Case Study

Casestudy5.1CaseStudyGovernmancecomplianceandoperatingperformance.pdf

Home >Business & Finance homework help >Accounting homework help >Case Study

Aecounting and Business Research. Vol. 39. No, 5, pp. 4 9 7 - 5 1 3 , 2009 497

On the relatíon between corporate governance compliance and operating performance Heidi Vander Bauwhede

Abstract — Better corporate perfonnance has been cited as one of the main benefits of adopting good corporate govemance structures within organisations. However, in contrast to theory, a prior European study (Bauer et al., 2004) reports evidence of a negative relationship between corporate govemance and corporate performance. This study re-examines this relationship, and reports evidence of a positive relationship between the extent of compliance with intemational best practices concerning board structure and functioning and operating perfonnance when operating perfonnance is measured by the retum on assets (ROA). This resuh is robust to controlling for the firms' compliance with best practices in other govemance areas, and holds for some other govemance dimensions, namely disclosure of corporate govemance and the range of takeover defences. Further tests indicate that greater compliance with intemational best practices conceming board stmcture and functioning is significantly associated with reporting less income from asset disposals and that studying a performance measure that includes this item obscures the inherently positive relationship between operating perfonnance and the extent of compliance with intemational best practices regarding board stmcture and functioning. The results provide some support for an often- cited motivation for the adoption of good govemance practices, and provide explicit evidence that the measure of operating perfonnance is cmcial in examining firm-level operating performance.

Keywords: corporate govemance; operating performance

1. Introduction This paper examines the relationship between corporate govemance compliance and operating performance for a set of large listed European companies. My focus is on compliance with intemational best practice in corporate govemance. Following Jensen (1993) and prior govemance research, I hypothesise that greater compliance with intemational corporate govemance best practices and, more specifically, best practices conceming the structure and functioning of the board, is associated with better operating performance, ceteris paribus.

I investigate the relationship between corporate govemance compliance and operating performance for a sample of European companies in 2000-2001, because, during that period, there remained consid- erable variation in corporate govemance practices

(see Wójcik, 2006; Bauer et a l , 2008), notwith- standing that there were pressures from, for example, institutional investors or cross-listings to comply with intemational corporate govemance best practices and that in some countries local codes were, de facto, mandatory."'^

The study focuses on operating performance, and not stock market performance, in order to investi- gate fiirther the result of a prior European study (Bauer et al., 2004) on the relation between compliance with best practices conceming corpor- ate govemance and operating performance which seems to conflict with both theory as well as prior American results. More specifically, Bauer et al. (2004) report evidence of a negative relationship between ratings on the extent of compliance with intemational best practices and firm operating

The author is at Maastricht University and at Ghent University. She also has an affiliation to Katholieke Universiteit Leuven. She gratefiilly acknowledges Ping-Sheng Koh, Kevin McMeeking, Piet Sercu, Konstantinos Stathopoulos, participants at the 2006 European Accounting Association Annual Conference (Dublin, Ireland), the editor and two anonymous reviewers for usefijl comments. She also thanks Deminor for providing the govemance data. The usual disclaimer applies.

Correspondence should be addressed to Dr Heidi Vander Bauwhede, Maastricht University, Department of Accounting & Information Management, P.O. Box 616, Maastricht, 6200 MD, Netherlands. E-mail: [email protected].

This paper was accepted for publication in July 2009.

' I refer to a study commissioned by the European commis- sion (Weil et al., 2002) and to the website of the European Corporate Govemance Institute (http://www.ecgi.org/codes/ all codes.php) for an overview of the corporate govemance codes in the European Union. Intemational govemance codes are, for example, those established by the Intemational Corporate Govemance Network (ICGN), and the Organisation for Economic Co-operation and Development (OECD).

^Some countries (such as the UK and Italy) required companies to disclose whether they complied with a (national) corporate govemance code under a 'comply or explain' approach. This approach requires firms to disclose whether (and to what extent) they comply with a particular corporate govemance code and, if they do not (fully) comply, to explain why they do not comply.

498 ACCOUNTING AND BUSINESS RESEARCH

performance, whereas theory (Jensen, 1993), pre- dicts a positive relationship^ and a prior American study (Larcker et al., 2005) finds some (albeit weak) evidence of a positive relationship. I primarily focus on board structure and functioning, and not on other dimensions of corporate governance (such as, for example, rights and duties of shareholders and range of takeover defences), because it is especially the structure and functioning of the board that can directly affect the operating efficiency and operat- ing performance of a company. However, for completeness, I also perform and report the results of some additional analyses on the relation between other dimensions of corporate governance and firm operating performance.

I use a sample of European listed companies for which a private rating agency issues a firm-level rating of the extent of compliance with inter- national best practices conceming board structure and functioning. Results of univariate and multi- variate tests indicate that the one-year ahead retum on assets (ROA) increases in the extent of compliance with intemational best practices con- ceming board structure and functioning. Tests show that the results are not affected by the potential endogeneity of the extent of govemance compliance. In addition, the results are robust to controlling for the firms' compliance with best practices in other govemance areas, such as rights and duties of shareholders and-range of takeover defences, and to controlling for country-level performance. Moreover, I also find a positive relation between the extent of compliance with recommendations in some other govemance dimensions, more specifically disclosure on cor- porate govemance and range of takeover defences, and firm operating performance.

Further, additional analyses indicate that greater compliance with intemational best practices con- ceming board stmcture and fiinctioning is signifi- cantly associated with reporting less income from asset disposals and that studying a performance measure that includes the income fi^om asset disposals, such as the retum on equity (ROE) or net profit margin (NPM) used by Bauer et al. (2004), instead of a performance measure which is not impacted by the income fi^om asset disposals, such as the retum on assets (ROA), obscures the inherently positive relationship between operating performance and the extent of compliance with

^ Bauer et al. (2004) find indications of a positive relationship between govemance ratings, and stock retums and firm value, respectively.

intemational best practices regarding board struc- ture and functioning.

This study contributes to the literature on the relation between corporate govemance and corpor- ate performance. A first contribution is that the study reports a positive relation between the extent of compliance with intemational best practices on various govemance dimensions (board stmcture and functioning, disclosure on corporate govem- ance) and the operating performance of European companies. A second contribution is that this study reports evidence which indicates that the unex- pected negative relationship between corporate govemance compliance and operating performance as reported by Bauer et al. (2004)̂ * is due to poorly- govemed companies using the available discretion over the timing of asset sales to cover up their inherently lower operating performance. The key difference between this study and that of Bauer et al. (2004) is that the retum on assets is introduced as the preferred measure of operating performance because the income measure used in computing the retum on assets, i.e. operating income, is less influenced by discretionary items than the income measure used to compute the retum on equity or net profit margin, i.e. income before extraordinary items. The retum on equity and net profit margin are the performance measures used by Bauer et al. (2004).

The remainder of the paper is organised as follows. The next section develops the main research hypothesis. Section 3 describes the sample and data. Section 4 presents the empirical model. Section 5 presents the empirical results. Section 6 concludes.

2. Hypothesis development The various corporate govemance codes that have been issued since the late 1990s oflen refer to better performance as one of the key benefits of adopting their corporate govemance recommendations. This performance can be understood as better market performance (i.e. higher stock retums or firm

Examples of other studies that have examined the relation between govemance and perfomiance using samples from other countries (for example, the US, Australia, and various Asian and some (individual) European countries), and using and focusing on a variety of govemance attributes and performance measures, are: Larcker et al. (2006), Black et al. (2006), Brown and Caylor (2006a), Brown and Caylor (2006b), Dumev and Kim (2005), Larcker et al. (2005), Alves and Mendes (2004), Bebchuk et al. (2004), Klapper and Love (2004), Drobetz et al. (2004), Kiel and Nicholson (2003), Bhagat and Black (2002), Yermack (1996), and Klein (1998).

