The Hazard Model for European SMEs: Combining Accounting and Macroeconomic Variables

Journal of Competitiveness 76 The Hazard Model for European SMEs: Combining Accounting and Macroeconomic Variables ▪Michal Karas Abstract Predicting the default of small and medium-sized businesses (SMEs) using the hazard model approach represents an area relatively neglected by mainstream literature. On the one hand, SMEs are regarded as the backbone of the economy; on the other hand, their specific features pose a challenge to the modelling process. This issue is further complicated by the fact that many modern structural approaches to default modelling are simply unsuitable for SMEs due to their limited size. Therefore, researchers only rely on accounting, non-financial, or macroeconomic data. The gap is especially noticeable in several studies on SME default prediction that employ the hazard model approach, which models the probability of default with respect to the time factor. A better understanding of the factors driving SMEs’ default might help in adopting policies that strengthen their competitiveness. The aim of this study is to introduce a hazard model for EU-28 SMEs and analyse the contribution of macroeconomic indicators and proxies of external financial obstacle factors. This model was derived using the Cox semiparametric proportional model, leaving the baseline hazard unspecified and employing macroeconomic variables as explanatory variables. By analysing a sample of 202,209 European SMEs over the period 2014–2019, the results indicated that factors of employment rate, personal cost per employee, and interest rate play significant roles in determining the survival of SMEs. Adding these macroeconomic variables significantly increased the area under curve values compared to the situation where only accounting variables were used. Keywords: SMEs, corporate default, Cox model, hazard model, competitiveness, baseline hazard rate. JEL Classification: C14, G17, G33 Received: March, 2022 1st Revision: June, 2022 Accepted: June, 2022 1. INTRODUCTION The current literature on the prediction of corporate default, especially the default of large businesses, is quite extensive, with two dominating approaches to the prediction of defaults: Altman’s (1968) approach utilising accounting ratios and Merton’s (1974) structural approach based on the employment of market-driven variables. Despite the extensive number of papers published on the prediction of corporate default, only a small fraction focus on the default of small and medium-sized businesses (SMEs). A better understanding of the factors that cause Karas, M. (2022). The Hazard Model for European SMEs: Combining Accounting and Macroeconomic Variables. Journal of Competitiveness, 14(3), 76–92. https://doi.org/10.7441/joc.2022.03.05 joc2022-3_v2.indd 76 29.9.2022 13:59:17 77 the default of SMEs could help them adopt appropriate financial decisions and preserve or even enhance their market competitiveness (Kliestik et al., 2020). This study aims to derive a hazard model for European SMEs and analyse the contribution of added macroeconomic variables to the model. Unlike previous studies, the Cox proportional model (Cox, 1972) was employed, while the baseline hazard rate was left undefined and macroeconomic variables were employed as explanatory variables. This approach makes it possible to employ more than one macroeconomic indicator, as would be the case when such a variable is used to specify the baseline hazard rate. A further difference lies in the analysis of EU28 data and the focus on the SME segment, as most hazard studies focus solely on national data (commonly the US and the UK) and few consider the SME segment rather than listed businesses. 2. THEORETICAL BACKGROUND The specifics of SMEs pose serious challenges to credit risk modelling (Karas & Režňáková, 2021; Kramoliš & Dobeš, 2020; Dankiewicz et al., 2020), as highlighted by Filipe et al. (2016) who noted that other contemporary credit risk modelling approaches such as Merton’s (1974) structural approach or the latest approach of modelling based on credit default swap (CDS) spreads (Trujillo-Ponce et al., 2013) could not be used for SMEs. This is because Merton’s approach requires daily stock price data to estimate the asset volatility of the company. At the same time, most European SMEs are small and do not satisfy the entry requirements of stock exchanges, and market data are simply not available. Furthermore, the CDS spreads approach is not applicable to SMEs, as these businesses do not have CDS information and do not issue bonds, as their main funding sources are their own resources and bank debt (Filipe et al., 2016). From this perspective, accounting data are the main source of usable information. It is worth mentioning that the accounting-based approach is criticised for the historic nature of the information it uses as input and does not consider the volatility of a firm’s assets when estimating the risk of default (Vassalou & Xing, 2004). In contrast, Agarwal & Taffler (2008) provide three arguments in favour of the accountingbased approach. According to their study, corporate distress is not a sudden event; there is a low probability that a firm with good profitability and a strong balance sheet will file for bankruptcy due to a sudden change in the economic environment. Furthermore, corporate failure is the culmination of several years of adverse performance and is generally recognised in the firm’s accounting statements. Moreover, the double-entry system of accounting ensures that windowdressing accounts or changing accounting policies will have a minimal effect on a measure that combines different facets of accounting information in financial ratios. From a general perspective, SMEs can be viewed as a rather specific segment of businesses. However, they play a highly significant role in the economy (Civelek et al., 2021; Ključnikov et al., 2021; Virglerova et al., 2021). Gupta et al. (2015), among others, point out that SMEs are considered the backbone of the global economy and are viewed as an important route to recovery in the aftermath of the global financial crisis of 2008–2009. De Moor et al. (2016) added that SMEs are regarded as an economy’s engine for sustainable growth and stable employment. According to Eniola et al. (2015), ‘the importance of SMEs in the evolution of the economy, joc2022-3_v2.indd 77 29.9.2022 13:59:17 Journal of Competitiveness 78 reduction in poverty, increase in employment, output, innovation in technology, and boost in social position and standard is globally proven and acknowledged in emerging as well as in developed economies’. The importance of SMEs in the economy is underlined by their numbers. By 2015, regarding the EU-28’s non-financial business, the vast majority (92.8%) of businesses employed fewer than 10 individuals. Conversely, just 0.2% of all enterprises had 250 or more employees and were therefore classified as large enterprises. Large businesses have greater weight in terms of value-added and work provided, as large businesses provided work to more than one third (33.7%) of the EU-28’s non-financial business economy workforce and generated 43.5% of its value added (Key figures on Europe, 2018). A review of the literature on predicting the default of SMEs shows that only a fraction of the studies focuses on default prediction issues in SMEs. However, there are several arguments highlighting the necessity of treating SME default risk specifically and separately from the risk related to large businesses (Virglerova et al., 2020). First, studies suggest that SMEs are more vulnerable to changes in economic conditions and financing and face more obstacles than larger and older businesses (Virglerova et al., 2021). According to Ullah (2019), financial constraints are the most significant obstacle that SMEs must face during their growth. The given level of constraints depends on the economic development of a given country (Gavurova et al., 2020). The more liquid the stock market, the more efficient the legal system, and the higher the GDP per capita, the lower the obstacles reported (Beck et al., 2006). Second, the importance of treating SME defaults separately from the defaults of large businesses was later accented by the change to the Basel regulatory framework (i.e., the Basel II Accord). Edmister (1972) published the first study on default prediction in SMEs, although his study did not highlight the need to treat the SME segment separately. Altman & Sabato (2007) showed that deriving a default prediction model specifically for SMEs may result in significantly higher accuracy than that achieved using a generic model. Altman et al. (2010) later addressed this problem by exploring the importance of adding non-financial data while predicting SME distress. Goncalves et al. (2016) and Belas et al. (2018) addressed the importance of non-financial data in the credit risk of SMEs. From the modelling perspective, there are several studies criticising the current approach, in which the significance of the predictors is proven for one period of time only as being static because the time factor is ignored. Grice & Dugan (2001) investigated Ohlson’s (1980) and Zmijewski’s (1984) models and concluded that the precision of both models degraded significantly