alali et al-2012-accounting & finance

22
The effect of corporate governance on firm’s credit ratings: further evidence using governance score in the United States Fatima Alali a , Asokan Anandarajan b , Wei Jiang a a California State University Fullerton, Fullerton, CA 92834, USA b School of Management, New Jersey Institute of Technology, University Heights, Newark, NJ 07102, USA Abstract We investigate whether corporate governance affects firms’ credit ratings and whether improvement in corporate governance standards is associated with improvement in investment grade rating. We use the Gov-score of Brown and Caylor (2006), the Gomper’s G index and an entrenchment score of Bebchuk et al. (2009) to proxy for corporate governance. Using a sample of US firms, we find that firms characterized by stronger corporate governance have a signifi- cantly higher credit rating, and that this association is accentuated for smaller firms relative to larger firms. We find that an improvement in corporate gover- nance is associated with improvement in bond rating. Key words: Corporate governance; Credit ratings; Changes in corporate governance; Changes in credit ratings JEL classification: G24, G32, G34 doi: 10.1111/j.1467-629X.2010.00396.x 1. Introduction In this study, we use the methodology developed by Ashbaugh-Skaife et al. (2006) to examine the influence of corporate governance on a firm’s credit ratings. The issue of corporate governance has become more important because of the highly publicized financial reporting frauds at Enron, Worldcom and Parmalat (Palmrose and Scholz, 2004). One of the most important functions of We are grateful to the anonymous reviewers and the editors for their helpful comments on the paper. Received 31 August 2009; accepted 16 December 2010 by Robert Faff (Editor). Ó 2011 The Authors Accounting and Finance Ó 2011 AFAANZ Accounting and Finance 52 (2012) 291–312

Upload: ferlyan-huang-

Post on 28-Jan-2016

219 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Alali Et Al-2012-Accounting & Finance

The effect of corporate governance on firm’s creditratings: further evidence using governance score in the

United States

Fatima Alalia, Asokan Anandarajanb, Wei Jianga

aCalifornia State University – Fullerton, Fullerton, CA 92834, USAbSchool of Management, New Jersey Institute of Technology, University Heights, Newark,

NJ 07102, USA

Abstract

We investigate whether corporate governance affects firms’ credit ratings andwhether improvement in corporate governance standards is associated withimprovement in investment grade rating. We use the Gov-score of Brown andCaylor (2006), the Gomper’s G index and an entrenchment score of Bebchuket al. (2009) to proxy for corporate governance. Using a sample of US firms, wefind that firms characterized by stronger corporate governance have a signifi-cantly higher credit rating, and that this association is accentuated for smallerfirms relative to larger firms. We find that an improvement in corporate gover-nance is associated with improvement in bond rating.

Key words: Corporate governance; Credit ratings; Changes in corporategovernance; Changes in credit ratings

JEL classification: G24, G32, G34

doi: 10.1111/j.1467-629X.2010.00396.x

1. Introduction

In this study, we use the methodology developed by Ashbaugh-Skaife et al.(2006) to examine the influence of corporate governance on a firm’s creditratings. The issue of corporate governance has become more important becauseof the highly publicized financial reporting frauds at Enron, Worldcom andParmalat (Palmrose and Scholz, 2004). One of the most important functions of

We are grateful to the anonymous reviewers and the editors for their helpful commentson the paper.

Received 31 August 2009; accepted 16 December 2010 by Robert Faff (Editor).

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Accounting and Finance 52 (2012) 291–312

Page 2: Alali Et Al-2012-Accounting & Finance

corporate governance is to ensure the quality of the financial reporting processand the reliability of financial information. This study has two objectives. First,we investigate whether corporate governance affects firm’s credit ratings. Second,we examine whether an improvement in corporate governance standards helpswith achieving an all important investment grade rating status for firms. We usethree different measures of corporate governance; the Gomper’s index (G-score),the Brown and Caylor’s score (GOV_SCORE) and a score developed by Beb-chuk et al. (2009).1 Our findings corroborate the conclusions of the extant litera-ture with specific reference to Ashbaugh-Skaife et al. (2006) that corporategovernance does affect credit ratings.The GOV_SCORE that is used in this study comprises 51 factors that span

eight categories; audit, board of directors, charter/bylaws, director education,executive and director compensation, ownership, progressive practices and stateof incorporation. These 51 governance provisions include, but are not limited to,the dimensions used by Ashbaugh-Skaife et al. (2006) and other governanceattributes used in prior empirical work. Each of 51 factors is coded 1 if the firm’sgovernance is considered to be minimally acceptable or 0 otherwise. The GOV_SCORE is computed as the sum of the firm’s binary variables. Thus, higher val-ues indicate stronger corporate governance. However, the GOV_SCORE haslimitations. For example, it is possible for a firm to ‘game’ its GOV_SCORE byaltering peripheral factors such as getting their directors to attend a day-long ses-sion of an ‘ISS-accredited program’. Hence, as robustness checks, we used theGomper’s index (G-Score) used by Ashbaugh-Skaife et al., which has 24 charac-teristics and a score referred to as the entrenchment index developed by Bebchuket al. (2009), which has six characteristics.In this study, we use the methodology of Ashbaugh-Skaife et al. and address

the same research question measuring corporate governance using the sameindex as Ashbaugh-Skaife et al. We find similar results. However, we contributeto the literature by showing that the results of Ashbaugh-Skaife et al. holdwhen we use two additional indices to measure corporate governance(Gov-score and the index developed by Bebchuk et al. More importantly, wefind that an improvement in corporate governance by a firm is associated withan improvement in investment grading. Our additional results also show thatimproving corporate governance is especially beneficial for small firms than forlarge firms.Section 2 provides a brief discussion of relevant prior literature and presents

our hypotheses. Section 3 provides our research design and sample selection.Section 4 discusses our results; Section 5 discusses additional tests. Section 6discusses our robustness tests, and our conclusions are provided in Section 7.

1 The entrenchment index is based on six provisions; staggered boards, limits to share-holder bylaw amendments, poison pills, golden parachutes and super majority require-ments for mergers and charter amendments.

292 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 3: Alali Et Al-2012-Accounting & Finance

2. Literature review and hypotheses

Prior researchers on corporate governance and its implications focused onthree key areas: (i) the influence of corporate governance on the quality ofreported earnings; (ii) the influence of corporate governance on the accuracy ofanalysts’ forecasts; and (iii) the influence of corporate governance on bond rat-ings. In the first category, Baxter and Cotter (2009) and Davidson et al. (2005)found that corporate governance (as measured by composition of the audit com-mittee and extent of non-executive directors on the board, respectively) was posi-tively and significantly associated with improved earnings quality. In the secondcategory, Kent and Stewart (2008) and Bhat et al. (2006) found that analysts’forecast accuracy was increased in the presence of greater corporate governance.In the third category, to which this study belongs researchers note that greatercorporate governance can help mitigate both agency risk and information riskthrough the establishment of mechanisms designed to deal with managerialbehaviour (Bhojraj and Sengupta, 2003; Cremers et al., 2004; Anderson et al.,2004; Klock et al., 2004; Ashbaugh-Skaife et al., 2006).Ashbaugh-Skaife et al. found that higher corporate governance is associated

with higher bond ratings. Hence, they concluded that the level of corporate gov-ernance mechanisms should influence the assessment of default likelihood andthereby credit ratings. Ashbaugh-Skaife et al. (2006) used four surrogates,namely, type of ownership structure and its influence, financial stakeholder rights(measured by an index developed by Gomper referred to as the Gomper’s indexthat has 24 characteristics) and extent of financial transparency; board structureand processes for decision making. They provide evidence that credit ratings arehigher for firms characterized by high accrual quality, earnings timeliness andboard independence, but are lower for firms with a large number blockholders,excessive CEO power and stockholder rights. Other studies have found similarresults using different surrogates to measure corporate governance.Prior research has used various surrogates for corporate governance including,

