Do financial ratios matter during corporate restructuring?

(1)

University of Amsterdam

Amsterdam Business School

Master in International Finance

Master Thesis

Do Financial Ratios Matter during Corporate

Restructuring?

Sergii Tymoshchenkov

August 2015

Prof. Chris Florackis

(2)

Abstract

This paper explores the methodology of successful prediction of restructuring of the companies, which happened to be in distress, and summarizes several relevant publications relating to which factors are most sufficient for successful restructuring. This work presents the analysis of financials of the companies, that have made an attempt for restructuring and we have analyzed the outcomes, which helped us to design a model with high rate of successful prediction of restructuring output. As the result, we present a handy tool for rough estimation of restructuring outcome, and potentially important outcome of what factors influence the accuracy of prediction. We expect to contribute to current studies regarding the different models that predict default and probability of success of restructuring.

(3)

Introduction

Two decades of major financial crises in countries from Latin America to Asia, with huge number of companies defaulted, such as AIG (insurance), Land resource LLC (Real estate), Ziff Davis (media), Mechel (steel production), have highlighted the importance of forecasting default risks of enterprises. Within the given period of time number of entities were at breach of financial distress and insolvency, however were restructured and brought to financial soundness throughout years. The experience with corporate restructurings in the aftermath of financial crises shows that successfully restructured firms can relatively quickly return to the pre-crisis performance.

Lots of research has been done with a focus on the performing companies and their probability of bankruptcy. Besides the widely known models such as Altman’s Z-score bankruptcy prediction model, there are researches that concentrate on numerous factors of financial distress, such as using stock returns, adjusted leverage ratios, KMV Merton predictions (Jin-Chuan, 2011), or using liquidity ratios as a key driver of bankruptcy predictions (Damijan, 2014). But, due to the turbulent market conditions and globalization of the economies, and increasing influence of different factors on the companies, it is still hard to take to consideration all the factors that are impacting the insolvency and default of the companies.

It is beyond the scope of this paper to discuss and apply a variety of forecasting methods, but rather to use some simplified and widely used financial ratios, such as ratio of working capital to total assets, current ratio, sales to total assets, etc, that help to assess the overall viability of individual firms. In this paper we majorly discuss the possibility of recovering the company from restructuring in the short-term (1 year) period of time, using the outputs of the designed model, which is based on observations and findings of the distressed debt market.

Since 2000 amount of distressed debt is becoming a huge market; These cases amount in total a staggering almost USD 3.0 trillion within last 15 years, requiring substantial efforts on the part of debtors and creditors, and their advisors, to be restructured so that firms can attempt to emerge from the process as a going-concern (Mead, 2013). In our research we focus on the companies that went through a default stage and tried

(5)

to perform a restructuring. The definition of the default stage is that the company has failed to fulfill their liabilities on any stage, like interest, principal or fee payments. As an outcome of the work we will try to predict the success of this restructuring, based on factors that in our mind has the vast impact on the performance of the company. The idea is to make a research, which is based solely on the officially reported financial data of the selected companies, where the designed model will show us the most important ratios, thus, financial indicators, which influence the outcome of the restructuring and provides the idea, what factors should be studied further to increase the correct prediction of the proposed model. Other incentive of taking only financial ratios of the companies, while restructuring, is the matter of secondary price of the assets in the market. As was discussed by Chen (Chen R.R., 2006) the asset pricing of the companies that are in default, but have not signed the restructuring yet are lower, than the companies, that have already signed restructuring, thus, the uncertainty in the market makes a difference in the asset pricing, therefore, we are considering this as a factor, that should be excluded from the research, because it is a market driven factor. In this work we are trying to create a prediction model of probability of successful restructuring based on recent historical data and identify key indicators to influence this probability.

According to researches there are qualitative and quantitative factors, which are used to measure probability of default. We will choose from the available quantitative models, which forecast the default and based on the selected model, we will create a model, which will show the most important factors in forecasting the success of restructuring.

(6)

Literature review

There is a vast literature on predicting financial distress and forecasting default rates of companies. The success of various methods is less satisfactory and remains a matter of dispute among financial economists.

As argued by A. Khaliq, et al (A. Khaliq, 2013), financial distress is the situation when a company cannot meet or face difficulty to pay off its financial obligations to the employees, state or creditors. The chances of causing economical distress increases when a company’s fixed costs are high, assets are illiquid, or revenues are too sensitive to economic recessions. A firm, which faces financial distress, can experience costs, such as high borrowing rates and opportunity costs of projects, which will affect the firm’s predicted probability of default in a widely used prediction models such as Z-score model. This model was created in 1967 and represented most current and valuable economical enterprise indicators of that time. Such as working capital, total assets, retained earnings, earnings before interest and tax, market value of equity, total liabilities, total assets and sales. According to the model, the higher the score, and the lower the likelihood of bankruptcy. For example, a score below 1.8 means the company is probably headed for bankruptcy, while companies with scores above 3.0 are not likely to go bankrupt. Since then, economical environment has changed a lot. It resulted in declining of the accuracy of the score model. In the paper by Altman (Altman, 2005), researchers tested the Z-score model on the enterprises, which went through distress, 86 companies in range of 1969-75, 110 bankruptcies from 1975-95 and 120 from 1995-99 was accurate in between 82 and 99 percent, meaning the cut-off of the Z-score model was below defined 2.675. However mainly after 2000, the number of companies, which were classified as distressed but didn’t go for bankruptcy has increased substantially with as much as 15-20 percent of all firms and 10 percent of the largest firms which were below 1.81 in the Z-score model.

