The Influence of Top-Management Characteristics on Corporate Credit Risk Measures

Table of Contents

1. Introduction

2. The general influence of CEOs on organizational outcomes
2.1. Empirical evidence on the influence of CEO characteristics
2.2. Relation between CEO characteristics, financial distress and bankruptcy

3. Measures of bankruptcy risk
3.1. Altman-Z-Score
3.2. Ohlson-O-Score
3.3. Comparative performance and predictive power

4. Definition of explanatory variables and hypotheses

5. Dataandmethodology
5.1. Acquisition of data & sample selection
5.2. Methodology
5.2.1. Econometricsetup
5.2.2. Controlvariables

6. Empirical analysis and findings
6.1. Description of data & sample
6.2. Empirical results of multivariate analysis
6.3. Robustness testing and sub-sample validation

7. Conclusion

8. Limitations and suggestions for further research

9. Appendix

10. References

List of tables and figures

Table 3.1: Altman-Z-Score - Intervals for interpretation

Table 3.2: Altman-Z’’-Score - Average score and equivalent US rating

Table 5.1: Data collection process

Table 5.2: SIC industry classification

Table 5.3: Number of firms included per year

Figure 6.1: Means ofbankruptcy indicators overtime

Figure 6.2: Kernel density estimates ofbankruptcy indicators

Table 6.3: Summary statistics

Table 6.4: Correlation matrix

Table 6.5: Multivariate analysis

Table A.1: Variable description

Table A.2: Kernel density estimates of explanatory variables

Table A.3: Correlation matrix: Z-Score

Table A.4: Correlationmatrix: Z’’-Score

Table A.5: Correlation matrix: O-Score probability

Table A.6: Multivariate regression - high risk

Table A.7: Multivariate regression - low risk

Table A.8: Multivariate regression - manufacturing

Table A.9: Multivariate regression - non-manufacturing

Table A.10: Multivariate regression - service

Table A.11: Multivariate regression - profitability/inverse leverage

ABSTRACT

Inspired by previous research this paper investigates whether personal CEO characteristics such as age, CEO tenure, gender, MBA and variable salary (%) have a significant effect on firm bankruptcy risk measured using the Altman-Z-Score (1968, 1983) and the Ohlson-O-Score (1980). This work is based on literature suggesting (i) CEO managerial characteristics such as overconfidence and optimism lead to higher leverage and increased risk-taking and that (ii) higher levels of debt and increased risk-taking behavior add to the likelihood of corporate financial distress. Using panel data on S&P 500 constituents during 1994-2014 our results provide evidence that CEO age and holding an MBA is positively associated with bankruptcy risk while CEO tenure and variable salary (%) seem to be negatively related to a firm’s propensity to default. Collectively, our results remain mostly unchanged over various robustness tests employing both pooled OLS and the least squares dummy variable (LSDV) model as well as year, industry and company fixed effects as control variables. Next to significant support that managerial attributes, traits, and style may help to understand organizational outcomes, this project also provides insights how available public information can be used to further explain style effects by disentangling them into separate, measurable impact factors.

Acknowledgements: First and foremost, we would like to thank our advisor and examiner Jeppe Christoffersen for his continuous support and guidance throughout the compilation of this work. Moreover, we are grateful to Bjoern Imbierowicz, Christian Wagner and Peter Raahauge as well as all contributors to the Stata community for sharing their expertise about the implementation of empirical methods and econometric applications.

Key words: CEO Characteristics, Bankruptcy, Corporate Governance, Empirical Analysis of Firm Behavior, Accounting, Distress prediction, Z-Score, O-Score

JEL Classification: D22, G33, G34, J16, M41, M52

1. Introduction

Does the level of the CEO’s education matter for determining bankruptcy risk? Are individual characteristics of the top management important to mitigate the risk of corporate failure? These and similar questions are raised in both academia as well as meetings of supervisory boards around the world. Pointing out the importance of leaders, the CEO and Chairman of the Board of General Electric, Jeffrey Immelt puts it this way during an interview:

“I’m responsible for this company. I stand behind the results. I know the details, and I think the CEO has to be the moral leader of the company.”^[1]

While credit risk and the prediction of corporate bankruptcy have emerged into one of the most actively examined research areas within quantitative finance, academia has devoted comparatively little resources to the influence that characteristics and attributes of a company’s top management have on the propensity of firm default.

