Internal and external corporate governance mechanisms and earnings management: an international perspective

Research question: This research aims to: 1. Determine which corporate governance mechanisms are most relevant in constraining earnings management and 2. Explain whether differences in results found in previous literature are attributable to moderating effects related to the legal system (common vs civil law). Motivation: One research topic of interest in the last two decades has been the effect of corporate governance mechanisms on earnings management. Idea: This research constitutes an extension of the study conducted by Zgarni and Haloui (2016) whose subject matter deals with only two corporate governance mechanisms, audit committee and audit quality, whereas the present paper contribution consists of testing the relevance of the findings achieved by Zgarni and Haloui (2016) after three years of study. Data: The study is based on a metaanalysis of 75 studies, the research approach being derived from Hunter et al. (1982). Research Findings: The empirical results show that the board of directors, the audit committee and audit size are statistically significant in reducing earnings management. Our findings emphasize the need to explicitly consider the influences of legal and institutional settings when analyzing the interactions between corporate governance and earnings management. The issues discussed in this article should be relevant to improve the supervision of companies and mitigating the earnings management. Especially, we provide some interesting insights to stakeholders, government regulators, banks, academics and research professionals. 1 Corresponding author: Department of Accounting, College of Economics and Administrative Sciences, Al Imam Mohammad Ibn Saud Islamic University, Riyadh, Office No: s216, Building 322, tel. +966580264912; e-mail address: fazehri@imamu.edu.sa Internal and external corporate governance mechanisms and earnings management: an international perspective 34 Vol. 19, No. 1