Meanwhile, Altman Z-score model takes to account sole financial endogenous indicators, measuring corporate default using financial ratios. Later in his work, Altman (Altman E. H., 1977) discusses the upgrade of the Z-model to ZETA model by adding ratios to the existing Z-score, such as introducing debt service coverage ratio and size ratio, which is turning the model into a more accurate model. However,

(7)

the model still uses only financial indicators without adding descriptive and qualitative variables.

In a paper by Chen et al (Chen R. S. Hu, 2006), researchers performed a comparison of corporate default models, namely, Olshon O-score model, which predicts default using financial ratios as well as market data and sector information, Moody’s KMV Merton default model, which studies returns of the assets verses the probability of default, as well as Merton, Longstaff and Schwartz, Flat Barrie, Black and Cox, and Geske models etc. All the discussed models are mainly focused on default prediction in the different time horizons. According to Chen et al. (Chen R. S. Hu, 2006), Altman (Altman E. a., 2013), and other researchers, the main factors, which are taken to the account, while discussing the probability of default and distressed debt, are focused on the secondary pricing of the distressed debt and the returns on the distressed portfolio. According to our observations, while trading the distressed debt, the returns from the distressed portfolio are artificially pushed down. If we observe the market of the distressed debt, we can come to the conclusion, that at the moment of signing the restructuring agreement the prices are going up, since it removes the uncertainty, however, the financial state of the company remains the same. According to Hotchkiss and Mooradian (Hotchkiss, 1997), the combined efforts of the restructuring specialists, including investment bankers and turnaround- management consultants, along with the coincident growth of institutional investors (buy-side) and broker-dealers (sell-side), have enabled this enormous amount of defaulted debt to be restructured reasonably effectively, providing significant returns.

The key problem in restructuring is the probability of the company recovery from restructuring meaning repayment of liabilities in full in short-term and long-term perspective. For the last 15 years trading of distressed debt increased significantly, as discussed by Altman (Altman E. a., 2013); To address the question of the impact and role that institutional investors (e.g., Hedge Funds) have had on the restructuring process, one can cite the impressive growth, scope and specific actions that these investors, and others, have played in the evolution of the reorganization methods and post-restructuring process over the last 20-plus years.

Together, with our estimate of the amount of distressed debt under management, discussed above, these statistics form the demand and supply dynamics, which is so

(8)

critical for any viable financial market. These dynamics have provided the incentive for a special breed of investors, experienced in distressed investing, to attract capital and, as mentioned earlier, provide a potential outlet for original investors to monetize their troubled assets over a period that can stretch from a year or more before the insolvency occurs, and lasting throughout the duration of the restructuring process. This liquidity is crucial to those other investors who do not have the resources expertise or desire to hold their claims until the resolution of the reorganization. Also, ability to estimate residual values in the event of default, is crucial for non-investment grade firms to raise capital from the so-called “junk bond and loan,” or leveraged-finance market, a market that is estimated now to be over USD 2 trillion; Mead and Natarjan (Mead, 2013).

Taking to account all possibilities of restructuring, there are few existing methods of restructuring and post-restructuring developments. Within last 15-20 years the global and local economies were hit by different crisis developments, which reveled inefficiency and financial weakness of the enterprises. The companies that experienced restructuring after the crisis were impacted by the turbulence in economies, however, the crisis revealed the weak positions in the companies, such as lack of equity, shortage of working capital, inability of matching assets with liabilities, as discussed by Hotchkiss and Mooradian (Hotchkiss, 1997).

According to researches of various hedge funds (Jain, 2012), (Casa T.D., 2008), major investors are looking for companies with rather strong equity, market position, sufficient working capital, which gives higher probability of success rate, or in other words less probability of default of the company. Given data in research represents companies with significant asset size, which are either world significant players in the market or have a strong local presence. Selected enterprises have performed the restructuring with the different outcome and the idea behind this work is to build a model, which is highlighting main quantitative ratios that are affecting performance and influences the outcome of restructuring. As discussed by Jiang, Li and Wang (Jiang, 2012), qualitative factors are also significant in the restructuring process, therefore, apart from the quantitative ratios in the model, we are introducing factors to the model which are, in our opinion, are mostly significant in course of result of restructuring after observing the market data and success rate of restructuring.

(9)

Following the research of Luan (Luan F., 2011), we can consider qualitative assessment method is a kind of artificial judgments by experts. Even though, the dataset in the research allows us to differentiate the qualitative factors and implement them in the model, the key objective is to focus on financial data, rather then having qualitative research of factors, which are difficult to measure. As it was discussed in the research of (Luan F., 2011), the aim of value assessment of enterprise bankruptcy by taking to account qualitative factors is to range the differences in approaches, and the qualitative factors used as ones which explain outliers in the researches.

