Applying, Updating and Comparing Bankruptcy Forecasting Models. The Case of Greece

Daskalakis, Nikolaos; Aggelakis, Nikolaos; Filos, John Article Applying, Updating and Comparing Bankruptcy Forecasting Models. The Case of Greece Journal of Accounting and Management Information Systems (JAMIS) Provided in Cooperation with: The Bucharest University of Economic Studies Suggested Citation: Daskalakis, Nikolaos; Aggelakis, Nikolaos; Filos, John (2022) : Applying, Updating and Comparing Bankruptcy Forecasting Models. The Case of Greece, Journal of Accounting and Management Information Systems (JAMIS), ISSN 2559-6004, Bucharest University of Economic Studies, Bucharest, Vol. 21, Iss. 3, pp. 335-354, https://doi.org/10.24818/jamis.2022.03002 This Version is available at: https://hdl.handle.net/10419/310834 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Accounting and Management Information Systems Vol. 21, No. 3, pp. 335-354, 2022 DOI: http://dx.doi.org/10.24818/jamis.2022.03002 Applying, updating and comparing bankruptcy forecasting models. The case of Greece Nikolaos Daskalakis 1 ,a, Nikolaos Aggelakisb and John Filosa a Department of Public Administration, Panteion University, Greece b Department of Economics, University of Athens MBA, Greece Abstract Research Question: This study examines whether bankruptcy prediction models work well during recessionary periods, on an advanced economy, and how their results can be improved, via a methodological approach to change the coefficients of their variables. Motivation: This is the first study to follow a methodological approach of a simultaneous comparison-update-comparison task, during a recessionary period, for an advanced economy. Idea: The paper explores, updates and compares the effectiveness of five of the most common bankruptcy prediction models on the listed companies of an advanced economy (Greece), covering the recessionary period of 2010-2019. Data: The study sample consists of Greek companies, listed in the Athens Stock Exchange, covering the period 2010-2019, classified into viable and non-viable, based on specific criteria. The final sample consists of fifty-two (52) companies, listed in the Athens Stock Exchange during the period from 2010 to 2019. Tools: We follow a two-stage analysis. First, we apply the original five bankruptcy prediction models of Altman (2000) and Grammatikos and Gloubos (1984), MDA and LPM models, Taffler (1983) and Dimitras et al. (1999) Next, we recalculate their coefficients, keeping the variables stable, and we again apply them to the same sample and compare them again. Findings: We find that the original models are significantly biased against viable companies, but predict with almost perfect accuracy non-viable companies’ bankruptcy. Once we update the variables’ coefficients, we get significantly improved results as regards correctly predicting viable companies, at the expense of slightly decreased, but still high, non-viable companies’ bankruptcy prediction rates. We suggest a similar methodology to be applied in other similar economies, to increase models’ accuracy. 1 Corresponding author: Nikolaos Daskalakis, Department of Public Administration, Panteion University, Greece, email addresses: [email protected] Accounting and Management Information Systems 336 Vol. 21, No. 3 Contribution: The contribution of the paper is threefold. First, we show how we can develop highly accurate bankruptcy prediction models that can be applied in the economic environment of a developed economy. Second, we show that these models work well during recessionary periods as well, and can also be improved when their coefficients are changed. Third, we suggest a methodology of applying, comparing and updating such models, thus showing in detail this improvement process per model. Keywords: bankruptcy prediction models, forecasting. JEL codes: G17, G33 1. Introduction Predicting bankruptcy has always been an attractive issue in the financial academic literature. Since the first attempts to predict bankruptcy in the early decades of the previous century (Fitzpatric, 1932; Smith & Winakor, 1935; Merwin, 1942; Jackendoff, 1962), to the widely known models of Altman (1968), Ohlson (1980) and Taffler (1983), lots of studies have been trying to develop a model that would forecast business bankruptcy as accurately as possible. Over time, forecasting models evolved into skills and complexity; nowadays bankruptcy models involve the application of neural networks in an attempt to offer more sophisticated solutions to the business viability forecasting issue. In time, researchers realized that there could not exist a perfect and universal model, to be applied in all economies worldwide, since economies are structured in different ways and their distinct features affect in turn the viability of