reduces the need for earnings management. Similarly, a negative relationship has been established between auditor size, specialization and earnings management.
This research constitutes an extension of the study conducted by Zgarni and Haloui (2016) whose subject matter deals with only two corporate governance mechanisms, audit committee and audit quality, whereas the present paper contribution consists of testing the relevance of the findings achieved by Zgarni and Haloui (2016) after three years of study. Seeing the considerable rising number of studies dealing with earnings management and corporate governance around the world ( Sun et al., 2014;Miko et al., 2015), we think it is interesting to take into account the appearance of a number of new researches dealing with this topic and apropos, a new variable is integrated in our meta-analysis namely board of directors. To resume, the present study extends Zgarni and Haloui (2016) in several ways. First, we take in focus three principal corporate governance mechanisms split into two parts: internal mechanisms (audit committee and board of directors) and external mechanisms (audit quality) whereas Zgarni and Haloui (2016) analyse only two corporate governance mechanisms namely audit committee and audit quality. Secondly, on a methodological level we improve two principal empirical aspects, the sample size which includes 75 studies in our research rather than 58 in Zgarni and Haloui (2016) one. Moreover, our meta-analysis is inspired by Hunter et al. (1982) whereas Zgarni and Haloui (2016) used a meta-analysis based on Wolf (1986). The main difference between Hunter et al. (1982) and Wolf (1986) is in the formula used to measure the size effect and in chi-squared test used to test the null hypothesis of no variation in sample correlations.
Our study has a dual purpose. First, we aim to apply a meta-analysis in order to summarize existing research findings dealing with the interactions between corporate governance mechanisms and earnings management quality. It has been established the efficiency of the meta-analysis in explaining mixed results in literature. It is an effective tool "to summarize, integrate, and interpret selected sets of scholarly works in the various disciplines" (Lipsey & Wilson, 2001). To test our hypotheses we conducted several independent meta-analysis. Each one examined the relationship between discretionary accruals (the proxy for earnings management) and one corporate governance variable.
Second, for the purpose of this research, we aim to reconcile the inconsistent study findings and draw a logical link between findings, which seem to be unclear from narrative reviews. Narrative reviews suffer from a lack of acceptable rules to extend the individual results in order to achieve generalized conclusions. Therefore, these limits can guide one to false conclusions because generally, the differences of the significance across researches are approved to the sampling errors.
This research suggests three novel contributions. First, we investigate the relation between the characteristics of the audit committee, boards of directors and substitutes of the audit quality on reducing the earnings management. This contribution could have an added value to the wide literature on this topic (García-Meca & Sánchez-Ballesta, 2009;Lin & Hwang, 2010;Zgarni & Halioui, 2016).
With reference to the study of García-Meca and Sánchez-Ballesta's (2010), we present several distinguishing features. Indeed, both the two previous researches focus on the independence of audit committee and ownership concentration, which are internal corporate governance mechanisms. Whereas Samaha et al. (2015) investigate the common effect of the board of directors and the audit committee characteristics. Our study, provides a more complete extension of this line of research by considering both internal (board size, board independence, CEO duality, audit committee size, independence and meetings of its members) and external corporate governance attributes (audit quality). Therefore, we discuss earnings management from two perspectives: Internal and external corporate governance structure.
We also contribute to talk about on the advantages of utilizing the meta-analysis technique, which is the application of statistical methods to a lot of results from existing individual works for the objective of including and evaluating the study findings. Furthermore, a meta-analysis should allow us to gather more conclusions that are significant from literature in order to achieve robust findings concerning the relation between internal and external corporate governance characteristics and earnings management.
Finally, we estimate that this synthesis of literature should eventually help regulators, stakeholders, auditors and boards of directors, who are concerned with improvement of the supervision of company and attenuating the opportunities specified to managers to engage in the earnings management. The mitigated results should attract their attention on the institutional and cultural countries' features to avoid the naive imitation around the world of these practices. This paper addresses multiple research questions. How important is the overall effect of the corporate governance mechanisms on earnings management? Are independent audit committees or shareholders more effective in reducing managerial accounting discretion around the world? Are empirical findings sensitive to the legal system as a moderator variable?
Relying on previous literature, we suggest to bring some elements of answer to above questions by studying the effect of two major corporate governance pieces, internal (via effectiveness of audit committee and board of directors) and external (via audit quality). The choice of variables is based on the corporate governance mechanisms "most popular" in empirical literature. Some internal and external mechanisms of corporate governance are excluded from our analysis due to the luck of research on such items, which constrains the implementation of a metaanalysis.
Although prior studies have provided some insights into the role of corporate governance, the findings of such researches are often contradictory and are limited to one specific mechanism of corporate governance rather than studying a whole structure of governance. Moreover, we should notice that the luck of studies using meta-analysis method to explore whether inconsistency in research is due to moderator effects, such as the way of measurement of earnings management or the effects of legal system (García-Meca & Sánchez-Ballesta, 2009;Lin & Hwang, 2010). The application of this quantitative review methodology in accounting sciences research should increase investors' confidence to the findings compared to conclusions drawn from narrative analysis (Hunter & Schmidt, 1990).
Our main results show that board independence, board size, audit committee independence and auditor reputation are the most significant variables in mitigating earnings management. Concerning the moderator effect of legal origin, we found that the common law system is negatively associated to earnings management. This paper is organized as follows. The second section presents an overview of prior studies on the association between board of directors, audit committee and audit quality variables and earnings management. Next, we developed the hypotheses, followed by a description of the research design especially the sample selection and variables used. Finally, we analyze and discuss the empirical findings to end with conclusion section.

Corporate governance and earnings management
Board of directors, audit committee and audit quality are considered as centerpieces of corporate governance. In this section, we analyze from theoretical view corporate governance upon three perspectives: board of directors, audit committee characteristics', audit quality.

Board of directors' effect
The board of directors is a crucial internal control mechanism and still play an important role in reducing agency problems (adverse selection and moral hazard). According to the agency perspective, the ability of the board to act as an effective monitoring mechanism depends on its members' independence (Beasley, 1996).
Several researches have proved the importance of the board of directors in the financial reporting process generally and in reducing earnings management particularly (Dechow et al., 2010). Specifically, a growing literature shows that board characteristics, such as board independence (Bradbury et al. 2006), board size (Okougbo et al., 2015), separate roles for the chairman and CEO (Francis et al., 2008) are associated with moderated occurrences of earnings management. In legislative level, the majority of codes of good practicesdefend the following issues: large size, directors' independence and separation in roles for the chairman and CEO. For these reasons, we suggest to give focus on the three following governance attributes: independence, size and duality functions.