According to Jiang, Li and Wang (Jiang, 2012), as well all as the market evidences and researches (Lemmon M., 2009), there are four possible outcomes from the restructuring. First outcome of the restructuring is a successful restructuring, where the company has repaid to the debt- and stakeholders all of the mentioned liabilities in full, being operational, without disposal of assets of the firm. The second outcome mentioned is a significant support from the third parties, namely shareholders or government. According to the research, major companies from Russia and China (USA is an example as well, however, government steps in in order to avoid systematic crisis) that faces insolvency, are supported by the government on many levels. Third outcome of the restructuring are merges and acquisitions, which are impacting the companies and the entity is no longer solely responsible for its liabilities. This outcome became very popular, due to significant leverage of the companies and introduced mechanism of debt to equity swaps. Even though, the M&A are also subject to default, it is no longer observed as a separate enterprise and the new merged companies are treated as a new enterprise with own risks of bankruptcy. The fourth outcome of restructuring is bankruptcy procedure. According to Altman (Altman E. a., 2013), and his work studying debt trading and Chapter 11 (Bankruptcy procedures in US) filings, the restructuring is possible even after the company has filed for bankruptcy and there are plenty of cases of successful restructurings after filing. It is beyond the scope of this work to study Chapter 11 filings, since the Chapter 11 includes moratorium on debt enforcement, disposal of assets and management restrictions of the company.

(10)

Data descriptive statistics

To assess the influence of financial indicators of the company on the results of success of restructuring of the international firms and subsequent empirical analysis, we make use of the financial statements and balance sheet data for the selected companies on the basis of quarterly statements, one year prior to default. All of the selected companies did attempt for debt restructuring and either failed or succeed to perform restructuring. Data come from the hedge funds and counter parties, which are involved in the distressed debt trading or in the restructuring of the loans of the companies, such as Deutsche Bank, Commerzbank, GarantiBank, SC Lowy Hedge Fund. We are focused on the companies, which cannot be influenced by the support of the government, hence the companies, which are filed for bankruptcy in US, are “ring-fenced” from the influence of the lenders and stakeholders can be treated as supported by the government.

While all enterprises are companies, that have either strong presence in the local market or have international business; their financials are recorded by Thomson Reuters.

Some of the financials were provided by above-mentioned parties on the disclosed basis. Based on the available public information, we choose to disregard companies, that were accused of fraud and/or illegal actions, which were connected to the operational or financial activities of the companies. The reason for the omission is twofold. First, data, which involves fraud cases, tends to be very noisy and often suffers from less reliable reporting. And second, the functioning of those types of firms is likely governed by non-economical motives. We are focused on the companies, which cannot be influenced by the support of the government, hence the companies, which are filed for bankruptcy in US, are “ring-fenced” from the influence of the lenders and stakeholders can be treated as supported by the government.

The scope of the information gathered by Thompson Reuters are provided according to IFRS accounting standards, however reported in the original currency and different digits; in order to provide information in one scale, we have converted all the figures into United States Dollars, using respective currency rates (where required), at the

(11)

moment of reporting.

This, in turn, presents some challenges in establishing precision of the required variables throughout the sample observation time-period and some compromises have to be made with respect to the level of detail extracted from the data.

This, in turn, presents some challenges in establishing precision of the observed variables throughout the sample period and some compromises have to be made with respect to the level of detail extracted from the data.

While we originally have 216 firm-year observations at our disposal, after we perform some data cleaning (dropping out firms which were convicted of fraud, companies, that didn’t report 4 quarters prior to default, as well as companies, which are considered to be small-size companies, with the assets size below USD 30 million) we are left with 200 firm-quarter observations. This translates into 50 enterprises from 1997 to 2013.

According to our estimation, the efficiency of restructuring and influence of certain financial indicators can be visible starting with the midsize and larger companies, therefore, based on the European commission definition of the mid size companies, they should have the assets size in the range of USD 30 million (European Commission, 2003). Thus, in the selected sample, the minimum amount of asset size is above the threshold of USD 30 million, which represents the above-mentioned companies that fit the criteria.

About quarter of the enterprises observed are representing FMCG sector, meaning production and retail sales of FMCG goods, another quarter represents different type of service firms. The rest companies are occupied in production, IT and media.

Data set represents 17 countries, where the one fifth of which comes from USA and other fifth is representing Spain. The rest companies are coming from different countries worldwide, which meet the requirements, specified above.

From the random sample of selected companies, 24 out of 50 have completed the restricting successfully, meaning that according to the public information, the company has managed to repay stated amount of liabilities to stake and debt holders.

(12)

Out of 26 companies, that failed to restructure, 10 companies were acquired by other market players, and continued to perform, but with sufficient support from the acquiring company, thus they didn’t complete the debt restructuring by their own means. 2 companies, from the selected sample required government support, which was granted, therefore we consider those companies failed restructuring. Even though the above-mentioned enterprises, which were acquired and received government support continued to operate successfully, we assume them as non-successful restructuring, since we can not predict what should have happened, if those companies were not influenced by third parties.

Table 1: Data characteristics of the sample of defaulted firms

This table depicts descriptive statistics of the selected companies. “Countries” (Column 1) is showing the origin/registration country of the company. “Sector” (Column 3) stands for the industry, in which the company is operating. “Original currency”(Column 5) is the currency, which company used for audited reporting. “Restructuring” (Column 7) represents the number of successful or failed restructurings.

