Executive compensation and corporate bankruptcy in the context of crisis

Transcription

1 68 Int. J. Business Governance and Ethics, Vol. 9, No. 1, 2014 Executive compensation and corporate bankruptcy in the context of crisis Sarra Elleuch Hamza* LIGUE Laboratory, ISCAE, Campus University, 2010, Manouba, Tunisia *Corresponding author Imen Lourimi ISCAE, Campus University, 2010, Manouba, Tunisia Abstract: This paper investigates the relationship between executive compensation and bankruptcy in the context of crisis. Selecting a sample of the 105 largest bankrupt US companies and the 105 largest solvent US companies for the periods 2000 to 2002 and 2007 to 2009, and using multivariate logit regressions, our results reveal a negative and significant relationship between CEO compensation and corporate bankruptcy two years prior to the occurrence of this event. In addition, the results show that the compensation levels received by CEOs in failed companies are not significantly different from those paid to its counterparts in healthy companies for a three-year period prior to the bankruptcy. This study suggests that, unlike many previous studies, CEO compensation is reduced when the firm is near bankruptcy. It supports the recent trend of a decrease in managerial power and the active role of creditors in fixing CEO remuneration. Keywords: bankruptcy; executive compensation; logit model. Reference to this paper should be made as follows: Elleuch Hamza, S. and Lourimi, I. (2014) Executive compensation and corporate bankruptcy in the context of crisis, Int. J. Business Governance and Ethics, Vol. 9, No. 1, pp Biographical notes: Sarra Elleuch Hamza is a Professor at High Institute of Accounting and Business Administration (ISCAE) in Tunisia, and holds a PhD from ISCAE and Jules Vernes University. She has published many articles on corporate governance and accounting management in refereed and ranked journals. Imen Lourimi is a Master Researcher at ISCAE. Her main areas of interest include accounting and corporate governance. Copyright 2014 Inderscience Enterprises Ltd.

2 Executive compensation and corporate bankruptcy in the context of crisis 69 1 Introduction The past decade has been marked by crises that have affected the most important financial markets and have led to many firms bankruptcies. The same period has also been accompanied by a rapid growth in executive compensation. Its level is even excessive according to Bebchuk and Grinstein (2005). Many researchers have suggested that this excessive pay is an important factor that amplified the magnitude of these crises and increased the number of corporate bankruptcies (Balachandran et al., 2010). In theory, two main views have been developed to explain this relationship: the managerial power view and the optimal contracting view. The managerial power model asserts that compensation contracts increase agency costs. Indeed, managers may take advantage of the informational asymmetry problem to act to further their own interests, rather than those of their shareholders. They can engage in manipulative activities such as earnings management in order to raise their remuneration. They can also make decision with short-term impacts rather than long-term ones, leading to firm value destruction (Faulkender et al., 2010). This power gives managers of financially distressed firms the possibility to extract rents just as in healthy firms (Henderson, 2007). The implications of this model are consistent with the high executive compensation incentives hypothesis, even in a bankrupt firm. In contrast, in the optimal contracting model, executive compensation contracts are positively associated with firm performance in order to increase shareholder wealth. Therefore, these incentive contracts serve as a tool to align managers interests with those of shareholders and thereby minimise agency costs of equity (Henderson, 2007). However, the role of executive compensation in aligning managerial incentives with the interests of shareholders changes for firms under financial distress. Firms going through bankruptcy are in a situation in which the separation between ownership and monitoring functions is reduced since owners have more power to monitor managers. Thus, agency costs of equity are diminished. In particular, the concept of cost agency must not be focused on the agency relationship between managers and shareholders. Other conflicts of interests arise between managers and creditors (John and John, 1993). Firms under financial distress generally have a high leverage ratio and are more controlled by creditors. The latter take an active role in developing executive compensation strategies. They can, for example, void the executive compensation contract or replace it. Such a model predicts a decrease in managerial pay sensitivity to shareholder wealth (Gilson and Vetsuypens, 1993). Empirically, the relationship between bankruptcy and executive compensation has not been analysed enough. To our knowledge, no attempt has been made to study this relationship and precisely identify its nature. A very few studies were limited to a clinical study approach or a simple descriptive analysis of executive pay levels for bankrupt firms (Gillan and Martin, 2007; Henderson, 2007; Fogarty et al., 2009). They did not seek to examine the association between remuneration and bankruptcy. It is therefore interesting to explore the relationship between these two variables in order to answer the question of whether or not executive compensation policies contribute to bankruptcy. Our objective is then to empirically examine the role of executive remuneration in corporate bankruptcy. The study of this relationship has important implications on the decisions of shareholders, regulators and investors. First, it allows shareholders a better design of an