when they were applied to different data samples. They postulated that the relationship between financial figures and bankruptcy might change over time. Only a limited number of studies, such as those by Holmes et al. (2010), Gupta et al. (2015), El Kalak & Hudson (2016), and Gupta et al. (2018), have addressed the application of hazard models for SME default modelling. Gupta et al. (2015) argued that the SME segment is not homogenous, with huge diversity in terms of capital structure, company size, access to external finance, management style, and the number of employees. Gupta et al. (2015) further highlighted that heterogeneity has been neglected in empirical studies on SME credit risk. The authors applied the discrete-time duration-dependent hazard rate to a large sample of non-financial UK SMEs during the period 2000–2009 while adopting the European Union definition of SMEs. joc2022-3_v2.indd 78 29.9.2022 13:59:17 79 Their model was separately developed for micro, small, and medium-sized businesses, while a comparison of the model’s version results suggests that the segment of micro-businesses should be treated separately from the rest of the SME segment. Gupta et al. (2015) used the logarithm of company age, insolvency rate, and industry ‘weight of evidence’ variables to control the survival time and macroeconomic conditions. El Kalak & Hudson (2016) applied the same approach as Gupta et al. (2015) to a sample of SMEs in the USA during the period 1980–2013, when the Small Business Administration (SBA) was in operation. El Kalak & Hudson (2016) confirmed the conclusion reached by Gupta et al. (2015) regarding the necessity of treating micro-businesses separately from the rest of the SME segment due to different (i.e., lower) survival probabilities. However, El Kalak & Hudson (2016) also pointed out that Gupta et al.’s (2015) approach of utilising an insolvency rate variable as a baseline hazard rate distorts the baseline hazard idea. 3. RESEARCH OBJECTIVE, METHODOLOGY AND DATA The sample consists of 202,209 SMEs from EU-28 countries, covering 2014–2019. Of these SMEs, 59,709 went legally bankrupt within one year, while financial statements from the prefinal period (a year prior to bankruptcy) were analysed. In this study, a business is considered a small company if its operating revenue is less than 1 million EUR, its total asset value does not exceed 2 million EUR, and the number of employees is less than 15. A business is considered a medium-sized company if its operating revenue does not exceed 10 million EUR, its total asset value does not exceed 20 million EUR and the number of employees is less than 150. The sample was randomly divided into a learning part (70% of all observations) and a testing part (30%), and later adopted receiver operating characteristic (ROC) curve methods; this approach was also employed in relation to the hazard model by Gupta et al. (2015). In line with Gupta et al. (2015), I tend to differentiate between small and medium-sized businesses, as the SME segment is not homogenous, and the reason behind treating small businesses separately from medium-sized businesses is based on their lower expected survival probability. A dummy variable (called the category of a company) was added to control for this. Further, I employ an industry dummy (‘IND’) to control the industry effect for two reasons. First, the analysed data comes from businesses in different industries. Second, it has been shown that industry specifics play a significant role in bankruptcy prediction (specifically in the case of a hazard model, see Chava & Jarrow, 2004 or, for a more general perspective, Grice & Dugan, 2001). In line with Chava & Jarrow (2004), the following industry grouping was employed: IND 1 – Miscellaneous industries, IND 2 – Manufacturing and mineral industries, IND 3 – Transportation, communications, and utilities, IND 4 – Finance, insurance, and real estate. The preliminary results of the data analysis showed that several variables clearly exhibit extreme outlier values; the variables under analysis were winsorised at the 1 or 99 percentile level to ensure that the results or the estimated parameters were not negatively influenced by this effect. joc2022-3_v2.indd 79 29.9.2022 13:59:17 Journal of Competitiveness 80 3.1 Potential company-specific variables Empirical studies dealing with the hazard approach and SMEs were reviewed and a list of potential variables was collected (see Table 1). Information on the expected variable sign was also drawn from these studies; in several cases, the authors explicitly stated the expected sign, while in other cases, the sign was drawn from the final