among others, proportion of outside directors on the board (Byrd and Hickman,1992; Beasley, 1996; Dechow et al., 1996; Subrahmanyan et al., 1997; Klein,2002); CEO serving as chairman of the board (Imhoff, 2003); size of board(Monks and Minow, 2001; Klein, 2002); extent of non-audit fees (Magee andTseng, 1990; Frankel et al., 2002). These studies, irrespective of the type of surro-gate, show that increased corporate governance has the potential to influencecredit ratings positively. Shleifer and Vishny (1997) taking a different track notedthat the relaxation of restrictions on shareholders’ rights increases the investors’ability to monitor and discipline managerial actions, thus reducing incentives formanagers to engage in opportunistic financial reporting.Elbannan (2009) found that low internal control quality adversely affected

bond ratings. Bradley et al. (2009) (using board stability, director liability indem-nification as surrogates for corporate governance) also found that lower corpo-rate governance had a negative influence on bond ratings.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 293

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 4: Alali Et Al-2012-Accounting & Finance

The above discussions suggest that, in the presence of higher levels of corpo-rate governance, the flexibility of managers to act in their own self-interest is lim-ited, resulting in more effective decision making and higher firm performance.Indeed, Fitch Ratings (2004) emphasizes the importance of corporate gover-nance in the rating process. Strong governance structures and practices improvethe reliability and validity of the reported accounting numbers used by ratingagencies in assessing a firm’s likelihood of default (Ashbaugh-Skaife et al., 2006).This leads to our two related hypotheses stated as follows:

H1: Higher level of corporate governance is associated with higher credit ratings.

H2: Improvement in levels of corporate governance leads to improvement in creditratings.

3. Research design and sample selection

3.1. Model specification and variable definitions

To test our first hypothesis, we estimate the following logistic regression:Model 1.

CREDITRATING¼ b0þ b1GOV SCOREþX

j

bjðControl VariablesjÞ þ e:

ð1ÞThe credit ratings data used in our study were based on the long-term issuer

credit ratings compiled by Standard and Poor’s. The Standard and Poor’s ratingsrange from AAA (highest rating) to D (lowest rating debt in payment default).We measured credit rating in two ways following Ashbaugh-Skaife et al. (2006).We first created an indicator variable, INVSTMENT_GRADE, representing 1 ifthe credit rating was, AAA, AA+, AA), A+, A, A) BBB+ and BBB); 0 other-wise. A firm is considered an investment grade if coded 1 and a speculative gradeif coded zero. A logistic regression model is estimated using this two-categoryclassification scheme. To increase the robustness of our results and to further testthe predicted relations between corporate governance and credit ratings, we alsoestimated an ordered logit model (CREDIT_RATING). The ordered logit model(also referred to as proportional odds model) is a regression model for ordinaldependent variables and an extension of the logistic regression model for dichot-omous dependent variables.2 We use ordered logit regression (OLR) because wehave different categories of credit ratings. As mentioned, OLR is an extension of

2 The model makes the proportional odds assumption: the odds ratio for being in a chosencategory or higher compared to being in a lower category is the same regardless of whichcategory is chosen. This implies that if the ordinal variable were collapsed into two catego-ries, the odds ratio would be the same regardless of the cut-off chosen for the collapsing.

294 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 5: Alali Et Al-2012-Accounting & Finance

the logistic regression model for dichotomous dependent variable allowing formore than two ordered response categories and is recommended as the mostappropriate technique (Ashbaugh-Skaife et al., 2006) when the dependent vari-able has multiple values that can be ordered from low to high. For the dependentvariable of the ordered logit model, we collapsed the multiple ratings into sevencategories of credit ratings, which convey ordinal risk assessments. Each categoryis mapped into a range of credit ratings as follows:

Rating category 1: D, C, CC, CCC, CCC+Rating category 2: B), B, B+

Rating category 3: BB), BB, BB+

Rating category 4: BBB), BBB, BBB+

Rating category 5: A), A, A+

Rating category 6: AA), AA, AA+

Rating category 7: AAA

Our independent variable of primary interest was GOV_SCORE, which ismeasured using the governance score developed by Brown and Caylor (2006).We also had a number of control variables that could affect the default risk (andhence, credit rating) of a firm. We subsequently use a number of control vari-ables that have been used in the literature to surrogate or act as proxy for afirm’s default risk. These are as follows:

3.1.1. Leverage (LEV)

Leverage of a firm proxies for default risk because the higher the proportionof debt in its capital structure, the greater the probability the firm could find dif-ficulty in paying off its creditors, (Ashbaugh-Skaife et al., 2006; Bhojraj andSengupta, 2003). Thus, we expect a negative relationship between a firm’s lever-age and credit ratings. Leverage is measured as total debt divided by total assets.

3.1.2. Firm’s operating loss (LOSS)

When a firm incurs operating losses, the chances of paying off creditors coulddiminish (Ashbaugh-Skaife et al., 2006). This is measured as a dummy variablerepresenting 1 if the net income before extraordinary items is negative in thecurrent year; 0 otherwise. The coefficient of LOSS is expected to be negative.

3.1.3. Times interest ratio (INTCOVR)

This ratio indicates the ability of a company to pay off the interest due on itsdebt. The lower this ratio, the greater the difficulty in paying off interest andhence a higher default risk (Ashbaugh-Skaife et al., 2006; Bhojraj and Sengupta,2003). Thus, we expect a positive relationship between INTCOVR and credit

F. Alali et al./Accounting and Finance 52 (2012) 291–312 295

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 6: Alali Et Al-2012-Accounting & Finance

ratings. This is measured as the operating income before depreciation divided bythe interest expense.

3.1.4. Subordinated debt (SUBORD)

Similar to Ashbaugh-Skaife et al. (2006), we also control for differences in afirm’s debt structure by including subordinated debt. The debt structure of a firmwith subordinated debt is considered to be more risky and increase the level ofdefault risk. This is because of the differential claims to assets by debt providers.As such, we expect a negative relationship between SUBORD and credit ratings.SUBORD is measured as a dummy variable; 1 if a firm has subordinated debt; 0otherwise. We use Compustat data item 80 used in the Ashbaugh-Skaife et al.’sstudy to measure presence or otherwise of subordinated debt.

3.1.5. Firm size (SIZE)

This is included as a control variable because smaller-sized firms are assumedto have greater default of risk than larger firms (and vice versa) (Bhojraj andSengupta, 2003). We expect a positive relationship between SIZE and credit rat-ings. Firm size is measured as the natural log of total assets.

3.1.6. Market value of equity (MVBV)

The greater the market value of equity relative to book value, the higher theprobability of default risk (Bhojraj and Sengupta, 2003). This is because firms withhigher MVBV represent high-growth firms that could be associated with greaterrisk. Thus, we expect a negative relationship between MVBV and credit ratings.This variable is measured as the market value of stock multiplied by the numberof shares outstanding divided by the book value of equity at the end of a period.

3.1.7. Firm’s capital intensity (CAPINTEN)

The firm’s capital intensity is included to control for differences in the firm’sasset structure where firms with lower capital intensity are stated to have higherrisk of default and thus lower credit ratings, and vice versa (Ashbaugh-Skaifeet al., 2006). This is measured by the gross book value of property, plant andequipment divided by total assets.