3 70 S. Elleuch Hamza and I. Lourimi optimal remuneration policy to avoid bankruptcy and to maximise their wealth. Second, it draws regulators attention to the best measures to be taken when issuing new rules for executive compensation. Finally, the results of this study help investors to better predict corporate bankruptcy, consequently excluding failed companies from their investment portfolios. For these reasons, we propose to extend the research by analysing a global sample of bankrupt firms instead of limiting it to one case. Furthermore, we adopt logistic regression models to test this relationship instead of using a clinical study approach. Using a dataset of 210 of the largest US-listed companies (105 bankrupt and 105 matching solvent firms), the purpose of our paper is to examine the relationship between bankruptcy and CEO compensation in the context of crisis. CEO compensation is measured by cash and equity pay as well as total CEO pay. To test this relationship, we use a logit model regressing bankrupt firms on CEO compensation. The empirical findings indicate a negative relationship between bankruptcy and executive compensation, corroborating the optimal contracting model rather than the managerial power one. This is particularly the case during the two-year period before entering into the bankruptcy process. However, this relationship is not significant prior to this period. This study is organised as follows: the second section develops the managerial power and optimal contracting models consistent with the agency theory. The third section summarises the previous studies linking bankruptcy to executive compensation. The fourth and fifth sections describe the objective, the hypotheses and the methodology of this study. The sixth section provides the empirical results and the last one is devoted to the conclusion and policy implications. 2 Relationship between bankruptcy and executive compensation: managerial power versus optimal contracting The agency theory framework focuses on the conflict of interests between different corporate stakeholders and, in particular, between managers and shareholders. The separation between ownership and management gives managers the opportunity to expropriate wealth from shareholders by acting in one s own interests (Jensen and Meckling, 1976). The design of optimal compensation contracts is one solution among others to align the interests of shareholders with those of managers (Core et al., 2003). Such contracts link the compensation of managers to corporate performance. According to Murphy (2000), optimal contracting is viewed as an efficient means to reduce agency costs. In contrast, other authors regard those optimal compensation schemes as incomplete because shareholders cannot perfectly observe the implications of managerial choices (John and John, 1993) and because of the greater bargaining power of the CEO over directors (Brick et al., 2006). In fact, when corporate governance mechanisms are weak, managers can influence the components of executive pay in their own interests to receive excessive compensation. For example, compensation contracts that focus on annual bonuses encourage managers to manipulate short-term results (Bowen et al., 2008). Option-based compensation also creates incentives to increase a firm s share price by managing the magnitude and the volatility of earnings (Bergstresser and Philippon,

4 Executive compensation and corporate bankruptcy in the context of crisis ). Managers can also manipulate the timing of disclosure by delaying the disclosure of positive news after the issuance of their options in order to increase their value (Yablon and Hill, 2000). All these accounting manoeuvres have led many firms to bankruptcy, as was the case of Enron, Xerox and WorldCom (Healy and Palepu, 2003). In addition, equity-based pay encourages managers to take higher levels of risk and does not give them the incentives to act in favour of increasing shareholder wealth that it is supposed to provide. Managers can take riskier decisions (aggressive acquisition strategies) by focusing on their short-term results at the expense of the long run. Therefore, this excessive risk-taking raises the probability of default (Balachandran et al., 2010). In this perspective, executive compensation policies can deviate from the objectives proposed by the optimal contracting model. The design of executive compensation is changed when firms become financially distressed and the probability of bankruptcy increases. Such firms are generally highly leveraged and the presence of large creditors has an effect on the agency relationship. In this case, executive compensation arrangements depend not only on the relationship between managers and shareholders, but also between managers and creditors. The role of creditors in monitoring the managers of distressed firms has fundamental effects on executive compensation levels and types. These effects are treated by the optimal contracting model as well as the managerial power one. 2.1 Executive compensation structures in distressed firms: optimal contracting model According to the optimal contracting model, the indexing of pay to firm performance (pay-for-performance sensitivity) is a good solution to minimise agency costs of equity. This is particularly true for solvent firms characterised by a great separation between ownership and control (Jensen and Murphy, 1990a, 1990b). The pay-performance sensitivity is altered when a firm becomes financially distressed. In fact, such compensation contracts can induce managers into risky projects to maximise the firm s value. However, these decisions put a company in greater peril and raise the probability of default (Coles et al., 2006). This problem is intensified further in highly leveraged firms (Becht et al., 2012). In this case, distressed firms try to restructure their debt. Creditors accept to help them by giving new credit facilities to avoid bankruptcy (John and John, 1993; Henderson, 2007). They agree to provide these facilities by negotiating new contracts that increase their role in the firm s corporate governance. Indeed, these debt re-contracting terms put creditors in a better position to access private information and to monitor managers. The active role of creditors in corporate reorganisations thus reduces monitoring costs. Directors are also more diligent in monitoring managers when firm performance approaches insolvency. They play a significant role in key decision-making, especially regarding compensation practices (Henderson, 2007). In particular, corporate boards, facing a substantial bankruptcy risk, try to structure compensation arrangements in order to reduce the agency problems that are amplified by bankruptcy risk. Since agency costs in such firms are minimal, there is no incentive to align the interests of managers with those of shareholders. The optimal managerial compensation structure must thus show lower pay-performance sensitivity (Kang and Mitnik, 2008).