the companies that operate within. It would thus make more sense to develop models that would fit economies with similar characteristics, rather than try to develop a unique and universal model. For example, Psillaki and Daskalakis (2009) refer to the importance of the institutional setting of each economy, as captured by individual features such as the bankruptcy law, fiscal treatment, ownership concentration and accounting standards, and develop a cluster of four European countries, namely France, Greece, Italy and Portugal, based on a specific set of financial and institutional indicators for several countries developed by Beck et al. (2008) Indeed, Psillaki and Daskalakis (2009) found that SMEs in these countries determine their capital structure in similar ways; and even though the capital structure field is not identical to the bankruptcy prediction field, researchers could find the idea of similar firm behaviour in similar contexts interesting in being applied in different contexts as well. In this context, we can assume that companies that operate in similar economies, are expected to behave in a similar way financially, so that a bankruptcy prediction model that works for one country may work for other similar countries as well. Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 337 Another dimension that should also be discussed is whether the prediction power of bankruptcy models is affected by differences in the macroeconomic states. Do these models work differently in growth states or in recessionary states? This is something that is highlighted in some studies in the bankruptcy prediction literature. For example, Khoja et al. (2019) denote that the financial health of firms should be examined in situ with the local macro environment. Giannopoulos and Sigbjornsen (2019) admit that one should take under consideration that the 2008 crisis should be taken under consideration when analyzing these models; their sample covers the period during 2002-2012, so that some cases they include take place after the crisis, but, still, the majority of the period they cover refers to a growth period. The purpose of this paper is to investigate the effectiveness of the five most important bankruptcy prediction models in a recessionary period (2010-2019) in a developed economy (Greece) Specifically, we first apply five bankruptcy prediction models as originally designed by their creators and we explore whether they predict bankruptcy during recession, and we then compare their success levels. Second, we update their coefficients, keeping the variables stable, apply them again and compare their effectiveness. To the best of our knowledge, this is the first study to follow this methodological approach of a simultaneous comparison-update-comparison task, during a recessionary period for an advanced economy. Following the discussion above, we suggest that the best model that comes up from our study could be tested in similar economic contexts in other similar economies as well. The remainder of the paper is structured as follows. Section 2 discusses the academic literature of business viability and business bankruptcy and describes in detail the existing bankruptcy models. Section 3 describes the definition of variables, the data used, and the econometric model employed. Section 4 discusses the empirical results and presents an inter-country comparison. Section 5 concludes the paper. 2. Literature review and hypotheses development One of the first questions to be answered in the business viability field is under which preconditions a company can be described as “non-viable”. Beaver (1966) was among the first to work on the business viability academic field and gave a broad definition of viability describing companies as "failed" when they are unable to meet their financial obligations. In particular when the following events occur: a. filing for bankruptcy, b. inability to repay a bond loan, c. overdraft in bank accounts and d. non-payment of dividends on preferred shares. Altman (1968) used the term "bankruptcy", including in his research only companies that were legally bankrupt. In Altman (1977), he extended this definition including cases of companies that had not gone bankrupt despite their high debt, either due to state intervention, or due to their acquisition by banks, or due to a forced merger with Accounting and Management Information Systems 338 Vol. 21, No. 3 other companies. In his later studies, Altman referred to the problems created by state intervention in the correct classification of companies with economic difficulties. The operational support of the companies for a long time after their financial bankruptcy, essentially prevented their legal bankruptcy. The support was mainly in the form of bank loan financing, under state guarantee, to