METHODOLOGY Proceeding E Book 4A Turky

398 years 1998 to 2005 were also excluded from the population of interest due to incomplete data. As consistent with prior research Davidson et al., 2005; Abdul Rahman and Mohamed Ali, 2006 industries with less than 8 firms were also eliminated from the analysis. Further, 139 companies were excluded as the required financial and corporate governance data was not available, resulting in a final sample of 277 companies from 2003 to 2005, giving a total of 831 firm-year observations with complete data for earnings quality and board of directors‘ characteristics. 3.2 Regression Model This study uses a linear multiple regression analysis to test the association between the dependent variable of earnings quality and the independent variable of board independence, board financial expertise, board governance expertise and board firm- specific expertise. EQ =  +  1 BIND +  2 BDFINEXP +  3 BDCROSS +  4 BDTENURE +  5 BDSIZE +  6 LNSALES +  7 LEV +  8 ROA +  9 BIG4 +  10 DUM_YR04 +  11 DUM_YR05 + … 1 Where: EQ = measured by accrual quality based on Dechow and Dichev 2002 model BIND = proportion of independent non-executive directors to the total number of directors on the board of the company BDFINEXP = proportion of directors on the board with financial expertise to the total number of directors BDCROSS = proportion of directors on the board, with directorships in other companies, to the total number of directors. BDTENURE = average number of years of board service of independent non- executive Directors BDSIZE = total number of directors on the board of company LNSALES = natural log of total sales LEV = ratio of total liabilities to total assets ROA = ratio of net income to total assets BIG4 = dummy variable, 1 if audited by Big 4 audit firms, 0 if otherwise As prior studies, this study includes board size, firm size, leverage, firm growth and audit quality as control variables in the regression model as these variables have been shown to have impact on earnings quality Wang, 2006; Jaggi et al., 2007. 399 3.3 Earnings Quality Variable To measure earnings quality, this study applies Dechow and Dichev‘s β00β accrual quality model hereafter DD that captures one aspect of the quality of accruals and earnings. This measure is based on the observation that accruals map into cash flow realizations and regardless of managerial intent, the accrual quality is affected by the measurement error in accruals. The nature of accruals that are frequently based on the assumptions and estimates create estimation errors that need to be corrected in the future. In the DD approach, the estimated residuals from firm specific regressions of working capital accruals, on past, present, and future cash flow from operation, captures the total accruals estimation error by management and are viewed as an inverse measure of earnings quality. The DD model does not distinguish between intentional and unintentional estimation errors. The approach taken is to assess accruals as a whole as both estimation errors imply a lower quality of earnings. TCA j,t =  0,j +  1,j CFO j , t-1 +  2,j CFO j , t +  3,j CFO j,t+1 +  j,t …………………2 Assets j,t Assets j,t Assets j,t Assets j,t Where: TCA j,t = Firm j ‘s total current accruals in year t, = CA j,t - CL j,t - Cash j,t + STDEBT j,t ; CA j,t = Firm j ‘s change in current assets between year t-1 and year t; CL j,t = Firm j ‘s change in current liabilities between year t-1 and year t; Cash j,t = Firm j ‘s change in cash between year t-1 and year t; STDEBT j,t = Firm j ‘s change in debt in current liabilities between year t-1 and year t; Assets j,t = Firm j‘s average total assets in year t and t-1; and CFO j,t = Firm j‘s net cash flow from operation in year t. For each firm-year, equation 2 is estimated cross-sectionally for all firms minimum 8 firms within each industry groups using rolling 7-year windows. These estimations yield five firm- and year-specific residuals,  j,t , t = t- 4,…t, which form the basis for accrual metric. Accrual Quality j,t =   j,t , is equal to the standard deviation of firm j ‘s estimated residuals. Larger standard deviations of residuals correspond to poorer accrual quality and vice versa. Following DeFond et al. 2007 the standard deviation score is multiplied by -1 so that higher score indicate higher earnings quality EQ. 400 4. RESULTS 4.1 DESCRPTIVE STATISTICS TABLE 1 DESCRIPTIVE STATISTICS FOR DEPENDENT AND INDEPENDENT VARIABLES EQ B IN D B D FINE X P B D C R O S S B D TE N U R E B D S IZE LN S A LE S LE V R O A B IG 4 Mean -0.765 0.414 0.190 0.546 6.6 7.9 19.471 0.487 0.029 0.743 Median -0.580 0.375 0.167 0.556 5.8 8.0 19.444 0.452 0.034 1.000 Std. Deviation 0.681 0.111 0.110 0.283 4.3 2.0 1.466 0.512 0.151 0.4375 Min -5.280 0.170 0.000 0.000 0.2 3.0 15.156 0.000 -2.310 0.000 Max -0.040 0.860 0.600 1.000 29.3 16.0 23.649 7.790 2.010 1.000 As reported in Table 1, the mean and median value of earnings quality is -0.765 and - 0.580, respectively. In terms of board composition, 87 percent of companies meet the recommendation of the MCCG 2000 to have at least one third of the board comprising independent non-executive directors. The average, 41.4 percent, of the proportion of independent non-executive directors indicates the domination of insiders in the board composition of companies in Malaysia. With respect to financial expertise, each company has at least 1 to 2 members of the board with financial expertise as represented by a median value of 0.167. In terms of board cross-directorship, more than half the board members 54.6 percent hold additional directorship in other firms. The average length of tenure for independent directors serving in companies in Malaysia is seven years with a maximum value of 29 years.

4.2 CORRELATION ANALYSIS

A Pearson product moment correlation r was computed to examine the correlation between the independent variables. As illustrated in Table 2, board independence, board cross-directorship, board tenure and firm size are significantly related to earnings quality  0.01. Other independent and controls variables are not correlated with earnings quality. The coefficient of correlation between board independence and earnings quality is however negative, which requires further explanation. With respect 401 to correlation among variables, the correlation matrix confirms that no multicollinearity exists between the variables since none of the variables correlates above 0.80 or 0.90. All variables have a correlation of less than 0.40. TABLE 2 CORRELATIONS AMONG VARIABLES EQ B IN D B D FINE X P B D C R O S S B D TE N U R E B D S IZE LN S A LE S LE V R O A B IG 4 EQ 1 BIND - .103 1 BDFINEXP -.019 .104 1 BDCROSS .130 .134 .167 1 BDTENURE .180 .023 - .166 .139 1 BDSIZE .067 - .267 - .141 .081 .061 1 LNSALES .159 -.020 .069 .321 .241 .288 1 LEV -.023 -.041 .035 .051 -.030 .022 .134 1 ROA -.004 .040 .051 .022 .097 .086 .169 - .133 1 BIG4 .013 .011 .065 .167 .081 .045 .130 .011 .033 1 Significant at the 0.01 level; Significant at the 0.05 level 4.3 MULTIVARIATE ANALYSES Table 3 presents the results from the multiple regression analyses. All the models tested in this study are highly significant at 0.01 percent level. TABLE 3 REGRESSION RESULTS Panel A Panel B Panel C t-Statistic t-Statistic t-Statistic Constant -4.699 -4.768 -4.629 BIND -2.299 2.665 -2.309 BIND 2 -2.852 BDFINEXP -0.054 0.020