Board independence
Agency theory recommends that the existence of independent directors on the board may have an important effect on the effectiveness of board's supervising activities. In the same context, García-Meca and Sánchez-Ballesta (2009) employ the 'domino effect' concept of independent directors to push inside board members to enhance corporate reporting policy.
Previous researches illustrate that independent non-executive directors have a benefit in improving the interests of the shareholders. Beasley (1996) found a negative relationship between the percentage of non-executive members on the board and the likelihood of financial fraud in the USA. Jaggi et al. (2007) and Bradbury et al. (2006) showed that an upper proportion of independent directors are related to more effective overseeing in reducing earnings management. Similarly, Visvanathan (2008) used pre-SOX data to investigate the relation between board independence and earnings management. He showed that board independence is negatively related to the occurrence of earnings management.
In contrary, Graven (2009) reveals that the boards of directors have no effective contribution in reducing earnings management. In Singapore, Bradbury et al. (2006) failed to find any association between earnings management and board independence. The conflicting results in literature can be explained by the differences in measures used to assess the independence of directors (percentage of independent members, dichotomous variable for independence etc.) or even by differences in the institutional contexts and without forgetting the definition, itself of directors' independence should matter. Therefore, we formulate the following hypothesis: H1.1: There is an association between board independence and the occurrence of earnings management.

Board size
The second attribute to assess the added value of the board of directors is its size. Dual contradictory positions prevail in literature: Some authors, such as Hidalgo et al. (2011), think that a large board plays a significant role in monitoring function. This opinion is often explained by the fact that a large board offers increasing expertise and experience. In the same perspective, Karamanou and Vafeas (2005) show that a larger board has better updates of management earnings forecasts. Indeed, they suggest that firms with larger boards show a higher quality of financial reporting.
Other authors, such as Xie et al. (2003) find that small boards are negatively associated with earnings management. This finding is consistent with Jensen (1993) proposing that small boards are more effective in monitoring mangers' discretionary behavior thanks to good communication and coordination which is difficult to achieve with large boards.
Further, Visvanathan (2008) and Graven (2009) find no significant relation between the size of the board and the extent of earnings management. We can underline that evidence from previous literature is not conclusive and therefore we suggest the following hypothesis:

H1.2:
There is an association between board size and the occurrence of earnings management.

CEO/Chairman duality of the board
CEO/Chairman Duality means that someone acts as a CEO and a chairman of a board at the same time. The functions of a CEO and a Chairman of the board should be separate since this characterizes a large concentration of power and this separation would guarantee that no person has unfettered powers of decision. A combined leadership structure reduces the significance of the separation between decision of management (CEO) and decision of control (chair of the board). Therefore, CEO duality may limit the desired system of checks, balances and compromise the boards' independence in supervising upper management behavior (Cerbioni & Parbonetti, 2007). Visvanathan (2008) demonstrates that boards that have the same individual in both functions is less likely to be effective overseeing and consequently are less likely to limit earnings management. Further, other studies (Graven et al., 2009) report a non-significant or a positive association between the duality of the board of directors and the occurrence of earnings management. Therefore, we formulate the following hypothesis: H1.3: There is an association between CEO duality and occurrence of earnings management.