However, younger economic history has demonstrated that individuals may have a considerable impact on the outcome of the organization they manage. For instance, the Financial Crisis Inquiry Commission of the US Government (2011) consternates in its final report that the recent financial turmoil and the subsequent economic downturn were materially reinforced by excessive corporate risk-taking and executive mismanagement.

This is in line with academic research by Hambrick and Mason (1984) who introduce the upper echelon theory which suggests that “organizational outcomes - strategic choices and performance - are partially predicted by managerial background characteristics”. Further research papers by, for instance, Bertrand and Schoar (2003) confirm the influence of individuals on organizational outcomes while providing a measurable impact of personal CEO characteristics on various fields of corporate business. The latter idea is taken up by Kallunki and Pyykko (2013) arguing that the probability of “[...] financial distress is affected not only by the firm characteristics but also by the personal characteristics of management”.

The aim of this paper is to investigate to which extent executive characteristics and biography may explain variations in the cross-section and time-dimension of corporate credit risk across firms. The venerable Altman Z-Score (1968), Altman-Z’’-Score (1983) and Ohlson-O-Score (1980) are chosen for the assessment of financial distress at individual firm level.

The high relevance of these bankruptcy models in practice is proven through their worldwide usage for e.g. business loan evaluation, internal control/monitoring purposes and investment decisions.

This work analyses 709 different S&P 500 firms and includes 1,433 CEOs over a period of 21 years between 1994 and 2014. Inspired by previous journal articles in this field of research, we focus on age, CEO tenure, gender, MBA and variable salary (%) as the explanatory variables.

We show that young CEOs tend to be employed in firms with lower bankruptcy risk, while a CEO holding an MBA typically leads firms with higher credit risk. Not surprisingly, short CEO tenure goes along with higher distress risk as well. We provide arrays of robustness tests to rule out several alternative hypotheses, but our results remain unchanged applying both pooled OLS and the least squares dummy variable (LSDV) model as well as year, industry and company fixed effects as control variables.

Our results demonstrate that already the inclusion of easily observable managerial characteristics may significantly enhance the understanding of financial distress and potential corporate failures. Accordingly, stakeholder may improve their evaluation whether to provide funding, supply materials or work force to a firm if personal characteristics are taken into account.

The remainder of this work is structured as follows. Section two provides theoretical background information coupled with a broad literature overview covering the field of research related to the influence of CEOs and personal characteristics on corporate outcomes and decision making. Section three introduces the measures of bankruptcy risk, especially the Altman-Z-Scores and the Ohlson-O- Score. The section concludes with an evaluation of their ability to forecast financial distress and a comparison with other models predicting corporate bankruptcy. Part four presents a definition of our explanatory variables and formulates our research hypotheses. Thereafter, section five points out the data acquisition procedure and sample construction. Moreover, the econometric set-up of this project is explained and motivated. Section six presents the results of our multivariate analyses and discusses the corresponding findings. Various robustness tests and sub-sample validations are employed to validate our inferences drawn. Section seven concludes while part eight discusses limitations of our findings and provides indications for future research.

2. The general influence of CEOs on organizational outcomes

A central premise for managerial characteristics to matter for a company’s propensity to default is that executives in general do influence firm performance. While more recent research mostly emphasizes the notion that “[...] decisions are made by people within firms, not just by firms as generic entities” (Graham et al. 2015) and that subsequently “[...] individual heterogeneity matters within corporate finance / governance” (Graham et al., 2013), the human factor has long played only a minor role in explanatory endeavors of organizational outcomes. As pointed out by Bertrand and Schoar (2003), standard neoclassical theory perceives top executives as a uniform and selfless production factor so that different managers can act as a perfect substitute for each other.

Also two of the most prominent management theories to explain inter-organizational relationships deny that the influence of an individual executive alone is large enough to consistently impact the overall organization: On the one hand, a population ecology perspective as proposed by Hannan and Freeman (1977) claims that corporate success or failure is mostly driven by the (in-) ability to successfully adapt to larger environmental forces. Managers may often have difficulties to foresee and to adapt to these drivers due to their firms’ organizational inherit age and consequently play only a minor role for corporate survival. Similarly, the institutional isomorphism view introduced by DiMaggio and Powell (1983) alleges that corporations adopt certain structures and behaviors to satisfy norms and values of stakeholders present in the institutional environment so that, again, the influence of the individual manager by herself is by default limited.

