Corporate governance and default prediction: a reality test

ABSTRACT

Default prediction has commanded the attention of researchers for at least 50 years. This paper addresses several testable hypotheses regarding the relations between corporate governance and default prediction. We employ the conventional logistic regression to provide empirical evidence from U.S. default data over the period of 2000 to 2015. Empirical results are consistent with the following notions: First, default firms are associated with high ownership concentration, low shareholder rights, low financial transparency and disclosures, and less board effectiveness. Second, in-sample and out-of-sample tests support the incremental contribution of corporate governance information on default prediction, when compared with the models involving just financial information.

KEYWORDS:

• Corporate governance
• default prediction
• accounting information
• market information

JEL CLASSIFICATION:

• C01
• G21
• C34
• G34
• C87

Disclosure statement

No potential conflict of interest was reported by the authors.

Data Availability

Data analysed in the study are collected from public sources

Notes

¹ Under the Basel requirements, banks need to link the capital requirements of the bank to the individual level of credit risk.

² We consider this as the compressive analysis of corporate governance information pertaining to the corporate governance literature. The framework includes four dimensions assumed to be necessary to reduce management opportunistic behaviour and information asymmetry.

³ In this paper the terms financial distress, failure, bankruptcy and liquidation are used interchangeably as each represents the situation where a firm is placed in default and investors suffer credit loss.

⁴ Even though there are provisions on executive remuneration we have considered the shareholder rights to approve remuneration because the act has implemented in 2010 (say-on-pay provision under Dodd-Frank Wall Street Reform and Consumption act 2010), however, our sample period convers 2000 to 2015.

⁵ There are no such legislative requirements for the appointment of external auditor in US context.

⁶ Altman and McGough (1974) ascertain companies had received a going-concern modified opinion before bankruptcy occurred. Lensberg, Eilifsen, and McKee (2006) find the most significant variable in their final model of bankruptcy prediction was the auditor’s opinion. Unmodified opinion (qualified and unqualified with explanatory language) shows a negative effect on bankruptcy prediction in Lensberg and others’ study (2006).

⁷ The ratio working capital to total assets and cash to market value of total assets are measures for liquidity. Profitability is measured by using three variables; sales to total assets, retained earnings to total assets, earnings before interest and tax to total assets. Market value of equity to book value of total debt is a proxy for the leverage.

⁸ Qualified includes qualified (scope limitation and different from GAAP) and unqualified opinion with explanatory language.

⁹ We define outside directors those who are not employed by the company. But they are not independent because outside directors might be prior employees of the company or may provide consultancy services to the company and independent directors as those who have no any material relationships with the company.

¹⁰ PCA provides five components; two components for ownership concentration with other three dimensions (board effectiveness, shareholder rights and financial transparency). However, we limit to four dimensions based on the S&P (2002) identification.

¹¹ Matched pairs design has been used by more than 70% of the studies in this area (Zmijewski 1984).

¹² The bankruptcy prediction results generally used to find the effect of selected variables on default likelihood of the companies which may go bankrupt, but not to generalize to the entire population (Ciampi 2015).

¹³ We assume five-year observations are necessary to find the signal of default risk among default and non-default companies and banks generally conduct 3 to 5 year analysis of their borrowers. Li and Miu (2010) used 10 year quarterly data for their analysis based on the USA.

¹⁴ As a solution to the outliers founded after applying the median absolute deviation (MAD). After detecting outliers, we use trimming (Taffler 1983; Barnes 1987) to avoid false positive results. For trimming, we apply winsorizing, which means changing an outlier’s value into the value of the closest non-outlier (Barnes 1987).

¹⁵ Shleifer and Vishny (1997), argued, the ownership concentration is an incentive for owners to monitor management, however, if the ownership exceed a certain threshold, the owners motivate to pursue their private benefits.

¹⁶ The result is in agreement with Jensen’s study (1993).

¹⁷ As per Ashbaugh-Skaife, Collins, and LaFond (2006), the higher shareholder rights and power enhance the power balance between management and stakeholders.

¹⁸ Audit committee quality represent the independency of the audit committee chair.

¹⁹ A study based on the USA by Deakin (1972), also found WCTA as the best predictor of potential distress re-classification. MVEBTD also found to be significant in Aziz and Lawson’s study in 1989. STA ratio was the least significant variable in Altman’s study (1968). However, the highest prediction has provided by EBITTA in his study. Campbell, Hilscher, and Szilagyi (2008) found CASHMTA variable as the most significant in their study.

²⁰ Excess return shows a negative coefficient, representing companies with higher excess return have less exposure to default risk (Shumway 2001).

²¹ Size of the company measured based on the relative market capitalization indicates a positive sign even though we expect a negative sign. However, this is in line with the results of Campbell, Hilscher, and Szilagyi (2008). One possible explanation would be larger companies have complex business processes and they are more exposed to the default risk due to this complexity.

²² The calculation of accuracy ratio is described in Li and Miu study in 2010.

²³ Results are available upon the request.