Audit committee effectiveness
Audit committees are usually considered a significant instrument of a firm's overall governance mechanism with specific importance on audit quality and supervision of financial reporting (He et al. 2008, Baxter & Cotter, 2009. Apart from the advantage that is gained from the creation of an audit committee, previous works propose that the size, independence, expertise, and meeting frequency of audit committee may have an effect on overseeing effectiveness (Carcello et al. 2011;Sun et al. 2014;Zgarni & Halioui, 2016).
Several empirical studies have shed light on the effectiveness of audit committee independence. For example, Baxter & Cotter (2009) explain that occurrence of earnings management decreases with the independence of the audit committee. Likewise, Yunos (2011) and Sun et al. (2014) show that audit committee independence is associated with less discretionary accruals.
In contrast, other authors report an insignificant association between audit committee independence and earnings management (Visvanathan, 2008;Garcia et al., 2010;Adiguzel, 2013). Moreover, the presence of a financial expert in the audit committee is regarded as necessary for the effectiveness of its role. Indeed, knowledge and experience in financial, auditing and accounting linked issues provide a major impact on audit committee efficiency. Thus, this know-how should obviously aid the audit committee members to be more rigorous when dealing with accounting discretion exercised by managers. Several empirical researches establish a positive relationship between the financial expertise of audit committee members' and earnings management (Xie et al., 2003;Choi et al., 2004;Dhaliwal et al., 2010). Contrarily, other studies such as, Ghosh et al. (2010) do not find any relationship between financial expertise and earnings management practices.
According to the resource dependence theory, a large audit committee means that the members can bring more resources to the firm, such as experience and expertise, which contribute to the audit committee's effectiveness to overseeing duties. Thus, leading to a high earnings quality. Vafeas (2005) demonstrates that a large audit committee reduced the performance of directors owing to the difficulties of organization and coordination and therefore this is considered as another reason for weak power in monitoring. In contrast, Yang & Krishnan (2005) report that audit committee size is negatively associated with earnings management (using abnormal accruals as proxy for earnings management). This result confirms that a minimum number of audit committee members may be relevant to improve the quality of financial reporting. Nevertheless, Baxter and Cotter (2009) and Adiguzel (2013) find an insignificant association between audit committee size and earnings quality. Krishnan and Visvanathan (2008) revealed that the idea in identifying optimal audit committee size is to have a committee small enough to be controllable but large enough to supervise well. Some researchers suggest an ideal average size of an audit committee between 3 and 4 members (Xie et al., 2003;Vafeas, 2005).
Another attribute to assess audit committee efficiency is meetings frequency of audit committee. From empirical studies, it seems that this attribute provides also mixed results. Visvanathan (2008), Ghosh (2010) and Lin and Wang (2010) showed that firms with a higher number of audit committee meetings, are associated with a lower occurrence of earnings management. These studies recommend that audit committees who meet regularly during the year are more effective in controlling process. The more frequently they meet, the more efficiently they discharge their oversight responsibilities. However, some others studies showed no relationship between frequency of audit committee meetings and earnings management (Bédard et al., 2004;Yang & Krishnan, 2005;Baxter & Cotter, 2009). In their meta-analysis, Lin and Hwang (2010) offer an insightful empirical examination on the influence of audit committees characteristics by analyzing numerous empirical researches published so far. They find that earnings quality is positively and significantly related to independent audit committee members, to frequency of audit committee meetings, to the size of audit committee and to the financial expertise of the audit committee members.
In summary, the above narrative review presented mixed evidence on the value relevance of audit committee characteristics to mitigate the earnings management. Therefore, we propose the following hypotheses for our empirical tests: H2: There is a positive relationship between effectiveness of audit committee and earnings management. H2-1: There is a negative relationship between audit committee independence and the occurrence of earnings management. H2-2: There is a negative relationship between the audit committee financial expertise and the occurrence of earnings management. H2-3: There is a positive relationship between audit committee size and the occurrence of earnings management. H2-4: There is a positive relationship between audit committee meetings and the occurrence of earnings management.

Audit quality effect
Numerous studies illustrate that a higher audit quality (auditor reputation, industry specialization and tenure) reduces managerial accounting discretion (Sun and Liu, 2013). Empirical evidence obtained with regard to these audit quality attributes has been mixed. While many prior works show that the retain of brand name (Big 4/5/6) auditors mitigates the occurrence of earnings management (Francis et al., 1999;Lin et al., 2006), many others do not succeed to confirm such results (Bédard et al., 2004;Davidson et al., 2005). These divergences in findings within research can be explained by the differences in measures retained to assess audit quality.
Concerning the specialization as an attribute for audit quality, Jaggi et al. (2012) find a positive association between industry specialization and earnings quality. Likewise, Chi et al. (2011) show that industry specialist auditors are related with earnings management. However, some other researchers obtain no significant difference in earnings quality and failed to find a relationship between industry specialist auditors and earnings management (Mascarenhas et al., 2010;Lawrence et al., 2011).
With regard to auditor tenure, Reichelt and Wang (2010) consider that long auditor tenure would reduce the occurrence of earnings management. Whereas, Chi et al. (2011) establish that auditor tenure is related to higher occurrence of earnings management. However, Jackson et al. (2008) find no relationship between audit tenure and the discretionary accruals.
Since the results related to the effect of audit quality are often heterogeneous, obviously it will be hard to draw substantive conclusions with regard to this corporate governance mechanism. Therefore, we think it is fruitful to use meta-analysis technique to review, evaluate and synthesize relevant empirical works that examine the relationships between internal and external corporate governance mechanisms and earnings management. Based on the narrative review of the association between audit quality and occurrence of earnings management, we suggest the following hypothesis: H3: There is an association between audit quality and earnings management. H3-1: There is a negative association between auditor size and the occurrence of earnings management H3-2: There is a negative association between auditor industry specialization and the occurrence of earnings management. H3-3: There is a positive association between auditor tenure and the occurrence of earnings management.