5 72 S. Elleuch Hamza and I. Lourimi However, Prendergast (2002) maintains that managers should receive higher compensation even if a firm is under financial distress. He argues that these managers assume more responsibilities and risks than in healthy firms and must thus be compensated by high rewards. This is particularly true for firms whose executives are more likely to be replaced by new ones. Such replacements considerably affect the reputation of CEOs and generate reputation costs (Fama, 1980; Gilson, 1989). Since these costs are larger for bankrupt companies executives, shareholders must compensate them with a high level of compensation. 2.2 Executive compensation structures in distressed firms: managerial power model The managerial power model argues that executive pay is not linked to firm performance. In contrast, compensation arrangements are viewed as a bargaining game between managers and boards. Managers profit from less effective board practices to influence their own pay (Brick et al., 2006). This theory designs two competing hypotheses to predict the level of executive compensation when a firm is going through a period of financial distress. The first one is the traditional view, which supposes that managers continue to profit from distressed firms and settle on contracts in line with their interests. The former still keep their bargaining power when fixing remuneration and are in a privileged position to set excessive pay, even if their firms become insolvent. According to this view, the board and the creditor agree to cooperate with existing executives because the latter has superior information about the business. They even tolerate higher compensation levels in order to find solutions to the firm s difficulties. In this instance, managers can make riskier decisions and this amplifies the probability of default. Many authors therefore argue that excessive compensation was a contributing cause to the financial crisis of 2008 (Balachandran et al., 2010). An increase in executive compensation is thus expected to be observed in financially distressed firms (Henderson, 2007). However, the second hypothesis predicts that managerial power is reduced when the firm is in difficulty. This current view argues that the bargaining game between managers and boards is dynamic. In fact, in such circumstances, the power shifts from executives to boards because fewer CEOs are the board s chairperson and fewer managers sit on the board. Executives who lose their power cannot influence their own remuneration. Boards have much freedom in determining a rational, instead of an excessive, amount of executive pay. The second hypothesis thus assumes a decrease in the level of compensation when a firm is near bankruptcy (Kang and Mitnik, 2012). 3 Previous studies on the relationship between bankruptcy and executive compensation Empirical works that have studied the relationship between executive compensation and bankruptcy are few. Several studies have shown this link through the management of risk taking. The excessive risk taking attitude of managers is thus deemed as a factor leading to bankruptcy (Cheng and Farber, 2008; Balachandran et al., 2010). In this sense, authors focused on the relationship between excessive risk taking and executive compensation. In particular, they examine one component of executive compensation, which is the

6 Executive compensation and corporate bankruptcy in the context of crisis 73 equity-based component. It involves both restricted stocks and stock options. This form of compensation induces more risk taking. Indeed, option pricing is positively associated with stock price. The latter depends on the volatility of the firms assets and this incites managers to undertake decisions that increase the level of asset volatility, such as the decision to invest more in research and development (Coles et al., 2006; Cheng and Farber, 2008; Landskroner and Raviv, 2010). Balachandran et al. (2010) provide evidence that equity-based compensation urges managers to choose riskier projects that increase the probability of default. This evidence, which was especially acute in the financial sector, shows risk-taking s contribution to the financial crisis. Using data from different sectors, Kadan and Swinkels (2008) also showed the positive and significant statistical relationship between bankruptcy risk and restricted stock compensation. By adopting a clinical analysis of some corporations, Gillan and Martin (2007) examine in-depth the factors contributing to the bankruptcy of Enron. They found that a significant amount of stock options were held by Enron executives. This is also true for the firm Nortel, which followed a compensation policy that was heavily based on option remuneration (Fogarty et al., 2009). The CEOs of these two companies not only received an excessive number of stock-options but also a large amount of bonus compensation. Indeed, since bonuses depend on short-term performance thresholds, executives prefer to engage in earnings manipulation activities at the expense of long-term performance, which accelerate their bankruptcy. Bebchuk et al. (2010) reached the same conclusion when studying the cases of Bear Stearns and Lehman. These findings are consistent with the managerial power hypothesis, suggesting that executives are in a better position to profit from distressed firms by increasing the level of their compensation. However, other studies do not support this conclusion. For example, Fahlenbrach and Stulz (2011) did not find any evidence that US financial institutions with higher option compensation had poor performance during the crisis of Thus, the compensation structure does not influence the economic situation of the bank during this period. Additionally, Erkens et al. (2012) who examined the compensation structure for top management of 306 financial institutions from 31 countries found the same results. By adopting a descriptive analysis of 76 distressed firms and a case study of WorldCom, Henderson (2007) also did not sustain a fundamental change in the compensation structure of CEOs. Some empirical work has even showed a significant reduction in total executive compensation when firms go through financial distress. A plausible explanation for this finding is that the relationship between risk-taking and executive pay depends on the level of leverage. Landskroner and Raviv (2010) pointed out that managers choose a riskier investment policy in order to maximise their stock remuneration when the level of leverage is low. On the contrary, the sensitivity to asset risk is low for highly leveraged firms. In this case, an optimal management compensation contract should have low payperformance sensitivity in order to reduce agency costs. The authors thus concluded that the relationship between asset risk and executive compensation is convex. John and Qian (2003) added that the pay-performance sensitivity in US banks is lower than it is in manufacturing firms and attributed this decrease to the high leverage of banks. Moreover, Hayes and Hillegeist (2006) argued that new CEOs of US firms, which face a substantial bankruptcy risk, receive lower levels of pay compared to those who manage low-risk firms. These findings are consistent with the results of Gilson and