avoid increases in unemployment due to the bankruptcy of large companies. Altman believed that it was wrong to think of the above companies as healthy and viable and also stated that the above state interventions make it difficult to determine the time of the actual bankruptcy of companies. Gloubos and Grammatikos (1988) later referred to the same subject for the case of Greece; they made special reference to the Greek state intervention showing that forecasting models were less accurate in the correct classification of the financial situation of companies in Greece, compared to similar companies in countries that operated without such state support. Deakin (1972) described companies as "failed" when the following events were observed: a. filing for bankruptcy, b. insolvency due to inability to service their financial obligations, c. liquidation of assets to service debts to creditors. In the Greek context, Vranas (1991) used a more general definition of financial failure, referring to the following: a. filing for bankruptcy, b. bank "takeover" through shareholding of debts following a respective decision of the general meeting of the shareholders of the company, c. severe inability to service financial obligations, such as accumulation of overdue debts and inability to pay interest or amortization, for a period of more than three years, d. all businesses that followed specific procedures under a specific Greek bankruptcy law (law 1386/83) After defining bankruptcy, the next step is to create relevant models to predict it. The first attempts to create such models were based on the comparison of financial data presented by two separate groups of companies; one group includes viable companies and the other non-viable. These first attempts belong to the general category of univariate analysis, which focuses on the study of economic indicators. According to Beaver (1966), an early study of the object was made by Fitzpatric (1932) The study included 19 pairs of failed or non-failed businesses and concluded that there are strong differences in the ratios presented by the two groups of companies, at least three years before the failure year. A few years later, Smith and Winakor (1935) expanded their research to 10 years before business failure. They found a steady deterioration in the average values presented by business ratios as they approached the failure year. A few years later, Merwin (1942) compared the average values of the indicators displayed by companies with continuous and discontinuous activity, during the period 1926-1936. He noticed differences between the results of the two groups, up to six years before the failure year in the companies of the second group. The tendency for differentiation was increasing as it approached the failure year. Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 339 The Beaver (1966) model Following these early attempts, the first effort to create a bankruptcy prediction model based on financial indicators was that of Beaver (1966) His sample included 79 pairs of failed and non-failed companies that belonged to 38 different industry branches. His research included 30 indicators which he selected based on various criteria, such as popularity in the scientific literature, success in previous research and correlation with cash flow indicators. Applying a series of tests, he calculated the predictive ability of each indicator individually to correctly classify businesses into failed or non-failed. The indicators that showed the lowest error term in ascending order were the following: • Cash flow / Total debt • Net income / Total Assets • Total Debt / Total Assets • Working capital / Total assets • Current assets / Current liabilities • Available + Receivables / Daily operating expenses Beaver concluded that single-variable analysis could provide reliable information on the course of business and thus business failure. The Altman (1968) model In 1968 Edward I. Altman published his research on "Financial ratios, discriminant analysis and the prediction of corporate bankruptcy" and essentially introduced the Multiple Discriminant analysis (MDA) into forecasting models. According to Altman (1968) a first application of MDA was made in 1936 by R. A. Fisher. Its use was then carried out mainly in the field of biological and behavioral sciences, aiming to predict results in problems where the dependent variable appeared in qualitative form. So, the first step in the MDA was to define specific classification groups, while the next step was to collect data for these groups and try to create a linear combination of their characteristics, that would lead to better distinction between groups. In the case of companies, Altman (1968) defined that these characteristics are their financial ratios, so that the MDA can determine a number