Countries Sector Original currency Restructuring

Australia 2 Construction 5 Australian Dollar 2 Fail 26 Canada 1 Engineering 2 Euro 25 Success 24 Denmark 1 Entertainment 3 British Pound 3

France 4 FMCG 12 Indian Rupee 2 Germany 2 IT 1 Israeli shekel 1 Greece 1 Logistics 5 US Dollar 22

India 2 Media 2 Italy 1 Metals&Mining 5 Israel 1 Services 12 Netherlands 1 Telecom 3 Norway 1 Portugal 1 Russia 1 Spain 11 UK 8 Ukraine 1 USA 11

One of the key features of the dataset, as shown in Table 1, is that the companies are selected randomly, providing the wide scope of activities, size and based in different countries worldwide, within the mentioned time period. The chosen companies, are giving the opportunity to focus mainly on the endogenous factors, like firm’s financials and restructuring team, disregarding concentration risk factors like industry downturns, legislation risks or location disadvantages.

(13)

Model construction background

The definition of company, which operates under the financial pressure, is a subject, discussed by numerous economists. Prior to the development of quantitative measures of company performance, economists established a quantitative measurement of companies’ performance, regardless of the company’s size, analyzing the company’s creditworthiness. One of the classic works in the area was performed by Altman, which has introduced a model, analyzing the companies, in different financial shape and was forecasting company’s bankruptcy. In his model Altman was comparing the enterprises and in his work he created a model, which was sufficient enough to determine the companies, which were likely to default using certain ratios.

The Altman’s model has an evaluation of future bankruptcy of the company, however, in this work we are discussing the recovery from the occurred event of default. Since Altman’s (Altman E. B., 2005) model is pretty accurate in prediction of corporate default, we are using the same ratios in the work.

The prediction of the Z-model is quite accurate; hence as a base of the model, we will take Altman’s Z-score ratios as base variables and tune the model with the selected variables from other default prediction models.

This study will explore the effect of each ratio of the Z-score model, having the linear regression formula of Z = 1.2X₁ + 1.4X₂ + 3.3X₃ + 0.6X₄ + 0.99X₅. As defined, in the model, each coefficient is measuring liquidity, profitability, efficiency and assets turnover, which are considered by Altman (Altman E. H., 1977) as the main drivers of financial solvency of the company:

X1 = Working Capital / Total Assets.

The ratio measures the liquidity of assets of the company, in relation to the size of the company. This shows the relation of the amount of liquid assets, as given the formula Current assets less current liabilities, and it’s scale on the operational influence of the company, explaining the inverse relationship as a driver of financial distress, e.g. the more negative Working Capital is, the higher probability of default is.

(14)

X2= Retained Earnings / Total Assets.

The ratio measures profitability that reflects the company's earning power. Likely, that the company which carries a significant loss in subsequent years, will have the higher insolvency index, which shows a direct relationship. Scaled on the asset size, the retained earnings reflect the compatible size of profit or loss on the size of the company.

X3 = Earnings Before Interest and Taxes / Total Assets.

The ratio explains operating efficiency apart from tax and leveraging factors, scaled on the total assets, which depicts good metrics in measuring profitability of the company, which tends to be the driver for insolvency prediction, which leaves out the cash required to fund working capital and the replacement of old equipment, which can be significant, but already present in the above indexes.

X4 = Market Value of Equity / Book Value of Total Liabilities.

The index is showing market dimension that can reflect security price fluctuation as a warning. Generally, the market value of equity is a sign for the management of their default risk. In the selected data, for the privately owned companies, but with the tradable debt securities, the value of equity will be adjusted by market value of the corporate debt, and the discount, which is applied to the tradable security will be applied to the booking equity level. X5 = Sales / Total Assets.

Given ratio measures total asset turnover to contribution to default state of company. This index helps to measure, by how the company is performing on the sales side and gives an opportunity to evaluate if the distress is occurred by poor performance of the given periods.

Besides the ratios of the model, there are other significant factors that influence the possibility of recovering, such as country of origin, industry sector, crisis affection of the market and influence of the third parties on the restructuring. According to Chen, (Chen R. S. Hu, 2006), mentioned factors are used in various models and have certain

(15)

level of significance. We tried to implement those factors in the model and they will be discussed further.

Before trying to create the prediction model we calculated the Z-score for the selected companies, in order to estimate the financial shape of the companies one year prior to default. According to Z-score calculation, for the sample of selected companies, which had experienced financial distress, 5 companies got the score above 3.0 points which states that the company is not likely to default, 32 companies, which have the Z-score less than 1.8, which are likely to default and 13 companies had a score between 1.8 and 3.0 and considered to be in the grey zone.

Table 2: Z-score of successfully restructured and non-successfully restructured companies

“Median Z-score” (Row 2) represents the median Z-score of restructured companies versus the companies that failed to restructure. “Average Z-score” (Row 3) shows the average Z-score of the companies, that succeeded in restructuring and those that failed respectively.

Restructured Non-restructured

Median Z-score 1,2314 1,1487

Average Z-Score 1,7366 1,4804

The median as well as the average of annually calculated Z-score of successfully restructured enterprises is expectedly higher, and they are in line with economical sense. The successfully restructured companies have higher coefficients, however, there are some examples, where the successfully restructured companies have a significantly lower than a default score and managed to restructure successfully and vice versa.

We calculated each of the ratios for the companies in pre-default year, 4 quarterly figures will be taken to account, both who managed to survive the distress and who didn’t, and having that data, will run a logistic regression in order to predict a default using the ratios mentioned above in the function, as well as qualitative variables. The initial model has the view as:

Probability of restructuring = Region + Crisis + Sector + Third parties + 𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟏𝟏)+ …+𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟒𝟒)+ … +𝐗𝐗_𝟓𝟓(𝐪𝐪_𝟒𝟒),

Probability of restructuring of the company was presented as a dummy variable as 1 if the company has succeeded in the restructuring and 0 if the company failed.