7 74 S. Elleuch Hamza and I. Lourimi Vetsuypens (1993) that shed light on an increase in the management turnover rate in distressed firms. The authors explain their findings by the theoretical models of John and John (1993) that put in evidence the active role of creditors in affecting CEO remuneration arrangements even before entering into bankruptcy. The conclusions of these studies thus corroborate the optimal contracting hypothesis that reduces agency costs. Other studies explain this negative relationship differently. In fact, Kang and Mitnik (2012) attributed the reduction of executive compensation to the board structure. They found that distressed firms are characterised by fewer executive members and more separation between CEO and chairperson. They concluded that such characteristics decrease the managerial influence over the board in fixing the level of CEO remuneration. Their results confirm the negative effect of managerial power on executive compensation. 4 Research objective, hypotheses and conceptual model Prior studies have shown a positive link between bankruptcy and executive compensation. Many case studies indeed revealed that firms award their CEO excessive remuneration before entering into bankruptcy. This evidence is not shared by all authors since others showed that the level of compensation diminishes when a firm becomes financially distressed. None of this research directly examines the relationship between executive compensation and bankruptcy. It analyses the level and the structure of remuneration of some bankrupt firms or only presents descriptive statistics of financially distressed firms. Our study is not limited to some clinical cases, but rather treats a large sample of firms that entered into bankruptcy by using a logit regression model. Our approach is also markedly different from prior works by constructing a control group of non-bankrupt companies that was used to match with bankrupt firms. Our objective is thus to analyse statistically the relationship between executive compensation and bankruptcy. To achieve this objective, we chose to select the largest listed firms which had declared bankruptcy in the USA. Following the results of previous clinical studies conducted among US firms, we can conclude that executive compensation contracts intensify earnings management and increase the bankruptcy risk of firms. Excessive remuneration contributed to the financial scandals of the 2000 s and to the current financial crisis. We can thus propose a positive and significant association between executive compensation and firm failure. H 1 Total executive compensation is positively associated with bankruptcy. In particular, we assume that bankrupt firms are more likely to pay their CEOs with restricted stock and stock options. Indeed, as indicated in these clinical studies, managers of bankrupt firms held a large amount of stock options. So, we can suppose that the latter is motivated to raise the stock price by taking an excessive risk that increases the probability of bankruptcy. In this regard, we can formulate the following hypothesis: H 2 Equity-based compensation is positively associated with bankruptcy. Moreover, the same studies reveal that CEOs of bankrupt firms receive a large cash payment for a bonus. This component of compensation generally influences earnings

8 Executive compensation and corporate bankruptcy in the context of crisis 75 management behaviour and accelerates the bankruptcy of many firms. We therefore presume that executives are more focused on short-term benefits and sacrifice shareholder value by abusing accounting discretion. A positive relationship between cash compensation and bankruptcy is hypothesised. Cash compensation is positively associated with bankruptcy. H 3 5 Research design and empirical strategy 5.1 Definition and measurement of variables Three kinds of variables must be defined in this empirical study. The first kind is related to bankruptcy. We chose a legal term that qualifies a firm going into bankruptcy when it is reorganised under Chapter 11 of the Bankruptcy Code. Firms filing under this chapter continue to run their business and do not file for liquidation. These firms must develop a substantial reorganisation plan that restructures their debt and cannot proceed to any payment of principal or interest to the creditors. Bankruptcy, which is the dependent variable of our study, is then measured by a dummy variable which will take the value 1 if the company files under Chapter 11 and 0 otherwise. The second kind of variables focuses on executive compensation. Since the CEO is the main manager who can influence major decisions, we limit our study to the CEO s compensation. This independent variable also measures the level of CEO remuneration in terms of its structure. Thus, we included the different components of executive compensation. We first examined total CEO compensation. This choice is motivated by the pronounced upward trend in total CEO pay during the last few years (Faulkender et al., 2010). Following the literature, this variable is calculated by the natural logarithm of the sum of annual compensation, long term compensation and all other compensation. These components are disclosed in accordance with Securities and Exchange Commission (SEC) rules. When the CEO is replaced during the year, executive compensation is measured by adding the pay of both old and new CEOs. We then analysed each CEO compensation component separately. Specifically, we used two sets of elements: equity-based compensation and cash compensation. We chose the equity-based component as an explanatory variable for many reasons. First, the literature reveals that equity-based incentives induce greater risk-taking and are therefore the biggest component that can increase default probability. Second, many authors attribute the rise in total pay to the high level of equity-based compensation (e.g., Frydman and Jenter, 2010). Third, equity-based compensation constitutes a prominent target of regulators who are interested in changing the structure of pay in response to the financial crisis (e.g., Balachandran et al., 2010). We defined equity-based compensation as the natural logarithm of the sum of stock options and restricted stock awards according to previous major studies. We also chose the cash compensation component as another explanatory variable because it is viewed in the literature as an incentive for earnings manipulation and a factor in value destruction. In fact, bonus compensation incentivises a CEO to only focus on short-term benefits, since the bonus, which is generally paid in cash, is based on annual earnings. Cash pay includes salary, bonus and other cash compensation and is measured by the natural logarithm of these elements.