On the other hand, members of the Carnegie School of Management have long argued that under bounded rationality, sophisticated decisions are mostly driven by behavioral forces instead of a pure economic optimization (e.g. March and Simon, 1958). Building on this view, Hambrick and Mason (1984) developed the upper echelon theory under which top executive characteristics such as psychological traits, but also more easily observable demographic indicators are a reliable predictor of strategic decisions and, eventually, organizational outcomes. Among the characteristics initially considered are, for instance, executive education, age, tenure, and financial position, which share the advantage that information upon them can be acquired without high costs. Subsequently, our project will examine the predictive power of the same or closely related measures when it comes to corporate financial distress.

As summarized by Hambrick (2007), since the introduction of upper echelons theory, scholars have proposed multiple extensions and refinements. In particular, two important moderators have been proposed to examine the predictive validity of the model. Hambrick and Finkelstein (1987) proclaim that the degree of managerial discretion, i.e. the scope of possible actions available to the decision maker and the presence of means-end ambiguity, positively relates to the importance of managerial characteristics for corporate outcomes. Hambrink et al. (2005) argue that a similar moderator can be found in executive job demands, which they describe as the relative difficulty of the executive’s job profile. The higher job demands are, the less can managers afford to employ comprehensive in-depth analysis tools and the more they have to rely on mental shortcuts and heuristics, eventually augmenting the importance of the personal background attitude on decision making.

Both managerial discretion and executive job demand are hypothesized to be - to some extent - a function of the environment of firm is operating in, which will allow us to test the power of the chosen set of characteristics to forecast financial distress across different industries in a later stage of this work.

2.1. Empirical evidence on the influence of CEO characteristics

Building on the upper echelon theory, the influence of CEO characteristics and their effects on corporate decision making and business outcomes has widely been researched.

The majority of studies focus on the link between managerial characteristics and firm performance. Cooper et al. (2010) investigate the influence of CEO compensation on shareholder wealth, Henderson et al. (2006) as well as McClelland et al. (2012) analyze the impact of CEO tenure in both a stable and rapidly changing business environment while Gottesman and Morey (2006, 2010) and Bertrand and Schoar (2003) examine whether the CEO’s educational background matters for corporate performance. Adams et al. (2005) find evidence that the CEO's structural power over the board and other top executives has also a measurable impact on corporate performance. Guided by the general question in how far top managers matter for organizational outcomes, Bennedsen et al. (2007) find that the death of the CEO or of a close relatives typically causes a decline in operating profitability, investment and sales growth.^[2] Similarly, Cronqvist and Yu (2015) show that CEOs who are father of a daughter typically increase the Corporate Social Responsibility ranking of the firm they lead. Further research on CEO characteristics and performance comprises but is not limited to Vintila et al. (2015), Peni (2014), Nelson (2005), Core et al. (1999) and Agrawal et al. (1991).

Besides firm performance the influence of CEO characteristics has been studied in various other fields of corporate outcomes and decision making. Barker III and Mueller (2002) link managerial characteristics to corporate R&D spending whereas Custodio and Metzger (2013) explore the relationship between CEO experience and acquisition returns. The latter is closely related to research made by Roll (1986) as well as Malmendier and Tate (2008) who relate managerial overconfidence to unsuccessful mergers and acquisitions. Moreover, Bebchuk and Grinstein (2005) analyze the influence of CEO compensation and firm expansion decisions.

Furthermore, Dyreng et al. (2010) propose that managerial characteristics of top executives significantly influence the level of corporate tax avoidance. Additionally, Graham et al. (2015) find that CEO characteristics (e.g. tenure, age, MBA, experience, variable compensation) influence the delegation of financial decision making and decision-making authority.

Cain and McKeon (2012) find that sensation-seeking CEOs who engage in personal risky activities consistently tend to engage in more acquisition transactions, employ higher levels of firm leverage and increase overall firm volatility. Similarly, Graham et al. (2013) explore that companies lead by more risk-tolerant CEOs tend to engage in more M&A activity while optimistic CEOs are more likely to use short term debt.