Sample selection
The meta-analytical process is essentially based on collection of different empirical researches on a particular topic. We search the desired studies done by downloading from the most popular and academic databases: Science Direct, Springer, JSTOR, ProQuest, EBSCO and Emerald. We use in our searching the following keywords: corporate governance mechanisms, earnings management, board of directors, audit committee and audit quality. We choose our sample based on most relevant articles funded on predetermined criteria and identifying variables that have been examined. Our research of papers leads us to the collection of 75 articles (see Appendix) limited to the effects of corporate governance by investigating (i) the effectiveness of board of directors; (ii) audit committee and (iii) audit quality in order to reduce the magnitude of earnings management.
We eliminate the following studies from our analysis: studies suffering from 1) The luck of empirical methodology, 2) Not English written studies 3) Studies conducted in specific context (crisis period, promulgation of new law etc.). The Selection of studies sample for Meta-analysis is detailed in Table 1.

Meta-analytic procedure
The meta-analysis conducted in our study is by following the method presented by Hunter et al. (1982). The author suggest a conversion formula to transform various reported statistics (F-fisher, t-statistics…) of individual studies into a common metric of effect size. Narrative literature analysis can be misleading because diverse studies may obtain different results about a set of individual research due to disparity in characteristics especially the sample size and the time period (Hunter & Schmidt, 1990).
The meta-analysis method is classic and usually consists of: 1. Converting the effect of individual studies into a common metric namely (r). The Statistical reports (r) will be employed to create collection, integration and evaluation. 2. Utilizing the statistical reports (r) as "size effect". It is the correlation coefficient between the substitute of corporate governance variables and earnings management. 3. If the size effect (r) isn't present, then test-statistics reports will be changed into another (r) via formula process of Lipsey & Wilson (2001) and Hunter &Schmidt (1990), as presented in Table 2.

Table 2. Transformation of Statistical Test to formula to (r) Statistic Test Transformation formula to (r) Statistic F √F/F+df (t) Statistic T √t 2 / (t 2 +df) P-value
Is converted firstly in t-statistics the nit converted in r Chi-square r=√X 2 /N

Where df is a degree of freedom
The meta-analytic process follows three phases to obtain mean correlation (r ̅ ) and estimate of population variance (Schmidt et al., 2004). The first phase is to collect and determine the average correlation coefficient (ᴦ), with formula: r ̅ =∑(Ni ri)/∑ Ni Where: Ni = Number of articles in the study. ri = Coefficient of Pearson Correlation for study i. Subsequently, the second phase is to estimate and correct the error variance: The calculation of sampling-error variance will be that subtracted from the observed variance, leaving the residual variance, which offer an unbiased estimation of the population variance:

( p 2 )= ( r 2 ) -( e 2 )
Where ( r 2 ) total observed variance is calculated by utilizing an average error weighted by sample size:

=∑ i ( i − ̅ ) /∑ i
The third phase is to investigate whether there are moderating variables that could influence the variables' correlations by construction 95% confidence interval and by calculating population mean ( ̅ ) and the standard deviation which used to assess the significance of association of interests. If a confidence interval comprises zero, the relevant mean correlation is supposed showing a non-significant population association. A powerful statistical analysis to estimate the moderating effects is to detect whether the observed variance is trivial (homogeneous) or is significantly greater than expected (heterogeneous), is to calculate the X 2 statistic (Hunter et al. 1990), with following formula: X 2 k-1=N r