Vol. 39, No. 5. 2009 499

value)^ or as better operating performance. The expected relationship between compliance with corporate govemance recommendations and oper- ating performance is based on the argument that firms with a better govemance structure operate more efficiently which increases their operating performance (see, for example, Jensen, 1993). However, results of previous studies on the relation between govemance and operating performance are mixed. Larckeret al. (2005), for example, find some evidence of a positive relationship between an overall govemance metric (The Corporate Library Board Effectiveness Rating) and the one-year ahead ROA for a set of large listed American companies. By contrast, Bauer et al. (2004) find a negative relationship between an overall govemance score and operating performance for large European companies.

As with any govemance study, a crucial element in examining the relationship between govemance and performance is how one defines and measures 'better govemance'. In this study, I use a rating, issued by a private rating agency (Deminor rating),^ that assesses the extent to which large listed European firms comply with intemational best practices conceming corporate govemance and, more specifically, the extent to which firms comply with intemational best practices conceming board structure and functioning.^ Higher compliance is implicitly assumed to be better govemance. However, this is not necessarily true. A first reason is that European companies may have adopted govemance mechanisms and practices that differ from the intemationally accepted best practices, but are better tailored to the specific context in which they operate. However, it is probably also true that

^ Examples of studies that have examined aspects of corpor- ate govemance and market performance in an American setting are Yemack (1996), Bhagat and Black (2002), Gompers et al. (2003) and Bebchuk et al. (2004). Beiner et al. (2006), Alves and Mendes (2004), Drobetz et al. (2004) and Kiel and Nicholson (2003) are examples of govemance-market perform- ance studies using samples of Swiss, Portuguese, German and Australian companies, respectively.

* In Section 3,1 provide more detail on the rating. ' M o s t govemance studies use either a single indicator of

govemance, or an 'arbitrary' index. Larcker et al. (2006) argue that measurement error in these govemance metrics may be partly responsible for the mixed results on the association between tiie typical measures of corporate govemance and accounting and economic outcomes. Nevertheless, I prefer to use the ratings issued by an independent rating agency as measures of govemance compliance since these are publicly available and easily accessible for market participants. The aim of the study is to see whether these publicly available measures of the extent of compliance with intemational best practices are related to fiiture operating performance and can as such signal future operating performance to market participants, who can, in tum, use this infonnation for decision making.

there is less need for govemanee practices tailored to local contexts for the largest companies in Europe, which operate globally instead of locally. Whether large listed European companies benefit fi-om compliance with intemational best practices, and then specifically in terms of higher operating performance, is ultimately an empirical question. Another reason why higher compliance is not necessarily better govemance is that the best practices identified by Deminor are not always unequivocally related to better govemance. For example, evidence on whether CEO duality is bad govemance and board diversity is good govemance, is mixed (see, for example, Sonnenfeld, 2004; Massa and Simonov, 2007).* In order to refine the analysis I focus on the dimension of corporate govemance which is particularly likely to directly infiuence operating efficiency and operating per- formance, i.e. the structure and functioning of the board of directors.^ As Jensen (1993: 862-863) puts it, 'The board, as the apex of the intemal control system, has the final responsibility for the function- ing of the firm. Most importantly, it sets the rules of the game for the CEO. The job of the board is to hire, fire, and compensate the CEO, and to provide high-level counsel' and ' . . . the very purpose of the intemal control mechanism is to provide an early waming system to put the organisation back on track before difficulties reach a crisis stage.' Jensen (1993) then also attributes the weak corporate performance from the early 1990s to problems with the intemal control activity (Jensen, 1993: 352) in the 1980s, which, in tum, stemmed from problems with the board of directors (Jensen, 1993: 862).

The major threat to a well-functioning board, and strong operating performance, is that the board is dominated by managers (especially in Anglo-Saxon countries) or majority shareholders (especially in continental European countries) who act in their ovra interest (instead of in the interest of all stakeholders), and cover up any underperformance by eamings management or manipulation to appease (minority) shareholders. Jensen (1993: 869) then also recognises that characteristics such as, for example, high-equity ownership by man- agers and board members, a small board, not many insiders on the board, and a CEO which is not the chairman of the board, are key elements of a well- functioning govemance system, which limits self- interested behaviour by managers, uncovers bad performance in time and takes the necessary actions

* I thank one of the anonymous reviewers for this observation. ' For completeness, I later expand the analyses to govemance

dimensions other than board structure and fiinctioning. The results are reported in Section 5.4.

500 ACCOUNTING AND BUSINESS RESEARCH

to 'put the organisation back on track' (Jensen, 1993: 863). These key elements of a well-function- ing board mentioned by Jensen (1993) are all covered by the intemational best practices concem- ing board structure and functioning. Therefore, I expect that higher compliance with intemational best practices conceming board structure and functioning is related to better operating perform- ance.

Although greater compliance with intemational best practices conceming rights and duties of shareholders and range of takeover defences may increase the pressure by investors and the market for corporate control on companies to perform well, it is less straightforward that this greater compliance with intemational best practices conceming rights and duties of shareholders and range of takeover defences is per se related to better underlying operating performance, for in the absence of a well- functioning board, managers and majority share- holders could still act in their own self-interest, underperform, and cover up weak performance by eamings management or manipulation.'"''' This leads to the following hypothesis: HI: A company's operating performance increases

in the extent of compliance with intemational best practices conceming board stmcture and functioning, ceteris paribus.

3. Sample and data This study uses ratings of compliance with inter- national best practices regarding board stmcture and functioning which are supplied by a private rating agency, Deminor Rating. Deminor Rating (a sub- sidiary of Deminor Intemational) releases, since March 2001, corporate govemance ratings on the companies of the FTSE Eurotop 300 index.'^''^ The ratings are based on over 300 corporate govemance indicators, which were identified after consulting

' " D e Angelo (1988), for example, reports that, during an election campaign, managers exercise accounting discretion to portray a favourable eamings picture to voters.

As concems disclosure on corporate govemance, it is straightforward that mere disclosure per se cannot improve the operating performance of a company. However, the level of disclosure is highly positively correlated with the quality of the structure and the fiinctioning of the board: companies with well- structured and -functioning boards have no problem in disclos- ing this information, while companies with badly-structured and -functioning boards are less transparent about this. A positive association between high disclosure and good operating performance is then probably also due to a well-structured and well-functioning board than to the level of disclosure per se.

'^On 25 May 2005, Deminor announced that it had sold its corporate govemance unit Deminor Rating to Institutional Shareholder Services (ISS).

'^ Some other studies that have used Deminor data are Bauer et al. (2008), Bauer et al. (2006), Wójcik (2006), Wójcik et al. (2005), and Bauer et al. (2004).

institutional investors. The indicators can be div- ided into four categories: rights and duties of shareholders, range of takeover defences, disclosure on corporate govemance and board stmcture and functioning. Deminor Rating issues a rating of each one of the four categories. This study focuses on the rating regarding board stmcture and functioning. This rating covers indicators on the election of members of the company's bodies, composition of the board, functioning of the board, remuneration of the company's bodies and committees of the board.

Ratings are assigned by senior analysts from the different European offices of Deminor after all the most recent publicly available information on a particular company (i.e. not only financial reports, but also articles of association, agendas, resolutions and minutes of ordinary and extra-ordinary general meetings, investor's handbooks and newsletters, intemet-sites and all other publicly available infor- mation) has been benchmarked against the best practice found in intemationally accepted stand- ards. Those intemationally accepted standards are established by, for example, the Intemational Corporate Govemance Network (ICGN) and the Organisation for Economic Co-operation and Development (OECD). A rating is measured on a scale of 5 to 1, with 5 representing the best practice (Deminor Rating, 2001: 9-10).