9 76 S. Elleuch Hamza and I. Lourimi The last kind of variables is the control variables. We introduced such variables in conformity with the bankruptcy prediction literature. As indicated by several researchers (Altman, 1968; Ohlson, 1980; Bellovary et al., 2007), many financial ratios can predict corporate bankruptcy. The most common ratios are size of the firm, liquidity, leverage and performance. Liquidity is generally measured by the ratio of current assets to current liabilities while leverage ratio is measured by long-term debt divided by total assets. Performance is commonly calculated by the ratio of net income to total assets (e.g., Bellovary et al., 2007). Finally, size is determined by the natural logarithm of total assets. 5.2 Period of study As mentioned in the previous literature, the last decade was characterised by two main financial crises. The earlier one began with the US financial bubble that reached its top in April 2000 and the recession that permeated the US economy a year later. Since that year, many corporate financial scandals occurred. The current crisis started with the symptoms of economic and financial distress which were officially declared in December 2007 (IMF, 2009) and reached their peak in The recession ended in the USA in mid These financial crises were characterised by a large number of bankruptcies (Landskroner and Raviv, 2010). We have therefore split our period of study into two sub-periods. The first begins with 2000 and includes the years 2001 and 2002 while the second extends from 2007 to the end of Data sources and sample collection The USA provides a particularly interesting context for many reasons. The first one is that the financial crisis began in this country before spreading to Europe and other countries. The study of the US context allows us to identify the initial reasons for numerous corporate failures marking this period. The second reason is linked to the evolution of executive compensation. Indeed, prior research has shown that the most important evolution of total executive compensation was noticed in this period (Conyon et al., 2013). The last reason is linked to the availability of data. Data about bankrupt US companies and CEO pay are, in fact, more available compared to other countries. The sample of US firms was selected according to the matching procedure and was composed of bankrupt firms and the same number of non-bankrupt firms. This procedure avoids statistical bias linked to the population s heterogeneity and minimises the error rate (Lennox, 1999). Each bankrupt firm was thus matched to one healthy firm that had the same criteria. According to many authors (e.g., Beaver, 1966; Ohlson, 1980; Crutchely et al., 2007), these criteria concern the industry sector, the firm size and the period of study. For each failed firm, we selected a non-bankrupt firm that operated in the same line of business and provided financial data for the same period. The financial sector was excluded due to the differences in its structure and the bankruptcy environment (Ohlson, 1980). In addition, as identified by the literature, a bankrupt firm was matched with a listed healthy company of similar size (in the same total asset tercile). The list of failed firms was tabulated from the UCLA-LoPucki bankruptcy research database (BRD) 1. This database provides information about all listed firms whose proceedings are before a court judge. As did Frydman and Saks (2010), as well as

10 Executive compensation and corporate bankruptcy in the context of crisis 77 Bergstresser and Philippon (2006), we chose large firms that have total assets superior to 1 billion dollars because the SEC obliges them to disclose their financial statements and information about CEO compensation. The sample of non-bankrupt firms was identified from the Fortune 500 classification 2 or from Forbes magazine 3. The sample covers data for the five-year period prior to the occurrence of bankruptcy. The choice of this period allows us to examine the evolution of CEO compensation not only one year before the bankruptcy, but also involves the period when the firm was more solvent. We cannot select a longer period due to the unavailability of data. We therefore identified 105 of the largest listed companies that filed bankruptcy and 105 healthy companies between and This sample includes most industrial sectors, such as manufacturing (38%), retail trade (27%) and services (21%). Since complete data was not available for each year of the five-year period, the sample size varies from 75 to 105, as seen in Table 1. Table 1 Firm size observed in five-year period Failed firms Healthy firms Total One-year before the bankruptcy Total Two-year before the bankruptcy Total Three-year before the bankruptcy Total Four-year before the bankruptcy Total Five-year before the bankruptcy Total The data related to CEO compensation and financial ratios were collected from electronic data gathering analysis and retrieval ( EDGAR ) database. This database, which was created in 1984, includes different statements of US and foreign companies listed on all US exchanges. These statements are established according to recommendations of the SEC 4. Data about CEO compensation were manually collected from proxy statements (FORM DEF 14A 5 ) while the remaining data were collected from annual reports (FORM 10-K).