of rates for the financial indicators, in order to classify companies into specific groups, such as bankrupt and non-bankrupt. This technique has the advantage that it takes into account a number of characteristics of companies as well as the interaction between them. In his research Altman uses a distinct function of the following form: Z = V1X1 + V2X2 +… +VnXn Where: V1, V2… Vn = Discrete coefficients X1, X2 ... Xn = Dependent variables The above discrete function converts the values of the individual variables into a distinct result called the Ζ value and is used to classify the companies. His sample Accounting and Management Information Systems 340 Vol. 21, No. 3 consisted of sixty-six companies divided equally into two groups. Businesses were classified into categories based on asset size and industry. For each bankrupt company, a non-bankrupt company belonging to the same size and activity category was selected to be included in the sample. In his research, twenty-two financial indicators were studied based on their popularity in previous surveys and their relationship with the conducted research. The variables under investigation were divided into five main categories of indicators such as liquidity, leverage, activity, profitability and solvency. Using various criteria, Altman concluded that from the initial list of variables, the indicators that can best help predict corporate bankruptcies were the following: X1Working capital / Total assets X2Retained earnings / Total assets X3Profits before taxes and interest / Total assets X4Market value of funds / Book value of total debt X5Sales / total assets so that the discrete function selected as the optimal for predicting business bankruptcy is the following: Z= 0.012X1 + 0.014X2 + 0.033X3 + 0.006X4 + 0.999X5 where Z is the total index. His model had a high prediction accuracy with an overall percentage of correct classification of 95%. Type 1 errors (bankrupt companies that were classified by the model as non-bankrupt) occured at 6%. Τype 2 errors (non-bankrupt companies that were classified by the model as bankrupt) occured at 3%. There was therefore a slight upward trend of bias in the group of bankrupt companies. Altman also studied the model's ability to predict bankruptcy up to five years before the event, but found deteriorating results while moving to the past at a rate where the model was unreliable after the second year. Thus, he came up with in a Z value zone that showed great inaccuracy of forecasting and was named "Zone of Ignorance" or "Grey Area". Its values ranged from 1.81 to 2.99 and the companies that lied within these values would be characterized as uncertain. Last, the value of 2.675 was set as the optimal price to distinguish between bankrupt and non-bankrupt companies. The Altman (2000) model In a later study, Altman (2000) reported that several researchers began using a simpler version of the original Z-Score model. Specifically, over the years researchers and practitioners have gradually moved to a more convenient specification of the model that uses an intuitively simpler set of coefficients, as shown below. Z= 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 1.0X5 Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 341 Specifically, the original coefficients of the first four variables were replaced with a x100 multiple where the coefficient of the last variable was rounded to 1.0 (from 0.99) Τhe critical values and indicators used for the classification of businesses remain the same as in the original model. This new model was named Z-Score 2000. The Altman (2010) model The rapid development of computers and their capabilities provided researchers with the tools to develop new methodological approaches. In the years that followed, neural networks emerged helping solve complex problems in the field of bankruptcy forecast as well, while sophisticated bankruptcy tools were developed leading to the creation of several categories of creditworthiness that credit rating agencies still use today. Altman could not fall behind these methodological developments. After his Altman (1994) model that used Neural Networks, his most recent model is that of 2011, which is called the Z-Metrics model. To create this model, Altman (2011) used data from private and public companies based in the US and Canada, covering the period from 1989 to 2009 (Altman E, 2011), and included thirteen variables in total, not only from financial statements but also from the stock market and from the wider macroeconomic environment. His logit model is: CSi,t = α + ΣβΧi,t + εi,t Where: CSi,t: The credit calibration value of each business i during time period t B: The parameters of the variables or otherwise the weights of the variables Xi,t: The variables of the model εi,t: The error terms The CSi,t