Audit quality and Earnings Management in Quoted Nigerian Banks
Abstract:
The objective of the study is to find out the impact of audit quality on earnings management. The study used a sample of all eighteen banks quoted on the stock exchange as at December, 2010. Data was gathered for the period 2005 to 2010. The crosssectional year by year regression analysis was performed. Audit quality is measured by using audit fees and auditor change, and abnormal loan loss provision is used to measure earnings management. Though the result was mixed, however, based on the frequency of results for the period of the study, both audit fee and auditor change were positively related to abnormal loan loss provision. This suggests that high audit fee and change in auditor tenure will aggravate earnings management. We recommend that auditor change should not be ceremonial but based on fact of inefficiency and audit fee from each auditor client should be monitored to enforce the five per cent maximum from each client as suggested by Institute of Chartered Accountants code of ethics.
1. Introduction
Banks generally have the incentive to engage in earnings management. There are three hypotheses that explain banks’ incentive to manage earnings (Diamond and Dybvig, 1983; Morgan, 2002). These include the customer confidence hypothesis, the asset substitution hypothesis, and the regulation hypothesis. The customer confidence hypotheses explains that banks manage earnings to gain customer confidence give the challenge of illiquidity that constantly confronts banks. The asset substitution hypothesis explains that because banks engage in risky asset substitution behaviour, they engage in earnings management to hide such risks. Finally, the regulation hypothesis state that banks engage in earnings management to beat or meet regulations, which are rife in the banking sector.
Arguably banks in Nigeria occupy a very important position in the economy and in the financial system in particular. Bank distresses in Nigeria are a recurring decimal. Adewakun (2010) noted that the Central Bank of Nigeria (CBN) has identified poor corporate governance and unethical practices as one of the major causes of distress in the nation’s banking industry. Unethical practices here include accounts manipulation.
The Prudential Guideline (2010) is meant to improve audit and financial reporting quality in banks. One of the focuses of this Prudential Guideline is to deal with accounts manipulation occasioned by earnings management. The Prudential Guideline (2010) regulated the tenure of external auditors. The guideline states that the tenure of the external audit should not exceed ten years from the date of first appointment. This is because there are empirical and practical evidences that long audit tenure compromise audit and quality leads corporate failure (Becker, DeFond, Jiambalvo and Subramanyam ,1998; Gerayli, Yanesari and Ma'atoofi, 2011). In particular, the demise of Enron and Anderson clearly shows the importance of audit quality in constraining accounts manipulation. The Cadbury and Akintola Williams Delloite (AWD) case of fraudulent financial reporting also demonstrates the relationship between audit quality and account manipulation. In that case Administrative Proceedings Committee (APC) found the parties guilty of preparation of fraudulent financial statements, outright fraud and gross negligence.
DeAngelo (1981) defined audit quality as the joint probability that the external auditor detects an anomaly in financial statements, and then reveals it to the users of these statements. This definition ascribes audit quality to both competence and integrity of auditor. It takes competence to detect anomaly and integrity to disclose it. Audit competence and auditor integrity are difficult to operationalize. However, prior research have used such proxies as auditor size, auditor specialization, length of auditorclient relationship and auditor reputation to measure audit quality (DeAngelo, 1981; Klein and Leffler, 1981; Knapp, 1991). Relating these measures to abnormal accruals, as a measure of earnings management has provided mixed results (Piot and Janin, 2005; Chen, Lin and Zhou, 2005; Gerayli, et al., 2011).
The main objective of the study is to ascertain relationship between abnormal loan loss provision (proxy for earnings management) to audit quality (measured by auditor tenure and audit fees. Abnormal loan loss provision has been adjudged as more suitable than abnormal accruals as a measure of earnings management for banks because banks have generally insignificant accruals (Fonseca and González, 2008; Kanagaretnam, Krishnan and Lobo, 2010; Oosterbosch, 2009). The study is unique because to our knowledge, there is no study that relates abnormal loan loss provision to audit quality using data from the Nigerian banking study. Beside, we use a dummy variable to capture auditor tenure.
Specific objectives of the study are to:
i. find out the effect of auditor tenure on earnings management; and
ii. ascertain whether the effect of audit fees on earnings management is significant. On the basis of these objectives, the study formulates the following hypothesis:
i. auditor tenure has no significant effect on earnings management and
ii. the effects of audit fees on audit tenure is not significant.