The sample studied in this paper consists of all companies from the FTSE Eurotop 300 for which there is a Deminor rating of board structure and functioning for the year 2000 and/ or 2001, as well as complete infonnation on the other variables in the model.'"* I exclude financial companies (FTSE industry sector code 80) because their financial stmcture is distinct from other companies and they are often subject to special mies and recommenda- tions. I delete observations with extreme observa- tions (i.e. values outside the 5* and 95* percentile) for the ratios in the model, namely leverage and the three measures of operating performance (i.e. ROA, ROE and NPM), for ratios easily take on extreme values. The final sample exists of 201 firm-year observations (from 118 different companies). Table 1, Panels A and B give a breakdown of the observations by industry sector and by country, respectively.

I obtain financial statement data from Worldscope.

' ' ' T h e item that is most frequently missing is the Deminor govemance rating. This rating is missing because not all FTSE Eurotop 300 firms are followed by Deminor.

Vol. 39, No. 5. 2009 501

Table 1 Sample description Panel A: Breakdown of sample by industry*

Industry code

4 7 11 13 15 21 24 25 26 31 34 41 43 44 47 48 49 52 53 54 58 59 63 67 72 73 78 93 97

Industry description

Mining Oil & Gas Chemicals Construction & Building Materials Forestry & Paper Aerospace Diversified Industrials Electronic & Equipment Engineering and Machinery Automobiles Household Goods & Textiles Beverages Food Producers & Processors Health

Number of firms % Number of firm-years

Personal Care & Household Products Pharmaceuticals Tobacco General Retailers Leisure, Entertainment & Hotels Media & Photography Support Services Transport Food & Dñig Retailers Telecommunication Services Electricity Gas Distribution Water Information Technology Hardware Software & Computer Services Total

* Following the FTSE Global Classification System.

Panel B: Breakdown of sample by country"*

Country

Belgium France Italy

2 6 9 6 1 3 3 7 7 8 4 2 4 1 2 4 2 7 2 9 3 1 4 6 8 2 1 2 2

118

Number of firms % Number of firm-years

3 2.54 25 21.19 6 5.08

The Netherlands 8 6.78 Portugal Spain Switzerland Germany Denmark Norway Sweden Finland Ireland UK Total

1 0.85 7 5.93 7 5.93

12 10.17 2 1.69 1 0.85 8 6.78 1 0.85 1 0.85

36 30.51 tl8 100

** All but two countries in the sample (Switzerland other countries and Monetary

, but the UK, Denmark and Sweden Union).

6 49 10 13

1 13 13 21

3 2

12 2 2

54 201

and Norway) are part of the

1.69 5.08 7.63 5.08 0.85 2.54 2.54 5.93 5.93 6.78 3.39 1.69 3.39 0.85 1.69 3.39 1.69 5.93 1.69 7.63 2.54 0.85 3.39 5.08 6.78 1.69 0.85 1.69 1.69 100

2.99 24.38 4.98 6.47 0.50 6.47 6.47

10.45 1.49 1.00 5.97 1.00 1.00

26.87 100

are member of the Eurozone or EMU

3 10 15 11 2 5 4

12 12 16 8 4 7 1 3 8 4

11 3

15 5 1 8 8

14 4 1 3 3

201

1.49 4.98 7.46 5.47 1.00 2.49 1.99 5.97 5.97 7.96 3.98 1.99 3.48 0.50 1.49 3.98 1.99 5.48 1.49 7.46 2.49 0.50 3.98 3.98 6.97 1.99 0.50 1.49 1.49 100

European Union. All (i.e. Europe's European

502 ACCOUNTING AND BUSINESS RESEARCH

4. Research design and model specification I test the relationship between the extent of compliance with intemational best practices con- ceming corporate govemance, and more specific- ally board structure and functioning, and the operating performance of large listed European companies by estimating the following operating performance model:

Performancei, = ßo + iS|CG.COMPi,

where:

i, +

(1)

Performanceit = ROA, where: ROA is one-year ahead retum on assets for firm i in year t;

CGCOMPLit = a rating proxying for the extent of compliance with intemational best practices regarding board structure and functioning for firm i in year t;

= leverage, as measured by the sum of short-term and long-term debt divided by total assets, for firm i in year t;

= the natural logarithm of total assets for firm i in year t;'^''^

Y2001it = indicator variable which takes one if the observation is from 2001, and zero if the observation is from 2000;

Xjt = a vector of industry dummies, i.e. indicator variables for the (two- digit) industry codes of the FTSE Global Classification system.

I measure the dependent variable in the operating performance model, i.e. one-year ahead firm-level operating performance, by the one-year ahead ROA. I use the one-year ahead, instead of the contemporaneous, ROA to make sure that the govemance systems described by the ratings are in place and operational at the moment that I start measuring operating perfonnance. Consistent with prior studies (e.g. Larcker et al., 2006), ROA is measured as operating income divided by average total assets.'^ For comparison, I also perform

'^ Total assets are measured in thousands of Euros. Values initially stated in a local currency are converted to Euros by using the exchange rate at the balance sheet date.

I use the natural logarithm because I do not expect a linear relationship between operating performance and firm size.

" T h e average is computed as the sum of the value at the beginning of the accounting period and the value at the end of the accounting period, divided by two.

analyses in which I replace the one-year ahead ROA with the one-year ahead ROE and the one-year ahead NPM, because the ROE and the NPM were used as performance measures in the study by Bauer et al. (2004). The ROE is measured as eamings before extraordinary items dividend by the average book value of stockholders' equity, and net profit margin is the ratio of eamings before extraordinary items divided by sales (see, for example, Gompers et al., 2003). As argued by Core et al. (2006) and Barber and Lyon (1996), the ROA is clearly the preferred measure of operating performance because it is less affected by discretionary items than the ROE and the NPM. This implies that I expect a stronger relationship between the extent of compliance with intemational best practices con- ceming board stmcture and functioning and the one-year ahead ROA, than between the extent of compliance and the one-year ahead ROE or the one- year ahead NPM. I will further refer to the three models as the ROA model, the ROE model and the NPM model.

The test variable is a measure of the level of compliance with intemational best practices con- ceming corporate govemance, and more specific- ally best practices conceming board stmcture and fiinctioning (CG COMPL). CG COMPL is prox- ied by Deminor's rating of board stmcture and functioning. The rating takes a value from 1 to 5 with 5 indicating the highest compliance with intemational best practice. A positive sign on CG_COMPL indicates that greater compliance with best practices conceming board structure and fimctioning is related to better operating perform- ance, and is consistent with the hypothesis. I further include in the regression leverage (LEV), computed as the sum of short-term and long-term debt over total assets, to control for the well-known impact of leverage on ROE, the natural logarithm of total assets (LNTA) as a measure of firm size, a year dummy (Y2001) to control for the impact of the general macro-economic context on individual firm performance, and a vector of industry dummies.

5. Descriptive statistics and results 5.1. Descriptive statistics Table 2, Panel A, presents descriptive statistics for the dependent and independent variables of the operating performance model. Table 2, Panel A, shows that the mean one-year ahead ROA is about 6.6% (median 6.4%). The mean one-year ahead ROE is higher, and about 7.3% (median 10.2%). The mean one-year ahead NPM amounts to 2.9% (median 3.7%). Mean and median leverage is about 28%.

Vol. 39, No. 5. 2009 503

Table 2 Descriptive statistics and correlations^ Panel A: Descriptive statistics

Variable

ROA ROE NPM LEV LNTA

Panel B:

201 201 201 201 201

Mean

0.0658 0.0729 0.0289 0.2773

16.5757

StdDev

0.0482 0.1507 0.0769 0.1135 1.0492

Min.

-0.0352 -0.5838 -0.4580

0.0470 13.7591

Pearson correlation coefficients

ROA ROE NPM CG COMPL LEV LNTA

ROA

1.0000 0.5095 0.5224 0.1823

-0.1993 -0.4028

ROE

0.8328 0.0271

-0.1724 -0.1337

NPM

0.0190 -0.1640 -0.2297

0.0314 0.0433 0.0110 0.1868

15.8908

Median Q3

0.0635 0.0946 0.1019 0.1588 0.0365 0.0684 0.2794 0.3642

16.3970 17.2708

CG COMPL LEV LNTA

Max.