3.1.8. Firm performance (ROA)

Generally, lower performing firms are associated with higher levels ofdefault risk, (Ashbaugh-Skaife et al., 2006; Bhojraj and Sengupta, 2003). Firmperformance is measured by return on assets which is the net income beforeextraordinary items divided by total assets.

296 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 7: Alali Et Al-2012-Accounting & Finance

3.1.9. Cumulative daily stock return over year t (CUMRET)

This variable captures the stock price performance of the firm over a period(Bhojraj and Sengupta, 2003). On the one hand, this could be positively associ-ated with future expected cash flows of the firm, suggesting a lower default riskand higher credit ratings. On the other hand, firms with good stock performancecould also be associated with higher default risk. Therefore, we do not predictthe sign for this coefficient.

3.1.10. Market risk (BETA)

This variable captures the systematic risk for the firm, and thus, increased lev-els of market risk could accentuate the risk of default and reduce credit ratingsof the firm (Bhojraj and Sengupta, 2003). We measure this by the equity betaobtained from Center for Research in Securities Prices (CRSP).

3.1.11. Type of industry in which a firm operates

To control for lower default risk for firms operating in regulated industries rel-ative to other industries. We create 13 indicator variables using the four-digitSIC industry classification (Frankel et al., 2002).3

The variable definitions are summarized in Table 1.To test the second hypothesis, we could not use ordinal logistic regression or

multivariate logistic regression.4 Hence, to test our second hypothesis regardingthe impact of change in corporate governance on credit ratings, we calculate thechange in probability of receiving an investment grade credit rating for a onestandardized unit change in the governance score and the control variables.5 Themarginal change in the probability of achieving an investment grade rating fromour logistic regression model is measured as follows:

3 Industry membership is determined by SIC code as follows: agriculture (0100–0999),mining and construction (1000–1999, excluding 1300–1399), food (2000–2111), textilesand printing/publishing (2200–2799), chemicals (2800–2824, 2840–2899), pharmaceuticals(2830–2836), extractive (1300–1399, 2900–2999), durable manufacturers (3000–3999,excluding 3570–3579 and 3670–3679), transportation (4000–4899), utilities (4900–4999),retail (5000–5999), services (7000–8999, excluding 7370–7379), and computers(3570–3579, 3670–3679, 7370–7379).

4 In our study, the cross tabs show many cells either empty or very small so that an ordi-nal logistic or multivariate logistic model would bias the results. The use of marginaleffects is appropriate as it is a common test used to determine the economic significance ineconomics research.

5 We standardize each non-binary variable by its mean and dividing by its standard devia-tion. Since the variables in our model are measured in different units, standardizationfacilitates the interpretation of our results.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 297

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 8: Alali Et Al-2012-Accounting & Finance

@pðXÞ=@xi ¼ bipðXÞ½1� pðXÞ�; ð2Þ

where b is the vector of coefficients, and X is the vector of independent variables.We employ two approaches to calculating the marginal effects. The firstapproach is to compute the marginal effect at each observation and then to cal-culate the sample average of individual marginal effects to obtain the overallmarginal effect. Under the second approach, the marginal effect is calculated atthe mean value of regressors. For instance, the marginal effect of the GOV_SCORE measures the change in the probability of receiving an investment gradebecause of a one standardized unit change in the governance score while holdingall independent variables at their mean values (Agresti, 2002). For large samplesizes, the two approaches yield similar results. However, for smaller samples,averaging the individual marginal effects is preferred when it is possible to do so(Greene, 2003, p. 668).

Table 1

Variable Definitions

Variable Definition

INVSTMENT_GRADE One if firm is an investment grade (S&P Debt Rating Compustat

no. 280 = 1 through 12 or BBB) or better), zero otherwise

CREDIT_RATING Assigned ordinal rating score

GOV_SCORE Corporate Governance Score computed based on Brown and

Caylor (2006)

HG One if GOV_SCORE ‡ 34, and zero otherwise

MG One if 29 £ GOV_SCORE £ 33, and 0 otherwise

LEV Total debt (Compustat no. 9 plus Compustat no. 34) divided by

total assets (Compustat no. 6)

ROA Net income before extraordinary items (Compustat no. 18) divided

by total assets (Compustat no. 6)

LOSS One if the net income before extraordinary items is negative in the

current and prior fiscal year, zero otherwise

INTCOVR Operating income before depreciation (Compustat no. 13) divided

by interest expense (Compustat no. 15) or (Compustat no. 339)

SUBORD One if company has subordinated debt, zero otherwise (Compustat

no. 80)

SIZE Natural log of total assets

CAPINTEN Gross PPE (Compustat no. 7) divided by total assets

CUMRET Cumulative daily stock return over year t

MVBV The market value of equity calculated as (Compustat no. 199)

multiplied by the number of shares outstanding (Compustat

no. 25); divided by the book value of equity at the end of period

(Compustat no. 60)

BETA The equity beta obtained from CRSP, (data item: BETAV)

298 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 9: Alali Et Al-2012-Accounting & Finance

3.2. Sample selection

Our initial sample consists of 8545 US companies for which GOV_SCOREdata are available from 2003–2005. We then exclude 5695 firms with missingcredit rating data (data item no. 280 Long-term Domestic Credit Ratings) andother financial data on Compustat. We further exclude 222 firms where there isinsufficient information to calculate the cumulative stock returns, stock marketbeta from CRSP.Traditionally, the argument could be made that firms with better corporate

governance have better performance especially when using the index developedby Gompers et al. (2003). Recent papers have questioned the findings of superiorperformance of so-called ‘democracies’ – firms with strong shareholder rightsover their polar opposite (dictatorships). For example, Johnson et al. (2009) findthat abnormal returns disappear once a fine classification of industry is used tocontrol for industry factors. The issue of controlling for industry factors isimportant. In this study, we use public company data and we control for indus-try effects using the industry classification of Frankel et al. (2002).

4. Results

4.1. Descriptive statistics and univariate tests

The descriptive statistics of our variables are shown in Table 2. The mean(median) credit rating is 3.788 (4) implying a crediting rating in the BBB+ toBBB) range. About 61 per cent of our samples have an investment grade creditrating. The mean Gov-score is 30.02. The mean Gov-score for the 25th percentileof our sample is 25, and the mean score is 35 at the 75th percentile. Untabulatedresults show that the average GOV_SCORE improved from 2002 to 2003 (22.7to 26.16) and further improved in 2004 (29.9). This indicates that the SarbanesOxley Act (SOX) significantly influenced corporate governance as measured bythe GOV_SCORE. The improvement could be either driven by SOX require-ments and the mandated provisions enacted by major US stock exchanges, or itcould stem from the fact that companies strengthened the corporate governanceon a voluntary basis in the post-SOX period.Table 2 also shows that firms in our sample are profitable (average ROA is

4.62 per cent). About 12 per cent of the firms in our sample incurred losses. Themean log of total assets is 22.15. On average, 65 per cent of a firm’s total assetsis invested in property, plant and equipment (CAPINTEN), and subordinateddebt (SUBORD) constitutes 16.93 per cent of total assets. Sample firms alsohave a market to book ratio (MVBV) of 2.95 and systematic risk (BETA) of1.22.Table 3 shows the correlations among the two credit rating measures and the

independent variables in our study. The upper right-hand portion of the tabledisplays the Pearson correlations, and the lower left-hand portion displays the

F. Alali et al./Accounting and Finance 52 (2012) 291–312 299

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 10: Alali Et Al-2012-Accounting & Finance

Spearman correlations. Both INVSTMENT_GRADE and CREDIT_RATINGare positively associated with the GOV_SCORE at the 1 per cent significancelevel, providing preliminary evidence supporting our hypothesis that strongercorporate governance is associated with higher credit ratings. While many of theindependent variables are correlated with the GOV_SCORE, none of the corre-lations are above 0.3, indicating multicollinearity is not likely a concern. We relyon the multivariate analyses to formally test our hypothesis in the next section.

