1
The effect of the financial crisis on the likelihood
of European and American firms going bankrupt
BSc Thesis
Stef Papadopoulos 10983147
Faculty of Economics and Business Specialization: Finance & Organization Supervised by Tolga Caskurlu
2 Statement of Originality
This document is written by Student Stef Papadopoulos who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in
creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of Contents
1. Introduction ... 4
2. Literature Review ... 7
3. Data and methodology ...12
3.1 Sample ...12
3.2 Methodology ...14
4. Estimation results ...15
5. Conclusion ...18
4 1. Introduction
Managerial decisions for investments can influence the health of an entity. Therefore, many research has been done on predicting which factors affect financial distress of companies. As a result, multiple models were created to predict how likely firms are to go bankrupt. More specifically, how likely firms are to be incapable of paying off their debt. Insolvency, which means the outstanding debt cannot be repaid at the time it becomes due, is an important cause of going bankrupt. Edward. I. Altman (1968) created the first model for predicting corporate bankruptcy based on financial measurements. This model had a 94% accuracy rate of predicting firms going bankrupt within one year. Prediction models for bankruptcy are key assets for stakeholders of a company in order to evaluate performance and risk.
The global financial crisis of 2008 had a big impact on the credibility of firms, not only in the United States but also globally (Deyoung, Gron, Torna, & Winton, 2015). After the financial crisis, it has become more important for firms to have a back-up plan for possible drawbacks or disadvantageous state of economies. The Basel Committee on Banking Supervision decided to change capital structures of banks by demanding that 4.5 % of the Risk-Weighted assets of banks will be funded by common equity. This requirement was meant to strengthen the liquidity of banks (Schmaltz, Pokutta, Heidorn, and Andrae, 2014). However, industrial firms do not require capital structure and it is therefore interesting to research if this change has impacted the capital structure of industrial firms after the financial crisis as well. Furthermore, a comparison will be made between firms from the United States and Europe to see in which continent a firm should have a better chance at being active for a long time.
During the financial crisis, the number of firms who filed for bankruptcy increased to the level of 60,0001 in the United States. After 2010, this number decreases once more, until now, which is due to the stabilized economic situation. The expectation is that firms will change their capital structure and other factors which influence the stability of the company. An economic state, like a financial crisis, should be reflected in the firm structure afterward and management decisions to cover potential losses.
5 After the financial crisis, the sovereign debt crisis occurred at the end of 2009 in the PIIGS countries, which refer to Portugal, Italy, Ireland, Greece and Spain. The cause of this crisis was mainly due to the disability of the states to repay their outstanding government debt. This also led to a downgrade of the bond ratings of the PIIGS countries, and the lenders of the government bonds saw their liability risk increasing (Allegret, Raymond & Rharrabti).
For corporations in Europe, some regulations regarding corporate tax rates were changed since 2006. In Figure 1, it is shown that there is a reduction of the tax rate in Europe within the time frame around the financial crisis, but for the United States, the corporate tax rate remained constant. So in terms of encouraging entrepreneurship by integrating low corporate tax rates, a division is created between Europe and the United States. Looking at corporate tax rates, it is expected that more businesses register in Europe than in the USA, but other factors play a role as well. Culture, for example, differs across continents and the economic freedom is more excessive in the US than in Europe.
People living in the United States are more eager to succeed in life and reflect this by owning a big corporate enterprise or earn a large amount of money. Brown, Dietrich, Nunez, and Taylor (2012) conducted a research in the US about the relationship between individual risk-taking and a successful business owner and they found a positive relationship, which means that people who are more risk-neutral, are more successful as a business owner. This should have a positive effect on the number of registered corporations in the United States per year.
Figure 1. Corporate tax rates. The y-ax shows the percentage of the corporate tax rate. All these factors might influence the current health of a firm across Europe and the United States. It is therefore an interesting topic to research whether the probability of bankruptcy
6 has changed after the financial crisis of 2008 and if this probability is higher in the US than in Europe. Thus, the research question will be if European and American firms have become less vulnerable to the likelihood of going bankrupt, after the financial crisis of 2008.