2. Literature Review
Prior studies, like Chen et al., (2005), Piot and Janin (2005) and Gerayli, Yanesari and Ma'atoofi, (2011) examined whether there is a nexus between earnings management and audit quality. Chen et al. (2011) examined the influence of audit quality on earnings management and cost of equity capital. They employed two sets of firms: stateowned enterprises (SOEs) and nonstateowned enterprises (NSOEs). They find that highquality auditors play governance role in China. The role is however, restricted to a subgroup of organisations and even under identical legal jurisdiction, the impact of audit quality (in the form of lower earnings management and cost of equity capital) fluctuate among firms with diverse ownership arrangements.
Piot and Janin (2005) did not find a substantial link between earnings management and audit quality. They used abnormal accruals to measure earnings management and presence of big five as a proxy for audit quality. Gerayli et al. (2011) considered the influence of audit quality on earnings management using data from quoted Iranian firms. They used auditor size, auditor industry specialization and independence to proxy audit quality, and discretionary accruals to proxy earnings management. Discretionary accruals showed a negative association with auditor size and auditor industry Specialization. Furthermore, they find negative association between auditor independence and discretionary accruals. Summarily, their study suggests that high audit quality is more likely to constrain earnings management than low quality. This result agrees with prior research (Teoh and Wong, 1993; Becker, DeFond, Jiambalvo and Subramanyam, 1998; Rusmin 2010; Francis, Maydew and Sparks, 1999; Li and Lin, 2005). Further indications of the negative link between audit quality and size of abnormal accruals is provided by the studies of Ebrahim (2001) and Tendeloo and Vanstraelen (2001).
Carcello and Nagy (2004) find that dishonest financial reporting arises early in an auditor client affiliation. Geiger and Raghunandan (2002) finds that corporate failure occurs considerably more often in the first five years of an auditorclient association. Myers, Myers and Omer (2003) find that risky accounting choices are inhibited more effectively by longer auditor term. In conclusion, Gosh and Moon (2005) find that investors and rating agencies depend on audited financial reports more strongly as auditor tenure rises. These studies suggest that longer auditor tenure increases audit quality and by extension the lowers earnings management propensity.
However, standard setters and regulatory authorities believe that longer audit tenure encourages earnings management. They therefore make auditclient rotation mandatory. In the United States, the SabaneOxley Act 2002 reduced the auditor tenure from seven to five years. While the Prudential Guideline 2010 in Nigeria limits auditor tenure to ten years and the European Commission limits it to sevenyear.
Audit fee is often used to proxy auditor independence and hence audit quality. Kanagaretnam et al. (2010) examine auditor independence in the banking industry by analysing the relation between fees paid to auditors and the extent of earnings management through loan loss provisions LLP. They find that unexpected audit fees are unrelated to earnings management for large banks. For small banks, they find greater earningsmanagementviaunderprovisioningofLLPbybanksthatpayhigherunexpected total and nonaudit fees to the auditor. Their results suggest that auditor fee dependence on the audit client is associated with earnings management via abnormal LLP and is a potential threat to auditor independence for smallbanks.
3. Data and Method
The data is drawn from preIFRS (International Financial Reporting Standard) adoption financial reports of all the 18 banks quoted on the Nigerian Stock Exchange as at 31 December, 2010. The data was sourced manually from publicly available annual reports for period 2005 to 2010. We use preIFRS period to exclude the impact of adoption of IFRS on the quality of financial.
The study uses a multiple regression model. Regression analyses were done year by year for the six year period, 20052010. Our result is based on the frequency of the relationship between the dependent and independent variables as revealed by the year by year regression analyses.