Finally, Benzing and Tuerk (2012) explore that CEO age, tenure and the CEO’s base salary are negatively associated with the volatility of the firm's equity, while CEO experience and education seem to be positively related.

2.2. Relation between CEO characteristics, financial distress and bankruptcy

Within the field of academic research related to CEO characteristics and their impact on leverage, financial distress and bankruptcy the vast majority is dedicated to the level of corporate debt.

Malmendier and Tate (2005, 2011, and 2015) examine the influence of CEO overconfidence on corporate investment and subsequent financing decisions. Their research findings suggest that overconfident managers tend to prefer sources of internal and debt financing over external equity financing. This is in line with research carried out by Hackbarth (2008) and Ben-David et al. (2007) advocating that optimistic and overconfident managers seem to choose higher debt levels and issue new debt more frequently.

Cronqvist et al. (2012) find a positive relation between firm’s corporate debt levels and their CEOs personal leverage decisions analyzing data on the CEO's private most current house acquisitions.

However, very little research can be found dealing with the relation of managerial characteristics and financial distress or bankruptcy. D'Aveni (1990) shows that prestigious high level managers may beneficially impact corporate survival as a high status gained through membership in elite social circles generally encourages creditors to keep the current executive team of a firm in financial distress in power and to eschew putting the company under the supervision of a bankruptcy court.

On the other hand, Hambrick and D’Aveni (1992) and Daily and Dalton (1994) analyze various top management team and board characteristics of bankrupt firms respectively. Their findings suggest that managerial deficiencies are responsible for corporate deterioration which in turn leads to team deterioration as a consequence of corporate failure.

Another strand of research builds on the idea that personal characteristics influence corporate decision making and can consequently improve the forecasting accuracy of bankruptcy.

Kallunki and Pyykko (2013) show that the predictive power of the bankruptcy models developed by Altman (1968) and Ohlson (1980) can be increased significantly by including personal information on the CEOs and board members personal credit record. Subsequently CEOs or directors with personal payment defaults substantially increase the probability of bankruptcy and insolvency. Their work builds on the idea presented earlier by Hackbarth (2008) that personal CEO characteristics influence corporate decision making in the field of leverage and risk-taking. Likewise Back (2005) and Laitinen and Laitinen (2009) confirm that models combining non-financial variables and financial ratios best predict financial distress. Those non-financial variables comprise information on the firm’s management and the board, i.e. management background and previous payment pattern and audit modification variables respectively.

These findings are in accordance with Platt and Platt (2012) who also suggest that the composition and characteristics of the firm's board are significantly related to the corporate solvency and bankruptcy. Given the scarcity of research dealing with the direct influence of CEO characteristics on corporate failure and financial distress the underlying work will serve as a valuable contribution to academic literature.

3. Measures ofbankruptcy risk

Academic literature has proposed various models to measure the risk of corporate bankruptcy, whereby the most common specifications can be clustered into accounting-based-, contingent claim- and hazard models. Further on, accounting-based models can be separated into univariate- and multivariate approaches with the Altman-Z-Score and the Ohlson-O-Score being the most popular within the latter category.

3.1.Altman-Z-Score

The assessment of firm’s financial and operating difficulties was ordinarily achieved by using traditional univariate ratio analysis. Amongst others Beaver (1966, 1968a, 1968b) extensively investigates the predictive power of corporate failure for various financial ratios including profitability, liquidity, solvency and turnover ratios. Generally, financial ratio analysis is based on the fact that failing firms show significantly different ratios than continuing companies (Altman, 1968; Beaver, 1966).

In 1968, Altman further develops a multivariate approach for predicting financial distress by combining various financial ratios into one single forecasting model of bankruptcy- the Altman-Z- Score. The original score is generated using multiple discriminant analysis (MDA) and is based on a total sample of 66 manufacturing corporations partitioned in two testing groups with 33 companies each: bankrupt and non-bankrupt firms. To estimate the coefficients of the model, Altman (1968) uses data from financial statements prepared one reporting period prior to the bankruptcy declaration. Considering a total of 22, Altman extracts five basic categories of financial ratios, which are found to be relevant in forecasting financial distress and in capturing a company’s characteristics such as liquidity, profitability, leverage, solvency and activity. Finally, the original model published reads as follows^[3]:

Zit = 1.2Xitt + 1.4X2jt + 3.3X3jt + 0.6X4jt + 0.999X5it (3.1)

where t represents the reporting period and i indicates a company index. The ratios are defined as:

Xl = Working Capital/Total Assets X2 = Retained Earnings/TotalAssets X3 = EBIT/Total Assets

X4 = MarketValue ofEquity/BookValue ofTotal Liabilities X5 = Sales/TotalAssets

The ratios (Xl — X5) at time t estimate the Altman-Z-Score at time t to evaluate the survival of the firm in t+1. The economic intuition can be summarized as follows:

Xl — Working Capital/Total Assets: This ratio constitutes the most important liquidity ratio^[4], which is consistent with Merwin (1942), who describes the ratio as the finest indictor of final discontinuance of business. “A firm experiencing consistent operating losses will have shrinking current assets in relation to total assets” (Altman, 1968). Consequently, Beaver (1966) finds that firms running into financial distress show a deterioration in the above mentioned liquidity measure.

X2 — RetainedEarnings/TotalAssets is a measure of cumulative (past) profitability over time. Thereby, the age of the firm is implicitly included given that an older company has had more time to pile up earnings. Likewise, Altman (1968) correctly suggests that younger firms are discriminated and more likely to be classified as bankrupt given their shorter period of operating activity/ revenue generation.

X3 — EBIT/Total Assets: The third ratio X3 is known as the return on total assets (ROTA) ratio and is a mark of current true productivity^[5] of total assets. Perhaps not surprisingly (as firm survival eventually depends on a firm’s ability to generate a positive earnings trajectory), Altman (1968) finds that this ratio exhibits the highest explanatory power for bankruptcy prediction.

X4 — MarketValue ofEquity/BookValue ofTotalLiabilities: By including market value of equity the model incorporates outside investors’ consensus on a firm’s viability, but however, limits its applicability to publicly listed firms. Moreover, the ratio constitutes an inverse measure of leverage and indicates how much the company’s total assets can be wiped out before the firm experiences financial distress.

X5 — Sales/TotalAssets: The last ratio is the capital-turnover ratio and measures the ability of the firm to handle competition in the market. This ratio is very industry specific and will therefore be omitted during later re-estimations and reformulations of the model such as the Z’’-version proposed by Altman (1983).

Considered individually all ratios are lower for bankrupt firms and secondly all coefficients in (3.1) show a strictly positive sign, thus, resulting in a low overall score for firms facing financial difficulties. For a general application and interpretation Altman suggests the following intervals:

Abbildung in dieser Leseprobe nicht enthalten

Table 3.1: Altman-Z-Score - Intervals for interpretation

Considering that (i) the Altman-Z-Score (1968) was exclusively estimated based on data from publicly held manufacturing companies and (ii) a very small sample size of only 66 entities plus (iii) the statistical requirements underlying the application of the MDA estimation procedure^[6] it is not surprising that other researchers have raised criticism regarding its’ reliability. Furthermore, Altman (1968) finds the accuracy of the model to be limited up to two years prior to bankruptcy, decreasing considerably as the lead time rises.

Despite the above mentioned limitations the Altman-Z-Score is widely used by both academics and practitioners including asset managers, banks, rating agencies and distressed firms themselves.

In the following years the model served as a prototype for other bankruptcy prediction models and received attention and appreciation worldwide. In light of Altman’s research, Ohlson (1980) suggests a logit model, Zmijewski (1984) applies a probit model and Dimitras et al. (1996) investigate several other models to forecast corporate failure.

In order to ensure the applicability of the Altman-Z-Score to smaller, privately held companies Altman publishes two revised Z-Scores in 1983. The first version does no longer rely on the market value of equity, which gets replaced by the book value of equity. Secondly, based on the original data the model gets re-estimated and subsequently is labelled Z’-Score:

Z'it = 0.717 Xltt + 0.847X2jt + 3.107X3it + 0A2X4it + 0.998X5it (3.2)

where t is defined as the year and i indicates a company index. The ratios are defined as:

Xl = Working Capital/ Total Assets X2 = Retained Earnings/Total Assets X3 = EBIT/TotalAssets

X4 = BookValueofEquity/BookValue ofTotal Liabilities X5 = Sales/TotalAssets

The second model aims at eliminating potential industry effects/distortions due to differences in asset turnover and, thus, excludes X5 = Sales/TotalAssets. In addition, all coefficients are re-estimated now based on a sample of privately held and publicly traded as well as manufacturing and non­manufacturing firms.