This research expands the existing literature about this topic by comparing the continents Europe and the United States in terms of probability of business failure. Up till now, research has only been conducted within a specific country and industry, so this research adds value to the existing literature in terms of comparing continents. I believe the economic freedom is more excessive in the US than in Europe and risk of failure is higher in the United States, thus firms located in the US are more likely to go bankrupt. Moreover, this thesis examines whether failure probabilities differ after the financial crisis of 2008 compared to the pre-crisis period, due to possible prepackaged managerial decisions after the crisis.
On the other hand, the crisis started in the US due to excessive risk-taking by banks through constructing mortgage-backed-securities. Consequently, American companies were more exposed during the crisis. As a result, it raised awareness about excessive taking which influenced more stable managerial decisions or more risk-averse.
Moreover, the determinants of the model that Altman introduced in 1968 are outdated so in this research, the validity of the variables will be checked for the model when looking at current firms.
Finally, the corporate tax rates differ across countries and they have been changed by the governments in the history. It has not been investigated whether tax rates might influence the weakness of the business and if this might induce extra risk of going bankrupt for firms. It could have a positive effect on the health of a firm because only successful firms can afford to pay higher tax rates and this shows that this liability can be paid by the firm. More well-performing firms will be started relative to bad performing ones if the corporate tax rate in a specific country is higher because successful start-ups can pay high taxes. However, a negative effect on the health of a firm might also be the result of the high expense. Higher tax rates will deduct more costs of the profit and are, therefore
7 an important factor for the success of a business. Taking all of this into account this research will contain the following hypotheses:
Hypothesis 1: European firms are less likely to go bankrupt relative to firms located in the United States
Hypothesis 2: Firms, after the financial crisis, are less likely to go bankrupt.
Hypothesis 3: European firms do not have the same probability of bankruptcy as American firms after the financial crisis from 2007 until 2009.
Hypothesis 4: The variables explaining Altman’s bankruptcy model are still relevant. Hypothesis 5: Corporate tax rates influence the likelihood of bankruptcy.
This thesis proceeds in the following order: the second section in which the related literature will be discussed and reviewed; in the third section, the methodology used for the statistical analysis is described; in the fourth section, the estimation results will be provided and in section 5, a conclusion will be given.
2. Literature Review
Failure of firms happens all the time and, in 1968, Edward Altman created the first model to predict failure of firms called the ‘Z-score model’. Bankruptcy can be interpreted either as referring to the net value of a firm or as a formal declaration of bankruptcy in a federal court, together with a petition to liquidate all total assets or file for a recovery program. Liquidating all of the firm’s assets can be defined as filing Chapter 7 and filing for a recovery program can be defined as filing for Chapter 11 (Altman Book p.20). Altman (1968) argued more academics were using ratio analysis for evaluating business performance. He linked ratio analysis with statistical techniques, which the majority of the academics began to use, even though it was not the right statistical technique in his opinion.
A multiple discriminant analysis is being used for predicting bankruptcies, which classifies groupings depending upon the individual attributes of the observation. It is applied for problems in which the dependent variable consists of two or more groups and
8 where the variable is qualitative. The group's characteristics will be conducted as the independent variables which have discriminant coefficients. These are linearly related to the groups, which in this paper are divided into two groups, namely bankrupted and non-bankrupted firms. This has an advantage over univariate analysis because the MDA technique can take into account a multivariable profile of characteristics which are related to the relevant group, instead of one at the time.
The MDA approach was not popular in the field of finance up to that point and it was mostly used for biological research,2 but Durand (1941) performed a research about consumer credit evaluation and he applied this method followed by a victorious result.
Beaver (1966) however, used a univariate structure of the MDA, to construct a bankruptcy prediction model, which had an 87% accuracy rate, compared to Altman’s model which predicted bankruptcies correctly by 94%.
Considering Altman’s Z-score model, he generated five ratio’s which had a high correlation with the probability of bankruptcy. He conducted the analysis by choosing a sample of 33 bankrupted companies within the manufacturing industry and 33 active companies within the same industry located in the United States, in the period of 1946-1965. Initially, Altman tried to test a correlation among 22 ratios, which were categorized into five groups, namely leverage, solvency, profitability, liquidity, and activity. Eventually, only five ratios were considered relevant and best predictable, which are shown in the following equation of the Z-score:
Z= 0.012X1 + 0.014X2 + 0.033X3 + 0.006X4 + 0.999X5.
A Z-score above 2.99 means a firm is defined as non-bankrupted, while a score below 1.81 is defined as bankrupted. Scores between those values cannot be defined due to error classification.