11 78 S. Elleuch Hamza and I. Lourimi 5.4 Empirical model To study the relationship between bankruptcy and CEO compensation, we conducted two kinds of analysis. The first one consists of a comparative analysis, which tests the difference between CEO compensation components of the two groups (bankrupt and non-bankrupt firms). We used the Student test as well as the Wilcoxon test to verify if this difference is statistically different from or equal to 0. The second analysis focuses on multivariate regressions to show the effect of the CEO fees components on bankruptcy in the presence of control variables. Indeed, the review of bankruptcy prediction studies identified numerous factors that might have an effect on bankruptcy. The most popular one is leverage. Earlier studies have shown that highly leveraged firms have a potential probability of default and have provided evidence on the positive relationship between firm liability and bankruptcy (e.g., Chatterjee et al., 1995). Liquidity is also another factor that is popularly associated with bankruptcy. A firm that cannot pay its short-term financial obligations at maturity is in a liquidity crisis and, if this crisis is not solved, it must declare bankruptcy (Beaver, 1966). We thus assume that a lack of cash flow induces bankruptcy. Moreover, existing literature has found that a company is most likely to go bankrupt when it has generated losses or when profitability is low. According to earlier studies, bankrupt firms are characterised by poor performance (e.g., Ohlson, 1980; Chatterjee et al., 1995). Finally, prior studies have confirmed the negative impact of firm size on bankruptcy prediction (e.g., Ohlson, 1980). In fact, large firms are less prone to bankruptcy since they are more diversified and can therefore borrow more easily. Since dependent variable is a dichotomous variable, we used a logistic-binomial regression to estimate the effect of the explanatory variables. Based on this dichotomous value, the model verified the relationships between the explanatory variables and the probability of default. All suggested hypotheses were tested by the following model: Bankruptcyi = b0 + b1log ( CEO compensationsi) + b2( Leveragei) + b3( Liquidityi) + b4( Performancei) + b5( Sizei) + εi where Bankruptcy is measured by a dichotomous variable taking the value 1 if the company (i) is failed and 0 otherwise. CEO compensation is measured by: a the total compensation defined as the sum of annual compensation, long-term compensation and all other compensation (Model 1) b the equity-based compensation defined as the sum of stock-options and restricted stock awards (Model 2) c the cash compensation defined as the sum of salary, bonus and other cash compensation (Model 3). Leverage ratio is measured by long-term debt divided by total assets. Liquidity is measured by the ratio of current assets to current liabilities.

12 Executive compensation and corporate bankruptcy in the context of crisis 79 Firm performance is measured by the ratio of net income to total assets (ROA). Firm size is measured by the natural logarithm of total assets. b k (k = 0,, 5) indicate the model parameters estimated using the maximum likelihood method. To test their significance, we used the Wald test, which is similar to the Student test in the linear regression. 6 Empirical results 6.1 Results of descriptive statistics Table 2 presents descriptive statistics of CEO compensation variables as well as firm characteristics variables one-year before the bankruptcy. As reported in this table, CEOs of healthy firms receive higher levels of compensation than CEOs of failed firms since the total CEO fees of the former exceed on average (median) $5,805,000 ($4,943,000), whereas CEOs of failed companies only receive $3,846,000 ($2,409,000). This is also true for all components of CEO pay. However, these variables present a wide difference between minima and maxima. Their standard deviations also indicate a wide variation within the sample. Table 2 Descriptive statistics of sample characteristics one-year before the bankruptcy Panel A: descriptive statistics of bankruptcy firms CEO compensation variables CEO cash CEO equity-based Total CEO Mean 2,126 1,720 3,846 Median 1, ,409 Minimum Maximum 25,973 16,934 25,973 Standard deviation 3,690 2,939 4,459 Firm characteristics Leverage (000s) Liquidity (000s) Performance (000s) Size (000s) Mean 6,234 1, ,772 Median 2, ,273 Minimum , Maximum 176,387 41,224 1, ,914 Standard deviation 19,484 4,548 1,359 14,800 Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long term compensation and all other compensation. Leverage is measured by the total liabilities. Liquidity is measured by the current assets. Performance is measured by the net income.

13 80 S. Elleuch Hamza and I. Lourimi Table 2 Descriptive statistics of sample characteristics one-year before the bankruptcy (continued) Panel B: descriptive statistics of non-bankruptcy firms CEO compensation variables CEO Cash CEO equity-based Total CEO Mean 3,260 2,545 5,805 Median 2,303 1,083 4,943 Minimum Maximum 19,494 19,154 34,754 Standard deviation 3,429 3,758 5,447 Firm characteristics Leverage (000s) Liquidity (000s) Performance (000s) Size (000s) Mean 3,697 1, ,933 Median 1, ,586 Minimum , Maximum 54,921 25,964 2,672 82,641 Standard deviation 6,903 2, ,209 Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long term compensation and all other compensation. Leverage is measured by the total liabilities. Liquidity is measured by the current assets. Performance is measured by the net income. It is important to signal that this situation was not the same three years prior to the bankruptcy. As seen in Table 3, the level of CEO compensation for failed firms and non-failed ones is indeed closer. The average of CEO compensation components for bankrupt firms is even higher. For example, total CEO compensation for these firms is on average $7,511,000, whereas CEOs of healthy firms only earn an average of $6,857,000. Table 3 Descriptive statistics of CEO compensation variables three-year before the bankruptcy Panel A: descriptive statistics of bankruptcy firms CEO cash CEO equity-based Total CEO Mean 2,271 5,240 7,511 Median 1, ,441 Minimum Maximum 26, , ,580 Standard deviation 3,470 20,350 21,904 Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long term compensation and all other compensation.