Table 2

Descriptive Statistics (N = 2628)

Variable Mean Median 25th Pctl 75th Pctl SD

Dependent Variable

INVSTMENT_GRADE 0.6123 1 0 1 0.4873

CREDIT_RATING 3.7881 4 3 4 1.0638

Independent Variable

LEV 0.4862 0.4725 0.3918 0.569 0.1431

ROA 0.0462 0.0436 0.0215 0.0742 0.0586

LOSS 0.1221 0 0 0 0.3275

INTCOVR 12.893 6.9279 3.9942 13.328 19.267

SUBORD 0.1693 0 0 0 0.3751

SIZE 22.268 22.147 21.284 23.198 1.2616

CAPINTEN 0.647 0.5874 0.3377 0.9388 0.386

CUMRET 3.5747 3.105 )0.84 8.945 10.788

MVBV 2.9484 2.2796 1.6279 3.4382 3.5593

BETAV 1.2247 1.1205 0.8212 1.5309 0.5748

GOV_SCORE 30.034 31 25 35 5.7427

HG 0.3204 0 0 1 0.4667

MG 0.2842 0 0 1 0.4511

INVSTMENT_GRADE equals one if firm’s credit rating is investment grade, and zero otherwise.

CREDIT_RATING is the ordinal ranking of the S&P LT Domestic Issuer Credit Rating (Compu-

stat no. 280). GOV_SCORE = Corporate Governance Score computed based on Brown and Caylor

(2006); HG = one if GOV_SCORE ‡ 34, and zero otherwise; MG = one if 29 £ GOV_

SCORE £ 33, and 0 otherwise; LEV = total debt (Compustat no. 9 plus Compustat no. 34) divided

by total assets (Compustat no. 6); ROA = net income before extraordinary items (Compustat no.

18) divided by total assets (Compustat no. 6); LOSS = one if the net income before extraordinary

items is negative in the current and prior fiscal year, zero otherwise; INTCOVR = operating income

before depreciation (Compustat no. 13) divided by interest expense (Compustat no. 15) or (Compu-

stat no. 339); SUBORD = one if company has subordinated debt, zero otherwise (Compustat no.

80); SIZE = natural log of total assets; CAPINTEN = Gross PPE (Compustat no. 7) divided by

total assets; CUMRET = cumulative daily stock return over year t; MVBV = the market value of

equity calculated as (Compustat no. 199) multiplied by the number of shares outstanding (Compustat

no. 25); divided by the book value of equity at the end of period (Compustat no. 60); BETA = the

equity beta obtained from CRSP, (data item: BETAV). All continuous variables are winsorized at

the top and bottom 1%.

300 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 11: Alali Et Al-2012-Accounting & Finance

Tab

le3

Pearson’san

dSpearm

an’sCorrelationCoeffi

cients

INVSTMENT_

GRADE

CREDIT

_

RATIN

GLEV

ROA

LOSS

INTCOVR

SUBORD

SIZ

ECAPIN

TEN

CUMRET

MVBV

BETAV

GOV_

SCORE

INVSTMENT_

GRADE

10.820

)0.282

0.350

)0.357

0.225

)0.424

0.470

)0.008

0.006

0.149

)0.420

0.172

CREDIT

_

RATIN

G

0.880

1)0.319

0.449

)0.398

0.336

)0.336

0.523

)0.047

0.010

0.219

)0.451

0.186

LEV

)0.274

)0.284

1)0.286

0.298

)0.395

0.197

)0.108

)0.051

)0.036

0.099

0.053

)0.071

ROA

0.341

0.430

)0.289

1)0.657

0.410

)0.136

0.093

)0.075

0.065

0.325

)0.191

0.134

LOSS

)0.357

)0.389

0.257

)0.567

1)0.188

0.036

)0.119

0.086

)0.104

)0.166

0.271

)0.120

INTCOVR

0.434

0.519

)0.489

0.755

)0.426

1)0.162

0.095

)0.125

)0.022

0.178

)0.055

0.077

SUBORD

)0.424

)0.370

0.190

)0.153

0.036

)0.279

1)0.170

)0.048

0.030

)0.098

0.101

)0.111

SIZ

E0.477

0.511

)0.069

0.046

)0.113

0.117

)0.173

10.042

)0.034

0.069

)0.207

0.276

CAPIN

TEN

0.022

)0.008

)0.075

)0.120

0.085

)0.152

)0.082

0.062

10.059

)0.058

0.040

)0.029

CUMRET

0.012

0.017

)0.040

0.107

)0.145

0.056

0.059

)0.024

0.034

10.018

0.017

)0.127

MVBV

0.206

0.307

0.050

0.513

)0.211

0.460

)0.112

0.111

)0.135

0.130

1)0.088

0.091

BETA

)0.401

)0.444

0.045

)0.165

0.229

)0.134

0.124

)0.238

0.007

0.022

)0.071

1)0.012

GOV_

SCORE

0.170

0.180

)0.064

0.137

)0.116

0.163

)0.109

0.267

)0.014

)0.188

0.156

0.010

1

Pearson’s

correlation

coeffi

cients

arereported

intheupper

Trian

gle,

and

Spearm

an’s

Correlation

Coeffi

cients

arereported

inthebottom

Trian

gle.

GOV_S

CORE=

Corporate

Governan

ceScore

computedbased

onBrownan

dCaylor(2006);LEV

=totaldebt(C

ompustat

no.9plusCompustat

no.34)

divided

bytotalassets

(Compustat

no.6);ROA

=net

incomebefore

extraordinary

item

s(C

ompustatno.18)divided

bytotalassets

(Compustatno.6);

LOSS=

oneifthenet

incomebefore

extrao

rdinaryitem

sisnegativein

thecurrentan

dpriorfiscal

year,zero

otherwise;

INTCOVR

=operatingincome

before

depreciation(C

ompustat

no.13)divided

byinterestexpense

(Compustat

no.15)or(C

ompustat

no.339);SUBORD

=oneifcompan

yhas

subordi-

nated

debt,zero

otherwise(C

ompustat

no.80);SIZ

E=

naturallogoftotalassets;CAPIN

TEN

=Gross

PPE(C

ompustat

no.7)

divided

bytotalassets;

CUMRET=

cumulative

daily

stock

return

overyeart;MVBV

=themarket

valueofequitycalculatedas

(Compustat

no.199)

multiplied

bythenumber

ofshares

outstanding(C

ompustat

no.25);divided

bythebookvalueofequityat

theendofperiod(C

ompustat

no.60);BETA

=theequitybetaobtained

from

CRSP,(dataitem

:BETAV).Bold

correlationcoeffi

cients

indicatesign

ificance

atthe1%

level.Allcontinuousvariab

lesare

winsorizedat

thetopan

d

bottom

1%.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 301

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 12: Alali Et Al-2012-Accounting & Finance

4.2. Multivariate tests

4.2.1. Test results for model 1

The estimation results of Model 1 are reported in Table 4. When we use theINVSTMENT_GRADE as our dependent variable to surrogate for credit

Table 4

Logistic regression results of the effects of corporate governance score on firms’ Credit Ratings