/(l-̅ ) 2
If Chi-square test result explains insignificant value, it means research's findings examined are homogeneous. It does not give indication of a moderating variable. If Chi-square test illustrates a significant value so we should conclude that there is an heterogeneity that require the existing of a moderated variables. Variations in research results are moderated by other variables. Moderation variables examined is done by grouping (sub-groups) of researches and estimated a correlation coefficient (r) and total observed variance ( r 2 ) for every hypothesized sub-groups. The organization is done by a sub-group divergences to measure dependent and independent variables. The purpose of this sub-group is to mitigate heterogeneity and to enhance explanatory power. Prior meta-analysis researches recommend that studies should be gathered according to dissimilarity in the measurement of the dependent and the independent variables to moderate the level of variance in findings. In this study, researchers are sub-grouped based on the disparity in mechanisms of corporate governance measures and differences in legal origin of each country retained in our study.

Results
The findings of our meta-analysis concerning the association between internal and external corporate governance mechanisms and earnings management are reported in Tables 3, 4 and 5. In each table, we summarize the results of the over-all metaanalysis then we detail results within sub-groups according features retained for each corporate governance mechanism. This step will allow us to test the hypothesis suggested previously. Table 3 shows the findings of the meta-analysis of the effect of board characteristics on earnings management. An overall meta-analysis is conducted for the 26 articles that study the association between board characteristics and earnings management. The results reveal a significant negative association with a (r) of -0.044 (z = 1.623, p 0.01) and a confidence interval between -0.024 and 0.075 with a moderate degree of homogeneity since the sampling error variance accounts for 69 percent of the total variance. This finding demonstrates that the characteristics of the board may limit the occurrence of earnings management. Furthermore, we establish a significant and negative association between board independence and earnings management with a (r) of -0.051 (z = 1.707, p 0.01) with a confidence interval between -0.032 and 0.358.
In terms of CEO duality, table 3 shows that the Chi-square test is non-significant. Thus, we cannot support the hypothesis (H1.3) that firms with CEO duality functions are concerned in more earnings management than companies whose CEO is independent of the chairman. These are due to the variance explained by sampling error (78%), and the real correlation between CEO duality and earnings management in the population has not significant effect (r)= 0.029 (z = 2.012, p 0.01). Thus, the homogeneity test is not rejected; we do not look for moderators, since the size of several moderator studies would be too small. Although most corporate practices provides strong evidence that separation between the CEO and chairman functions is preferable as it enhances the effectiveness of the board monitoring role. Although, this finding could be contrary to Jensen (1993) suggesting that the existence of a chairman who is also the chief executive of one firm could override the benefit of having independent directors on the board and the supervising of the CEO by the chairman signals stronger governance and improve a firm value.
The findings of the meta-analysis of 19 studies show a negative and significant effect (r) = -0.062 (z = 1.56, p 0.01) of board size on earnings management. This results support that this mechanism of governance reduces managerial discretion behavior; that is, larger boards are more effective in limiting earning management. While, this result could be contrary to Okougbo and Okike (2015) who showed that small board is more effective in overseeing managers compared to large one. Moreover, as boards become larger, they are likely to comprise more independent directors (Karamanou & Vafeas, 2005) and varied experiences to improve the board's monitory function and subsequently reduce the occurrence of earnings management (Hidalgo et al., 2011).
We study the moderator effect played by the various legal origins (common and civil law). We find that for studies belonging to common law countries, the (r) is -0.056 (Z = 0.454) so we deduce that common law countries are effective in limiting the occurrence of earnings management. However, from focus within civil law countries, we do not find any significant association with earnings management. However, we should notice that such conclusion is obtained from a little number of correlations in civil law countries (11 studies).  Table 4 displays the findings of the relationship between audit committee characteristics and earnings management. The results of the overall meta-analysis is undertaken for the 47 studies are significant (r) = -0.069 (z = 1.812, p 0.01). Therefore, we can deduce that audit committee characteristics are effective mechanisms of governance in reducing earnings management. Furthermore, When we separate the sample according measures of characteristics of audit committee in order to look for moderators, we find a negative effect (r) = -0.026 (z = -1.943, p 0.01) of independence of audit committee on earnings management. These findings show also a significant negative association between financial expertise and earnings management with a (r) of -0.063 (z = -1.530) and a confidence interval between -0.024 and 0.084 with a moderate degree of homogeneity since the sampling error variance accounts for 33.27 percent of the total variance.
The independence of audit committee is showed to be a major corporate governance mechanism in constraining earnings management. This governance characteristic is likely to supply shareholders with the most protection in preserving the reliability of a firm's financial reporting. Seeing researches included in our meta-analysis, on average, the majority of audit committee members are independent directors. These results confirm both the previous literature and governance recommendations which support that audit committee should contain solely non-executive or independent directors to be efficient. Our results are consistent with Baxter and Cotter (2009);Carcello et al. (2011) and Sun et al. (2014). Moreover, our results are consistent with Zgarni and Halioui (2016) showed that the independence of the audit committee, its size, expertise and the number of meetings have a negative relationship with earnings management. However, our findings do not corroborate with those of Ghosh et al. (2010) and Adiguzel (2013) who find an insignificant relation between independence of audit committee and earnings management.
Moreover, the presence of a financial expert in the audit committee is observed as an important characteristic for the effectiveness of its functions. Really, the experience and knowledge in finance, auditing and accounting play a relevant role in enhancing earnings quality. Our results are coherent with Krishnan and Visvanathan (2008), Dhaliwal et al. (2010) and Lin and Hwang (2010).
