Audit partner style and financial statement comparability: New evidence from the US market
Abstract
We explore the influence of individual audit partner style on financial statement comparability in a US setting using newly available data. We find evidence of an audit partner style effect on comparability incremental to audit firm and audit office effects. Our results are consistent across several comparability measures and are economically significant. For example, the audit partner effect is associated with a 12.59% decrease in the mean difference in abnormal accruals compared with a 4.4% decrease for firm pairs audited by the same Big 4 audit office. We also find that audit partner expertise and experience have a greater influence on comparability than audit office expertise and experience. Our results are consistent in endogeneity and robustness tests including (1) auditor switching tests, (2) entropy balancing, (3) placebo tests and (4) controlling for clients’ economic similarity. Our results suggest that the association between audit partner style and financial statement comparability persists in highly regulatory environments and corroborate the PCAOB's intuition that the role of the engagement partner “is of singular importance” to the outcome of the audit (PCAOB 2011).
1 INTRODUCTION
Motivated by evidence from inspections that uncovered disparities in audit quality within audit firms, the PCAOB observed that the role of the engagement partner is “of singular importance” to the outcome of the audit (PCAOB, 2011). This recognition, in part, led to the introduction of Rule 3211 by the PCAOB in 2015, mandating the disclosure of audit partner names for public company audits on Form AP (PCAOB, 2015). However, prior research has not explored whether the influence of the audit partner on the audit supersedes that of the audit office or the audit firm. Moreover, the PCAOB's decision to require audit partner identification was influenced by research conducted in jurisdictions where partner identification was mandatory (e.g., Knechel et al., 2015; Aobdia et al., 2015; PCAOB, 2015). These studies indicate that disclosing partner names provides valuable information to financial statement users. However, it is unclear whether the evidence from these jurisdictions, which in some cases have different regulatory and legal environments, also applies to the United States with its highly regulated and litigious institutional framework.
We investigate whether an audit partner style effect, distinct from previously explored audit office and audit firm effects (e.g., Francis et al. 2014; Kawada, 2014), exists in the United States where legal and regulatory institutions are highly developed and where investor protections and litigation risk work to increase auditors’ conformity with firm policies and methodology. We investigate the existence of an audit partner style effect by exploring whether client firms’ financial statement comparability is more sensitive to Big 4 audit partner style than to audit office or audit firm style. Recent studies carried out in other jurisdictions find evidence suggesting that engagement signing auditors1 exhibit unique audit styles that affect earnings comparability (e.g., Chen et al., 2020; Jiu et al., 2020; Li et al., 2021), although some evidence in these studies casts doubt on whether the engagement signing auditor style effect supersedes that of the audit office.2 Identification of an audit partner style effect in the United States would corroborate the PCAOB's intuition of the singular importance of the audit partner to the audit in highly regulated jurisdictions and would also shed light on how institutional differences could influence audit partner style.
Prior research finds evidence suggesting an association between audit style and financial statement comparability in the United States at the firm (Francis et al., 2014) and office level (Kawada, 2014). However, individual auditors carry out audits, potentially making their characteristics far more relevant at the auditor level than at the firm and office levels (Lennox & Wu, 2018). Indeed, evidence from China suggests that engagement signing auditor style has a greater effect on comparability than audit firm or audit office style, though Chen et al. (2020) caution against generalizing their results to the United States due to differences in institutional environments.
Although prior research in other countries finds evidence suggesting a relation between engagement signing auditor style and financial statement comparability, individual auditor “style” may be influenced by regulatory and institutional features, making it unclear whether this type of effect exists in the United States. Prior research finds evidence that the institutional context affects the factors that shape audit style3—individual auditors’ unique risk preferences (Bik & Hooghiemstra, 2018), ethical orientation (Baïada-Hirèche & Garmilis, 2016), education (Che et al., 2018; Gul et al., 2013) and judgement and decision making (Thorne et al., 2003). Exactly how institutional setting might influence the association between audit partner style and comparability, however, is unclear. For example, weaker regulatory oversight might allow greater flexibility in exercising judgment and decision making for partners, which would argue for greater style effects in weaker regulatory settings. In addition, the high degree of regulatory oversight in the United States and the emphasis on comparability by the Securities and Exchange Commission and the Financial Accounting Standards Board (FASB) may overcome the tendency of individual auditors in the United States to exhibit a style effect. In this case, we would expect to find a weak relation between audit partner style and financial statement comparability in the United States. On the other hand, higher education requirements and experience associated with practicing in more highly regulated environments may lead to a greater style effect in more regulated environments. For example, better educated partners may have a more advanced understanding of the audit process and a stronger grasp of auditing standards, which would make their interpretations more consistent with industry norms (Wang et al., 2015). Recent evidence related to audit partner fixed effects in the United Kingdom (Cameran et al., 2020) suggests that the audit partner is more important than firm or office in explaining audit quality, even in a highly regulated setting. Therefore, whether there is an incremental effect of audit partner style on clients’ financial statement comparability in the US setting, and the potential magnitude of the effect, remains an empirical question.