0.1966 0.3015 0.1626 0.4955

19.1507

-0.0092 1.0000 -0.0273 0.2066 1.0000

* Variable definitions:

ROA = one-year ahead return on assets for firm i in year t, and is measured as operating income divided by average total assets,

ROE = one-year ahead retum on equity for firm i in year t, and is measured as eamings before extraordinary items divided by the average book value of stockholders' equity,

NPM = one-year ahead net profit margin for firm i in year t, and is measured as eamings before extraordinary items divided by sales revenues

LEV = ratio of short-term debt plus long-term debt over total assets for firm i in year t LNTA = natural logarithm of total assets for firm i in year t CG_COMPL = score fi^om 1 to 5 with higher scores indicating greater compliance of firm i in year t

with intemational best practice conceming board stmcture and fiinctioning

Table 2, Panel B, presents the Pearson correlation coefficients between the dependent and independ- ent variables of the operating performance model. The dependent variables, one-year ahead ROA, one-year ahead ROE and one-year ahead NPM are all positively correlated with CG_CO]V[PL. However, only the correlation of ROA and C G C O M P L is significant. The highest absolute value of the correlations among the independent variables is 0.21, which indicates that the regression results are not affected by multicollinearity.

5.2. Regression results I estimate the performance model using ordinary least squares (OLS). A concem in testing the relation between the extent of compliance with intemational best practices conceming corporate govemance, and more specifically board stmcture and fiinctioning, and the operating performance of large listed European companies is that firms with

good prospects may self-select into the group with stronger govemance stmctures, while firms with poor prospects may self-select into the group with weaker govemance stmctures. If this is indeed tme, the OLS parameter estimates are inconsistent. However, the results of a Hausman-like test for endogeneity as described in Gujarati (2003: 713) show that CG_COMPL is not endogenous with respect to any of the dependent variables (p on the fitted value of CG_COMPL > 0.10, two-sided). For the endogeneity test, I used the following first stage model (govemance compliance model), which is based on prior empirical disclosure and govemance studies (see, for example, Pincus et al., 1989 and Willekens et a l , 2004):

CG_COMPLi, = 70 + y,FFLOATi, +

(2)

504 ACCOUNTING AND BUSINESS RESEARCH

where: FFLOATit= the free float of firm i in year t; ROA_Cit = the contemporaneous retum on assets for firm i in year t; Yit = a vector of country dummies, i.e. indicator variables for the country of domicile of the firms in the sample; Z^ = a vector of other exogenous variables from the operating performance model (i.e. second stage regression). The other variables are as defined in Equation (1). Re-performing the endogeneity tests (1) deleting the contemporaneous retum on assets in the first stage regression, or (2) replacing the contemporan- eous retum on assets with the past retum on assets in the first stage regression confirm that CG_COMPL is not endogenous with respect to any of the dependent variables. This contrasts with results in Renders and Gaeremynck (2006), and is most likely due to differences in research design. More specif- ically, to be sure that the corporate govemance systems are in place and operational at the moment that I start measuring operating performance, I use one-year ahead operating perfomiance measures (instead of contemporaneous operating perform- ance measures).

As there are two years of data, there are repeated observations on some companies. Although obser- vations are still independent across firms, they are no longer independent within firms. Therefore, the t-values are adjusted to control for within-company dependence by using clustered robust standard errors.'*

Table 3 reports the results of the OLS regression analyses. Columns 3, 5, and 7 show the results of models that use the one-year ahead ROA, the one- year ahead ROE and the one-year ahead NPM as the dependent variable, respectively. Table 3 shows that the performance models have explanatory power (adjusted R^ of 49.30%, 23.41% and 22.11%, respectively). The coefficient on CG_COMPL is positive and significant at the 1% level (one-sided) in the ROA model. The magnitude of the CG_COMPL coefficient (0.0054) suggests a dif- ference of 2.16% in the realised one-year ahead ROA between firms with the lowest and the highest rating of CGCOMPL (i.e. 4*0.54%). This result supports the hypothesis that operating performance is higher for firms that comply to a greater extent with intemational best practices conceming board stmcture and functioning. Consistent with the argument that the one-year ahead ROA is the preferred measure of operating performance, the coefficient on CG_COMPL is not significant for the ROE and NPM models.

'̂ The results are qualitatively similar when not adjusting for within company dependence.

Table 3 fiirther shows that the one-year ahead ROA and one-year ahead NPM decrease in size (LNTA). In addition, the industry dummies (not reported) are significant predictors of all three performance measures.'^

5.3. Additional analyses Other measures of operating performance To confirm the evidence from the ROA model on the positive and significant relation between the extent of compliance with intemational best prac- tices conceming board stmcture and fiinctioning and operating performance, I replace the one-year ahead ROA in the operating performance model with two altemative measures of operating per- formance, i.e. the one-year ahead retum on cash- adjusted assets and the one-year ahead retum on sales (ROS) (see Barber and Lyon, 1996). The retum on cash-adjusted assets is measured by dividing operating income by the average cash- adjusted assets, i.e. total assets minus cash and cash equivalents. The ROS is operating income divided by sales. Table 4 reports the results of the operating performance regressions when using these altema- tive operating performance measures. Table 4 also reports the results of the regression when using the one-year ahead ROA as the dependent variable for comparison. Table 4 shows that the results on the test variable (CG COMPL) when using the alter- native operating performance measures are qualita- tively similar to the result when using the one-year ahead ROA.

Examining the difference in results between performance measures The results in Tables 3 and 4 report a significantly positive relationship between the extent of compli- ance with intemational best practices conceming board stmcture and fiinctioning (CG_COMPL) and performance for some performance measures (i.e. the one-year ahead ROA, the one-year ahead retum on cash-adjusted assets and the one-year ahead ROS, all measures which use operating income in the numerator), but not for other performance measures (i.e. the one-year ahead ROE) and one-year ahead NPM, two measures which use income before exfraordinary items in the numerator). Especially the difference in results when using the ROS and the NPM is striking, since the only difference between these two measures is the numerator. (The ROS uses operating income

" Deleting the industry dummies from industries with only one observation (i.e. health, transport, and water) or all industry dummies does not change the results on the test variable.

Vol. 39, No. 5. 2009 505

Table 3 Regression

Variable

results

Pred. sign

ROA

Coef. estimate (t-statistic)

Pred. sign

ROE

Coef. estimate (t-statistic)

Pred. sign

NPM

Coef. estimate (t-statistic)

Intercept

CG COMPL

LEV

LNTA

Y2001

Industry dummies

Adj. R-squared N Evidence of endogeneity

0.4252*** (7.14) 0.0054***

(2.38) -0.0291

(-0.90) -0.0197***

(-5.33) -0.0060*

(-1.82) Included

49.30% 201 No

0.4511* (1.70) -0.0034

(-0.39) -0.1930

(-1.32) -0.0144

(-0.88) -0.0258

(-1.59) Included

23.41% 201 No

regression model:

[TAj,-1-/Î4 72001 ¡,

+ ß5

0.4142*** (2.87) -0.0023

(-0.61) -0.0702

(-1.2) -0.0179**

(-2.04) -0.0106

(-1.10) Included

22.11% 201 No

¡̂ it + Si,

" This table reports the results of the OLS estimation of the following regression model:

Perfotroanceu = ßo+ ^ i C G _ C O M P L ¡ , -I- ftLEVj, +

Where: PerfonTiance;, = ROA, ROE or NPM. ROA = one-year ahead retum on assets for firm i in year t, and is measured as operating income divided by average total assets; ROE = one-year ahead retum on equity for firm i in year t, and is measured as eamings before extraordinary items divided by the average book value of stockholders' equity; NPM = one-year ahead net profit margin for firm i in year t, and is measured as eamings before extraordinary items divided by sales revenues; CG_COMPL¡, = score fi-om 1 to 5 with higher scores indicating greater compliance of firm i in year t with intemational best practice conceming board stmcture and fiinctioning; LEV¡, = ratio of short-term debt plus long-term debt over total assets for firm i in year t; LNTAj, = natural logarithm of total assets for firm i in year t; Y2001 ¡t = year dummy, =1 when an observation is fi'om 2001, zero otherwise; X¡t = a vector of industry dummies based on the FTSE Global Classification System two-digit code for the industry sector.