The resulting Altman-Z’’-Score (1983) allows for the widest range of application and calculates as follows:

Z"it = 3.25 + 6.56Xiit + 3.26X2it + 6.72X3it + 1.05X4it (3.3)

where t is defined as the year and i indicates a company index. The ratios are defined as:

Xx = Working Capital/Total Assets X2 = Retained Earnings/TotalAssets X3 = EBIT/Total Assets

X4 = Book Value of Equity/Book Value ofTotal Liabilities

Again, profitability (X3 = EBIT/Total Assets) carries the highest discriminant power, which is natural given that a firm that runs profitable operations is very unlikely to experience financial distress.

For the correct interpretation of the Z’’-Score Altman and Saunders (1998) as well as Altman and Hotchkiss (2006) propose the following sections paired with the equivalent US rating:

Abbildung in dieser Leseprobe nicht enthalten

Table 3.2: Altman-Z’’-Score - Average score and equivalent US rating

This paper employs both the original Altman-Z-Score (1968) and the adjusted Altman-Z’’-Score (1983) as measures for bankruptcy among firms of the S&P 500 between 1994 and 2014.

Since our analysis is solely based on publicly listed firms from both the manufacturing and non­manufacturing industry we find this selection most appropriate. Finally, given that the market value of equity captures the instant market consensus we assume the original model to reflect changes or differences in CEO related performance, managerial style and general firm attributes faster than a purely accounting-based, i.e. backwards-looking, version so that we will also employ this specification as an independent variable later on.

3.2.0hlson-0-Score

Ohlson (1980) suggests another model for the prediction of bankruptcy, which is based on partially different financial ratios and employs the maximum likelihood estimation of the commonly named conditional logit model rather than the MDA approach used by E. Altman.

According to Ohlson (1980) the MDA approach can only be applied if, for instance, the variance- covariance matrices of the predictors are the same for both groups, bankrupt and non-bankrupt firms, and if all predictors are jointly normally distributed, which would exclude the usage of dummy variables. “The use of the conditional logit analysis, on the other hand, essentially avoids all of the problems discussed with respect to MDA” (Ohlson 1980) and thereby also allows for the inclusion of binary variables.

Ohlson’s study is based on a sample of 105 bankrupt and 2,058 non-bankrupt firms between 1970 and 1976. Ohlson defines bankruptcy as the act of filing for bankruptcy with regards to Chapter X or Chapter XI of the United States Bankruptcy Code. Moreover, the O-Score covers exclusively publicly traded firms. The average lead-time between the date of fiscal year oflast relevant financial report and the event of bankruptcy is approximated to be 13 months. Finally, utilities, transportation companies, and firms from the financial service industry are excluded from the sample due to structural differences and a dissimilar bankruptcy environment.

The main finding was the identification of four basic factors, which significantly influence the probability of bankruptcy. The relevant factors of the Ohlson-O-Score are (a) the size of the firm, (b) a measure of the capital structure, (c) a performance measure and (d) a liquidity measure.^[7] The best known “Model 1” of the compendium presented by Ohlson (1980) forecasts bankruptcy within one year and comprises one intercept and nine independent variables of which two are dummy variables.

The Ohlson-O-Score (Model 1) calculates as follows:

Oit = -1-32 - 0.407Wlit + 6.03W2it - 1.43W3it + 0.0757W4it (3.4)

—2.37W5jt - 1.83W6it + 0.285W7it - 1.72W8it - 0.521W9it

where t is defined as the year and i indicates a company index. The variables are defined as:

Wl = log (Total assets/ GNP price - level index) (a)

W2 = Total liabilities/ Total assets (b)

W3 = Working capital/ Total assets (d)

W4 = Currentliabilities/ Currentassets (d)

W5 = Netincome/Totalassets (c)

W6 = Funds from operations/Total liabilities (d)

W7 = Dummy variable, 1 if net income was negative for the last two years, 0 otherwise (d)

W8 = Dummy variable, 1 if total liabilities exceed total assets, 0 otherwise (b)

W9 = (Netincomet - Net incomet_l)/ (|Net Incomet| + |Net Incomet_l|) (c)

Besides size (W-jJ and the usage of dummy variables (W7, W8) the Ohlson-O-Score uses familiar financial ratios from similar categories as the Altman-Z-Score: liquidity, profitability, leverage, solvency and activity. Moreover, the score relies entirely on accounting-based information. The GNP price level index used in Wx assumes a base value of 100 in 1968. While W9 is intended to measure the change in net income, W8 is supposed to serve as a discontinuity correction for W2.