model details (i.e. parameter estimates published in the studies). Tab. 1 – List of analysed ratios. Source: 1 – Altman et al. (2010); 2 – Campbell et al. (2008); 3 – El Kalak & Hudson (2016); 4 – Gupta et al. (2015); 5 – Gupta et al. (2018); 6 – Hillegeist et al. (2004); 7 – Chava & Jarrow (2004); 8 – Shumway (2001) Abbr. Description Ex. sign. Abbr. Description Ex. sign C/TA cash/total assets3;4 (-) QA/TA quick assets/total assets4(-) CA/CL current assets/ current liabilities1;7;3 (-) QR quick ratio3;5 (-) CA/S current asset/ sales3(+) RE/TA retained earnings/ total assets8;7;6;4;3;5 (-) CashR cash ratio5(-) S/TA sales/total assets8;7;6 (+) CE/TL capital employed/ total liabilities1;4;3;5 (-) S/TTA sales/tangible assets5(-) CL/E short-term debt/ equity4;3;5 (+) SHP stock holding period5(+) CL/TA current liabilities /total assets3(+) size ln(total assets/ GDP price level index)6(-) DCP debtor collection period5(+) ST/TA stock/total assets4(+) EBIT/CE EBIT/capital employed5(-) St/WC stock/working capital1(+) EBIT/S EBIT/sales5(-) T/TA taxes/total assets4;5 (-) EBIT/TA EBIT/ total assets 8;7;6 (-) TC/TA trade creditors/total assets4;3 (+) EBITDA/ IE EBITDA/ interest expenses4;3;5 (-) TC/TD trade creditors/trade debtors1(+) EBITDA/ TA EBITDA/total assets4;3;5 (-) TC/TL trade creditors/total liabilities1;4 (+) FE/S financial expenses/sales5(+) TCPP trade creditors payment period5(+) FE/TA financial expenses/ total assets3;5 (+) TD/TA trade debtors/total assets4(+) IA/TA intangible assets/ total assets4(+) TL/NW total liabilities/net worth5(+) joc2022-3_v2.indd 80 29.9.2022 13:59:17 81 Ln(age) natural logarithm of age (number of days)7 (-) TL/QA total liabilities/quick assets1(+) log (CA/ CL) log (current assets/ current liabilities)4 (-) TL/TA total liabilities/total assets8;7;2;3 (+) NI/E net income/equity3;5 (-) TL/TTA total liabilities/ tangible total assets5(+) NI/S net income/sales3;5 (-) WC/S working capital/sales3(-) NI/TA net income/ total assets8;7;6;2 (-) WC/TA working capital/total assets8;7;6;5 (-) 3.2 Research methods The Cox semiparametric proportional model approach was employed to derive the model, which was first adopted by Lando (1998), who was the first to model default with the Cox model. Further seminal work in this field was conducted by Shumway (2001), who demonstrated the superiority of the hazard model approach in predicting business defaults over the static approach model (i.e., not considering the multi-period nature of the data). The superiority of the hazard approach has also been confirmed by other authors, such as Chava & Jarrow (2004) and Berent et al. (2017). According to Gupta et al. (2015), ‘the discrete hazard modelling technique is well suited to analyse data that consists of binary dependent variables and exhibits both time-series and cross-sectional characteristics, such as bankruptcy data’. The advantage of the Cox semiparametric hazard model is that its estimation is possible even when the baseline hazard function is left unspecified, which ‘offers a considerable advantage when we cannot make a reasonable assumption about the shape of the hazard’ (Cleves et al., 2008). Applications of the hazard model are most often inspired by Shumway’s (2001) seminal study, which showed that the discrete-time hazard model is equivalent to a multi-period logit model. In contrast, authors tend to specify the baseline hazard rate. Generally, there are two main approaches to the specification of the baseline hazard rate. The first is to use time dummies, as shown by Beck et al. (1998), or employ macroeconomic variables, as suggested by Nam et al. (2008). They argue that indirect measures, such as time dummies, are less effective in capturing time-varying macro dependencies. In this study, the Cox semiparametric model was used, leaving the baseline hazard rate unspecified and employing macroeconomic variables as explanatory variables. Therefore, this approach is different from that of other studies (e.g. Nam et al., 2008). The main difference is that, with this approach, the macroeconomic variables influence the hazard rate through a shift of baseline hazard (as other explanatory variables), which seems to be useful as the analysis deals with panel data. joc2022-3_v2.indd 81 29.9.2022 13:59:17 Journal of Competitiveness 82 3.3 Selecting model variables To select the variables, I employed the same two test procedures used by El Kalak & Hudson (2016). The starting point of the procedure was the derivation of a univariate model for each of the analysed variables, while the variables that exhibit significant estimates and had the expected sign were retained for further analysis. This