credit calibration value is then converted into a business credit probability as follows: 𝑃𝐷𝑖,𝑡 =1 1 + exp(CS𝑖,𝑡) Based on the above credit risk probability, companies are classified into a total of fifteen credit reliability categories. The highest category in reliability is called "ZA" and the lowest respectively "ZF". These classifications were performed with estimation time horizons from one to five years. Compared to the ratings of international rating agencies, Z-Metrics presented results with greater accuracy in business ratings. The Taffler (1983) linear discriminant model based on industrial firms in the UK Taffler (1983) also used discriminant analysis to build a model which was applied on financial data of English industrial companies, listed on the stock exchange market. His sample consisted of 23 failed firms and 45 non failed firms during the period 1968-1973. The criteria that he used for the definition of the failed firms was Accounting and Management Information Systems 342 Vol. 21, No. 3 receivership by court order and voluntary liquidation by creditors. The model is as follows: Y = 3.2 + 12.18X1 + 2.5X2 – 10.68X3 + 0,029X4 Where: X1: profit before tax / current liabilities. X2: current assets / total liabilities. X3: current liabilities / current assets X4: (quick assets − current liabilities) / ((sales − profit before tax)/365) The cut-off value for the classification of companies between bankrupt and nonbankrupt is set at -1.95. According to Giannopoulos and Sigbjørnsen (2019) the model was highly accurate of predicting bankruptcy at 80% for one year before bankruptcy reducing to 70% and 58,97% for two years and three years respectively. The Grammatikos and Gloubos (1984) MDA model for Greek companies Grammatikos and Gloubos (1984) built the first MDA model which was based on financial data of Greek industrial companies. They used data from companies’ published financial statements for the period 1977-1981. The model has the following format: Z = -0.863 – 2.461X1 + 5.33X2 – 0.022X3 + 3.676X4 + 3.543X5 + 4.23X6 Where: X1: Current Assets / Total Assets. X2: Working capital / Total assets. X3: Inventories / Working Capital. X4: Bills payable / Total assets. X5: Profits after taxes / Total assets. X6: Gross Profit / Total assets. The cut-off value for the classification of companies between bankrupt and nonbankrupt, is set at 0, while the grey area values were between -0.4754 and 0.2747. The above model was highly accurate of predicting bankruptcy at 91%, for one year before bankruptcy, reducing to 78% and 70% for two years and three years respectively. The model also shows a tendency for more inaccurate classification of bankrupt companies than non-bankrupt ones (type 1 error) The Grammatikos and Gloubos (1984) LPM model for Greek companies Apart from the above mentioned MDA “Z” model, Grammatikos and Gloubos (1984) also built A Linear Probability Model (LPM) “Y”. In the LPM model, linear regression is used to determine the relationship between the dependent quality variable (business viability) and a series of independent variables (economic indicators) The probability of a company to be classified in a given group is a linear function of its financial characteristics. The model has the following format: Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 349 for Y and 42.31% for T) are dramatically improved turning 100% for models Z, Y and T and 83.33% for X. Another interesting result is that the success rates for viable companies slightly improve for the year -2 case, which is not uncommon in the literature (Grammatikos and Gloubos, 1984) Last, we also observe significant improvement in the total accuracy percentages, which are now all statistically significant, except for models Z, X and T, and only for year -3. The high prediction success rates reflect the good adaptation level of the respective observations. The financial data of the remaining twelve companies, that were not selected for the basic sample of Logit regression, were used to determine the accuracy of the forecast models (see appendix III) Table 2 – Success rates of updated models (in percentages) Models Ζ Χ Υ T Actual classification of companies Non-viable Viable Nonviable Viable Nonviable Viable Nonviable Viable Year Non-viable 83.33% 16.67% 83.33% 16.67% 83.33% 16.67% 66.67% 33.33% -1 Viable 0.00% 100.00% 16.67% 83.33% 0.00% 100.00% 0.00% 100.00% Total accuracy 83.33%** 91.67%** 91.67% ** 83.33% ** Non-viable 100.00% 0.00% 100.00% 0.00% 100.00% 0.00% 83.33% 16.67% -2 Viable 0.00% 100.00% 16.67% 83.33% 0.00% 100.00% 0.00% 100.00% Total accuracy 100.00%** 91.67%** 100.00% ** 91.67% ** Non-viable 50.00% 50.00% 50.00% 50.00% 66.67% 33.33% 33.33% 66.67% -3 Viable 0.00% 100.00% 0.00% 100.00% 0.00% 100.00% 0.00% 100.00% Total accuracy 75.00% 75.00% 91.67% ** 66.67% 