4. Model Specification
Our model relates audit quality to earnings management. Audit quality is the independent variable. Audit fees and auditor change is used as measures of audit quality (See Gerayli et al., 2011, and CBN Prudential Guideline, 2010). Following Kanagaretnam et al. (2010), we define proxy earnings management using abnormal loan loss provision and define abnormal loan loss provision as follows:
Model 1 Abnormal loan loss provision
$\begin{aligned} & \text { Where LLAB = Loan Loss Provision at Beginning } \\ & \mathrm{NPLAB}=\mathrm{Non}\text { performing Loans at the Beginning } \\ & \triangle \mathrm{NPL}=\text { Change in Non } \text { performing Loan } \\ & \text { NBLW }=\text { Net bad loans written } \text { off } \\ & \Delta \mathrm{TOTL}=\text { Change in total loans } \\ & \text { TOTL }=\text { total loans } \\ & \mathrm{E}_{\mathrm{t}}=\text { errorterm } \\ & \mathrm{a}_2 \mathrm{a}_3, \mathrm{a}_4 \text { and } \mathrm{a}_6 \text { are parameter }>0 \\ & \end{aligned}$
$\mathrm{a}_2 \mathrm{a}_{3,}, \mathrm{a}_4$ and $\mathrm{a}_6$ are parameter $>0$
We specify our model as follows:
Model 2: Earnings management and audit quality
The model for relating earnings management to audit quality is given below:
Where: $A B L L=$ Abnormal Loan loss Provision
AUDFEE $=$ Total Audit Fees
AUDCH = Auditor Change
$\mathrm{E}=$ error term
$\emptyset_1$, and $\emptyset_2$ are parameters
$E=$ error term
Table 1 defines the variables for the study and their measurements.
ABLL  Abnormal Loan Loss Provision  Difference between total loan loss and normal loan loss  Kanagaretnam et al. (2010) 
AUDFEE  Total Audit Fees  Natural log total audit fees paid by bank as disclosed in annual report  Gerayli et al., 2011 
AUDCH  Auditor cℎange  Dichotomous variable, 1 if Auditor was change in the year and 0 otherwise  CBN Prudential Guideline, 2010 
5. Results and Discussion
The results of the regression are shown in table 1 and 2. The relationship between auditor change and abnormal loan loss provision is positive except for 2005 and 2010, where it is negative. Our result is mixed, though the frequency of positive relationships between abnormal loan loss provision and audit change is more. A positive relationship between abnormal loan loss provision and auditor change, as the frequency of the results show, implies that a change in auditor increases tendency for earnings management. A positive relationship between auditor change and earning management is supported by Gul, Chen, Tsui, and Judy (2003) and Abbott, Paker and Peters (2000). Johnson, Daily and Ellstrand (1996) suggest that auditor change lowers audit quality. They find that short audit tenure lowers audit quality; implying that a change in auditor is not desirable.
The results, although mixed, is not significant. This study therefore supports the first hypothesis that:
Auditor tenure has no significant effect on earnings management
 2005  2006  2007  
VARIABLES:  Coef  pvalue  Coef  pvalue  Coef  Pvalue 
AUDCH   8.65E+08  0.27  5.44E+08  0.25  6.81E+08  0.38 
AUDFEE  22883153  0.44  1210292  0.01  38709  0.87 
R2 
 0.17 
 0.40 
 0.17 
Adjusted R2 
 0.02 
 0.26 
 0.02 
Fstatistic 
 0.88 
 2.83 
 0.91 
Pvalue 
 (0.47) 
 (0.07) 
 (0.46) 
DWstat 
 1.89 
 2.16 
 2.03 
Audit fee is positively related to abnormal loan loss provision for the period studied except for 2007 and 2009. A positive relationship between audit fee and earning management implies that higher audit fees tend to aggravate earnings management. This is explainable by the concept of economic bonding. When the auditor receives enormous fees from the client, there is a tendency for the auditor to acquiesce when the client adopts unacceptable accounting rules to prepare financial statements. Srinidhi and Gul (2007) find that there is a positive relationship between audit fees and accrual quality. Moreover, Gosh and Moon (2005) posits that audit fee has a negative impact on audit quality.
The study provides mixed results of the relationship between earnings management and audit fee. However, the result is not significant. The study there supports the second hypothesis that:
the effects of audit fees on audit tenure is not significant.