CREDITRATING ¼ b0 þ b1GOV SCOREþP

j bjðControlVariablesjÞ þ e

Predicted

Sign

Dependent Variable =

INVSTMENT_GRADE

Dependent Variable =

CREDIT_RATING

Coeff. P > v2 Coeff. P > v2

Intercept ? )26.62 <0.0001 Not reported

GOV_SCORE + 0.0844 <0.0001 0.0471 <0.0001

LEV ) )4.898 <0.0001 )2.672 <0.0001

ROA + 15.766 <0.0001 11.114 <0.0001

LOSS ) )1.127 <0.0001 )1.019 <0.0001

INTCOVR + 0.0098 0.0634 0.0155 <0.0001

SUBORD ) )2.797 <0.0001 )1.367 <0.0001

SIZE + 1.3658 <0.0001 0.9853 <0.0001

CAPINTEN + 0.1161 0.5827 0.0366 0.7819

CUMRET ? )0.002 0.7211 0.0003 0.9299

MVBV ) 0.0146 0.4769 0.054 <0.0001

BETA ) )1.862 <0.0001 )1.435 <0.0001

Year Indicators Included Included

Industry Indictors Included Included

Pseudo-R2 0.5353 0.6274

Likelihood v2 2013.8 <0.0001 2594.4 <0.0001

Wald v2 602.08 <0.0001 1614.4 <0.0001

Sample size 2,628 2,628

Bold indicates governance variable(s) of interest. INVSTMENT_GRADE equals one if firm’s credit

rating is investment grade, and zero otherwise. CREDIT_RATING is the ordinal ranking of the

S&P LT Domestic Issuer Credit Rating (Compustat no. 280). GOV_SCORE = Corporate Gover-

nance Score computed based on Brown and Caylor (2006); LEV = total debt (Compustat no. 9 plus

Compustat no. 34) divided by total assets (Compustat no. 6); ROA = net income before extraordi-

nary items (Compustat no. 18) divided by total assets (Compustat no. 6); LOSS = one if the net

income before extraordinary items is negative in the current and prior fiscal year, zero otherwise;

INTCOVR = operating income before depreciation (Compustat no. 13) divided by interest expense

(Compustat no. 15) or (Compustat no. 339); SUBORD = one if company has subordinated debt,

zero otherwise (Compustat no. 80); SIZE = natural log of total assets; CAPINTEN = gross PPE

(Compustat no. 7) divided by total assets; CUMRET = cumulative daily stock return over year t;

MVBV = the market value of equity calculated as (Compustat no. 199) multiplied by the number of

shares outstanding (Compustat no. 25); divided by the book value of equity at the end of period

(Compustat no. 60); BETA = the equity beta obtained from CRSP, (data item: BETAV). All

continuous variables are winsorized at the top and bottom 1%.

302 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 13: Alali Et Al-2012-Accounting & Finance

rating, the coefficient on the Gov-score is positive and highly significant(P < 0.0001). Consistent with our hypothesis, this indicates that better corporategovernance is associated with a higher likelihood of receiving an investment gradecredit rating. Table 4 also presents the results when we used CREDIT_RATINGas the dependent variable in our regression. We find that this measure is also pos-itively and significantly associated with the GOV_SCORE. This adds to therobustness of our results and provides further support for our hypothesis thatposits a positive effect of corporate governance on firms’ credit ratings. Theseresults support our H1. Note that both models are highly significant with a likeli-hood ratio of 2013.8 and 2594.4, respectively, and both exhibit high explanatorypower with adjusted R squares of 0.5353 and 0.6274, respectively.Turning to control variables, we note that the coefficients of the control vari-

ables proxying for default risk (LOSS, LEV, SUBORD, MVBV and Beta) areall negative and significant (P < 0.0001). This indicates that crediting ratings arelower for firms reporting a loss, with a higher leverage level, higher systematicrisk, having subordinated debt, or experiencing high growth. We also find theperformance variable measured by ROA is positive and significant (P < 0.0001)indicating that higher performance is associated with higher credit ratings asexpected. The positive and significant (P < 0.0001) coefficient of the size vari-able indicates that, overall, the larger the firm, the greater the probability ofreceiving a higher crediting rating. These findings are in accordance with the pre-dictions based on the literature (Sengupta, 1998; Ashbaugh-Skaife et al., 2006;Bhojraj and Sengupta, 2003). We did not find a significant association betweenour rating measures and CAPINTEN and CUMRET. The coefficient of CUM-RET even though insignificant is negative. This is consistent with Bhojraj andSengupta (2003). Traditionally, we would expect CUMRET to be positivelyassociated with future expected cash flows suggesting lower default risk. How-ever, Bhojraj and Sengupta suggest that the alternative could be true as well,namely, that firms with superior stock performance could also be associated withhigher risk perhaps because of the ‘superior’ performance representing marketover valuation that could be possibly be adjusted in the future.

4.2.2. Test results for model 2

The marginal effects calculated using both approaches discussed in Section 3.1.are reported in Table 5. The marginal effect represents the change in probabilityof receiving an investment grade credit rating because of a one standardized unitchange in the variable of interest. The marginal effect of variable Xi is computedas ¶p(X)/¶xi = bip(X)[1 ) p(X)] where b’x is evaluated either 1) at individualobservations and averaged across the sample, or 2) at mean values of X.The result of key interest indicates that one unit increase in the GOV_SCORE

increases the probability of receiving an investment grade rating by 4.28 per centand 4.89 per cent, respectively, depending on how the marginal effect is calcu-lated. This result provides further evidence that governance mechanisms have

F. Alali et al./Accounting and Finance 52 (2012) 291–312 303

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 14: Alali Et Al-2012-Accounting & Finance

significant implications for assessing a firm’s credit rating and that improvementin GOV_SCORE is associated with improvement in credit ratings. This resultsupports our H2. In addition, the results for the other variables with respect tocoefficient signs and significance remained the same including the coefficient ofINTCOVR, which is positive and significant at P-value 0.06 level.

5. Additional analyses

In this section, we conduct additional analyses to examine the role of the gover-nance mechanism in explaining firms’ credit ratings. The raw GOV_SCORE used

Table 5

Marginal changes in probabilities of receiving an investment grade credit rating dependent variable:

INVSTMENT_GRADE (N = 2628)

Expected

sign

Marginal effect

at individual obs.

Marginal effect

at mean values

GOV_SCORE ? 0.0428*** 0.0489***

LEV ) )0.0619*** )0.0588***ROA + 0.0817** 0.0695***

LOSS ) )0.0996** )0.1047***INTCOVR + 0.0166* 0.0061*

SUBORD ) )0.2473*** )0.2495***SIZE + 0.15236*** 0.1526***

CAPINTEN + 0.00396 )0.0062CUMRET ? )0.0023 )0.0092MVBV ) 0.00458 0.0051

BETA ) )0.0946** )0.0948***

Bold indicates governance variable(s) of interest. The marginal effect represents the change in proba-

bility of receiving an investment grade credit rating because of a one standardized unit change in the

variable of interest. The marginal effect of variable Xi is computed as: ¶p(X)/ ¶xi = bip(X)[1 ) p(X)]where b’x is evaluated either 1) at individual observations and averaged across the sample, or 2) at

mean values of X, Greene (2003). *,**,*** indicates statistical significance at the 1%, 5% and 10%

level or better, respectively. INVSTMENT_GRADE equals one if firm’s credit rating is investment

grade, and zero otherwise. GOV_SCORE = Corporate Governance Score computed based on

Brown and Caylor (2006); LEV = total debt (Compustat no. 9 plus Compustat no. 34) divided by

total assets (Compustat no. 6); ROA = net income before extraordinary items (Compustat no. 18)

divided by total assets (Compustat no. 6); LOSS = one if the net income before extraordinary items

is negative in the current and prior fiscal year, zero otherwise; INTCOVR = operating income

before depreciation (Compustat no. 13) divided by interest expense (Compustat no. 15) or (Compu-

stat no. 339); SUBORD = one if company has subordinated debt, zero otherwise (Compustat no.