14 Executive compensation and corporate bankruptcy in the context of crisis 81 Table 3 Descriptive statistics of CEO compensation variables three-year before the bankruptcy (continued) Panel B: descriptive statistics of non-bankruptcy firms CEO cash CEO equity-based Total CEO Mean 2,287 4,570 6,857 Median 1,352 1,562 3,055 Minimum Maximum 27, , ,210 Standard deviation 3,590 16,591 17,341 Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long term compensation and all other compensation. With regard to firm characteristics, it appears from Table 2 that, compared to non-bankrupt firms, failed companies are highly leveraged (on average $6,234,000 versus $3,260,000) and have poor profitability (on average $ 779,000 versus $62,000). However, we cannot observe a discernible difference in liquidity between the two types of firms (on average $1,612,000 versus $1,676,000). 6.2 Results of comparative analysis Table 4 compares CEO compensation components of failed firms with those of healthy firms during the five-year period prior to the bankruptcy. Results reported in this table show that one-year before the bankruptcy, the total CEO compensation of failed firms is statistically lower than the one of healthy firms at the 1% level (mean of differences = 0.564). This can be explained by the decrease of both cash and equitybased compensation. Furthermore, Table 4 indicates that the decline in CEO compensation begins two years before the bankruptcy, with a significant decrease in the equity-based compensation component at the 1% level (Student test is and Wilcoxon test is 3.687). This conclusion is valid for results obtained in the third year prior to the bankruptcy but only at the 5% level. However, these interpretations cannot be maintained when we analyse results presented four years before the bankruptcy. The decrease in CEO compensation is in fact no more significant at the 5% level. It therefore appears that bankrupt firms tend to diminish the level of CEO compensation as the year of bankruptcy draws near. 6.3 Results of logit regressions To test the relationship between bankruptcy and CEO pay, we carried out multivariate logistic regressions to show the effect of the CEO compensation components on bankruptcy in presence of control variables.

15 82 S. Elleuch Hamza and I. Lourimi Table 4 Results of comparative analysis One-year before the bankruptcy Mean value Mean of differences Student test Wilcoxon test CEO cash compensation * 3.454*** Failed firms Healthy firms CEO equity-based compensation * 2.184* Failed firms Healthy firms Total CEO compensation *** 3.983*** Failed firms Healthy firms Two-year before the bankruptcy CEO cash compensation * 2.096* Failed firms Healthy firms CEO equity-based compensation ** 3.687*** Failed firms Healthy firms Total CEO compensation * 2.818** Failed firms Healthy firms Three-year before the bankruptcy CEO cash compensation *** Failed firms Healthy firms CEO equity-based compensation * 2.726** Failed firms Healthy firms Total CEO compensation * 2.527* Failed firms Healthy firms Notes: Cash compensation is measured by the logarithm of the sum of salary, bonus and other cash compensation. Equity-based compensation is measured by the logarithm of the sum of stock-options and restricted stock awards. Total compensation is measured by the logarithm of the sum of annual compensation, long-term compensation and all other compensation. ***, **, *, Denote significance at the.001,.01,.05,.10 level.

16 Executive compensation and corporate bankruptcy in the context of crisis 83 Table 4 Results of comparative analysis (continued) Four-year before the bankruptcy Mean value Mean of differences Student test Wilcoxon test CEO cash compensation Failed firms Healthy firms CEO equity-based compensation Failed firms Healthy firms Total CEO compensation * Failed firms Healthy firms Five-year before the bankruptcy CEO cash compensation Failed firms Healthy firms CEO equity-based compensation * 2.365* Failed firms Healthy firms Total CEO compensation * 2.144* Failed firms Healthy firms Notes: Cash compensation is measured by the logarithm of the sum of salary, bonus and other cash compensation. Equity-based compensation is measured by the logarithm of the sum of stock-options and restricted stock awards. Total compensation is measured by the logarithm of the sum of annual compensation, long-term compensation and all other compensation. ***, **, *, Denote significance at the.001,.01,.05,.10 level. The results of multivariate logistic regressions are reported in Table 5. These regressions seem to have good explanatory power (Nagelkerke R²s exceed 60%) and goodness of fit (chi-square tests are significant at 1% level). We can therefore conclude that the regression models are well specified. Table 5 reveals a deterioration of the total CEO compensation level only one year prior to the bankruptcy since the coefficient of this variable (coefficient = 0.964) is significant at 0.1% level (Wald test value is 11.54). This coefficient is not significant for the remainder of the period. It is also important to note the significant lower equity-based compensation for the CEOs of failed companies. This is particularly true two years prior to the bankruptcy where the coefficient is significant and negative (coefficient = 0.117) at the 1% level (Wald test value is 5.7). However, Table 5 does not show any significant difference in cash compensation levels between bankrupt and healthy firms.

