Governance Role of Auditors and Legal Environment: Evidence from Corporate Disclosure Transparency

Abstract

In a sample of firms originating from 20 countries, we examine whether and how auditor size (our proxy for audit quality) associates with corporate disclosure transparency. While prior studies examine the relation between auditor size and several aspects of financial reporting quality (e.g. discretionary accruals, restatements, etc.), there is limited evidence on how auditor size relates to disclosure transparency. There is also mixed evidence on how auditor size relates to reporting quality in different legal environments. We find that auditor size is positively associated with disclosure transparency around the world and that the association is stronger in code law regimes than in common law regimes. The latter finding supports the view that audits play a greater governing role in weaker legal environments.

Acknowledgements

We thank Martin Baumann, Julia Higgs, Ole-Kristian Hope, Eric Rapley and seminar participants at 2008 AAA annual meeting for helpful comments. Han acknowledges the financial support of the Korea University Business School Research Grant.

Notes

Information in financial statements is reliable if it is representationally faithful (corresponding to what it purports to measure), verifiable (providing assurance that some consensus in the reported measurement exists) and free from bias. We argue that higher quality audits yield more reliable accounting numbers, including the summary measure, such as earnings.

Here we use the term ‘quality’ to refer to managers' commitment to reduce information asymmetry with corporate outsiders by using mandatory and/or voluntary disclosure.

We use the term ‘Big N’ to refer to the Big 4 auditors and their predecessor auditors.

For instance, Bushman et al. (2004) show that country-level disclosure levels are positively related to the Big N audit market share in a country. However, this result has little implication for firm-level associations. Consider for instance a simple two-country setting where there are two firms in each country. In country A, firm A1 (A2) has the disclosure score of 70 (50) and hires a non-Big N (Big N) auditor. In country B, firm B1 (B2) has the disclosure score of 75 (40) and both hire a non-Big N auditor. In this case, coding a Big N (non-Big N) auditor as 1 (0), the country-level correlation between the disclosure scores and Big N auditor market share is positive, but the firm-level correlation is negative with −0.35.

Specifically, they note that the US Securities and Exchange Commission (SEC) adopted reporting transparency as an important criterion for the acceptability of IAS for US securities offerings, along with comprehensiveness and full disclosure.

Representational faithfulness refers to the agreement between a measure and a real-world phenomenon that the measure is supposed to represent. Thus, intentional reporting bias (e.g. discretional accruals) decreases representational faithfulness.

However, Lee et al. (2003) conclude that Australian IPO firms issuing management earnings forecasts are more likely to choose Big 4 auditors to provide a signal of auditors' high quality attestation, but not vice versa.

However, Dunn and Mayhew (2004) do not address the question of whether auditor size (Big N) is associated with disclosure transparency, as their entire sample consists of Big N auditees.

While several prior studies examine the effect of legal environments on the association between audit quality and financial reporting quality, our study differs from those studies as follows. First, Francis et al. (2003) and Choi and Wong (2007) examine the effect of legal environments on auditor choice (Big N vs. non-Big N), the proxy of the equilibrium demand for high quality audits. However, our study examines the direct effect of auditor choice on corporate disclosure transparency across varying legal environments, rather than the determinants of auditor choice as in those studies. Second, Francis and Wang (2008) examine the effect of auditor choice on earnings quality across countries with different legal environments. However, our study examines the effect of legal regimes on auditors' governance role for another aspect of financial reporting quality (i.e. disclosure transparency).

While we do not conduct any validity test for the CIFAR index to prove its reliability, Bushman and Smith (2001) refer to the CIFAR index as an obvious candidate for the quality of financial accounting regime. Salter (1998) also points out that the strengths of CIFAR scores are (1) it is based on actual annual reports; (2) the data have been audited by external sources; and (3) the information is clearly provided. In addition, Cooke and Wallace (1989) audited the CIFAR database and concluded that no biases or errors were present in the data. Hope (2003a) further verifies the CIFAR scores by conducting extensive validity tests, such as re-scoring a sample of annual reports, comparing total disclosure scores with various domestic disclosure rankings, comparing subsets of CIFAR scores with similar rankings by other sources and testing the internal consistency of the scores.

While CIFAR scores may inevitably entail measurement errors by simply rating the annual reports for inclusion or omission of 85 specific items, we believe that this issue would in fact bias against documenting our finding. In addition, although we can consider sub-scores created by choosing the components of CIFAR scores which seem to capture the issues being studied, we still use the composite CIFAR scores. This is because prior literature points out that breaking out subsamples of CIFAR scores is likely to add noise to the ability to identify the true disclosure level (e.g. Miller, 2004; Hope, 2003a).

Size and sales variables are translated into Special Drawing Rights (SDRs). SDRs were created by the International Monetary Fund (IMF) as an international reserve asset and also as the unit of account of the IMF. By dividing those variables based on local currency by the exchange rate of SDR (local currency units per SDR), the variables are based on a common unit of currency.