X1 is the variable Working Capital/ Total Assets. This is a ratio in the category of liquidity, where working capital means the current assets minus the current liabilities of a firm. The ratio reflects how well a firm can pay off their short-term liabilities relative to the firm size. A higher ratio means more liquidity, which has a positive effect on the health of
2 For a comprehensive review of studies using MDA see W. G. Cochran, "On the Performance of the
9 the firm.
X2 measures the ratio Retained Earnings / Total Assets. Retained earnings are the earnings built up since the establishment of the firm. Firms that exist for a longer period have built up more retained profits and this has a positive effect on the health of a firm. Firms which just started are more likely to go bankrupt in the earlier years of existence since they did not have a lifetime to build up the retained earnings.
X3 is the ratio Earnings Before Interest and Taxes / Total Assets. This ratio shows how profitable the firm is with the available total assets. By not taking into account interest and taxes, the Earnings Before Interest and Taxes value shows the productivity of the firm better. A higher ratio results in a higher productivity which has a positive effect on the health of a firm.
X4 is the ratio Market Value of Equity / Book Value of Total Debt. The market value of equity is defined as the total sum of all common and preferred shares and common and preferred stocks represented by the market value. The book value of total debt is measured by current and long-term debt. This ratio reflects the solvency of a firm. A higher solvency ratio has a positive effect on the solvency of the firm, thus a positive correlation with the health of a firm.
X5 is the ratio Sales / Total Assets. This shows how high the bargaining power under competitive conditions is with the available assets. A higher ratio shows a relative better-positioned market position, which has a positive effect on the health of a firm.
Altman’s model could be biased in multiple ways. It was tested for firms that operate in the manufacturing industry. Therefore, it is debatable if it can be generalized to other industries and other countries as well. Industries have different risks because in some industries the competition is very high and other industries could be heavily regulated by the government (Altman, 1993). Kshirsagar (1971) discussed that the sample should be drawn randomly from the population in order to regress the MDA analysis optimally. Altman (1968) used a non-random sample by choosing companies from the same industry and he only chose firms with a total assets size between $1 and $25 million. Firms outside of this range were omitted.
Since this method was biased and the MDA was not used properly due to the sample bias, new methods to predict bankruptcy for corporations were introduced. Ohlson
10 (1980) predicted bankruptcy by using a logit model and Zmijewski (1984) introduced a bankruptcy prediction model by using the probit probability model. Shumway (2001) created hazard models to consider time-changing variables and market data.
Ohlson (1980) took a different approach than Altman to estimate corporate bankruptcy by using a probability function model. Moreover, Ohlson had included 105 bankrupted companies and 2,058 active companies, so his sample size was relatively big. He argued that the MDA approach contained some faults. A requirement for the MDA model is that the covariance-variance matrices had to be the same for both bankrupted and active firms. Another requirement is that the independent variables are normally distributed. Ohlson used dummy variables instead to avoid this requirement. The results of the MDA model are hard to interpret because it only shows a ranked score instead of probabilities. The logit model prevents all these problems. His research showed a predictive accuracy of 96.30%.
Zmijewski (1984) also argues that the MDA analysis is biased, because of non-random samples. He categorizes two different biases, namely choice-based sample biases and sample selection biases. According to him, the first bias is a result of sampling too many distressed firms relative to active firms and the second bias is a result of using complete data. Furthermore, he criticizes previous research about their sample selection which is not representative of the population since the financial distress ratio of the United States has never been more than 0.75 %. Altman’s research sample contains a ratio of 50-50. Zmijewski analyzed bankruptcy probability by using the probit model. His sample met the financial distress population ratio of under 0.75 % for firms with a SIC code under 6000 and between the period 1972-1978.
Lennox (1999) researched whether probit and logit models predict bankruptcy better than an MDA approach. He added industry dummy variables as explanatory variables and his sample consisted of firms in the UK between 1987-1994. Moreover, Lennox tests if heteroscedasticity is present in his research and omitted variable bias. The tests came out positive for the coefficients cashflow and leverage, so they have a non-linear effect on bankruptcy. He avoided over-fitting problems, by not using a sample of one industry and by not matching bankrupt firm characteristics with active firm characteristics. The conclusion of Lennox’s paper was that probit and logit models are a
11 better method to estimate bankruptcy probability.