approach has also been widely adopted by other authors (e.g. Altman et al., 2010; Gupta et al., 2015; Nam et al., 2008). The categorical or dummy variable categories of company and industry were formed to control for the heterogeneity of the SME segment and the heterogeneity among the different industries. Before adding these variables to the otherwise univariate model or first-step model, the KaplanMeier procedure was run, together with a log-rank test, to test the equality of survival functions to gain insight into survival functions for all these categorical variables. In the case of unequal survival functions, the common approach of deriving the univariate model must be adjusted to control for differences between the groups. The next step was to run a multicollinearity check. If a highly correlated pair of variables was identified, the covariate with the lowest Wald statistic value was excluded from the final multivariate model. 3.4 Potential macroeconomic variables In line with Nam et al. (2008), I employed macroeconomic variables to capture time-varying macro dependencies and, most importantly, as this study deals with panel data, to capture differences between European countries arising from different levels of economic development. The choice of potential macroeconomic variables was inspired by previous studies on hazard models or other studies dealing with default risk or SME financial constraints. These are expected to reflect the specific features to which SME survival is sensitive. The analysed macroeconomic variables are listed in Table 2, and the potential link to the survival probability of a business is discussed later. Data on macroeconomic variables were drawn from the EUROSTAT database. Tab. 2 – Overview of hazard model literature employing macroeconomic variables. Source: own research Macroeconomic variable Literature Ex. sign Exchange rate Holmes et al. (2010), Nam et al. (2008) (+) Interest rate Christidis & Gregory (2010), Tinoco & Wilson (2013), Holmes et al. (2010), Nouri & Soltani (2016), Hillegeist et al. (2004) (+) Gross value added (GVA) per employee Holmes et al. (2010) (-) Personal cost (PC) per employee Holmes et al. (2010) (+) Inflation Christidis & Gregory (2010), Nouri & Soltani (2016), Tinoco & Wilson (2013) (+) Employment Holmes et al. (2010) (-) joc2022-3_v2.indd 82 29.9.2022 13:59:17 83 GDP annual growth rate Simons & Rolwes (2009), Nouri & Soltani (2016) (-) GDP per capita Beck et al. (2006) (-) According to Holmes et al. (2010), the exchange rate factor may be particularly important for SME survival, as they are more likely to ‘face competition from abroad and be involved in exports and imports’. Exchange rate changes are expected to have an adverse effect on company survival as change ‘implies a worsening of the competitive position relative to overseas competitors’ (Holmes et al., 2010). The exchange rate from local currency to EUR was employed in this study, while the data were drawn from the Amadeus database, which quotes the exchange rate based on data from the International Monetary Fund (IMF) website; the exchange rates refer to the closing date of the statement. Interest rates influence a company’s survival probability through its capital structure; low-interest rates are incentives for companies to make investments, and the expected return on investments is higher when interest rates are lower than when interest rates are high. However, high interest causes rising costs of debt capital, with firms having to pay more to their lenders (Tinoco & Wilson, 2010). Higher interest rates are, for this reason, expected to increase the probability of a company’s failure. In this study, the yield on government bonds with a maturity of 10 years was adopted as the interest rate variable; such interest rates are used to define the Maastricht criterion for long-term interest rates. Gross value (GVA) added is expected to have a positive influence on a company’s survival (decreasing the probability of failure) since increasing GVA is associated with a growing market. In contrast, any wage increase (personal costs) means a rise in costs and is therefore expected to increase the probability of failure (Holmes et al., 2010). Inflation is expected to indirectly affect the probability of a company’s default by serving as an incentive to invest savings rather than seeing their purchasing power erode further in the future due to inflation. Inflation, therefore, increases investors’ risk-taking capacity and, as a result, reduces the default probability (Tinoco & Wilson, 2013; Qu, 2008). However, as acknowledged by Qu (2008), the direction of the inflation effect on default