*,** and *** denote statistical significance at the 10%, 5% and 1% level respectively and is based on a one- tailed test if the sign of the coefficient is in the predicted direction, and is based on a two-tailed test otherwise. T-values are adjusted for within-company dependence by using clustered robust standard errors. Results on the two-digit industry dummies are not reported for parsimony.

whereas the NPM uses eamings before extraordin- ary items in the numerator. Both measures have sales revenues in the denominator.) This suggests that something that causes the difference between operating income and eamings before extraordinary items can explain why there is a significant relationship between performance and the extent of compliance with intemational best practices conceming board structure and functioning when using some performance measures, but not when using other performance measures. Some of the difference between operating income and eamings before extraordinary items is in interest payment and taxes. Moreover, given the Worldscope data definitions of operating income and eamings before

extraordinary items, some of the difference between the two income measures stems fi'om allocations to and/or fi-om reserves, irom minority interests, from equity in eamings,̂ *^ and from other non-operating income and expenses. These other non-operating income and expenses include items such as: non- operating interest income, non-operating dividend income, and the gain/ loss on disposal of assets, i.e. the income from asset disposals.

To further explore what causes the difference in the relationship between corporate govemance

^^ This represents the 'pretax portion of the eamings or losses of a subsidiary whose financial accounts are not consolidated with the controlling company's accounts' (see Thomson Financial, 2003).

506 ACCOUNTING AND BUSINESS RESEARCH

Table 4 Regression results using alternative operating performance measures"

ROA Retum on

cash-adj. assets Return on sales

Variable Pred. sign Coef. estimate Pred. sign Coef. estimate Pred. sign Coef. estimate (t-statistic) (t-statistic) (t-statistic)

Intercept

CG_COMPL

LEV

LNTA

Y2001

Industry dummies

Adj. R-squared N

0.4252*** (7.14) 0.0054***

(2.38) -0.0291

(-0.90)

-0.0197*** (-5.33) -0.0060*

(-1.82) Included

49.30% 201

0.4722*** (7.11) 0.0050**

(2.02) -0.0399

(-1.15)

0.0220*** (-5.42) -0.0056

(-1.57) Included

50.05% 201

0.4829*** (4.55) 0.0062**

(2.09) 0.0901*

(1.93)

-0.0212*** (-3.57) -0.0048

(-1.09) Included

47.90% 201

This table reports the results of the OLS estimation of the following regression model:

Performancei, = î o + ;9iCG_C0MPLi, + + + Where: Performance;, = ROA, Retum on cash-adjusted assets or Retum on sales. ROA = one-year ahead retum on assets for firm i in year t, and is measured as operating income divided by average total assets; Retum on cash-adjusted assets = one-year ahead retum on cash-adjusted assets for firm i in year t, and is measured as operating income divided by average (total assets minus cash and cash equivalents); Retum on sales = one-year ahead retum on sales for firm i in year t, and is measured as operating income divided by sales revenues; CGCOMPLit = score from 1 to 5 with higher scores indicating greater compliance of firm i in year t with intemational best practice conceming board structure and functioning; LEVi, = ratio of short- term debt plus long-term debt over total assets for firm i in year t; LNTAj, = natural logarithm of total assets for firm i in year t; Y2001i, = year dummy, =1 when an observation is fi-om 2001, zero otherwise; X;, = a vector of industry dummies based on the FTSE Global Classification System two-digit code for the industry sector.

compliance (CG_COMPL) and performance when using the NPM instead of the ROS as the measure of performance, I compute the pairwise correlations between each one of the identified items which make up the difference between operating income and eamings before extraordinary income (scaled by sales) and the measure of corporate govemance compliance (CGCOMPL).^' I fiirther adjust oper- ating income for each one of the items at a time and regress each one of the new income measures

^' As the performance measures used in the primary analyses, these items and the newly computed perfonnance measures are on a one-year ahead basis.

(scaled by sales) on the measure of corporate govemance compliance (CGCOMPL) and the control variables, i.e. I replace the dependent variable in the performance model with a new performance meastire.

The results of these detailed analyses show that income fi-om asset disposals^^ (scaled by sales) is significantly negatively related to the extent of corporate govemance compliance (CGCOMPL) (correlation coefficient = -0.1869, p-value < 0.01),

•̂•̂ The sample companies report a gain on disposal of assets in 64% of the firm-years, a loss on disposal of assets in 22% of the firm-years, and neither a gain nor a loss on disposal of assets in 14% of the firm-years.

Vol. 39, No. 5. 2009 507

and that, once operating income is adjusted for the income from asset disposals and this measure scaled by sales is used as the measure of performance (i.e. the 'new ROS'), the significant relationship between performance and corporate govemance compliance (CG_COMPL), established when using the original ROS as the measure of performance, disappears. The coefficient is 0.0025, p-value > 0.10, one-sided, when using the new ROS, com- pared to a coefficient = 0.0062, p-value < 0.05, one- sided, when using the original ROS (see Table 4).

Moreover, although some other items which make up the difference between operating income and eamings before extraordinary items are also negatively related to corporate govemance com- pliance, none of the adjustments to operating income for these items makes the relationship between performance and the corporate govem- ance compliance insignificant like the adjustment for the income from asset disposals does. Taken together, the results of the additional analyses suggest that the income from asset disposals plays a major role in the disappearance of the significant relationship between performance and corporate govemance compliance, once the NPM is used instead of the ROS as the measure of perform- ance.

Management has some discretion over the timing of asset sales and previous studies (Herrmann et al., 2003; Bartov, 1993) have shown that the timing of assets sales is used as an instrument to manage eamings. More specifically Bartov (1993) finds that companies use the timing of asset sales to smooth income, i.e. report a higher (lower) income from asset sales in years in which (pre-managed) income is lower (higher) compared to the previous year. Further analyses^^ indicate that this study's sample companies present similar behaviour. Moreover, the extent of corporate govemance compliance (CG_COMPL) is significantly negatively correlated with the occurrence of a lower (pre-managed) income compared to the previous year (correla- tion coefficient = - 0 . 2 0 7 6 , p-value <0.01). That is, poorly governed companies underperform more often than well-governed companies. Consequently, one expects that poorly govemed companies record a higher income from asset disposals than well-govemed companies. The sig- nificantly negative correlation between the income from asset disposals (scaled by sales) and the

^̂ Consistent with the results of Bartov et al. (1993), the OLS estimation results of a regression of the income from asset disposals (scaled by sales) on: (1) the difference between the ROS and the prior period NPM; and (2) firm leverage show a significantly negative coefficient on the perfonnance difference.

extent of corporate govemance compliance ( C G C O M P L ) is consistent with this expectation. In sum, the results indicate that companies which comply less with intemational best practices con- ceming board structure and fiinctioning try to make up their inherently lower performance by reporting a higher income from asset disposals (scaled by sales). The net result of this eamings management behaviour is that, once the income from asset disposals is included in the numerator of the performance measure, one no longer observes a significant relationship between operating perform- ance and the extent to which firms comply with intemational best practices conceming corporate govemance.

To further examine whether the income from asset disposals can help explain that there is a significant relationship between performance and the extent of compliance with intemational best practices conceming board stmcture and fiinction- ing when using some performance measures but not others, I replace in the performance model the original ROA, which uses operating income in the numerator, by a new ROA measure which uses operating income adjusted for the income from asset disposals in the numerator. The result of replacing the original ROA with this new ROA measure is that the coefficient on CG_COMPL becomes insignificant. The coefficient is 0.0027, p- value > 0.15, one-sided, when using the new ROA, compared to a coefficient = 0.0054, p-value <0.01, one-sided, when using the original ROA (see Table 3). This confirms the idea that the income from asset disposals plays a major role in explaining the differences in the significance ofthe relationship between operating performance and the extent of compliance with intemational best practices con- ceming board stmcture and fianctioning when using other performance measures.