5. Conclusion In this paper, we apply, compare, update and compare again five popular bankruptcy prediction models. We apply this methodological approach for a developed country (Greece), during a mainly recessionary period (2010-2019); we thus test whether prediction models work well in a recessionary macroeconomic state, plus we then show how we can improve these models, by changing their coefficients. We find that the original models are significantly biased against viable companies (type 2 error), while the non-viable companies’ bankruptcy prediction rates are very high. Once we update their coefficients without changing the variables used, we get significantly improved results as regards correctly predicting viable companies, at the expense of slightly decreased, but still high, non-viable companies’ bankruptcy prediction rates. We thus show, in detail, how the original models are improved, providing Accounting and Management Information Systems 350 Vol. 21, No. 3 researchers with substantial information as regards this improvement process from the original to the updated models. Conclusively, we show that the main bankruptcy prediction models seem to work well during recessions, a concern that was raised by researchers in the field (Khoja et al., 2019; Giannopoulos & Sigbjornsen, 2019) Additionally, we also show how these models can be improved, within this recessionary environment, by applying a coefficient change methodology, that captures these macroeconomic conditions endogenously. Comparing our results with those of previous studies, we observe that we get relatively better results compared to those of Giannopoulos and Sigbjornsen (2019), which is the study closest to ours. Specifically, they show that their overall accuracy rates range from 70 between 70% - 90% one-year prior bankruptcy, 40% - 72% two-years prior bankruptcy and 40% - 67% three-years prior bankruptcy, while our respective rates are 83%-92% (year 1), 92%-100% (year 2) and 66%-92% (year 3) Earlier studies (Dimitras et al., 1999; Grammatikos & Gloubos, 1984) show similar results to the ones of Giannopoulos and Sigbjornsen (2019), but it should be noted that the overall business environment of their studies differs dramatically to the one we use. Overall, we get slightly better results than all previous studies we cite, and this might be attributed to the fact that we apply these models during a mainly recessionary period; this implies that the models seem to be working better during recessionary periods. The aim of the paper is threefold. First to come up with highly accurate bankruptcy prediction models that can be applied in the economic environment of a developed economy (Greece), so that they can be used from practitioners as an additional tool in their fundamental analysis. Second, to show that these models work well during recessionary periods as well, and can also be improved when their coefficients are changed. Third, to suggest a methodology of applying, comparing and updating such models, thus showing in detail this improvement process per model. We believe that a similar methodological process can be applied as such in other countries with similar institutional characteristics, in the context discussed in the paper introduction (Psillaki and Daskalakis, 2009), where businesses in similar economies seem to operate in similar ways. We believe that the paper has fulfilled all research objectives and subsequently contributes to the respective specific fields of the academic literature. Future studies could test the idea of applying-updating and comparing bankruptcy prediction models simultaneously in similar countries. References Altman, E. (1968) “Financial ratios, discriminant analysis and the prediction of corporate bankruptcy”, Journal of Finance, vol. 4: 589-609. Altman, E. (1977) “Zeta analysis-A new model to identify bankruptcy risk of corporations”, Journal of Banking and Finance, vol. 1: 29-54. Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 351 Altman, E. (1994) “Corporate distress diagnosis: comparisons using linear discriminant analysis and neural networks”, Journal of Banking and Finance, vol. 18: 505-529. Altman, E. (2000) “Predicting financial distress of companies: Revisiting the ZScore and ZETA models”, in Bell, A., Brooks, C., & Prokopczuk, M., Handbook of Research Methods and Applications in Empirical Finance, ch. 17: 428-456, Edward Elgar Publishing. Altman, E. (2011) “Ζ-metrics methodology for estimating company credit ratings and default risk probabilities”, Journal of Applied Corporate Finance, vol. 23 no. 1: 20-32. Beaver, W. (1966) “Financial ratios as predictors of failure”, Journal of Accounting Research, vol. 4: 77-111. Beck, T. Demirguc¸-Kunt, A. & Maksimovic, V. (2008) “Financial patterns around the world: Are small firms different?”, Journal