80); SIZE = natural log of total assets; CAPINTEN = Gross PPE (Compustat no. 7) divided by

total assets; CUMRET = cumulative daily stock return over year t; MVBV = the market value of

equity calculated as (Compustat no. 199) multiplied by the number of shares outstanding (Compustat

no. 25); divided by the book value of equity at the end of period (Compustat no. 60); BETA = the

equity beta obtained from CRSP, (data item: BETAV). All continuous variables are winsorized at

the top and bottom 1%.

304 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 15: Alali Et Al-2012-Accounting & Finance

in our main analysis is an ordinal measure. Bebchuk et al. (2009) notes that usingthe raw score measures imposes a linearity constraint on the ordinalmeasures.6 This constraint may or may not be appropriate. We therefore performa second set of analyses by relaxing this constraint. More specifically, we split oursample into three levels of corporate governance. We define HG (strong corporategovernance) to take on the value of 1 if GOV_SCORE ‡ 34, and 0 otherwise,while MG (medium corporate governance) takes on the value of 1 if 29 £ GOV_SCORE £ 33, and 0 otherwise. Our choices for interval width are designed to cap-ture unambiguously the high, medium and low levels of the governance factor.Roughly, 32 per cent of the firms are classified as having strong corporate gover-nance regimes, 28 per cent as having weak corporate governance.7 Using the parti-tioning scheme, we estimate the following logistic regression model:Model 3.

CREDITRATING ¼ b0 þ b1HGþ b2MGþX

j

bjðControlVariablesjÞ þ e

ð3Þ

where HG and MG represent the partitioning indicator variables defined previ-ously, and all other variables are as defined in Model 1. As specified, the modelalso allows us to estimate the incremental effect of the level of corporate gover-nance on firm’s credit ratings. The results are reported in Table 6.Table 6 presents the regression analyses for the partitioned corporate gover-

nance. In the column where INVSTMENT_GRADE is the dependent variable,the coefficients on HG and MG are 0.8814 and 0.1467, respectively, but the

6 Specifically, it would implicitly assume that the quantitative effect of going from, forexample, a low score (LG) to a medium score (MG) is the same as the quantitative effectof going from a medium score (MG) to a high score (HG). There is no particular reasonthat this should be the case. In other words, using the raw score implicitly imposes a ‘lin-ear’ restriction on the relation between the level of credit ratings and the Gov-score. Bypartitioning our sample using indicator variables, we relaxed this constraint, and let thedata determine whether the effect of moving from LG to MG is the same as moving fromMG to HG.

7 We use alternative definition of HG and MG using the 75th and 25th quartiles. That is,HG is defined as one if a firm has a GOV_SCORE of 35 or higher and zero otherwise.We redefine MG to take on the value of 1 if 25 £ Gov-Score £ 35, and 0 otherwise. Wefind that the coefficient of HG is positive and significant at the 1per cent level and that thecoefficient of MG is positive and significant at the 1 per cent level but the magnitude ofthe coefficient of HG (1.02) is more than twice as large compared to the coefficient of MG(0.45). In the OLR model, the coefficient of MG is positive and significant at 5 per centlevel, while the coefficient of HG is positive and significant at the 1 per cent level. Themagnitude of the coefficient of HG is more than twice higher than the coefficient of MG.These results further support that only truly strong corporate governance is associatedwith higher credit ratings.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 305

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 16: Alali Et Al-2012-Accounting & Finance

coefficient on MG is statistically insignificant (P-value = 0.4349). Note that thecoefficient of HG is considerably more positive in magnitude than that of MG.When we run the regression using CREDIT_RATING as our dependent vari-able, the results are similar except that the coefficient on MG has turned margin-ally significant. These results appear to suggest that only firms in the highestcategory of corporate governance enjoy significantly higher credit ratings relative

Table 6

Logistic regression results of the effects of high and medium governance scores on firms’ credit rat-

ings CREDITRATING ¼ b0 þ b1Hþ Gb2MGþPj

bjðControlVariablesjÞ þ e

Predicted sign

Dependent Variable =

INVSTMENT_GRADE

Dependent Variable =

CREDIT_RATING

Coeff. P > v2 Coeff. P > v2

Intercept ? )25.23 <0.0001 Not reported

HG ? 0.8814 <0.0001 0.5434 <0.0001

MG ? 0.1467 0.4349 0.3611 0.0716

LEV ) )4.809 <0.0001 )2.695 <0.0001

ROA + 15.407 <0.0001 11.086 <0.0001

LOSS ) )1.149 <0.0001 )1.01 <0.0001

INTCOVR + 0.0106 0.0464 0.0156 <0.0001

SUBORD ) )2.798 <0.0001 )1.374 <0.0001

SIZE + 1.3922 <0.0001 0.9952 <0.0001

CAPINTEN + 0.1498 0.4765 0.0544 0.6807

CUMRET ? )0.002 0.8172 0.0006 0.8704

MVBV ) 0.0189 0.3633 0.0562 <0.0001

BETA ) )1.848 <0.0001 )1.44 <0.0001

Year Indicators Included Included

Industry Indictors Included Included

Pseudo-R2 0.5347 0.6268

Likelihood v2 2010.5 <0.0001 2590.3 <0.0001

Wald v2 600.19 <0.0001 1611.9 <0.0001

Sample size 2628 2628

Bold indicates governance variable(s) of interest. INVSTMENT_GRADE equals one if firm’s credit

rating is investment grade, and zero otherwise. CREDIT_RATING is the ordinal ranking of the

S&P LT Domestic Issuer Credit Rating (Compustat no. 280). HG = one if GOV_SCORE ‡34, andzero otherwise; MG = one if 29£ GOV_SCORE £33, and 0 otherwise; LEV = total debt (Compu-

stat no. 9 plus Compustat no. 34) divided by total assets (Compustat no. 6); ROA = net income

before extraordinary items (Compustat no. 18) divided by total assets (Compustat no. 6); LOS-

S = one if the net income before extraordinary items is negative in the current and prior fiscal year,

zero otherwise; INTCOVR = operating income before depreciation (Compustat no. 13) divided by

interest expense (Compustat no. 15) or (Compustat no. 339); SUBORD = one if company has sub-

ordinated debt, zero otherwise (Compustat no. 80); SIZE = natural log of total assets; CAPIN-

TEN = gross PPE (Compustat no. 7) divided by total assets; CUMRET = cumulative daily stock

return over year t; MVBV = the market value of equity calculated as (Compustat no. 199) multi-

plied by the number of shares outstanding (Compustat no. 25); divided by the book value of equity

at the end of period (Compustat no. 60); BETA = the equity beta obtained from CRSP, (data item:

BETAV). All continuous variables are winsorized at the top and bottom 1%.