Since the selected studies are heterogeneous, we extend our analysis to determine moderators' variables. Thus, we test whether the heterogeneity in the selected studies is due to the legal system. Therefore, we split researches retained in our meta-analysis according to two groups: common law countries and civil-law countries. We find that common law countries have a (r) of -0.041and a confidence interval between -0.155 and -0.024 and a higher degree of homogeneity since the error variance accounts for 51.32 per cent of the observed variance. Furthermore, studies within civil law countries do not show any significant relationship with earnings management. These findings should be interpreted with caution since there are few researches in civil law countries group (14 researches). These findings seem to be coherent with Garcia-Meca and Sanchez-Ballesta (2009) and Zgarni and Haloui (2016).  Table 5 provides the findings of the meta-analysis of the association between audit quality and earnings management. The overall meta-analysis of 45 researches does not show any significant association between the variables. In order to limit heterogeneity, we first look for moderators according to the operational definition of audit quality (auditor size, auditor industry specialization and auditor tenure).
The findings of the meta-analysis of 40 studies show a negative and significant effect (r) = -0.141 (z = 0.594, p 0.01) of auditor size on earnings management. This result support that the retained of brand name (Big 4/5/6) auditors mitigates the occurrence of earnings management (Francis et al., 1999;Lin et al., 2006;Zgarni & Halioui, 2016).
Concerning the association between auditor industry specialization and earnings management, we find a non-significant association. However, Zgarni and Halioui (2016) find a negative relationship between auditor specialization and earnings management. Moreover, the effect of the auditor tenure in reducing the earnings management is positive which confirm then H3.3 for audit quality. This result supported that long auditor tenure would reduce the occurrence of earnings management (Reichelt & Wang, 2010;Chi et al., 2011;Lin & Hwang, 2010).
Finally, we divide the sample according to the legal origins of countries (common and civil law) in order to find more relevant explanation for heterogeneity. However, this subdivision does not show any significant relationship neither between common law and earnings management nor within civil-law countries. Table 4 displays the findings of the relationship between audit committee characteristics and earnings management. The results of the overall meta-analysis is undertaken for the 47 studies are significant (r) = -0.069 (z = 1.812, p 0.01). Therefore, we can deduce that audit committee characteristics are effective mechanisms of governance in reducing earnings management. Furthermore, When we separate the sample according measures of characteristics of audit committee in order to look for moderators, we find a negative effect (r) = -0.038 (z = -1.752, p 0.01) of independence of audit committee on earnings management. These findings show also a significant negative association between financial expertise and earnings management with a (r) of -0.023 (z = -1.422) and a confidence interval between -0.022 and 0.084 with a moderate degree of homogeneity since the sampling error variance accounts for 40.27 percent of the total variance.
The independence of audit committee is showed to be a major corporate governance mechanism in constraining earnings management. This governance characteristic is likely to supply shareholders with the most protection in preserving the reliability of a firm's financial reporting. Seeing researches included in our meta-analysis, on average, the majority of audit committee are independent directors. These results confirm both the previous literature and governance recommendations, which support that audit committee, should contain solely nonexecutive or independent directors to be efficient. Our results are consistent with Baxter & Cotter (2009), Carcello et al. (2011) and Sun et al. (2014). However, our findings do not corroborate with those of Ghosh et al. (2010) and Adiguzel (2013) who find an insignificant relation between independence of audit committee and earnings management.
Moreover, the presence of a financial expert in the audit committee is observed as an important characteristic for the effectiveness of its functions. Really, the experience and knowledge in financial, auditing and accounting play a relevant role in enhancing earnings quality. Our results are coherent with Krishnan and Visvanathan (2008), Dhaliwal et al. (2010) and Lin & Hwang (2010).
Since the selected studies are heterogeneous, we extend our analysis to determine moderators variables. Thus, we test whether the heterogeneity in the selected studies is due to the legal system. Therefore, we split researches retained in our meta-analysis according to two groups: common law countries and civil-law countries. We find that common law countries have a (r) of -0.041and a confidence interval between -0.155 and -0.024 and a higher degree of homogeneity since the error variance accounts for 51.32 per cent of the observed variance. Furthermore, studies within civil law countries do not show any significant relationship with earnings management. These findings should be interpreted with caution since there are few researches in civil law countries group (eight researches). Our results seem to be coherent with Garcia-Meca &Sanchez-Ballesta (2009) and Zgarni & Haloui (2016).  Table 5 provides the findings of the meta-analysis of the association between audit quality and earnings management. The overall meta-analysis of 45 researches does not show any significant association between the variables. In order to limit heterogeneity, we first look for moderators according to the operational definition of audit quality (auditor size, auditor industry specialization and auditor tenure).
The findings of the meta-analysis of 39 studies show a negative and significant effect (r) = -0.234 (z = 0.594, p 0.01) of auditor size on earnings management. This result support that the brand name (Big 4/5/6) auditors mitigates the occurrence of earnings management (Francis et al., 1999;Lin et al., 2006). Similarly, Zgarni and Halioui (2016) showed that auditor size has a significant negative relationship with earnings management.
Concerning the association between auditor industry specialization and earnings management, we find a non-significant association. The effect of the auditor tenure on reducing the earnings management is positive which confirm then H3.3 for audit quality. These results do not support that long auditor tenure reduced the occurrence of earnings management (Reichelt & Wang, 2010;Chi et al., 2011 andLin &Hwang, 2010). However, our findings do not corroborate with the results of Zgarni and Halioui (2016).
Then, we divide the sample according to the legal origins of countries (common and civil law) in order to find more relevant explanation for heterogeneity. However, this subdivision does not show any significant relationship neither between common law and earnings management nor within civil-law countries.