(16)

Where Region is the region variable, which divides the sample companies by the region, into 3 groups: Europe (31), US (12) and Other countries (6),

Crisis – is the crises variable, which depicts the companies (10), which went to distress in time when there was a crisis in the financial market in 2008-2009.

Sector – is the sector variable, which reflects the company’s operating sector and their odds ratio in the probability of recovery from default depending on the industry. Third parties – is the variable, which reflected the influence on restructuring the company such as merge and acquisition or government influence over the enterprise.

Variables X₁ – X₅ represent the ratios from the Altman Z-score model on the quarterly basis one year prior to default. All in it was 20 ratios for the period of interest.

We used the logistic model to predict the success of restructuring. After trying several cut offs values, we decided to stick to mathematical rounding, which resulted in the success values, which greater than 0.5 and the non-success values which are less than 0.5.

In order to estimate the accuracy of the model prediction, we used the confusion matrix, or error matrix approach, which allows visualizing accuracy of the algorithm performance. It is often used to present information about observed and predicted values.

Following the potential limitations of the model, due to limited dataset, we had to restrict our number of variables within the model to a maximum of ten, thus we tried to restrict amount of initial variables doing backwards selection based on p-values. Therefore we ended up with three models, based on different variables, testing the economical plausibility of the financial ratios and model variables, which mostly affect the consequences of the restructuring.

We run three models with different data to ensure the best prediction outcome based on the below mentioned assumptions. In all the models we used the EU companies as the benchmark, since it is the largest group of companies in the dataset. After running several models, with different variables, we noticed that the factors like sector

(17)

variable, crisis variable and third party influence are not significant in the model, meaning that the factors of the variable doesn’t influence the outcome of the prediction and the values and p-values are not significant.

In the first model we used the assumption that the factors, which affect the success of restructuring should be the first quarter of the year prior to financial distress, due to the understanding that the company is still operating and have decent financial indicators. According to our Z-score quarterly calculations, the score observed the first quarter of observed period is, in predominant number of observations, the highest score within the selected periods. We took fourth quarter, meaning the last quarter before default, to estimate the affect of weak financial indicators. Following the same logic, the Z-score of the last quarter prior to insolvency is the lowest score, which depicts the higher probability of default, which doesn’t contradict the financial sense. The prediction model is as follows:

Probability of restructuring = Region + 𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟏𝟏)+…+ 𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟒𝟒)+ … +𝐗𝐗_𝟓𝟓(𝐪𝐪_𝟏𝟏)+…+ 𝐗𝐗_𝟓𝟓(𝐪𝐪_𝟒𝟒), (Model A)

Where names of variables are mentioned above in the base model.

The second model takes into account the ratios of quarter three and quarter four of the companies, meaning two consequent quarters prior to default. The rationale behind the selection is to estimate the pace of situation and how it escalated closer to default. According to our calculation, the Z-score of the companies in the third quarter is predominantly higher than the last quarter prior to default, therefore we might also see the downward trending slope in the model, reflecting the situation in the enterprises closer to default.

The second prediction model is as follows:

Probability of restructuring = Region + 𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟑𝟑)+𝐗𝐗_𝟏𝟏(𝐪𝐪_𝟒𝟒)+ … +𝐗𝐗_𝟓𝟓(𝐪𝐪_𝟑𝟑)+ 𝐗𝐗_𝟓𝟓(𝐪𝐪_𝟒𝟒), (Model B)

Where names of variables are mentioned above in the base model.

In the third model we made an assumption, that the success of the restructuring is based on the company’s performance first and second quarter, meaning the half a year to default date as the last observation point. According to our estimation, the

(18)

enterprises which performed better before the default have the higher margin of safety, therefore, will be more likely successfully restructured.

The third prediction model is as follows:

(19)

Model output

All models that are listed in the material section represent all range of prediction probability, starting from 62% to 86%. The initial model has the highest prediction probability, which equals to 86%, but bears the insignificant p-values of the variables, probably due to small sample size.

Table 3: Output of initial model

“Estimate of coefficient” (Column 2) is the value of each coefficient, predicted by the model. “Standard error” (Column 3) shows the standard error of the coefficient in the linear regression model. “Wald Statistics” (Column 4) represents the z value (Wald statistic) for testing the hypothesis that the corresponding parameter (regression coefficient) is zero, where under the null hypothesis it has an approximately N(0,1) distribution. “P(>|z|)” (Column 4) is the tail area in a 2-tail test, i.e. the test within a 2-sidedouter hypothesis. “(Intercept)” (Row 2) stands for constant term (the mean of the responce) in linear regression. “x.Region.US”(Row 3) is the dummy variable for the countries, which are registered in US. “x.Region.Other” (Row 4) is the dummy variable for the countries, which are registered elsewhere, other than US and EU. “X_1…5(q_1…4)” (Rows 5-25) stands for ratios of the respective quarter, i.e. X₂(q₄) stands for ratio of Retained earnings to Total assets, using the financials of fourth quarter.

