficiency of economic safety. The parity of the manufacturers of commodity production. Main problems of the size of pension of common people.
  
  статья [15,4 K], добавлен 12.04.2012
  

• Agrarian relations and land rent

  The essence of agrarian relations: economic structure and specificity. The land rent, land price as a capitalized rent. History of the formation of agricultural sector of Ukraine, its reforms. Assessment of the investment attractiveness of AIC of Ukraine.
  
  курсовая работа [1,1 M], добавлен 04.01.2016
  

• Облачные вычисления: оценка экономической привлекательности

  Концепции облачных технологий как удаленного вычислительного центра, к которому предоставляется доступ на основе оплаты Pay-As-You-Go. Рассмотрение облачных технологий применительно к "Business-to-business" модели. Экономический взгляд на "облака".
  
  реферат [30,7 K], добавлен 10.12.2014
  

• How competitive is Russian economy?

  Concept of competitiveness and competition, models. Russia’s endowment. Engendered structural dominance and performance. The state of Russian competitiveness according to the Global Competitiveness Index. Place in the world, main growth in detail.
  
  курсовая работа [1,2 M], добавлен 28.05.2014
  

• Predicting completion of cash acquisitions using option implied risk-neutral probabilities

  Estimate risk-neutral probabilities and the rational for its application. Empirical results of predictive power assessment for risk-neutral probabilities as well as their comparisons with stock-implied probabilities defined as in Samuelson and Rosenthal.
  
  дипломная работа [549,4 K], добавлен 02.11.2015
  

• Transfer Pricing

  Special features of multinational corporations. Out the main objectives of a transfer pricing system. Modernisation of business processes of enterprise, use of innovative technologies. Preparing the profit and loss account of the company of Crystal ltd.
  
  курсовая работа [28,6 K], добавлен 16.02.2014
  

• главная
• рубрики
• по алфавиту
• вернуться в начало страницы
• вернуться к началу текста
• вернуться к подобным работам