Again, the ratios worsen the closer one moves to the year of bankruptcy. If the Ohlson-O-Score Oit is one, the firm i is predicted to go bankrupt during the next year, zero indicates the opposite. The ultimate probability of default for firm i in t+1 can be calculated using the logistic function:

Prob = e°lt/{ 1 + e[0]it) (3.5)

A narrative for a limitation of the O-Score (which, of course, is also valid for Altman’s set-up) was already developed by Deakin (1972), who argues that companies with a positive change in earnings are inclined to increase debt, which will lead inevitably to a higher probability of default. Furthermore, Ohlson himself points out “[...] that size is an important predictor of bankruptcy”, but however, this statistical significance may simply result from the strong size bias towards large corporations present inthe COMPUSTAT universe.

Internationally, the Ohlson O-Score was used by several researches around the globe, either to verify/modify or to use it solely as a robustness check. Selected examples are studies by Swanson and Tybout (1988) - Argentina, Knight (1979) - Canada, Baetge et al. (1988) - Germany, Bhatia (1988) - India, Ta and Seah (1988) - Singapore and Pascale (1988) - Uruguay.

In addition to the above mentioned Altman-Z-Scores this work will make use of the Ohlson-O-Score to measure the likelihood of financial distress and be able to cross-validate our empirical findings.

3.3. Comparative performance and predictive power

While Altman (1968) and Ohlson (1980) made important pioneering contributions to the multivariate forecasting of corporate failure, alternative models to predict financial distress have been proposed more recently.

These can be broadly clustered into contingent claim- and hazard models. The former is purely market-value based and, building on seminal work of Black and Scholes (1973) and Merton (1974), applies option pricing theory to corporate debt and equity securities. Hazard models for bankruptcy predictions typically combine both market- and accounting-based variables and employ the time period over which a firm survives as the dependent variable. In contrast to the more static accounting- and contingent claim based approaches, hazard models are hence able to explicitly account for a company’s development within the forecasting horizon (for detailed background information and implementation, see Shumway, 2001 and Campbell et al., 2008).^[8] Even if not used in our further analysis, these models provide a meaningful benchmark against which the accuracy of the Altman (1968, 1983) and Ohlson (1980) framework can be evaluated: A vast body of literature has examined and compared the predictive power and applicability of the models introduced above.^[9]

Tinico and Wilson (2013) and Acosta-Gonzalez and Fernandez-Rodrigues (2014) find that the Altman set-up still performs comparatively well at forecasting bankruptcies but suffers from a high Type II Error, i.e. tends to predict business failure for firms which eventually survive. Moreover, Reisz and Perlich (2007) show that the Altman (1968) Z-Score as well as the Altman (1983) Z’’-Score outperform the predictive power of market-based models in the short-term.

A broad body of research often finds - studying different time periods and geographical locations - that each model captures different, unique facets of default risks. However, ratio based approaches are consistently outperformed by either contingent claim (see e.g. Hillegeist et al., 2004 and Xu and Zhang, 2009) or more recent hazard set-ups (e.g. Chava and Jarrow, 2004; Wu et al., 2010; or Bauer and Agarwal, 2014) with regard to the overall information content subsumed.

While an extensive study comprising company data from a total of 34 countries conducted by Altman et al. (2016) suggests that the original Z-Score is very robust across country and time, earlier research (see e.g. Grice and Ingram (2001)) largely demonstrates that its predictive power diminished over time and thereby highlights the need to re-calibrate the coefficients. However, empirical work by Begley et al. (1996) finds that even after re-estimation both the Ohlson- and the Altman-Score did not work as well for more recent sample periods as they did when originally formulated.