Estimate of coefficient

Standard Error Wald statistic (z-value) Pr(>|z|) (Intercept) 0.66129 0.92975 0.711 0.477 xRegion.US -1.9199 1.38139 -1.39 0.165 xRegion.Other 0.18435 1.41114 0.131 0.896 X1(q1) 1.79871 3.48004 0.517 0.605 X1(q2) 0.48308 3.43933 0.14 0.888 X1(q3) 1.10939 3.05799 0.363 0.717 X1(q4) 1.45513 2.80725 0.518 0.604 X2(q1) 10.4337 6.56223 1.59 0.112 X2(q2) -15.3868 11.68356 -1.317 0.188 X2(q3) 6.80936 8.14662 0.836 0.403 X2(q4) -1.18699 3.09506 -0.384 0.701 X3(q1) -21.0065 14.56258 -1.442 0.149 X3(q2) 49.31046 24.13322 2.043 0.041 X3(q3) -12.1362 14.27259 -0.85 0.395 X3(q4) -5.35859 6.81819 -0.786 0.432 X4(q1) 0.07704 4.47714 0.017 0.986 X4(q2) -2.92157 4.77258 -0.612 0.54 X4(q3) 5.2583 3.69138 1.424 0.154 X4(q4) -3.92632 3.52539 -1.114 0.265 X5(q1) 6.03159 12.92197 0.467 0.641 X5(q2) -15.6543 12.84235 -1.219 0.223 X5(q3) 7.01535 8.6689 0.809 0.418 X5(q4) 1.3828 8.95422 0.154 0.877

(20)

The only variable, which is statistically significant (p-value less 0.05), is X₃(q₂). There were other certain variables, which were close to be statistically significant, but probably due to sample size, they couldn’t breach p-value 0.05 level. The results of the model is expected to be as aforementioned, since we observer high inner correlation in the model. Even though, we can state that state that due to negative values in the X₄ and X₅coefficients, which reflects the size of the company, or mobility of the company are depicting the hypothesis, that for the smaller the company is, the easier for to change the track of business and to apply certain changes to business model to overcome distressed periods. On the other hand, predominantly positive coefficients of the variables X₁, X₂ and X₃ show, that the company should bare high liquidity in order to resist shortage of working capital and EBIT.

However, according to the confusion matrix, the insample correct prediction of the model is 86%, with the following breakdown.

Table 4: Initial model confusion matrix

“P’(Predicted success)” (Column1) stands for the predicted positive outcome of the model; “N’ (Predicted default)” (Column 2) stands for predicted negative outcome of the model. “p’(Actual success)”(Row 2) shows the actual success of restructuring; “n’(Actual Default)” (Row 3) shows the actual default, hence non-succesfull restructuring.

P' (Predicted Success) N' (Predicted Default)

p' (Actual Success) 21 2

n' (Actual Default) 5 22

We can state that the model is pretty accurate in prediction of successful restructuring with only 2 false negative and 5 false positive predictions.

(21)

The first tested model outcome is as follows:

Table 5: Output of model A

Standard Error Wald statistic (z-value) Pr(>|z|) (Intercept) 0.3505 0.5911 0.593 0.553 xRegion.US -0.9505 0.8253 -1.152 0.249 xRegion.Other 0.1011 1.1214 0.09 0.928 X1(q1) 0.1328 1.8732 0.071 0.943 X1(q4) 2.3208 1.7629 1.316 0.188 X2(q1) 1.6489 2.3968 0.688 0.491 X2(q4) -1.3426 1.988 -0.675 0.499 X3(q1) 1.3346 5.4817 0.243 0.808 X3(q4) 1.9558 3.6555 0.535 0.593 X4(q1) 0.8415 1.8682 0.45 0.652 X4(q4) -1.2902 1.7302 -0.746 0.456 X5(q1)) -2.6092 4.3219 -0.604 0.546 X5(q4) 2.098 4.6596 0.45 0.653

According to the output of the model, there are no significant p-values, which can be explained by the fact that those variables are not significant in defining the probability of succesfull restructuring. The confisuon matrix shows, that insample accuracy of the prediction model is 62%, with 10 false negative prediction and 9 false positive predictions.

(22)

Table 6: Model A confusion matrix

P' (Predicted Success) n' (Predicted Default)

P' (Actual Success) 16 9

The second tested model output is as follows:

Table 7: Output of model B

Standard Error Wald statistic (z-value) Pr(>|z|) (Intercept) 0.3543 0.5876 0.603 0.547 xRegion.US -1.0316 0.8541 -1.208 0.227 xRegion.Other -0.3779 1.1331 -0.333 0.739 X1(q3) 1.2372 2.1529 0.575 0.566 X1(q4) 0.7538 2.0086 0.375 0.707 X2(q3) 1.9635 2.3859 0.823 0.411 X2(q4)) -1.6006 2.4024 -0.666 0.505 X3(q3) 7.8086 6.7312 1.16 0.246 X3(q4) 0.5796 3.7123 0.156 0.876 X4(q3) 2.0701 2.188 0.946 0.344 X4(q4) -2.7164 2.4482 -1.11 0.267 X5(q3) -4.7659 4.6066 -1.035 0.301 X5(q4) 4.3535 5.207 0.836 0.403

According to the model output, there are values, close to significant p-values, which can be explained by the fact that those variables are closer to statistically significant and have more influence on the prediction. The confisuon matrix reflects, that

(23)

insample accuracy of the prediction model is 74%, with 10 false negative prediction and 9 false positive predictions.

Table 8: Model B confusion matrix

The third tested model output is as follows.