Conclusion
Using the meta-analysis approach, we reconcile the mixed findings of 75 empirical studies investigating the effect of internal and external corporate governance on reducing the earnings management. Our meta-analysis highlights that a country's corporate governance system significantly influences earnings quality. Especially, we demonstrate that both board independence and board size present a great power to constrain the opportunistic behavior of managers. Further, the findings show that the inclusion of independent and financial experts is very effective in audit committee.
In addition, this meta-analytic review proposes that auditor size can be seen as an effective attribute of audit quality to mitigate the occurrence of earnings management. Therefore, we estimate that the results obtained from this metaanalysis could have several practical issues for both: 1. Regulators to emphasize the requirements for audit committee, board of director's structures, profile and liabilities of external auditors. Researchers to develop and increase efforts in order to achieve a strong methodological approach to assess earnings management behavior.
Although this meta-analysis contributes to accounting literature, it is subject to several limitations. First, the sample used include only studies published in recognized databases whereas it exists many both non published and published articles in other databases such as universities publications' or congress acts'. Such limit may lead to different findings. Second, this research analyzes mainly studies, which use discretionary accrual proxy to detect magnitude of earnings management. Whereas, we can considered other earnings management tools such as real earnings management. Another limitation is the bias of the sample to civillaw countries.
A fruitful avenue of further researches would be to incorporate real earnings management as the proxy for measurement earnings management. In addition, the new studies would also facilitate us to estimate the moderating functions played by other variables that could affect the effectiveness of corporate governance mechanisms on reducing earning management, such as changes in governance regulation and comparable acts in other countries.