Bauer et al. (2004) suggest that lacking man- agerial discretion to manipulate eamings is respon- sible for their observation that well-govemed companies report lower performance (ROE and NPM) than poorly govemed companies, but do not provide evidence to support this suggestion. The results ofthe above analyses confirm that manager- ial discretion is an important factor to consider when examining the relationship between corporate govemance compliance and operating performance. More specifically, the results of the analyses are evidence: (1) that, in contrast to the finding of Bauer et al. (2004), well-govemed companies perform better than poorly govemed companies, once examining performance measures which are less prone to managerial discretion (ROA or ROS); and

508 ACCOUNTING AND BUSINESS RESEARCH

(2) that poorly governed companies exploit their discretion to manage bottom line eamings upwards to a level which is (relatively) similar to that of well- governed companies.

Code law versus common-law countries Various elements of the institutional infrastructure, such as investor protection, differ between com- mon-law and code-law countries. These differ- ences were, in tum, found to be related to differences in, for example, the size of the capital markets (both equity and debt markets) (La Porta et al., 1997), ownership concentration (La Porta et al., 1998, and La Porta et al., 1999), information provision (see, for example. La Porta et al., 1998, Bail et al., 2000, Francis et al., 2003, Leuz et al., 2003), and the demand for auditing (Francis et al., 2003). I test whether the observed relation between the extent of compliance with intemational best practices regarding board structure and functioning and operating performance is present in both types of countries by estimating the operating perform- ance model on the subsamples of the code-law and common-law countries separately. The common- law countries in the sample are the UK and Ireland. The code-law countries are all other countries in the sample. The results for the code- law subsample (145 firm-year observations, 81 different companies) and common-law subsample (56 firm-year observations, 37 different com- panies) (not reported) are qualitatively similar to the results reported in Table 3, i.e. I find a positive, albeit somewhat less significant, relation between the extent of compliance with intemational best practices conceming board stmcture and function- ing and the one-year ahead ROA.

5.4. Robustness checks The robustness of the results is tested to model specification. I first replace the 28 two-digit industry dummies, which consume a lot of degrees of freedom given the size of the sample, with dummies for broader industry classes with at least 10 firms per class. I formed the broader classes by using the one-digit classification and by redistributing the observations from one-digit classes with less than 10 firms, i.e. mining, oil & gas (one-digit 0) and information technology (one-digit 9) to the other one-digit classes. The results of estimating these new operating performance models (not reported) are quahtatively similar to the results reported in Table 3.

As an additional check, I replace the dependent variables, i.e. the performance variables, with the industry median adjusted values, and deleted the

two-digit industry dummies from the operating performance model. The medians were computed per broader industry class. Results (not reported) are qualitatively similar to the results reported in Table 3.

Next, I test whether the results are robust to controlling for country-level performance differ- ences by including country dummies in the original operating performance model. I also estimate an operating performance model with inclusion of country dummies and replacement of the two-digit industry dummies by dummies for broader industry classes, and an operating performance model with inclusion of country dummies, exclusion of two- digit industry dummies and replacement of the dependent variables with their industry median adjusted values. The estimation results of all these altemative operating performance models show that the results on the test variable CGCOMPL (not reported) are qualitatively similar to the results reported in Table 3 for all three performance measures.

In the primary analyses, I use one-year ahead instead of contemporaneous performance meas- ures in order to make sure that the govemance systems described by the ratings are in place and operational at the moment that I start measuring performance. To test whether the results depend on this specification, I reran the operating performance model using contemporaneous per- formance measures. The results on the test variable are similar to the results reported in Table 3.

To test whether the results are dependent on the measure of firm size, I replace the natural logarithm of total assets (LNTA) by the natural logarithm of market capitalisation. I explore the impact of firm history on the corporate govemance-operating performance relation by including in the regression model firm age or book-to-market. The results on the test variable are qualitatively similar to the results reported in Table 3.

I further test whether the results for the operating performance model are not biased by correlated omitted corporate govemance variables. If a vari- able which is related to both performance and the extent of compliance with intemational best prac- tices conceming board stmcture and functioning is omitted from the analyses, then this variable may actually cause the observed relation between per- formance and the extent of compliance with inter- national best practices conceming board stmcture and functioning, and we may wrongly conclude that the extent of compliance with intemational best practice regarding board stmcture and fiinctioning

Vol. 39, No. 5. 2009 509

per se is related to firm performance. A correlation matrix (not reported) shows that the extent of compliance with best practices in various other govemance areas, such as rights and duties of shareholders,'^'' range of takeover defences,^^ and disclosure on corporate govemance^^ are signifi- cantly and positively related to the firms' compli- ance with best practices conceming board stmcture and fijnctioning. Moreover, the results of estimating (using OLS) the basic regressions in which the variable board stmcture and fiinctioning is replaced sequentially by, first, the Deminor rating of rights and duties of shareholders, next, the rating of takeover defences, and finally, the rating of disclo- sure on corporate govemance (see Table 5, Panel A) show that the rating of the range of takeover defences and the rating of disclosure on corporate govemance are significantly correlated with the ROA. Consequently, the question arises whether the primary results suffer fiom correlated omitted variables bias, for the other dimensions of govem- ance are clearly linked to the test variable and the dependent variable ROA, but not included in the main analyses. To address this concem I expand the operating performance model with the rating of rights and duties of shareholders and the rating of takeover defences, in addition to the rating of board stmcture and fiinctioning (i.e. the expanded oper- ating performance model). I do not include the rating of disclosure of corporate govemance because the correlation coefficient between the disclosure rating and the rating of board stmcture and fiinctioning is high (0.7927). This high correl- ation does not only suggest that both ratings contain very similar information, but also implies that the results of a regression including both ratings suffer fiom multicollinearity, which makes interpretation of the coefficient estimates difficult. Table 5, Panel B, reports the results of the expanded operating performance model. These results closely parallel the results fiom the primary operating performance

•̂ '' The rating of rights and duties of shareholders covers indicators conceming the respect of the one-share one-vote one- dividend principle, voting right restrictions, voting issues, shareholder proposals and voting procedures (Deminor, 2001: 7).,^

The rating of range of takeover defences covers indicators conceming the presence and strength of anti-takeover devices such as poison pills, golden parachutes, core shareholdings, extensive cross-shareholdings and co-option systems that could be used to protect the company from a hostile takeover and to disenfranchise shareholders (Deminor, 2001: 7).

^* The rating of disclosure on corporate govemance covers indicators regarding the quantity and quality of non-financial infomiation, such as the diversity and independence of board members, board committees, director remuneration, accounting standards, information on major shareholders of the eompany, and environmental information (Deminor, 2001: 7).

model as reported in Table 3, which suggests that the primary results are not biased by correlated omitted govemance variables. More specifically, the rating of board stmcture and fiinctioning remains positive and significant in the ROA model, while it remains insignificant in the ROE and NPM models. The results on the other two ratings, i.e. rights and duties of shareholder and range of takeover defences, parallel the results of the regressions in which each rating was included separately as reported in Table 5, Panel A.