probability has not been unequivocally established due to the complexity of the effect of inflation on the economy. Mare (2012) noted that a high inflation rate is a sign of weak macroeconomic conditions, during which there is also an elevated risk of a bank crisis. In this study, we adopted the Harmonised Index of Consumer Prices (HICP), specifically the annual average rate of change, as the inflation rate. Within this study, and based on the arguments above, it is expected that an increase in the inflation rate is related to a rise in a company’s hazard probability. The employment rate is expected to lower the probability of failure. Employment is a proxy for demand; the higher the level of employment, the higher the expected demand (Holmes et al., 2010). The employment rate was drawn from the EUROSTAT database and referred to the percentage of employed individuals between the ages of 15 and 64, expressed as a proportion of the total population. Studies on SMEs often regard them as vulnerable to changes in the economic environment. Simons & Rolwes (2009) reported a significant negative relationship between GDP growth and company default rate. Beck et al. (2006) found that businesses in countries with a higher level of intermediary financial development, a more liquid stock market, a more efficient legal system, joc2022-3_v2.indd 83 29.9.2022 13:59:17 Journal of Competitiveness 84 and a higher GDP per capita report fewer financing obstacles. Ullah (2020, p. 121) highlights that ‘among all the business environment constraints affecting firm growth, financial constraints have been identified as one of the most detrimental growth obstacles’. GDP per capita might serve as a proxy for the financial obstacles a company has to face in its country; however, growth obstacles seem to affect survival probability indirectly. A negative relationship between GDP per capita and company survival may be expected for these reasons. 4. RESULTS AND DISCUSSION According to the results of the log-rank test, there are significant (p <0.001) differences between small and medium-sized businesses (SB variable), while there are also significant differences between businesses operating in different industries (IND variable). Initial discrimination analysis indicated that, of the 42 tested variables, only 25 exhibited significant coefficient estimates while also taking the expected sign. A similar procedure was followed for the macroeconomic factors under analysis. All the analysed macroeconomic variables are significant at the 1% level, except for the GDP annual growth rate. A possible explanation might be that the analysed period was a relatively stable period for EU SMEs, with only the Greek economy turning into recession in 2015 and 2016 and Croatia, Cyprus, Finland, and Serbia experiencing negative annual GDP growth in 2015. In terms of country-year GDP data, 99% of observation values were positive and did not, therefore, significantly trigger business defaults. For example, Nouri & Soltani (2016) analysed the impact of the GDP growth rate, interest rate, and inflation on the bankruptcy of businesses listed on the Cyprus stock exchange and found that these variables have no significant impact. However, it should be noted that their results were based on a different methodology. Regarding the expected sign of the analysed variables, only interest rates, personal cost (PC) per employee, and employment rate variables had the expected sign, which will be kept for further analysis. The next step was to conduct a correlation check. Pearson’s correlation coefficient and the variance inflation factor were used for this purpose. Four variables (NI/TA, CA/S, EBIT/S, and CashR) were excluded from further analysis for collinearity reasons. 4.1 Estimating the model coefficients The model was estimated in two forms: Model 1, combining macroeconomic and accounting variables, and Model 2, employing only accounting variables (the same set as Model 1). Both models were tested using ROC curves, and area under curve (AUC) values were compared using the procedure suggested by DeLong et al. (1988). Model 1 was estimated in a stepwise manner by employing a backward elimination procedure using conditional likelihood ratio statistics, which are considered the least prone to error measurement, as a criterion. As a result, the model was significant at the 1% level. Model 2 was derived using the same variables (except for macroeconomic variables) as Model 1. This model was also significant at the 1% level. The details of the estimated coefficients for Model 1 are presented in Table 3. joc2022-3_v2.indd 84 29.9.2022 13:59:17 91 28. Karas, M., & Režňáková, M. (2021). The