Shumway (2001) introduced the hazard model into the bankruptcy prediction literature. This model considers that firms change over time and a difference exists in the financial distress period. Some firms file for bankruptcy when they are exposed to risk and other firms file after many years of being exposed to risk. Single-period models are referred to as static models and they do not account for time which the hazard model does. The hazard model also has the advantage that it uses more data observations, because available data for a single firm from a specific year is accounted for as a single observation, meanwhile the average available years of a firm is 10 years, the sample data extends by 10 times. Besides, the hazard model controls for period at risk and it includes covariates that are time-varying. The result was that previously used accounting variables were not significant and market-based variables were more accurate.
Chava and Jarrow (2004) extended the hazard model of Shumway by introducing industry effects and adapting the time intervals from yearly observations to monthly observations. Furthermore, the sample size of bankrupted companies is 1461, which is larger than usual research on this topic. The contribution of the paper validates the better predicting hazard model over the Altman model and Zmijewski model. Next, Industry effects occur and the variables that explain bankruptcy for non-financial firms can also be applied to explain bankruptcy of financial firms.
Another method used to predict bankruptcies is the neural network method. Lippman (1987) defined this model as the best model for forecasting. The neural network model examines correlations between predictive variables which were used for the bankruptcy predicting model. Neural networks are not restricted by the requirement of normality. A study conducted by Odom and Sharda (1990) showed better predictive results than the MDA model.
Taking into consideration all the mentioned methods to predict bankruptcy and literature, the MDA model is not suitable, and it is not used anymore nowadays. The restrictions make it harder as well to analyze bankruptcy. McFadden (1976) showed that the MDA model and the logit model are closely related and can be substituted for research. Since the Z-score is based on an MDA model, we can approach this by using the logit model, which doesn’t have the restrictions of the MDA model.
12 3. Data and methodology
3.1 Sample
All the data has been retrieved from Bureau van Dijk Osiris. Only European industrial firms and industrial firms from the United States are included in the sample, so they can be compared in the analysis. Portugal, Italy, Ireland, Greece, and Spain are excluded in this sample due to the impact of the sovereign debt crisis in these countries. Next, all the required variables of the Z-score model had to be available for the firms in the sample. Furthermore, the time frame is outside the financial crisis between 2007-2009 and the period before the crisis are the fiscal years 2005 and 2006 (Craigwell, Lorde, and Moore, 2013). The period after the crisis is 2010-2012. Now firms can be filtered by location, before and after crisis period, and available ratios.
Next, the firms were filtered by their status. A distinction must be made between distressed firms and non-distressed firms. In Osiris, the company status is presented by the following statuses: Active, Active (receivership), no longer with accounts in OSIRIS, Bankruptcy, Dissolved, dissolved (demerger), Dissolved (merger), Dissolved (Bankruptcy), Dissolved (In liquidation), In liquidation, and Inactive (no precision). Distressed firms are defined as firms with the status Bankruptcy, Dissolved (Bankruptcy) and dissolved (In liquidation). Non-distressed firms are firms with an active status.
Another filter is the Industry Classification Benchmark. In table 1 these are shown by their industry level. This reflects the first digit of the code. The second digit represents the super sector the firm operates in, the third digit is the sector and the fourth digit is the subsector. For this research only industry levels are filtered, to estimate industry effects. Next, the firms with a Total Asset value of zero, Book Value of Total Debt value of zero, and active firms with a sales value of zero were omitted from the sample because otherwise, the analysis will be biased and cannot be computed. Furthermore, firms with extreme values for the ratios were omitted as well from the sample to prevent skewed results.
13 Table 1.
Industry classification code sorted by the first digit
Industry code Industry
0-999 Oil and Gas
1000-1999 Basic Materials 2000-2999 Industrials 3000-3999 Consumer Goods 4000-4999 Health Care 5000-5999 Consumer Services 6000-6999 Telecommunications 7000-7999 Utilities 8000-8999 Financials 9000-9999 Technology
After all the filter requirements, the sample size contains 240 bankrupted companies, and the active firm's sample contains 5392 firms. The distressed firm's ratio is 4.26 %. This percentage is higher than in previously mentioned literature in which this percentage did not exceed 1%. However, Zmijewski (1984) did his research in another period. In table 2 the differences are shown between bankrupt and active firms. All the ratios of Altman’s model are higher for active firms, which on forehand shows expected results. In section 4 the actual significances will be presented.