17 84 S. Elleuch Hamza and I. Lourimi Table 5 Results of multivariate logit regression models CEO cash compensation CEO equity-based compensation Total CEO compensation Panel A: one-year before the bankruptcy Leverage Liquidity Performance Size Coefficient Wald test *** *** * 0.2 R 2 Nagelkerke 60.2% Percent correctly predicted 83.3% Chi-square test 104.2*** Panel B: two-year before the bankruptcy Coefficient Wald test ** *** * R 2 Nagelkerke 55.7% Percent correctly predicted 83.3% Chi-square test *** Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long-term compensation and all other compensation. Leverage is measured by the total liabilities. Liquidity is measured by the current assets. Performance is measured by the net income. ***, **, *, Denote significance at the.001,.01,.05,.10 level.

18 Executive compensation and corporate bankruptcy in the context of crisis 85 Table 5 Results of multivariate logit regression models (continued) CEO cash compensation CEO equity-based compensation Total CEO compensation Panel C: three-year before the bankruptcy Leverage Liquidity Performance Size Coefficient Wald test *** R 2 Nagelkerke 57.8% Percent correctly predicted 82.3% Chi-square test *** Panel D: four-year before the bankruptcy 3.608* Coefficient Wald test *** R 2 Nagelkerke 62.6% Percent correctly predicted 82.1% Chi-square test *** Panel E: five-year before the bankruptcy Coefficient Wald test *** R 2 Nagelkerke 63.4% Percent correctly predicted 85.3% Chi-square test *** 4.261* Notes: Cash compensation includes salary, bonus and other cash compensation. Equity-based compensation includes stock-options and restricted stock awards. Total compensation is the sum of annual compensation, long-term compensation and all other compensation. Leverage is measured by the total liabilities. Liquidity is measured by the current assets. Performance is measured by the net income. ***, **, *, Denote significance at the.001,.01,.05,.10 level.

19 86 S. Elleuch Hamza and I. Lourimi Moreover, it appears from Table 5 that the leverage level and the performance of the firm are two important determinants of bankruptcy. The significant positive coefficient on the leverage variable (the coefficient is and the Wald test value is according to panel A) shows that a firm is more likely to go bankrupt when the level of leverage is high. The significant negative coefficient on the performance variable (the coefficient is and the Wald test value is according to panel A) implies that a decrease in performance enhances the probability of bankruptcy. However, the findings indicate that there is no relationship between liquidity and bankruptcy (the coefficient is 0.05 and the Wald test value is according to panel B), nor between size and bankruptcy (the coefficient is and the Wald test value is according to panel B). As a result, findings reveal that a firm is more likely to enter into bankruptcy when it is suffering performance difficulties. The negative coefficient of performance variable shows how a decrease in profit affects the probability of bankruptcy. A firm is also more likely to fail when leverage increases. These results therefore are consistent with earlier studies (Ohlson, 1980; Lennox, 1999). However, contrary to previous studies, liquidity and size do not affect significantly the probability of default. The insignificant effect of size factor can be explained by the preliminary choice of the largest bankrupt and healthy firms. Thus, we can conclude that the total CEO compensation of the failed and healthy firms is the same three years before the bankruptcy. Executives of failed companies do not receive higher fees than those who manage healthy ones. However, two years prior to this event, the equity-based compensation level of failed companies diminishes notably. This decline continues until the year before the failure and includes the total CEO compensation. Hence, executives experience an important decrease in their pay when their firms become financially distressed. The relationship between CEO compensation and bankruptcy becomes negative and significant in the period after a firm s distress and near its bankruptcy. This result is at odds with the first and second hypotheses in the sense that during the two-year period prior to the failure, both total compensation and equity-based remuneration are negatively and not positively associated with the bankruptcy. In addition, the absence of a relationship between cash pay and bankruptcy does not confirm the third hypothesis. The negative relationship between equity-based pay and bankruptcy is evoked by John and John (1993) who argued that bankrupt firms are in general highly leveraged and CEO remuneration in such firms is less reliant on stock-based compensation than in other firms, due to the greater volatility in their stock prices and the high agency costs of debt. Options are in this sense an inefficient way to remunerate managers. This explanation sustains the optimal contracting hypothesis that assumes low pay-performance sensitivity due to the reduction in monitoring costs. There is another possible explanation for this observation that is linked to the managerial power hypothesis. Since executives are more likely to be replaced or dismissed when their firms are experiencing problems (Gilson and Vetsuypens, 1993), new CEOs cannot negotiate the level of their compensation in their favour. They cannot benefit from the entrenchment mechanism that allows them to extract rents from their firm. In this sense, the managerial power hypothesis supposes a decrease in managerial influence when fixing the level of executive compensation.