By excluding the rating of disclosure on corpor- ate govemance in the expanded operating^ perform- ance model there is still a possibility that the results in Table 3 and Table 5, Panel B are biased because of correlated omitted variables. In order to fiirther address this concem, I also performed a principal component factor analysis^^ on the four different ratings disclosed by Deminor, i.e. the rating of board stmcture and fiinctioning, the rating of rights and duties of shareholders, the rating of takeover defences and the rating of disclosure on corporate govemance. The solution is rotated using an orthogonal VARIMAX rotation. This analysis shows that the four ratings can be summarised in two underlying factors. (All factors with an eigenvalue greater than unity are retained.) As expected, the rating of board stmcture and fiinc- tioning and the rating of disclosure on corporate govemance load on factor 1. This confirms that both ratings contain very similar infonnation. The rating of rights and duties of shareholders and the rating of takeover defences load on factor 2. The cones- ponding factors scores were computed (using the regression method) and the two factors scores were then introduced together in the basic regression (i.e. Regression (2)) for CG_COMPL. Table 5, Panel C, reports the results of these tests. As concems the one-year ahead ROA, the results show that factor 1, but not factor 2, is positively and significantly correlated with this performance meas- ure (p<0.01). As concems the one-year ahead ROE and the one-year ahead NPM, the results show that neither factor 1 nor factor 2 are significantly correlated with these performance measures. The results on the factor scores confirm the results reported in Table 3 and Table 5, Panels A and B.

hi order to fiirther investigate the overall impact of firms' govemance stmctures, I estimate an operating performance regression in which I replace the rating of board stmcture and fiinctioning by a total rating (i.e. a total across the different govem-

^'Results are qualitatively similar when using a principal factor method.

510 ACCOUNTING AND BUSINESS RESEARCH

Tahle 5 Additional regression results"

Panel A: OLS coefficient estimates one at a time

Board structure and functioning

Rights and duties of shareholders

Range of takeover defences

Disclosure on corporate govemance

Panel B: OLS coefficient estimates

Board structure and functioning

Rights and duties of shareholders

Range of takeover defences

Other regressors

Adj. R-squared N

Panel C: OLS coefficient estimates

Factor 1

Factor 2

Other regressors Adj. R-squared N

Panel D: OLS coefficient estimates

Total rating

Adj. R-squared N

(and t-statistics) on the

ROA

0.0054*** (2.38) -0.0016

(-0.55) 0.0054**

(2.16) 0.0056**

(2.27)

different governance

ROE

-0.0034 (-0.39)

0.0224* (1.67) 0.0012

(0.16) -0.0097

(-1.18)

(and t-statistics) on ratings included together

ROA

0.0046** (2.27) -0.0066**

(-2.04) 0.0060**

(2.20) Included

51.56% 201

ROE

-0.0073 (-0.80)

0.0280** (2.00) -0.0052

(-0.64) Included

25.54% 201

(and t-statistic) for factor scores

ROA

0.0094*** (2.69) 0.0008

(0.25) Included

49.82% 201

ROE

-0.0178 (-1.42)

0.0238 (1.63) Included 25.46%

201

(and t-statistic) for overall governance rating

ROA

0.0018** (2.03)

49.12% 201

ROE

-0.0002 (-0.06)

23.32% 201

ratings, included each

NPM

-0.0023 (-0.61)

0.0086 (1.42) 0.0007

(0.20) -0.0043

(-1.09)

in one model

NPM

-0.0041 (-0.98)

0.0107 (1.56) -0.0013

(-0.30) Included

22.96% 201

NPM

-0.0082 (-1.47)

0.0097 (1.58)

Included 23.33%

201

NPM

-0.0004 (-0.29)

21.99% 201

* This table reports the results of the OLS estimation of the following regression model:

Performancei, = ^o + ^iCG_COMPLi, + i?2LEVit -f ftLNTAu + jß4y200

Where: Performance!, = ROA, ROE or NPM. ROA = one-yea and is measured as operating income divided by average total

r ahead retum on assets

li, + ^5A'i,+ ei,

for firm i in year t, assets; ROE = one-year ahead retum on

Vol. 39, No. 5. 2009 511

Table 5 Additional regression results {continued) equity for firm i in year t, and is measured as eamings before extraordinary items divided by the average book value of stockholders' equity; NPM = one-year ahead net profit margin for firm i in year t, and is measured as eamings before extraordinary items divided by sales revenues; CG_COMPLi, = score JTom 1 to 5 with higher scores indicating greater compliance of firm i in year t with intemational best practice conceming board structure and functioning; LEV;, = ratio of short-term debt plus long-term debt over total assets for firm i in year t; LNTAi, = natural logarithm of total assets for firm i in year t; Y2001i, = year dummy, =1 when an observation is fi-om 2001, zero otherwise; X¡, = a vector of industry dummies based on the FTSE Global Classification System two-digit code for the industry sector.

Panel A reports the results on CG_COMPL in four different regressions. The four regressions differ in the measure used for CGCOMPL. In the first regression, CG COMPL equals the Deminor rating of board stmcture and fiinctioning (cf the model and the results reported in Table 3). In the second regression, CG_COMPL is the Deminor rating of rights and duties of shareholders. In the third regression, CG_COMPL is the Deminor rating of range of takeover defences. And in the fourth regression, CG_COMPL is the Deminor rating of disclosure on corporate govemance. Panel B reports the results on CG_COMPL of a regression in which CG_COMPL is measured by including three separate Deminor ratings (i.e. rating of board stmcture and fiinctioning, rating of rights and duties of shareholders and the rating of takeover defences) together in one regression. Panel C reports the results of a performance regression in which CG_COMPL is replaced by two factor scores. These two factors scores are obtained (using the regression method) fi-om a principal component factor analysis (using an orthogonal VARIMAX rotation) on the four sub-ratings disclosed by Deminor. The rating of board stmcture and fiinctioning and the rating of disclosure on corporate govemance load on factor 1. The rating of rights and duties of shareholders and the rating of takeover defences load on factor 2. Panel D reports the results on CG_COMPL in a regression in which CG_COMPL is measured by the total govemance rating as reported by Deminor. The total govemance rating summarises firms' performance in the various govemance dimensions covered by Deminor.

anee dimensions) as disclosed by Deminor. The and the operating performance of large listed estimation results for that model are included in European companies. Although theory suggests Table 5, Panel D. These results show that the total that greater compliance with intemational best rating is significantly and positively related with the practices conceming board structtire and fiinction- one-year ahead ROA, but is not significantly related ing is positively related with operating perform- with the one-year ahead ROE nor with the one-year anee, a prior European study (Bauer et al., 2004) ahead NPM. reports evidence of a negative association between

Note that, notwithstanding that the performance corporate govemance compliance and corporate model was carefiilly constmcted based on theory performance. I have re-examined the relation and prior empirical evidence, and that the above between corporate govemance compliance and tests indicate that the results ofour primary analysis operating performance in a setting where there are robust to correlated omitted govemance vari- remained considerable diversity in the extent of ables, one can never exclude that there exists an compliance with intemational best practices regard- unknown correlated factor which is omitted from ing corporate govemance. Consistent with expect- the analyses and actually causes the relationship ations, results of univariate and multivariate tests between perfonnance and the extent of compliance show that greater compliance with intemational best with intemational best practices conceming board practices conceming board stmcture and fiinction- stmcture and functioning. ing is significantly and positively correlated with

the one-year ahead ROA. The results are robust to 6. C o n c l u s i o n controlling for firms' compliance with best practices This paper has examined the relation between the in other govemance areas, such as rights and duties extent of compliance with intemational best prac- of shareholders and range of takeover defences, and tices conceming board stmcture and ñinctioning to controlling for country-level performance.

512 ACCOUNTING AND BUSINESS RESEARCH

Results of additional tests show that greater com- pliance with intemational best practices in other govemance dimensions, namely disclosure of cor- porate govemance and range of takeover defences, is significantly and positively correlated with the one-year ahead ROA.

Further tests show that greater compliance with intemational best practices conceming board struc- ture and functioning is significantly associated with reporting less income from asset disposals and that studying a performance measure that includes this item, as Bauer et al. (2004) do, obscures the inherently positive relationship between operating performance and the extent of compliance with intemational best practices regarding board stmc- ture and functioning. The measure of operating performance is thus crucial when examining firm- level operating performance.