Table 2.
Descriptive Statistics for active and bankrupted firms
Bankrupted companies Active companies
Variable Obs Mean S. dev Obs Mean S. dev
WC/TA 240 1.54 0.84 5392 4.08 0.76
RE/TA 240 11.56 3.78 5392 36.22 5.38
EBIT/TA 240 0.82 1.66 5392 1.75 0.70
14
Sales/TA 240 0.87 0.06 5392 5.77 4.00 Note: WC= Working Capital, TA=Total Assets, RE=Retained Earnings, EBIT=Earnings before taxes and interest, E=Market Value of Equity, D=Total debt and liabilities. Stdv means Standard deviation
3.2 Methodology
For this research, the logistic probability model will be used. This model is being used the most together with the probit model, which is closely related. The logit model does not contain problematic restrictions and the dependent variable is easy to interpret because it has a binary outcome. The function is given by P (Xi, β) = 1
(1+𝑒𝑌) where P stands for
the probability of going bankrupt with a value of 1 if firm(i) is bankrupt and 0 if firm(i) is active, with the condition 1 ≥ P ≥ 0.
Y = Σβ(i)X(I,j) = β’X(i). and β can be interpreted as the unknown parameters of the variables that will be implied.
The dataset includes unbalanced panel data because the active firms have data available for multiple years, while the bankrupted companies do not have all years available. Either random effects or fixed effects can be applied, while according to Sohn and Kim (2007) random effects have a better predicting accuracy when looking at corporate bankruptcy. Therefore, I will use random effects for the regression.
The variables chosen for this research are given by the Z-score model from Altman plus other variables that could influence bankruptcy based on previous literature and the hypotheses. The regression model is given by:
(𝑌 = 1| 𝑋) = β0 + β1European + β2(WC/TA) + β3(RE/TA) +β4(EBIT/TA) + β5(E/L) + β6(Sales/TA) + β7Post-crisis + β8(Post-crisis*European) + β9(Industry1)-β18(Industry10) + β19(Corporate tax rate) + 𝜀i.
The explanatory variable β1 has a value of 1 if the company is located in the European zone and 0 otherwise. The data has been filtered such that zero will be equal to an American firm
15 Variable β7: Dummy variable with value 1 if the given financial data of firm(i) is stated after 01/01/10 and 0 for financial data before 01/01/07.
Variable β8: Interaction term of a European firm with financial data after the financial crisis of 2008. This variable measures if a European firm after the financial crisis of 2008 has a different likelihood of bankruptcy than firms from the United States after the financial crisis of 2008.
Variable β9- β18: Industry dummies to control for different industries, because some industries are riskier than others, therefore these effects must be controlled. In table 1 the different industries are mentioned.
Variable β19: Numerical variable for the corporate tax rate percentage for firm (i) given the country it is located in and the fiscal year it belongs to.
My expectations of the analysis are as followed:
β1< 0 European firms less likely to go bankrupt relative to US firms.
β2-β7<0 The explanatory variables from Altman’s model are still relevant and have a negative effect on the bankruptcy probability for corporations.
β7<0 Firms after the financial crisis of 2008 are less likely to go bankrupt.
β8<0 European firms after the crisis are less likely to go bankrupt after the financial crisis of 2008 but this digit is less significant compared to the pre-crisis period.
β9-β18 ≠0 Industry effects for different bankruptcy probabilities appear across industries. β19≠0 Corporate tax rates across countries and fiscal years have a significant effect on the probability of going bankrupt for a firm.
4. Estimation results
In this section, the regression will be performed, after having checked the correlation between variables. If variables are highly correlated than multicollinearity is present, which increases the bias of the results. The standard errors of the coefficients will increase with insignificant coefficients of the explanatory variables as a result. In table 3, the correlations between variables are presented. The correlation of crisis and
post-16 crisis*EU is 0.84 which indicates a strong correlation. Therefore, two separate analyses will be executed with either the independent Post-crisis dummy variable included, or the post-crisis*EU variable included.