PCAOB Rule 3211 requires that registered public accounting firms file a Form AP providing the name of the engagement audit partner for all public company audit reports issued after January 31, 2017. We use these newly available data over the period 2016–2020 to explore whether and to what extent individual audit partner style affects the comparability of financial statements in the United States. Following Francis et al. (2014), we measure financial statement comparability in three primary ways: (1) the similarity of earnings accruals between two firms with the same audit partner in the same year and industry; (2) the comovement of earnings between two firms sharing the same audit partner and industry; and (3) audit partner fixed effects to examine whether they are a statistically significant determinant of accruals (Chen et al., 2020; DeFranco et al., 2011; Francis et al., 2014).4
Our results suggest that the effect of audit partner style on financial statement comparability persists in the highly regulated environment of the United States. Empirically, we provide evidence consistent with individual partners having a greater effect on financial statement comparability than audit firms or audit offices. Our results suggest that two firms in the same industry-year who share the same audit partner have more comparable financial statements than two firms that share the same audit firm or audit office but have different audit partners. The results are consistent across our comparability measures and are economically significant.
We first explore whether firms audited by the same Big 4 audit partner have greater accruals similarity than firms audited by different partners. In particular, we find that firms with the same Big 4 audit partner experience a decrease of 12.59% from the mean value of differences in abnormal accruals for firm-pairs in our sample.5 This decrease is especially salient because it is larger than the economic effect found in prior studies (e.g., Chen et al., 2020) in weaker institutional settings and is important evidence that indicates that client firms’ accruals comparability is sensitive to audit partner style, even in a setting with a high degree of regulatory oversight.
We next investigate whether firm-pairs with the same Big 4 audit partner exhibit a higher degree of earnings comovement than firm pairs with different audit partners, and we find that firm pairs audited by the same Big 4 audit partner have greater earnings comovement. Moreover, our results suggest that audit office and audit firm style effects are dominated by the audit partner style effect. Tests of differences of coefficients reveal that the difference between the effect of audit partner and audit firm (office) is statistically significant. Finally, we find that adding audit partner fixed effects to the model containing audit firm and office fixed effects increases the explanatory power of the main model.6 Together, our results corroborate the PCAOB's observation that the identity of the audit partner provides more information about audit quality than the name of the audit firm or audit office alone (2015).
We perform several cross-sectional tests to examine whether the influence of audit partner style on clients’ financial statement comparability varies with audit office and audit partner characteristics. We find that experience and expertise at the audit partner level has a stronger influence on financial statement comparability than experience and expertise at the audit office level. In addition, to explore our prediction that high education requirements and experience of partners in the US audit market may be associated with a strong audit partner style, we utilize hand-collected data from LinkedIn and other public sources to examine whether audit partner education and experience are associated with higher levels of comparability. We find that clients whose audit partners have master's degrees and more industry expertise have more comparable financial statements than clients with less educated and experienced audit partners. In addition, while we limit our main tests to Big 4 firms, additional tests indicate that clients of the same non-Big 4 partner do not exhibit financial statement comparability, together suggesting that education and experience are factors that influence audit partner style.7 Last, we provide evidence that audit partners’ unique styles are associated with audit effort among their own clients, measured using audit fees and report lag.