 2008  2009  2010  
VARIABLES:  Coef  pvalue  Coef.  pvalue  Coef  pvalue 
AUDCH  2.71E+09  0.42  1.70E+10  0.32  7.53E+09  0.31 
AUDFEE  6982058  0.89   61964660  0.61  9.31E+07  0.33 
R2 
 0.04 
 0.12 
 0.16 
Adjusted R2 
 0.08 
 0.09 
 0.00 
Fstatistic 
 0.34 
 0.57 
 0.81 
Pvalue 
 (0.71) 
 (0.64) 
 (0.50) 
DWstat 
 1.69 
 1.86 
 1.90 
6. Conclusion and Recommendations
The study shows the inconclusiveness of the relationship between audit quality and earnings management. This inconclusiveness suggests the need for improving method of analysis and developing new proxies for audit quality and earnings management. Using audit fee and auditor change as proxy for audit quality has provided researchers with mixed results. Probably a better measure of audit quality would be from the viewpoint of the audit firm. A measure of audit quality that combines auditor competence and auditor independence is recommended. Moreover, abnormal loss provision may not capture earnings management properly. Doing the same study using post IFRS data, may improve the consistency of results as IFRS is meant to narrow the latitude in accounting rules.
We recommend that based on our findings that audit fees should be monitored to ensure that no audit firm has a client whose audit fee exceeds more than five per cent of total audit fee as suggested by the Institute of Chartered Accountants of Nigerian code of conduct for professional accountants. Significantly high audit fees from an audit client will result in economic bonding which will jeopardize auditor independence and audit quality.
1 REGRESSION RESULTS
REGRESSION RESULT Earnings management and Audit Quality (Abnormal loan loss provision as dependent variable) 2005
Dependent Variable: ABLL Method: Least Squares
Date: 09/07/14 Time: 16:15 Sample (adjusted): 2 18
Included observations: 17 after adjustments
Convergence achieved after 2 iterations
Variable  Coefficient  Std. Error  tStatistic  Prob. 
AUDCH  8.65E+08  7.46E+08  1.158917  0.2673 
AUDFEE  22883153  28761306  0.795623  0.4405 
C  2.02E+08  1.13E+09  0.17811  0.8614 
AR(1)  0.219795  0.310683  0.707458  0.4918 
Rsquared 
0.169562 
Mean dependent var 
27788498  
Adjusted Rsquared  0.022078  S.D. dependent var  1.01E+09  
S.E. of regression  1.02E+09  Akaike info criterion  44.53569  
Sum squared resid  1.37E+19  Schwarz criterion  44.73174  
Log likelihood  374.5534  HannanQuinn criter.  44.55518  
Fstatistic  0.884794  DurbinWatson stat  1.888908  
Prob(Fstatistic)  0.474565 

 
Inverted AR Roots  0.22 


REGRESSION RESULT Earnings management and Audit Quality (Abnormal loan loss
provision as dependent variable) 2006
Dependent Variable: ABLL Method: Least Squares
Date: 09/07/14 Time: 17:32 Sample (adjusted): 2 18
Included observations: 17 after adjustments Convergence achieved after 21 iterations
White heteroskedasticityconsistent standard errors & covariance
Variable  Coefficient  Std. Error  tStatistic  Prob. 
C  8.71E+08  3.87E+08  2.249074  0.0425 
AUDCH  5.44E+08  4.53E+08  1.201248  0.2511 
AUDFEE  1210292  359377.8  3.367744  0.005 
AR(1)  0.425735  0.181602  2.344323  0.0356 
Rsquared 
0.39518 
Mean dependent var 
3.96E+08  
Adjusted Rsquared  0.255606  S.D. dependent var  7.75E+08  
S.E. of regression  6.68E+08  Akaike info criterion  43.68119  
Sum squared resid  5.81E+18  Schwarz criterion  43.87724  
Log likelihood  367.2901  HannanQuinn criter.  43.70067  
Fstatistic  2.83133  DurbinWatson stat  2.156609  
Prob(Fstatistic)  0.079605 

 
Inverted AR Roots  0.43 


REGRESSION RESULT Earnings management and Audit Quality (Abnormal loan loss
provision as dependent variable) 2007
Dependent Variable: ABLL Method: Least Squares
Date: 09/07/14 Time: 18:59 Sample (adjusted): 2 18
Included observations: 17 after adjustments Convergence achieved after 17 iterations
White heteroskedasticityconsistent standard errors & covariance
Variable  Coefficient  Std. Error  tStatistic  Prob. 