Table 9: Output of model C

“Estimate of coefficient” (Column 2) is the value of each coefficient, predicted by the model. “Standard error” (Column 3) shows the standard error of the coefficient in the linear regression model. “Wald Statistics” (Column 4) represents the z value (Wald statistic) for testing the hypothesis that the corresponding parameter (regression coefficient) is zero, where under the null hypothesis it has an approximately N(0,1) distribution. “P(>|z|)” (Column 4) is the tail area in a 2-tail test, i.e. the test within a 2-sidedouter hypothesis. “(Intercept)” (Row 2) stands for constant term (the mean of the responce) in linear regression. “x.Region.US”(Row 3) is the dummy variable for the countries, which are registered in US. “x.Region.Other” (Row 4) is the dummy variable for the countries, which are registered elsewhere, other than US and EU. “X_1…5(q_1…4)” (Rows 5-14) stands for ratios of the respective quarter, i.e. X₂(q₂) stands for ratio of Retained earnings to Total assets, using the financials of second quarter.

Standard Error Wald statistic (z-value) Pr(>|z|) (Intercept) 0.4897 0.7746 0.632 0.5273 xRegion.US -1.1647 0.9345 -1.246 0.2126 xRegion.Other -0.4569 1.1376 -0.402 0.688 X1(q1) -1.446 2.2047 -0.656 0.5119 X1(q2) 4.9911 2.8013 1.782 0.0748 X2(q1) 7.8035 5.4929 1.421 0.1554 X2(q2) -6.8445 4.8932 -1.399 0.1619 X3(q1) -17.333 11.4746 -1.511 0.1309 X3(q2) 29.4672 14.8842 1.98 0.0477 X4(q1) 3.7518 3.2896 1.141 0.2541 X4(q2) -4.2776 3.3198 -1.289 0.1976 X5(q1) 3.7232 6.5066 0.572 0.5672 X5(q2) -5.7834 5.7037 -1.014 0.3106

Following the third model output, we can observe the best outcome of the p-values, with the results which are close to statistically significant and the highest rate of the of

(24)

the accuracy of the prediction according to the confusion matrix of 78%, with 4 false negative and 7 false positive predictions.

Table 10: Model C confusion matrix

(25)

Discussion of the results.

The initial model (table 2), which is the base model, which is using 𝑞𝑞₁- 𝑞𝑞₄ (4 quarters) and the X₁ - X₅ratios have the insample accuracy prediction of 86%, with EU as a benchmark. Since we observe internal correlation within the variables of the initial model, probably due to the fact that we are using same financial indicators (i.e. Revenues/Total assets and Current liabilities/total assets) as well, small sample size, we were running another 3 models with maximum amount of variables concerning sample size (5 per 1). Therefore after using backwards selection of the model we have come up with 3 models, which have a prediction rate in the range of 62-78%.

The first tested model (A) uses all Altman score ratios from 𝑞𝑞₁, 𝑞𝑞₄ and EU as a benchmark, which gives a 62% probability of successful prediction. In the model (table 2) we have only one variable which is close to significant value, which is X1(𝑞𝑞4) and has the coefficient of 2.32, the ratio of working capital to total assets reported in the 𝑞𝑞₄ (1 quarter prior to default). According to this observation, we can state that the company’s need of working capital is mostly affecting the outcome of restructuring. We observe linear dependence between the successful restructuring and the higher ratio of working capital to assets, which make economical sense. As mentioned before, within the sample, the Z-score of the fourth quarter is predominantly lowest score among all the quarters, therefore the enterprise with the high default prediction rate will hardly get the cheap funding.

The second tested model (B) uses all Altman score ratios from 𝑞𝑞₃, 𝑞𝑞₄ and EU as a benchmark, which gives a 74% probability of successful prediction. From the model we can state that the values of variables X₃ of the third quarter are close to statistical significance levels. X₃ variable depicts ratio of EBIT to total assets, which reflects linear dependence, so the higher EBIT in 𝑞𝑞₃, the higher chances of the company to be successfully restructured. Compared to the first model, this variable as well explains liquidity of the company and its influence on the success of the restructuring. If we compare the confusion matrixes outputs, the model is more accurate with almost the same amount of false-negative predicted outcome (10 versus 9), but with lower rate of false-positive predictions (9 versus 4). This shows that the parameters, which were taken in 𝑞𝑞₁ and 𝑞𝑞₄ and in 𝑞𝑞₃and 𝑞𝑞₄ has almost same accuracy in prediction positive

(26)

result of the restructuring, however, when the predicting the negative outcome of the restructuring, the data in 𝑞𝑞₃ and 𝑞𝑞₄ are more accurate in the prediction the result which is a logical assumption, since 𝑞𝑞₃ and 𝑞𝑞₄ provide the model dynamics of the ratios, which are getting lower, as mentioned above, while comparing Z-score’s results.