of Financial Economics, vol. 89 no. 3: 467–487. Deakin, E. (1972) “A discriminant analysis of predictors of accounting research”, Journal of Accounting Research, vol. 10, no. 1, p.p. 167-179. Dimitras, A.I. Slowinski, R. Susmaga, R. and Zopounidis, C. (1999), “Business failure prediction using rough sets”, Eureopan Journal of Operational Research, vol. 114: 263-280 Fitzpatrick, F. (1932) “A comparison of ratios of successful industrial enterprises with those of failed firm”, Certified Public Accountant, vol. 6: 727-731. Giannopoulos, G. & Sigbjørnsen, S. (2019) “Prediction of bankruptcy using financial ratios on the Greek market”, Theoretical Economic Letters, vol. 9, no. 4: 1114-1128. Gloubos, G. & Grammatikos, T. (1988) “The success of bankruptcy prediction models in Greece”, Studies in Banking and Finance vol. 7: 37-46. Grammatikos, T. & Gloubos G. (1984) “Predicting bankruptcy of industrial firms in Greece”, Journal of Economics and Business, vol. 34, no. 2-4: 421-433. Jackendoff, N. (1962) “A study of published industry financial and operating ratios”, Philadelphia: Temple University, Bureau of Economic and Business Research. Khoja, L. Chipulu, M. & Jayasekera, R. (2019) “Analysis of financial distress cross countries: Using macroeconomic, industrial indicators and accounting data”, International Review of Financial Analysis, vol. 66: 101379. Merwin, C.L. (1942) “Financing small corporations in five manufacturing industries, 1926-1936: A Dissertation in Economics”, National Bureau of Economic Research. Ohlson, J.A. (1980) “Financial ratios and the probabilistic prediction of bankruptcy”, Journal of Accounting Research, vol. 18: 109-131. Psillaki, M. & Daskalakis, N. (2009), “Are the determinants of capital structure country or firm specific? Evidence from SMEs”, Small Business Economics, vol. 33, no. 3: 319-333. Accounting and Management Information Systems 352 Vol. 21, No. 3 Smith, R. & Winakor, A. (1935) “Changes in the financial structure of unsuccesful corporations”, University of Illinois, Bureau of Business Research, Bulletin No. 51 Taffler, R.J. (1983) “The assessment of company solvency and performance using a statistical model”, Accounting and Business Research, vol. 13, pp. 295-308. Vranas, A. (1991) “Probability models for predicting the financial failure of Greek industrial enterprises”, Journal of Economics and Business, vol. 41, no. 4: 431-448. Appendix: Sample of Greek listed firms during the period 2010-2019 I. Initial sample 1 Altec Holdings SA 27 Profile A.E.B.E 2 Alpha Grissin SA 28 Byte Computer A.B.E.E. 3 Microcomputer Systems SA 29 Quality A.B.E.E. 4 Compucon Α.Β.Ε.Ε. 30 Ilida SA 5 Marak Electronics SA 31 Space Hellas A.E. 6 Hellenic Fish Farms SA 32 Galaxidi Thal. Crops SA 7 Crete Farm SA 33 Kri Kri Biom. Milk SA 8 Chatzikranioti SA 34 Stelios Kanakis SA 9 Hellenic Sugar Industry SA 35 Loulis Mills SA 10 Nutriart SA 36 Karaolegos Bakery SA 11 I.Boutaris & Son Holding SA 37 Estate K. Lazaridi SA 12 Hellenic Textile SA 38 ELVE. Α.Β.Ε.Ε. 13 ΑΤΤΙ-ΚΑΤ Α.Τ.Ε. 39 I.Kloukinas SA 14 Engineering 40 Intracom Constructions SA 15 Folli-Follie A.B.E.T.E. 41 A.S. A.E. 16 Sfakianakis SA 42 Motodynamics SA 17 Euromedica A.E. 43 Iasso SA 18 Alco Hellas SA 44 Elval SA 19 Pegasus Publishing SA 45 Attiki Publications SA 20 N.E.L SA 46 Kyriakoulis Shipping SA 21 Spider A.E. 47 Mevaco Metal. A.B.E.E. Applying, updating and comparing bankruptcy forecasting models. The case of Greece Vol. 21, No. 3 353 22 M.I. Maillis SA 48 Karatzi SA 23 Selman SA 49 Iktinos SA 24 Alsinco A.E.E. 50 Douros SA 25 Hatziioannou SA 51 Selected SA 26 Tech. Publications SA 52 Kathimerini SA II. Basic sample used in Logit regression 1 Altec Holdings SA 21 Profile A.E.B.E 2 Alpha Grissin SA 22 Byte Computer A.B.E.E. 3 Microcomp. Systems SA 23 Quality A.B.E.E. 4 Marak Electronics SA 24 Space Hellas A.E. 5 Hellenic Fish Farms SA 25 Galaxidi Thal. Crops SA 6 Crete Farm SA 26 Kri Kri Biom. Milk SA 7 Chatzikranioti SA 27 Stelios Kanakis SA 8 Nutriart SA 28 Karaolegos Bakery SA 9 I.Boutaris & Son SA 29 Estate K. Lazaridi SA 10 ΑΤΤΙ-ΚΑΤ Α.Τ.Ε. 30 I.Kloukinas SA 11 Engineering 31 Intracom Constructions SA 12 Folli-Follie A.B.E.T.E. 32 A.S. A.E. 13 Euromedica A.E. 33 Iasso SA 14 Pegasus Publishing SA 34 Attiki Publications SA 15 N.E.L SA 35 Kyriakoulis Shipping SA 16 Spider A.E. 36 Mevaco Metal. A.B.E.E. 17 Selman SA 37 Iktinos SA 18 Alsinco A.E.E. 38 Douros SA 19 Hatziioannou SA 39 Selected SA 20 Tech. Publications SA 40 Kathimerini SA III. Sample of companies chosen to determine the accuracy of the models 1 Compucon Α.Β.Ε.Ε. 7 Ilida SA Accounting and Management Information Systems 354 Vol. 21, No. 3 2 Hellenic Sugar Ind. SA 8 Loulis Mills SA 3 Hellenic Textile SA 9 ELVE. Α.Β.Ε.Ε. 4 Sfakianakis SA 10 Motodynamics SA 5 Alco Hellas SA 11 Elval SA 6 M.I. Maillis SA 12 Karatzi SA