C  4.74E+08  6.67E+08  0.71037  0.49 
AUDCH  6.81E+08  7.55E+08  0.902446  0.3832 
AUDFEE  38709  243043  0.15927  0.8759 
AR(1)  0.280727  0.253904  1.10564  0.2889 
Rsquared 
0.173359 
Mean dependent var 
99198670  
Adjusted Rsquared  0.017405  S.D. dependent var  1.05E+09  
S.E. of regression  1.05E+09  Akaike info criterion  44.59341  
Sum squared resid  1.45E+19  Schwarz criterion  44.78946  
Log likelihood  375.044  HannanQuinn criter.  44.61289  
Fstatistic  0.908762  DurbinWatson stat  2.030198  
Prob(Fstatistic)  0.463526 

 
Inverted AR Roots  0.28 


REGRESSION RESULT Earnings management and Audit Quality (Abnormal loan loss
provision as dependent variable) 2008
Dependent Variable: ABBL Method: Least Squares
Date: 09/07/14 Time: 20:11 Sample: 1 18
Included observations: 18
Variable  Coefficient  Std. Error  tStatistic  Prob. 
C  3.15E+09  5.85E+09  0.537991  0.5985 
AUDCH  2.71E+09  3.27E+09  0.828903  0.4202 
AUDFEE  6982058  53382096  0.130794  0.8977 
Rsquared 
0.04391 
Mean dependent var 
2.63E+08  
Adjusted Rsquared  0.083569  S.D. dependent var  4.95E+09  
S.E. of regression  5.15E+09  Akaike info criterion  47.7127  
Sum squared resid  3.98E+20  Schwarz criterion  47.86109  
Log likelihood  426.4143  Fstatistic  0.344449  
DurbinWatson stat  1.690528  Prob(Fstatistic)  0.714071 
REGRESSION RESULT Earnings management and Audit Quality (Abnormal loss
provision as dependent variable) 2009
Dependent Variable: ABLL Method: Least Squares
Date: 09/07/14 Time: 21:43 Sample(adjusted): 2 18
Included observations: 17 after adjusting endpoints Convergence achieved after 8 iterations
White HeteroskedasticityConsistent Standard Errors & Covariance
Variable  Coefficient  Std.Error  tStatistic  Prob. 
C  9.40E+09  1.13E+10  0.82846  0.4224 
AUDFEE   61964660 
1.19E+08 
0.52106 
0.6111 
AUDCH  1.70E+10  1.65E+10  1.033195  0.3204 
AR(1)  0.349967  0.449102  0.779259  0.4498 
Rsquared 
0.116779 
Mean dependent var 
3.05E+08  
Adjusted Rsquared  0.087042  S.D. dependent var  1.30E+10  
S.E. of regression  1.35E+10  Akaike info criterion  49.69534  
Sum squared resid  2.38E+21  Schwarz criterion  49.89139  
Log likelihood  418.4104  Fstatistic  0.572949  
DurbinWatson stat  1.859165  Prob(Fstatistic)  0.642759  
Inverted AR Roots  0.35 


REGRESSION RESULT Earnings management and Audit Quality (Abnormal loss
provision as dependent variable) 2010
Dependent Variable: ABLL
Method: Least Squares
Date: 09/07/14 Time: 19:54 Sample(adjusted): 2 18
Included observations: 17 after adjusting endpoints Convergence achieved after 6 iterations
White HeteroskedasticityConsistent Standard Errors & Covariance
Variable  Coefficient  Std. Error  t Statistic  Prob. 
C  3.30E+10  3.90E+10  0.8475  0.4121 
AUDFEE  9.31E+07  9.25E+07  1.006  0.3328 
AUDCOM  7.53E+09  7.12E+09  1.058  0.3094 
AR(1)  0.238377  0.294821  0.8085  4.33E01 
Rsquared 
1.59E01 
Mean dependent var 
1.11E+09  
Adjusted Rsquared  3.52E02  S.D. dependent var  1.62E+10  
S.E. of regression  1.65E+10  Akaike info criterion  50.09396  
Sum squared resid  3.54E+21  Schwarz criterion  50.29001  
Log likelihood  421.7987  Fstatistic  0.818791  
DurbinWatson stat  1.859095  Prob(Fstatistic)  0.506312 