Although the results show only correlation between one-year ahead operating perfonnance and the extent of compliance with intemational corporate govemance best practices, and do not prove causation, the results suggest that firms may benefit from complying with intemational corporate govemance best practices, which may help to convince companies to adopt good govemance practices. In addition, the results ofthe investigation may support regulators in motivating and defending their corporate govemance decisions, and may prove useful for investors when assessing and evaluating company perfonnance. One has to bear in mind, however, that not all individual govemance characteristics identified by the independent rating agency as being best practices can unequivocally be classified as such. Also, as with most other govemance studies, tests were performed on a set of large listed non-financial companies. Therefore, the results do not necessarily hold in a sample of smaller companies, privately-held companies, or financial companies. An examination ofthe impact of govemance on the operating perfonnance of smaller companies, privately-held companies, or financial companies is then also an interesting avenue for fiiture research.

References Alves, C. and Mendes, V. (2004). 'Corporate govemance policy and company performance: the Portuguese case'. Corporate Govemance: An International Review, 12(3): 290-301.

Ball, R., Kothari, S. P. and Robin, A. (2000). 'The effect of intemational institutional factors on properties of account- ing eamings'. Joumal of Accounting and Economics, 29(1): 1-51.

Barber, B. M. and Lyon, J. D. (1996). 'Detecting abnormal operating perfonnance: the empirical power and specifica-

tion of test statistics'. Journal of Financial Economics, 41(3): 359-399.

Bartov, E. (1993). 'The timing of asset sales and eamings manipulation.' The Accounting Review, 68(4): 840-855.

Bauer, R., Braun, R. and Clark, G. L. (2008). 'The emerging market for European corporate govemance: the relation- ship between govemance and capital expenditure, 1997- 2005'. Joumal of Economic Geography, 8(4): 4 4 1 ^ 6 9 .

Bauer, R., Clark, G. L. and Wójcik, D. (2006). 'Geographically dispersed ownership and inter-market stock price arbitrage - Ahold's crisis of corporate govem- ance and its implications for global standards'. Journal of Economic Geography, 6(3): 303-322.

Bauer, R., Giinster, N. and Otten, R. (2004). 'Empirical evidence on corporate govemance in Europe: the effect on stock retums, firm value and performance'. Joumal of Asset Management, 5(2): 91-104.

Bebchuk, L., Cohen, A. and Ferrell, A. (2004). 'What matters in corporate govemance?' Olin Paper No. 491, Harvard Law School.

Beiner, S., Drobetz, W , Schmid, M. M. and Zimmermann, H. (2006). 'An integrated framework of corporate govern- ance and firm valuation'. European Financial Management, 12(2): 249-283.

Bhagat, S. and Black, B. (2002). 'The non-correlation between board independence and long-term firm perform- ance'. Journal of Corporate Law, 27(2): 231-273.

Black, B. S., Jang, H. and Kim, W. (2006). 'Does corporate govemance predict firms' market values? Evidence fi-om Korea'. Journal of Law, Economics, and Organization, 22 (2): 366-413.

Brown, L. D. and Caylor, M. L. (2006a). 'Corporate govemance and firm valuation'. Journal of Accounting and Public Policy, 25(4): 409-434.

Brown, L. D. and Caylor, M. L. (2006b). 'Corporate govemance and firm operating performance'. Working paper, Georgia State University and University of South Carolina.

Core, J. E., Guay, W. R. and Rusticus, T. (2006). 'Does weak govemance cause weak stock retums? An examination of firm operating performance and investors' expectations'. Joumal of Finance, 61(2): 655-687.

De Angelo, L.E. (1988). 'Managerial competition, informa- tion costs and corporate govemance: the use of accounting performance measures in proxy contests. Joumal of Accounting and Economics, 10: 3-36.

Deminor Rating (2001). Corporate Govemance Rating Service Europe, http://www.deminor.com (22pp.).

Drobetz W , Schillhofer, A. and Zimmermann, H. (2004). 'Corporate govemance and expected stock retums: evi- dence fi-om Germany'. European Financial Management, 10(2): 267-293.

Dumev, A. and Kim, E. H. (2005). 'To steal or not to steal: firm attributes, legal environment, and valuation'. Journal of Finance, 60(3): 1461-1493.

Francis, J. R., Khurana, I. K. and Pereira, R. (2003). 'The role of accounting and auditing in corporate govemance and the development of financial markets around the world'. Asia-Pacific Joumal of Accounting and Economics, 10: 1—30.

Gujarati, D. (2003). Basic Econometrics. Intemational Edition. McGraw-Hill Companies, Inc.

Gompers, P., Ishii, J. and Metrick, A. (2003). 'Corporate govemance and equity prices'. Quarterly Joumal of Economics, 118(1): 107-155.

Vol. 39, No. 5. 2009 513

Herrmann, D., Inoue, T. and Thomas, W. B. (2003). 'The sale of assets to manage eamings in Japan'. Journal of Accounting Research, 41(1): 89-108.

Jensen, M. C. (1993). "The modem industrial revolution, exit, and failure of intemal control systems'. Journal of Fma/îce, 48(3): 831-880.

Kiel, G.C. and Nicholson, G. J. (2003). 'Board composition and corporate performance: how the Australian experience informs contrasting theories of corporate govemance'. Corporate Govemance: An Intemational Review, 11(3): 189-205.

Klapper, L. F. and Love, I. (2004). 'Corporate govemance, investor protection and performance in emerging markets'. Joumal of Corporate Einance, 10(5): 703-728.

Klein, A. (1998). 'Firm perfomiance and board committee structure'. Joumal of Law and Economics, 41: 275-303.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1997). 'Legal determinants of extemal finance'. Joumal of Einance, 52(3): 1131-1150.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998). 'Law and finance'. Journal of Political Economy, 106(6): 1113-1155.

La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (1999). 'Corporate ownership around the world'. Joumal of Einance, 54(2): 411-5\7.

Larcker, D.F., Richardson, S. A. and Tuna, I. (2005). 'Ratings add fire to the govemance debate'. Financial Times Mastering Corporate Governance Part Two.

Larcker, D.F., Richardson, S. A. and Tuna, 1. (2006). Corporate govemance, accounting outcomes, and organi- zational performance. Working paper. Graduate School of Business Stanford University and The Wharton School, University of Pennsylvania.

Leuz, C , Nanda, D. J. and Wysocki, P. D. (2003). 'Earnings management and investor protection: an intemational comparison'. Journal of Einancial Economics, 69(3): 505-527.

Massa, M., and Simonov, A. (2007). Shareholder homo- geneity and firm value: the disciplining role of non- controlling shareholders. Working paper

Pincus, K., Rusbarsky, M. and Wong, J. (1989). 'Voluntary formation of corporate audit committees among NASDAQ ñrms'. Joumal of Accounting and Public Policy, 8(4): 239- 265.

Renders, A., and Gaeremynck, A. (2006). 'Corporate govemance and performance: controlling for sample selection bias and endogeneity'. FETEW Research Report AFI06-06, KU.Leuven, 35pp.

Sonnenfeld, J. (2004). Good govemance and the misleading myths of bad metrics. Academy of Management Executive, 18(1): 108-113.

Thomson Financial (2003). Worldscope Data Definitions Guide.

Weil, Gotshal and Manges LLP (2002). 'Comparative matrix of corporate govemance codes relevant to the european union and its member states', http://ec.europa.eu.int/comm/ internal market/company/docs/corpgov/corp-gov-codes- • rpt-part3_en.pdf, 379pp.

Willekens, M., Vander Bauwhede, H. and Gaeremynck, A. (2004). 'Voluntary audit committee formation and practices among Belgian listed firms'. Intemational Joumal of Auditing 8(3): 207-222.

Wójcik, D., Clark, G. L. and Bauer, R. (2005). 'Corporate govemance and cross-listing: evidence from European companies'. Working paper.

Wójcik, D. (2006). 'Convergence in corporate govemance: evidence from Europe and the challenge for economic geography. Joumal of Economic Geography, 6(5): 639- 660.

Yermack, D. (1996). 'Higher market valuation of companies with a small board of directors'. Joumal of Einancial Economics, 40(2): 185-211.

may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express

written permission. However, users may print, download, or email articles for individual use.