Next to mixed empirical results, the Altman (1968) and Ohlson (1980) set-up is often animadverted by theorists for its weak theoretical foundations such as the missing relation to a consistent asset pricing framework. On the other hand, the strong reliance on accounting numbers may impair the economic validity of the model: For example, Agarwal and Taffler (2008) criticize that accounting figures are prone to manipulation by management and, due to their reliance on general accounting principles, have only limited ability to represent the true business value of an asset. Hillegeist et al. (2004) allege that accounting information is by definition backward looking and based on the going-concern assumption so that its application for predicting the future in a context where the continuation of business activities is questionable may be tremendously flawed. Moreover, both Altman’s and Ohlson’s model ignore asset volatility, which is, however, crucial for assessing the likelihood that a firm may become unable to honor its debt obligations.

Despite these theoretical and empirical limitations, the Z-Score and the O-Score are still actively employed: Agarwal and Taffler (2007) provide a broad summary on academic publications where the measures are used as a proxy for credit risk. Moreover, as pointed out by Altman (2000), the scores are widely applied by practitioners, e.g. for capital allocation, early warning systems or managerial evaluations.

Taking these remarks into considerations, we conclude that the Altman Z-Score and Ohlson O-Score are at least well known to and understood by a broad audience. In addition to the comparatively easy process of data gathering, this advantage encourages us to use both scores as a measure for corporate default risk in the subsequent parts of the analysis.

4. Definition of explanatory variables and hypotheses

Inspired by earlier research^[10] we employ age, CEO tenure, gender, MBA and variable salary (%) as our CEO characteristics to explain changes in bankruptcy risk measured using the Altman-Z-Sore and the Ohlson-O-Score. Throughout the following section we agree with and consequently build on the reasoning developed by Kallunki and Pyykko (2013) that “[...] unjustified risk-taking may lead a firm into financial distress” and miserable managerial decision making will first be reflected in deteriorating financial ratios and finally lead to corporate failure.

Bankruptcy risk and age

There are various research papers suggesting that the CEO's age has a measurable impact on corporate risk-taking so that we assume this variable may subsequently influence the likelihood of corporate failure.

For instance, Graham et al. (2013) argue that age reflects experience and perspective, which enable top managers to take on more risk. These results are in line with McClelland et al. (2012) who find that CEOs with a shorter present career experience measured by their age tend to employ risk-averse strategies which on average may impact corporate performance adversely.

Building on the conclusion reached by Huang et al. (2012) that older CEOs can be linked to higher quality of financial reporting we expect the above relation to be reflected in financial accounting figures and consequently measurable by the employed bankruptcy measures.

[...]

---

^[1] Jeffrey Immelt, CEO and Chairman of the Board of General Electric. 2002. Interview with CNNMoney about the Q2 forecast 2002.

^[2] Actually, the study finds that this effect holds true for all family members except for the mother-in-law whose disease typically triggers a positive (yet insignificant) effect on corporate performance.

^[3] Notation of coefficients in line with Altman (2000)

^[4] Other liquidity ratios analyzed have been the current ratio and the quick ratio.

^[5] The term “true productivity” refers to the fact, that EBIT indicates earnings before taxes and the influence of corporate leverage decisions.

^[6] “MDA is based on the ordinary least squares (OLS) method and thus requires assumptions of multinormality, homoscedasticity, and linearity, which are not often met in empirical financial ratio analysis” (Altman et al. 2016). The interested reader may also refer to Ohlson (1980) and section 3.2 for additional discussion and criticism of the model.

^[7] Classification according to Wu et al. (2010)

^[8] In a survey on intelligent and statistical techniques to forecast bankruptcy, Ravi Kumar and Ravi (2007) find that tools and models building on neural network structures obtained broad coverage in both academia and practice. As these approaches, however, originate from operations instead of finance or accounting research, we will omit them within this review.

^[9] As a fully comprehensive overview on the performance of different bankruptcy prediction models is beyond the scope of this project, we refer to Grice and Ingram (2001) (for articles published until 2000) and Altman et al. (2016) (for articles published thereafter) for a more complete coverage.

^[10] Hambrick and Mason (1984), Berger et al. (1997), Barker III and Mueller (2002), Bertrand and Schoar (2003), Gottesman and Morey (2006, 2010), Huang and Kisgen (2013), Graham et al. (2015)