306 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 17: Alali Et Al-2012-Accounting & Finance

to other categories of firms. While firms with the strongest corporate governancehave a high likelihood of receiving an investment grade credit rating or bettercredit ratings, firms having moderately strong corporate governance may not.Thus, although efforts by firms to strengthen their corporate governance are per-ceived favourably by rating agencies, only those firms adopting strong corporategovernance regimes achieve a significant improvement in credit ratings. Theseresults provide additional insight into the relation between corporate governanceand credit ratings.We also examine whether governance mechanisms have a differential impact

on firms of different sizes. Small firms tend to be more risky. For high-risk firms,traditional indicators such as past profitability and debt–equity ratio may not beparticularly informative about future cash flows. Small firms also have greaterinformation asymmetry relative to large firms and are perceived to have a higherlikelihood of withholding unfavourable information with regard to firm-specificrisk. Therefore, rating agencies would rely more on the firm’s governance struc-ture such as the board’s monitoring of management actions and oversight of thefinancial reporting process in evaluating the default risk for small firms. As aresult, higher corporate governance in small firms may be perceived more posi-tively by rating agencies and thus have a relatively greater impact on these firms’credit ratings relative to larger firms.To test whether firm size affects our results, we partition the firms in our sam-

ple into terciles based on firms’ fiscal year-end total assets. The large firms aredefined as those falling into the top tercile, and the small firms are those in thebottom tercile. We ran separate regressions for the two subgroups, andthe results are reported in Table 7. Panel A presents the results usingINVSTMENT_GRADE as our rating measure. Similar to our main analysis, wefind a positive association between the GOV_SCORE and INVSTMENT_GRADE for both groups. However, while the coefficient on the GOV_SCOREis statistically significant at the 0.01 level for the small firms, it is only marginallysignificant for the large firms (P-value = 0.085). In addition, the magnitude ofthe coefficient estimate for the small firms is more than twice that for the largefirms (0.1574 versus 0.0711). Panel B reports similar results for the regressionbased on the CREDIT_RATING measure. These results lend support to ourconjecture that an increase in corporate governance has a greater effect (andhence is more important) for smaller firms relative to larger firms. In addition,we note that the control variables have the expected sign and significance. Note-worthy is the coefficient of INTCOVR that is marginally significant in Panel Afor both small and large firms and is significant at <0.001 in panel B.8

8 We consistently find the OLR that uses CREDIT_RATING as dependent variableprovides stronger and more significant results than the logistic regression that usesINVESTMENT_GRADE as dependent variable. This can be attributed to the nature ofthe OLR that it allows the intercepts to vary across different categories of the creditratings.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 307

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 18: Alali Et Al-2012-Accounting & Finance

Table 7

Logistic regression results of the effects of corporate governance score on firms’ credit ratings for

subsamples based on firm size

Predicted

sign

Large* Small*

Coeff.

estimate P > v2Coeff.

estimate P > v2

Panel A: Effect of GOV_SCORE on investment grade credit rating for large and small firms INVST-

MENT_GRADE = b0 þ b1GOV SCOREþPj

bjðControl VariablesjÞ þ eIntercept ? 1.8533 0.9184 1.4737 0.1664

GOV_SCORE ? 0.0711 0.085 0.1574 0.003

LEV - )4.999 0.0115 )3.416 0.0003

ROA + 9.543 <0.0001 11.51 <0.0001

LOSS ) 1.373 0.076 )1.52 0.0087

INTCOVR + 0.0166 0.3953 0.0053 0.4257

SUBORD ) )2.305 <0.0001 )4.132 <0.0001

CAPINTEN + )0.439 0.5377 1.4428 0.0001

CUMRET ? )0.043 0.0467 )0.003 0.7556

MVBV ) )0.028 0.6015 0.1413 0.0001

BETA ) )4.287 <0.0001 )1.479 <0.0001

Year Indicators Included Included

Industry Indictors Included Included

Pseudo-R2 0.4361 0.4557

Likelihood v2 501.89 <0.0001 532.77 <0.0001

Wald v2 103.9 <0.0001 166.64 <0.0001

Sample size 876 876

Panel B: Effect of GOV_SCORE on credit rating ranking for large and small firms CREDIT_

RATING = b0 þ b1GOV SCOREþPj

bjðControlVariablesjÞ þ eIntercept ? Not reported Not reported

GOV_SCORE ? 0.0642 0.0752 0.0731 0.0002

LEV ) )2.669 0.0005 )3.578 <0.0001

ROA + 2.409 <0.0001 3.2196 0.0517

LOSS ) )0.136 0.6959 )1.824 <0.0001

INTCOVR + 0.0171 0.0005 0.008 0.0639

SUBORD ) )1.964 <0.0001 )0.803 <0.0001

CAPINTEN + 0.087 0.7652 0.0909 0.675

CUMRET ? )0.015 0.0201 )0.005 0.5212

MVBV ) )0.063 0.0091 0.1374 <0.0001

BETA ) )2.243 <0.0001 )0.903 <0.0001

Year indicators Included Included

Industry indicators Included Included

Pseudo-R2 0.6492 0.4819

Likelihood v2 917.53 <0.0001 575.98 <0.0001

Wald v2 532.69 <0.0001 391.29 <0.0001

Sample size 876 876

308 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 19: Alali Et Al-2012-Accounting & Finance

6. Robustness checks

6.1. Tests using additional surrogates for corporate governance

As mentioned, the GOV_SCORE has limitations. An important limitationraised by Bebchuk et al. (2009) is that the ‘kitchen sink’ approach might be mis-guided. They note that, among a large set of provisions, the provisions of realsignificance are likely to constitute only a limited and possibly small subset. As aresult, an index that gives weight to many provisions that do not matter, and asa result underweighs the provisions that do matter is likely to provide a less accu-rate measure of governance quality than an index that focuses only on the latter.Furthermore, when the governance index include many provisions, firms seekingto improve their index rankings might be induced to make irrelevant or evenundesirable changes and might use their improved rankings to avoid making thefew small changes that do matter. To check the robustness of our results, we re-ran the regressions using the Gompers G-Score. The results corresponding toTables 4, 5 and 6 are not shown but available on request. The results usingthe Gompers G-Score are consistent with the main results of the study. We alsore-ran the regressions using the Bebchuk et al. (2009) index.9 The results are stillconsistent. However, we note that for the Bebchuk et al. index, data were onlyavailable for the year 2004.10 Hence, owing to smaller sample size, the tests havea limitation in that they lacked power.

Table 7 (continued)

Bold indicates governance variable(s) of interest. INVSTMENT_GRADE equals one if firm’s credit

rating is investment grade, and zero otherwise. CREDIT_RATING is the ordinal ranking of the

S&P LT Domestic Issuer Credit Rating (Compustat no. 280). GOV_SCORE = Corporate Gover-

nance Score computed based on Brown and Caylor (2006); LEV = total debt (Compustat no. 9 plus

Compustat no. 34) divided by total assets (Compustat no. 6); ROA = net income before extraordi-

nary items (Compustat no. 18) divided by total assets (Compustat no. 6); LOSS = one if the net

income before extraordinary items is negative in the current and prior fiscal year, zero otherwise;

INTCOVR = operating income before depreciation (Compustat no. 13) divided by interest expense

(Compustat no. 15) or (Compustat no. 339); SUBORD = one if company has subordinated debt,

zero otherwise (Compustat no. 80); SIZE = natural log of total assets; CAPINTEN = Gross PPE

(Compustat no. 7) divided by total assets; CUMRET = cumulative daily stock return over year t;

MVBV = the market value of equity calculated as (Compustat no. 199) multiplied by the number of

shares outstanding (Compustat no. 25); divided by the book value of equity at the end of period

(Compustat no. 60); BETA = the equity beta obtained from CRSP, (data item: BETAV). *Large

indicates that firm size falls into the top tercile, and Small indicates that firm size falls into the

bottom tercile. All continuous variables are winsorized at the top and bottom 1%.