The third tested model (C) uses all Altman score ratios from 𝑞𝑞₁, 𝑞𝑞₂ and EU as a benchmark, and has the highest (out of three models), 78% probability of successful prediction. Another compatible advantage of the model, that it has more statistically significant results, which are X₁(𝑞𝑞₂), X₂(𝑞𝑞₁) and X₃(𝑞𝑞₂). The most significant result is the X₃(𝑞𝑞₂) with the value of 29.4, evidencing a linear dependence on the success of restructuring, which shows EBIT to total assets ratio, the economical rationale of which was explained above. The X₁ ratio, showing the necessity of working capital in the company is also confirmed in the first tested model. The third ratio, with a significant p-value is X₂(𝑞𝑞₁), which is a ratio of retained earnings to total assets, which shows the margin of safety of enterprise. The more cash the company collected within their lifecycle, the longer the company would stand the tough economical environment. Comparing the outputs of confusion matrixes, we can see that the rate of correct prediction increased due to decrease in the false-negative predictions. We can explain it by the fact, that the structure of financials of the company’s are more stable in longer period prior to default, which is proved by the Z-score calculation as well. We can notice as well, that the wrong prediction of success of restructuring, which logically should be more visible in the later periods (sooner to default), is not the correct way of estimating the success of the restructuring. We refer to PwC report (PwC, 2012), where the auditors explain the irregularities of the accounting statements prior to bankruptcy by the will of the management of the companies to produce financial reporting, which doesn’t reflect the reality. According to auditor’s research, the companies are trying to re-evaluate value of certain items of balance sheet such as Intangible assets, Provisioning items, Long term and Short term payables and Receivables, which result in non-fair reporting, thus leads to unfair judgments of the company’s financial performance and margin of safety. Hence, following the third model, the companies, which are expected to have restructuring, should be observed not from the later periods, but starting relatively far from the restructuring date.

(27)

Conclusion

After conducting the research on the probability of prediction of the restructuring outcome, we designed a model, which can be used as a handy tool for a rough estimate of the probability of success of restructuring, with high probability of correct prediction. Also, we noticed several patterns, which can be used for further researches on this topic.

Firstly, we noticed that even using only qualitative factor model we could reach the correct insample prediction level of the model, which is 86%. That proves the earlier cited work of Chen (Chen R. S. Hu, 2006), who discusses the qualitative factors in the model, which in his opinion is not significant in the model and used mostly to explain outliers and sharpen the prediction rate of the model, which indirectly proven by the high prediction rate of the solely quantitative model.

Secondly, we discovered the interconnection between the factors, which are mostly affecting the successful prediction of restructuring. The output of the model showed that the company with high liquidity, meaning “rich in cash” companies, have high success rate in the positive outcome of restructuring. Stated that, the further research, which will be conducted on this topic, should take to account the free cash generating ability of the companies is highly important in terms of success in restructuring and success in the prediction of outcome of restructuring, where the equity of the company, as well as their turnover is less important.

Third, we explained, why we should pay more attention to the earlier financial statements, meaning, using the ratios from the earlier quarterly reporting prior to default rather than later quarters, which are closer to default date in order to predict restructuring outcome precisely. Since we have an evidence of financial manipulations in accounting, which were proven by researches of the auditing firms, namely PwC, we would suggest to study the performance of the company in earlier periods prior to default, to catch the trends and incentives of the businesses of the company that might lead to a successful or not-successful restructuring and neglect the later periods.

(28)

We would like to outline that the model itself has certain limitations, like limited sample size, thus further research on the topic is required, however, the model gives a general idea of what are the key factors of the correct prediction of the restructuring outcome.

(29)

References

Altman, E. (1991 ). Distressed Securities, Analyzing and Evaluating Market Potential and Investment Risk. Probus Publishing .

Altman, E. (1992). The Market for Distressed Securities and Bank Loans. Foothill Group , Los Angeles.

Altman, E. a. (2013). Defaults & Returns in the High-Yield Bond and Distressed

Debt Markets: 2012 in Review . NYU Salomon Center Special Report &

Paulson & Company .

Altman, E. B. (2005). The Link Between Default Rate and Recovery Rates:

Implications for Credit Risk Models and Procyclicality. Journal of Business,

78(6), 2203.

Altman, E. H. (1977). ZETA Analysis: A New Model to Identify Bankruptcy of Corporations. Journal of Banking and Finance.

Casa T.D., R. M. (2008). Hedge fund investing in distressed securities. MAN Investments.

Chen R. S. Hu, G. P. (2006). Default prediction of various structural models.

Working Paper.

Chen R.R., F. F. (2006). Sources of credit risk: evidence from credit default swaps.

The Journal of Fixed Income , 16, 7-21.

Damijan, J. P. (2014). Corporate financial soundness and its impact on firm performance. European Bank of Reconstruction and Development Journal. European Commission. (2003). Commission Recommendation 2003/361/EC .

Official Journal of the European Union , 124, 36.

Hotchkiss, E. a. (1997). Vulture Investors and the Market for Control of Distressed Firms. Journal of Financial Economics, 43(2), 401.

Jain, S. (2012). Investing in Destressed Debt. Alternative Investment Analyst

Review, 43-46.

Jiang, W. K. (2012). Hedge Funds and Chapter 11 . The Journal of Finance , LXVII (2), 573.

Jin-Chuan, D. (2011). Multiperiod Corporate Default Prediction; A Forward

Intensity Approach . Singapore: National University of Singapore .

Khaliq, A. (2013, January). Determinants of Financial Distress Evidence. Business

Review, 8(1), 7-9.

Lemmon M., M. Y. (2009). Survival of the fittest? Financial and economic distress

and restructuring outcomes in Chapter 11. University of Utah, Department

of Finance, Utah.

Luan F., L. J. (2011). The Construction of Assessment System: Bankruptcy Reorganization Value of Corporate. Orient Academic Forum, (pp. 229-232). Beijing.

Mead, C. a. (2013, August 26). Leveraged Debt Exceeds $2 Trillion in Repression: Credit Markets . Bloomberg News .

PwC. (2012). A look at current financial reporting issues. PricewaterhouseCoopers LLP. PwC.