Table 3. Correlation matrix Tax rate Eu Post-crisis Post-crisis*Eu
WC/TA RE/TA EBIT/TA E/D Sales/TA
Tax rate 1 Eu -.62 1 Post-crisis -.58 0.52 1 Post-crisis*Eu -.72 0.69 0.84 1 WC/TA -.04 0.05 -0.03 0.04 1 RE/TA -.05 0.06 -0.04 0.04 0.67 1 EBIT/TA -.02 0.02 -0.00 0.02 -0.26 0.09 1 E/D -.00 -.01 -0.04 0.04 0.00 0.00 0.00 1 Sales/TA -.08 -.01 -0.01 -0.01 -0.00 -0.05 0.14 0.00 1
In table 4, the regression results are presented with two different regressions. The first regression excludes the post-crisis*EU interaction variable and the second regression excludes the post-crisis dummy variable. Corporate tax rates are only significant in the second regression at a 10% significance rate. The coefficient is positively related to bankruptcy in both regressions, which can imply that the effect of higher tax expenses is greater than the effect of more establishments of start-ups when corporate tax rates are low.
The coefficient of the EU dummy variable is significant at 10% significance level in the first regression and even significant at 1% significance level in the second regression. This implies that European firms are less likely to go bankrupt than firms located in the US. This confirms the expectation of hypothesis 1.
17 The first three variables (WC/TA, RE/TA, and EBIT/TA) from Altman’s Z-score model are contrary to the expectations, positively related to bankruptcy. However, these coefficients are insignificant, so the explanatory power of the three ratio’s has decreased since the establishment of the model in 1968. On the contrary, Altman reported significant results for the first four ratios at 1% significance level. The fourth ratio equity/debt is significant at 10% significance level, and in line with the expectation, because the expectation was that the coefficient would be negative. The results show a negative relation towards bankruptcy. The fifth ratio Sales/Total Assets was expected to have a negative coefficient and the results confirm the expectation. It is significant in both regressions at 10% significance level. Taking into account the significance levels of 10%, it can be concluded that only two out of 5 ratios are significant, but it is not strongly significant and thus the ratios are not relevant anymore.
The coefficient of the dummy variable Post-crisis is negative which is in line with expectations, however, this sign is not significant, so these results cannot be relied on. As a result, hypothesis 2 has to be rejected because the effect is not significant.
The last explanatory interaction variable Post-crisis*EU shows a positive coefficient in the results of table 4, which confirms hypothesis 3. Table 4 shows a significant result at 5% significance level. After the financial crisis of 2008, European firms are more likely to go bankrupt and the cause could be that the crisis had a bigger impact on firms located in the US, that take preparations to increase the leverage rate for example so an economic situation in the future will not influence the existence of a firm. Table 4.
Estimation results
Dependent Variable: (Dummy variable Bankruptcy) Logit
(1)
Logit (2)
Corp. tax rate 0.29
(4.61) 9.42* (5.51) EU -3.48* (1.88) -3.91*** (0.76)
18 WC/TA 0.002 (0.01) 0.002 (0.01) RE/TA 0.001 (0.003) 0.001 (0.002) EBIT/TA 0.01 (0.01) 0.007 (0.005) E/D -0.003* (0.002) -0.003* (0.001) Sales/TA -0.35* (0.75) -0.25* (0.15) Post-crisis -0.30 (0.44) Post-crisis*EU 1.92** (0.83) Industry dummies Random effects N Yes Yes 5632 Yes Yes 5632
Note. *=p≤0.10, **=p≤0.05, ***=p≤0.01. Between parentheses, standard errors are given.
5. Conclusion
The main objective of this thesis was to research differences in the likelihood of going bankrupt between firms located in Europe and the United States. Furthermore, the impact of the financial crisis in 2008 was tested against the likelihood of going bankrupt, and the determinants of Altman’s Z-score model were used as main variables. This research tests if these variables are still relevant, given the fact that the model was established in 1968. Furthermore, based on the literature, industry effects were added to the model, and the influence of corporate tax rate differences across countries was tested on significance.
The literature on this topic shows different models and different variables being used to predict bankruptcy. The models which can be interpreted the best and do not rely on many assumptions like the multiple discriminant analysis are the logit and probit
19 model. Most of the research was performed in the United States within an industry, or by comparing industries. However, this thesis adds value to the literature by comparing the probability of going bankrupt between the United States and Europe.
The results do not confirm all hypotheses. Corporate tax rates have a positive coefficient on bankruptcy, but it is only significant at 10% significance level. European firms are less likely to go bankrupt than US firms overall, which confirms the hypothesis, but after the financial crisis in 2008 firms located in the United States are less likely to go bankrupt.