To provide additional assurance of the validity of our inferences, we perform several supplementary tests. Importantly, our results are robust to including an array of control variables (e.g., same audit office and same audit firm) that are associated with comparability. To further mitigate concerns regarding potential reverse causality or endogeneity, we perform an audit partner switching test, and we find that two client firms with different audit partners who switch to the same audit partner have more comparable financial statements following the switch. We also find that our results are not a function of client firms’ economic similarities by controlling for clients’ product similarity, and we perform a placebo analysis to provide additional comfort that our inferences are warranted. Finally, our inferences are unchanged using entropy balanced and propensity score matched samples. Together, our results indicate that in the highly regulated US setting characterized by strong institutional investor protections and high litigation risk, clients’ financial statement comparability varies with audit partners’ personal characteristics. One explanation for our evidence is that the existing protections and threat of litigation in the United States may increase incentives for audit firms to elevate auditors with extensive experience and education, which in turn leads to individual audit partners who are more confident in their individual audit styles.
We make several contributions to the literature. First, our results contribute to the audit style literature. The US audit market differs significantly from other countries in market competition and litigation risk, and to the extent that these country-level institutional factors affect how individual auditors exert influence on audit outcomes, the results of other audit partner style studies may not be readily generalizable to the United States. Important institutional differences between the United States and the rest of the world, especially emerging market countries, provide an opportunity to increase our knowledge about the incremental importance of audit partners over firms and offices in influencing financial statement comparability.8 Prior to our study there was little evidence about how engagement auditor style effects might vary among institutional environments. In addition, studying the US audit market is important in its own right because it accounts for approximately one-third of the global audit market (SelectUSA, 2018).
Second, we corroborate the PCAOB's determination that Rule 3211 would provide transparency about who performs the audit and promote accountability for audit failures. Rule 3211 is intended to help investors, audit committee members and other financial statement users track audit partner reputation over time. Our results suggest that Rule 3211 may indeed provide decision-useful information for users. Since partners have an audit style that extends across engagements, knowing the partner's other engagements may provide audit committees with additional context in the choice of engagement partner and may help investors evaluate the audit quality of the financial statements.
Relatedly, we corroborate the PCAOB's intuition that the role of the engagement partner “is of singular importance” to the outcome of the audit (PCAOB, 2011). As PCAOB Rule 3211 became effective for all public company audits on or after January 31, 2017, the literature examining partner effects on audit outcomes in the United States is still developing. We add to evidence of the importance of the audit partner from Burke et al. (2018), Downey et al. (2019) and Cunningham et al. (2019), among others, by providing evidence that newly available partner identification information is useful in assessing how partners influence the comparability of audited financial statements.
2 BACKGROUND AND HYPOTHESIS DEVELOPMENT
Regulators have long recognized the benefits of comparability between financial statements. The FASB identifies comparability as an enhancing, qualitative characteristic that makes information about a reporting entity more useful because it allows comparison across entities (FASB, 2010). The International Accounting Standards Board similarly argues that high quality global accounting standards are necessary to provide financial market participants with comparable financial statements and help them make economic decisions (IASC Foundation, 2005).
Complementing the practical recognition of the importance of financial statement comparability among financial statement users and stakeholders is a growing body of empirical research regarding financial statement comparability. Recent research suggests that financial statement comparability may be influenced by audit firm “style” that arises from the unique internal working rules that audit firms adopt in implementing auditing standards and enforcing GAAP (Francis et al., 2014). Related evidence also suggests that audit offices have in-office rules and latitude to determine how to apply audit standards and accounting regulations to their local clients (Kawada, 2014). Office-level studies indicate that local office “style” has an incremental effect on comparability over audit firms (Kawada, 2014).9
Although prior evidence suggests that systematic differences in audit approaches and interpretation and enforcement of accounting standards among audit firms and audit offices influence financial statement comparability, audits are performed and overseen by individual engagement teams and partners. Prior research does find that audit outcomes are associated with individual auditors’ ability and incentives (Nelson & Tan, 2005; Nelson, 2009; Stice et al., 2022). Recently, several studies carried out in non-US settings provide evidence suggesting that individual signing auditors (often senior managers or partners) have unique styles, which systematically affect the way that they apply accounting standards (Chen et al., 2020; Li et al., 2021; Jiu et al., 2020; Shi et al., 2021).10 However, because of institutional differences between the United States and other jurisdictions, it is unclear whether a partner style effect distinct from audit firms and offices may exist in the United States.