role of financial constraint factors in predicting SME default. Equilibrium. Quarterly Journal of Economics and Economic Policy, 16 (4), 859–883. https://doi.org/10.24136/eq.2021.032 29. Kennedy, P. (2005). Oh No! I Got the Wrong Sign! What Should I Do? The Journal of Economic Education, 36 (1), 77–92. https://doi.org/10.3200/JECE.36.1.77-92 30. Key figures on Europe, 2018. https://ec.europa.eu/eurostat/web/products-statistical-books/-/ KS-EI-18-001 31. Kliestik, T., Valaskova, K., Lazaroiu, G., Kovacova, M., & Vrbka, J. (2020). Remaining Financially Healthy and Competitive: The Role of Financial Predictors. Journal of Competitiveness, 12 (1), 74–92. https://doi.org/10.7441/joc.2020.01.05 32. Ključnikov, A., Civelek, M., Fialova, V., & Folvarčná, A. (2021). Organizational, local, and global innovativeness of family-owned SMEs depending on firm-individual level characteristics: evidence from the Czech Republic. Equilibrium. Quarterly Journal of Economics and Economic Policy, 16 (1), 169–184. https://doi.org/10.24136/eq.2021.006 33. Kramoliš, J., & Dobeš, K. (2020). Debt as a financial risk factor in SMEs in the Czech Republic. Equilibrium. Quarterly Journal of Economics and Economic Policy, 15 (1), 87–105. https://doi.org/10.24136/eq.2020.005 34. Lando, D. (1998). On Cox Processes and Credit Risky Securities. Review of Derivatives Research, 2 (2), 99-120. https://doi.org/10.1007/BF01531332 35. Mare, D. S. (2012). Contribution of macroeconomic factors to the prediction of small bank failures. Journal of International Financial Markets, Institutions and Money, 39, 25–39. https://doi.org/10.1016/j.intfin.2015.05.005 36. Merton, R. (1974). On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance, 29 (2), 449–470. https://doi.org/10.1111/j.1540-6261.1974.tb03058.x 37. Nam, C. W., Kim, T. S., Park, N. J., & Lee, H. K. (2008). Bankruptcy prediction using a discrete-time duration model incorporating temporal and macroeconomic dependencies. Journal of Forecasting, 27 (6), 493–506. https://doi.org/10.1002/for.985 38. Nouri, B. A., & Soltani, M. (2016). Designing a bankruptcy prediction model based on account, market and macroeconomic variables (Case Study: Cyprus Stock Exchange). Iranian Journal of Management Studies, 9 (1), 125–147. 39. Ohlson, J. A. (1980). Financial Ratios and the Probabilistic Prediction of Bankruptcy. Journal of Accounting Research, 18 (1), 109–131. https://doi.org/10.2307/2490395 40. Qu, Y. (2008). Macroeconomic factors and probability of default. European Journal of Economics, Finance and Administrative Sciences, 13, 192–215. 41. Shumway, T. (2001) Forecasting Bankruptcy More Accurately: A Simple Hazard Model. Journal of Business, 74 (1), 101–24. https://doi.org/ 10.1086/209665 42. Simons, D., & Rolwes, F. (2009). Macroeconomic Default Modelling and Stress Testing. International Journal of Central Banking, 5 (3), 177–204. joc2022-3_v2.indd 91 29.9.2022 13:59:17 Journal of Competitiveness 92 43. Tinoco, M. H., & Wilson, N. (2013). Financial distress and bankruptcy prediction among listed companies using accounting, market and macroeconomic variables. International Review of Financial Analysis, 30, 394–419. https://doi.org/10.1016/j.irfa.2013.02.013 44. Trujillo-Ponce, A., Medina, R., & Cardone Riportella, C. (2013). Examining what best explains corporate credit risk: accounting-based versus market-based models. Journal of Business Economics and Management, 15 (2), 1–24. https://doi.org/10.3846/16111699.2012.720598 45. Ullah, B. (2020). Financial constraints, corruption, and SME growth in transition economies. The Quarterly Review of Economics and Finance, 75, 120–132. https://doi.org/10.1016/j.qref.2019.05.009 46. Vassalou, M., & Xing, Y., (2004). Default risk in equity returns. Journal of Finance, 59, 831–868. https://doi.org/10.1111/j.1540-6261.2004.00650.x 47. Virglerova, Z., Dvorsky, J., Kozubikova, L., & Cepel, M. (2020). Perception of non-financial risk determinants in SMEs in Visegrad countries. Oeconomia Copernicana, 11 (3), 509–529. https://doi.org/10.24136/oc.2020.021 48. Virglerova, Z., Ivanova, E., Dvorsky, J., Belas, J., & Krulický, T. (2021). Selected factors of internationalisation and their impact on the SME perception of the market risk. Oeconomia Copernicana, 12 (4), 1011–1032. https://doi.org/10.24136/oc.2021.033 49. Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22, 59–82. https://doi.org/10.2307/2490859 Contact information assoc. prof. Michal Karas, Ph.D. Brno University of Technolog y Faculty of Business and Management Department of Finance Czech Republic E-mail: [email protected] ORCID: 0000-0001-8824-1594 joc2022-3_v2.indd 92 29.9.2022 13:59:17