Working Capital/Total Assets, Retained Earnings/Total Assets, and EBIT/Total Assets show a positive insignificant relationship with bankruptcy, which contradicts the expectation. Earnings/Debt and Sales/Total assets show a significant negative relationship with bankruptcy which is in line with the expectation. The dummy variable Post-crisis shows a negative coefficient which is in line with the expectations, but the result is insignificant.
The explanatory variables of the Z-score model could be biased in this research due to the reason that Altman’s sample contained only firms that have a total asset size up until $25 million in the manufacturing industry. In this thesis, the total asset size exceeds this number and firms from multiple industries are included in the sample. Recent literature confirms that different variables predict bankruptcy better since 1968.
20 6. References
Allegret, J., Raymond, H., & Rharrabti, H. (2017). The impact of the European sovereign debt crisis on banks stocks. Some evidence of shift contagion in Europe. Journal
of Banking & Finance, 74, 24-37. doi:10.1016/j.jbankfin.2016.10.004
Altman, E. I. (1993). Corporate financial distress and bankruptcy: a complete guide to
predicting and avoiding distress and profiting from bankruptcy. New York: Wiley.
Altman, E., (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The Journal of Finance, 23(4), 589-609.
Beaver, W. H. (1966). Financial Ratios As Predictors of Failure. Journal of Accounting
Research,4, 71-111. doi:10.2307/2490171
Brown, S., Dietrich, M., Nuñez, A. O., & Taylor, K. (2013). Business ownership and attitudes towards risk. Applied Economics,45(13), 1731-1740. doi:10.1080/00036846.2011.636026
Chava, S., & Jarrow, R. A. (2004). Bankruptcy Prediction with Industry Effects. Review
of Finance,8(4), 537-569. doi:10.1093/rof/8.4.537
Craigwell, R., Lorde, T., & Moore, W. (2013). Fiscal policy and the duration of financial crises. Applied Economics,45(6), 793-801. doi:10.1080/00036846.2011.613769 Deyoung, R., Gron, A., Torna, G., & Winton, A. (2015). Risk Overhang and Loan
Portfolio Decisions: Small Business Loan Supply before and during the Financial Crisis. The Journal of Finance,70(6), 2451-2488. doi:10.1111/jofi.12356
Gentry, J. A., Newbold, P., & Whitford, D. T. (1987). Funds Flow Components, Financial Ratios, and Bankruptcy. Journal of Business Finance & Accounting, 14(4), 595-606.
Kshirsagar, A. M. (1971), Advanced theory of multivariate analysis. Marcei Dekker, Ini, New York.
21 Lennox, C. (1999). Identifying failing companies: A re-evaluation of the logit, probit and
MDA approaches, Journal of Economics and Business, 51(4), 347-364.
Lippmann, R. (1987). An introduction to computing with neural nets. IEEE ASSP
Magazine,4(2), 4-22. doi:10.1109/massp.1987.1165576
Mc Fadden, D. (1976). A Comment on Discriminant Analysis "Versus" Logit Analysis.
Annals of Economic and Social Measurement, 5(4), 511-523.
Odom, M., & Sharda, R. (1990). A neural network model for bankruptcy prediction. 1990
IJCNN International Joint Conference on Neural Networks.
doi:10.1109/ijcnn.1990.137710
Ohlson, J., (1980). Financial ratios and the probabilistic prediction of bankruptcy,
Journal of Accounting Research, 18, 109-131.
Schmaltz, C., Pokutta, S., Heidorn, T., & Andrae, S. (2014). How to make regulators and shareholders happy under Basel III. Journal of Banking & Finance,46, 311-325. doi:10.1016/j.jbankfin.2014.05.031
Shumway, T. (2001). Forecasting bankruptcy more accurately: a simple hazard model,
The Journal of Business, 74, 101-124.
Sohn, S. Y., & Kim, H. S. (2007). Random effects logistic regression model for default prediction of technology credit guarantee fund. European Journal of Operational
Research,183(1), 472-478. doi:10.1016/j.ejor.2006.10.006
Sowers, D. C., & Durand, D. (1942). Risk Elements in Consumer Instalment Financing. Journal of Marketing,6(4), 407. doi:10.2307/1246534
Zmijewski, M.E. (1984). Methodological issues related to the estimation of financial distress prediction models, Journal of Accounting Research, 22, 59-86.