9 For expediency, the results using the entrenchment index are not reported but are avail-able on request.

10 The Risk Matrix database does not report governance variables in years 2003 and2005, and as a result, our usable sample is small and results lack power.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 309

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 20: Alali Et Al-2012-Accounting & Finance

6.2. Issue of endogeneity

One of the fundamental problems that papers on corporate governanceencounter is the issue of endogeneity. Specifically, here the analysis could be sub-ject to an omitted variable bias, where an omitted variable could impact bothcredit ratings and the firm’s governance. To address endogeneity, we re-ran themodels using the simultaneous equation approach in which the endogenous vari-ables are debt rating and governance score. The results are consistent. The coeffi-cient of the GOV_SCORE is positive and significant at P-value 0.013 when weuse INVESTMENT_GRADE as a proxy; the coefficient of the GOV_SCORE ispositive and significant at P-value 0.007 when we use CREDIT_RATING.

7. Summary and conclusions

In this study, we provide evidence linking corporate governance to credit rat-ings. Governance mechanisms can reduce default likelihood by mitigating theagency risk through effective monitoring of management actions and by attenu-ating the information asymmetry between the firm and creditors. In the priorliterature, Bhojraj and Sengupta (2003) used outside board member compositionto surrogate for corporate governance. Similarly, Anderson et al. (2004) usedboth board member and audit committee composition to surrogate for corporategovernance. Using a single dimension to surrogate for corporate governance wascriticized as causing a potential correlated omitted variables problem that couldbias the results. Klock et al. (2004) subsequently used a score comprising 24dimensions of corporate governance. Ashbaugh-Skaife et al. (2006) subsequentlyused four surrogates including three single-dimension and one multidimensionsurrogate (i.e. G-score).As a further advancement, to measure corporate governance, we use a score

developed by Brown and Caylor (2006), the GOV_SCORE; a score developedby Gompers et al., the G-Score and an index developed by Bebchuk et al. Over-all, our results show that increased corporate governance does influence creditratings, namely higher levels of corporate governance are associated with highercredit ratings. Furthermore, we find that only firms in the highest category ofcorporate governance are associated with an increase in credit ratings. We findevidence that increased governance has a marginally greater impact on smallerfirms. Finally, we find that information related to changes in corporate gover-nance are value relevant since an improvement in corporate governance stan-dards help with achieving an all-important investment grade rating status changefor the firms in our sample. The findings in this research add to the extant litera-ture on whether the level of corporate governance is factored in credit ratingdecisions. Our results are important because they show that it is important forcompanies to enhance corporate governance in all its possible dimensionsbecause this information is assimilated by bond rating agencies and translates tolower cost of borrowing for companies.

310 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 21: Alali Et Al-2012-Accounting & Finance

References

Agresti, A., 2002, Categorical data analysis (John Wiley and Sons Inc., New York, NY).Anderson, R. C., S. A. Mansi, and D. W. Reeb, 2004, Board Characteristics, accountingreport integrity, and the cost of debt, Journal of Accounting and Economics 37, 315–342.

Ashbaugh-Skaife, H., D. W. Collins, and R. LaFond, 2006, The effects of corporate gov-ernance on firms’ credit ratings, Journal of Accounting and Economics 42, 203–243.

Baxter, P., and J. Cotter, 2009, Audit committees and earnings quality, Accounting andFinance 49(2), 267–290.

Beasley, M., 1996, An empirical analysis of the relation between the board of directorcompensation and financial statement fraud, The Accounting Review 71, 443–465.

Bebchuk, L., A. Cohen, and A. Ferrel, 2009, What matters in corporate governance?Review of Financial Studies 22(2), 783–827.

Bhat, G., O. K. Hope, and T. Kang, 2006, Do corporate governance transparency affectaccuracy of analyst forecasts? Accounting and Finance 46(5), 715–732.

Bhojraj, S., and P. Sengupta, 2003, Effect of corporate governance on bond ratings andyields: the role of institutional investors and the outside directors, The Journal ofBusiness 76, 455–475.

Bradley, M., C. Dong, G. Dallas, and E. Snyderwine, 2009, The effects of corporate gover-nance attributes on credit ratings and bond yields, Working paper (Duke University, IN).

Brown, D., and M. L. Caylor, 2006, Corporate governance and firm valuation, Journal ofAccounting and Public Policy 25, 409–434.

Byrd, J. W., and K. A. Hickman, 1992, Do outside directors monitor managers? Evidencefrom tender offer bids Journal of Financial Economics, 32(2), 195–221.

Cremers, M., V. B. Nair, and C. J. Wei, 2004, The impact of shareholder control onbondholders, Working paper (New York University, NY).

Davidson, R., J. G. Stewart, and P. Kent, 2005, Internal governance structures andearnings management, Accounting and Finance 45(2), 241–267.

Dechow, P., R. G. Sloan, and A. P. Sweeney, 1996, Causes and consequences of earningsmanipulation: an analysis of firms subject to enforcement actions by the SEC, Contem-porary Accounting Research 13(1), 1–36.

Elbannan, M. A., 2009, Quality of internal control over financial reporting, corporategovernance and credit ratings, International Journal of Disclosure and Governance 6(2),127–149.

Fitch Ratings, 2004, Credit policy special report, evaluating corporate governance:the bondholders’ perspective (Manual, New York, NY). Available from: ‘http://www.fitchratings.com’ ,retrieved June 12, 2009>.

Frankel, R. M., N. F. Johnson, and K. K. Nelson, 2002, The relation between auditor’sfees for non audit services and earnings management, The Accounting Review 77,71–105.

Gompers, P., J. Ishii, and A. Metrick, 2003, Corporate governance and equity prices,Quarterly Journal of Economics 118, 107–155.

Greene, W., 2003, Econometric Analysis (Prentice Hall, Englewood Cliff, NJ).Imhoff, E. Jr, 2003, Accounting quality, auditing, and corporate governance, AccountingHorizons 17(Suppl.), 117–128.

Johnson, S. A., T. C. Moorman, and S. Sorescu, 2009, A reexamination of corporate gov-ernance and equity prices, Review of Financial Studies 22(11), 4753–4786.

Kent, P., and J. Stewart, 2008, Corporate governance and disclosures on the transition tointernational financial reporting standards, Accounting and Finance 48(4), 649–671.

Klein, A., 2002, Audit committee, board of director characteristics, and earnings manage-ment, Journal of Accounting and Economics 33, 375–400.

F. Alali et al./Accounting and Finance 52 (2012) 291–312 311

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ

Page 22: Alali Et Al-2012-Accounting & Finance

Klock, M. S., S. A. Mansi, and W. F. Maxwell, 2004, Does Corporate Governance Matterto Bondholders?, Working paper (The Pennsylvania State University, PA).

Magee, R., and M. C. Tseng, 1990, Audit pricing and independence, The AccountingReview 65, 315–336.

Monks, R. A., and N. Minow, 2001, Corporate Governance (Blackwell Publishing, Lon-don, UK).

Palmrose, Z. V., and S. Scholz, 2004, The circumstances and legal consequences of non-GAAP reporting: evidence from restatements, Contemporary Accounting Research21(1), 130–190.

Sengupta, P., 1998, Corporate disclosure quality and the cost of debt, The AccountingReview 73, 459–474.

Shleifer, A., and R. W. Vishny, 1997, A survey of corporate governance, Journal ofFinance 52(2), 471–517.

Subrahmanyan, V., N. Rangan, and S. Rosenstein, 1997, The role of outside directors inbank acquisitions, Financial Management 26, 23–36.

312 F. Alali et al./Accounting and Finance 52 (2012) 291–312

� 2011 The AuthorsAccounting and Finance � 2011 AFAANZ