A priori, the directions of association between DISC and ln Coverage/FDisp are not clear. On the one hand, it is possible that firms disclose more when there is less private information production (low ln Coverage) and higher demand for information (high FDisp). On the other, it seems also possible that corporate disclosure level is higher when analysts' monitoring is stronger (high ln Coverage).

For example, Dye and Sridhar (1995) show that industry membership can be an important factor in firms' disclosure practices.

We conduct a sensitivity test controlling for the possible endogeneity in the interaction between Auditor and Legal. Our inference doesn't change.

Our main results are qualitatively similar when replacing investor protection (Invpro) by legal origin (Legal), which is another proxy for legal environments used in this study.

While ln Size and ln Sales are strongly correlated with each other, we include both in the regression. We do so since prior studies note that they represent distinct dimensions of audit complexity, enhancing the comparability of our findings.

This criterion follows Frankel and Lee (1999).

To mitigate the effects of outliers, we winsorise ln Size, FDisp, Short, Long, ln Sales, ln BM, InvRec, Lev and ROE at the 1st and 99th percentiles of pooled distributions. Other variables are categorical in nature and do not exhibit extreme observations.

Although more recent data is available, we intentionally restrict the sample up to 2003 to coincide the time periods of CIFAR scores (i.e. year 1995) and other firm characteristics. In particular, we conjecture that the introduction of Regulation of Fair Disclosure (Regulation FD) around 2002 by many countries may make the CIFAR scores outdated after 2003. This is because Regulation FD may affect firms' disclosure practices around the world. However, since we do not have alternative disclosure scores incorporating the potential effect of Regulation FD, we rather choose to restrict the sample up to 2003. Although we do not use more recent data, we believe that our empirical analyses, using firm-level variables contemporaneous with CIFAR scores, are still able to contribute to the literature by reporting the significant joint effect of auditor choice and legal environments, which will be stable for a long time, on corporate disclosure transparency at least during 1990s.

Our requirement of firm-level CIFAR scores bias our sample toward large firms, which tend to select Big 4 audit firms. Thus, the proportion of Big 4 audits to non-Big 4 audits seems to be overstated.

Our firm-level analyses may partly address a potential bias due to no variance of auditor size for these five countries by considering each firm-year observation as an individual observation after controlling for other firm characteristics. Nonetheless, we repeat our analysis excluding the sample from these five countries. Our main results are robust.

To mitigate the collinearity between Legal and Culture, we orthogonalise Culture with respect to Legal by regressing it on Legal and taking the residuals.

Opposite to the implication of our main results, Francis and Wang (2008) suggest that higher litigation risks in stronger investor protection countries may induce further Big N auditors, who have more to lose from litigation than non-Big N, to prevent earnings management. However, such an effect of litigation risks on auditors' governing incentive for clients' disclosure transparency will be less. This is because investors are less likely to sue auditors at least when the minimum disclosure requirement is met by firms. Rather, Big N auditors (relative to non-Big 4 auditors) may be able to aid further their clients to provide more transparent disclosure within weaker legal environments where auditors serve as the substitute mechanism for the protection of outside shareholders. Thus, although Lobo and Zhou (2001) report a positive association between disclosure transparency and earnings quality, it is possible that legal environments may have differing effects on the association between auditor choice and auditors' governance role for the prohibition of earnings management and for the improvement of disclosure transparency. In addition, it is important to note that we examine the effect of auditor choice on the availability of relevant information on financial statements while Francis and Wang (2008) examine that on the earnings quality, so the two studies are not directly comparable. Lastly, neither study directly observes the effect of legal environments on auditors' judgment, decision-making process or efforts. Thus, the opposite results of these two studies may indicate the need for further investigation to better understand the effects of legal environments on auditing practices around the world.

The results of first-stage regressions are not reported for brevity.

We also note that the USA is heavily represented in the sample, which may drive our empirical results. Thus, as an additional sensitivity test (the results are not tabulated) we re-estimate the DISC model after excluding the USA from the sample. Our main findings remain the same.

To err on the side of caution, however, we re-estimate the two-stage model and apply Newey–West (1987) correction to the standard errors. We obtain very similar coefficient estimates and t-values under this approach.

We repeat our analyses using other alternative specifications, such as Ordinary Least Squares (OLS), Generalised Method of Moments (GMM) and Heckman's two-stage method (1979). Our main empirical results do not change.

To measure discretionary accruals, we estimate the following Jones (1991) model in each country-year, and use the absolute value of the residuals: TAcc _t  = δ ₁ + δ ₂ GPPE _t  + δ ₃ΔREV _t  + ν _t . Here TAcc _t is the total accruals in year t; GPPE _t is the gross property, plant and equipment at the end of year t; ΔREV _t is the change of revenue in year t. All variables, including the constant term (δ ₁), are scaled by total assets at the beginning of year t. ν _t is the error term (i.e. discretionary accruals).

Additional information

Notes on contributors

Sam Han

Paper accepted by Salvador Carmona.

Tony Kang

Paper accepted by Salvador Carmona.

Yong Keun Yoo

Paper accepted by Salvador Carmona.