Prior research suggests that the elements that have been hypothesized to foster style—such as individual auditors’ unique risk preferences (Bik & Hooghiemstra, 2018), ethical orientation (Baïada-Hirèche & Garmilis, 2016), education (Che et al., 2018; Gul et al., 2013) and judgment and decision making (Thorne et al., 2003)—may be influenced by regulatory and institutional context. Comparing the United States and Canada, Thorne et al. (2003) determine that the stronger regulations and more litigious environment of the United States better encourages auditors to employ moral reasoning to identify solutions. Others have found that the ethical judgment of auditors in the United States is more conservative than that of French auditors, due to the US's stricter disciplinary system (Baïada-Hirèche & Garmilis, 2016) and that auditors in common law countries (reflecting an environment of higher litigation risk) are more likely to conform with firm policies and methodology that ensure consistency of audits for global audit firms (Bik & Hooghiemstra, 2018). The language of accounting standards (Hronsky & Houghton, 2001) and the level of auditor education (Michas, 2011; Che et al., 2018; Gul et al., 2013), both of which are influenced or determined by regulation at the country level, have also been found to influence auditor decision making. In short, differences in regulatory and institutional settings may affect audit style through their influence on the decision-making patterns of partners (Brown et al., 2014; Choi & Wong, 2007; Frost & Ramin, 1996; Van der Plaats, 2000). Institutional and regulatory differences between the United States and the settings of other studies raise the question of whether a partner style effect is observable in the United States with its highly developed regulatory and litigious setting or whether these results are dependent upon contexts in which regulatory oversight is relatively weaker.
It is unclear how institutional setting might influence auditor style. For example, weaker regulatory oversight might allow greater flexibility in exercising judgment and decision making for partners, which would argue for greater style effects in weaker regulatory settings. The US audit market has the strictest independence rules in the world,11 and the United States has well developed investor-protection and litigation systems. Nevertheless, prior research suggests that auditors in these environments are more likely to conform strictly with firm and regulatory policies (e.g., Bik & Hooghiemstra, 2018; Thorne et al., 2003), which would dampen a partner style effect. By comparison, China and Taiwan, where most partner studies have been carried out and where evidence for a partner effect is most prominent, have weaker institutional protections (Firth et al., 2012; He et al., 2016; Hsu et al., 2018; Liao et al., 2012; Simunic & Wu, 2009).12 Consistent with this view, evidence regarding the effect of partner signature requirements on audit quality are notably mixed in stronger regulatory environments, suggesting that a strong institutional setting may dampen the effect of individual auditors on audit outcomes.13
On the other hand, higher education requirements and experience associated with practicing in more highly regulated environments may lead a to greater style effect in more regulated environments. Evidence consistent with audit partner style effects in strong regulatory settings includes recent evidence from the United Kingdom, which suggests that even in a highly regulated environment, partner effects explain more of the variation in audit quality than the combined effect of firms and offices (Cameran et al., 2020). Better educated partners may have higher financial statement comparability because they have more insights into the audit process and a stronger grasp on auditing standards (Wang et al., 2015), which allows them to gravitate toward the industry norm.14 Similarly, professional experience is key to forming auditor style because it facilitates development of mental models for understanding and applying evidence systematically during the audit process (Bonner & Lewis, 1990).
Based on the existing empirical evidence of partners’ influence on financial statement outcomes, in general and with regard to comparability, we may find a partner style effect in a US audit setting; however, for all the reasons noted previously, the outcome is far from clear. It is also possible that a partner style effect may not be affected by institutional setting at all. Because our setting departs from prior studies in ways that are likely to influence partner style, our results, regardless of direction, contribute substantially to current knowledge. Formally, our hypothesis, stated in the null, is:
HYPOTHESIS 1.Financial statement comparability of companies audited by the same audit partner will be no different than the financial statement comparability of companies audited by different audit partners, controlling for the determinants of financial statement comparability including audit firm and audit office.
3 RESEARCH DESIGN, DATA SOURCES AND SAMPLE SELECTION
3.1 Measures of financial statement comparability
Prior research investigating financial statement comparability most often analyzes either cross-sectional similarities in accruals or earnings covariance over time (e.g., Joos & Lang, 1994; Liao et al., 2012). We adopt both of these measures to assess earnings comparability following Francis et al. (2014). In addition, we use an audit partner fixed effect model, comparable to that put forth by Bertrand and Schoar (2003), to test for earnings similarities within audit partner clienteles.
3.2 Sample selection and data sources
We construct our sample using data from the PCAOB, Audit Analytics and Compustat. Table 1 presents our sample selection procedures. We first obtain 53,173 partner-firm-years with clients’ CIK number information from the PCAOB Audit Partner (AP) Filings for the period 2016 to 2020. We next exclude 27,688 observations for investment companies, employee benefit plans and foreign firms. With the remaining 25,485 partner-firm-years, we form 1,086,415 unique firm-pairs in which both firms have the same fiscal year and industry classification based on two-digit SIC code. We then remove 488,900 observations with insufficient information to calculate accrual similarity, 300,103 observations without enough information to calculate control variables, and 63,600 observations not audited by Big 4 firms. We limit our focus in primary tests to Big 4 firms because prior studies find that financial statement comparability is attributable to Big 4 firms (Francis et al., 2014).18 We remove 95,547 observations that lack information to calculate comovement variables. Our final sample for the accrual analyses includes 233,812 observations, whereas the comovement sample includes 138,265 observations.
| Firm-pairs in the same industry | 1,086,415 |
| Less: observations missing information to calculate accrual similarity | (488,900) |
| Less: observations missing information to calculate control variables | (300,103) |
| Less: firm-pairs audited by non-Big 4 firm | (63,600) |
| Sample for accrual similarity tests | 233,812 |
| Less: observations missing information to calculate earnings comovement | (95,547) |
| Sample for earnings comovement tests | 138,265 |
Table 2, Panel A reports descriptive statistics of our variables. The variable of interest, SameB4Partner, is equal to one for about 0.3% of the observations in the sample. The variable Sameoffice_DiffPartner is equal to one for approximately 2.0% of the observations in the sample, and the variable Samefirm_Diffoff_DiffPart is equal to one for approximately 26.3% of the observations in the sample. For the Accrual Quality metrics of Equation 4, the mean (median) difference in total accruals (Tacc_Diff) between firm-pairs is 8.8% (5.2%) of total assets. The mean (median) difference in discretionary accruals (Dacc_Diff) is 13.5% (8.5%). These statistics are similar to the mean (median) reported in Francis et al. (2014). Turning to ECOMP_COV, the mean is 0.175, slightly greater in magnitude than the sample average of 0.114 reported by Francis et al. (2014).
| Panel A: Summary statistics | |||||
|---|---|---|---|---|---|
| Variable name (N = 233,812) | Mean | Standard deviation | 25th percentile | Median | 75th percentile |
| Tacc_Diff | 0.088 | 0.103 | 0.021 | 0.052 | 0.114 |
| Dacc_Diff | 0.135 | 0.144 | 0.034 | 0.085 | 0.187 |
| ECOMP_COV (138,265) | 0.175 | 0.179 | 0.058 | 0.175 | 0.175 |
| SameB4Partner | 0.003 | 0.053 | 0.000 | 0.000 | 0.000 |
| Sameoffice_DiffPartner | 0.020 | 0.139 | 0.000 | 0.000 | 0.000 |
| Samefirm_Diffoff_DiffPart | 0.263 | 0.440 | 0.000 | 0.000 | 1.000 |
| Tenure_Diff | 2.351 | 1.076 | 1.609 | 2.398 | 3.045 |
| Tenure_Min | 2.174 | 0.730 | 1.609 | 2.303 | 2.708 |
| IndustrySpec | 0.716 | 0.451 | 0.000 | 1.000 | 1.000 |
| Tacc_Min | −0.114 | 0.091 | −0.148 | −0.088 | −0.054 |
| Dacc_Min | −0.088 | 0.093 | −0.121 | −0.064 | −0.035 |
| Size_Diff | 1.690 | 1.295 | 0.662 | 1.410 | 2.424 |
| Size_Min | 6.814 | 1.606 | 5.689 | 6.855 | 7.994 |
| Lev_Diff | 0.237 | 0.268 | 0.044 | 0.171 | 0.344 |
| Lev_Min | 0.135 | 0.162 | 0.000 | 0.062 | 0.245 |
| MB_Diff | 8.616 | 24.439 | 0.341 | 1.877 | 5.858 |
| MB_Min | 2.275 | 2.261 | 0.855 | 1.793 | 2.968 |
| CFO_Diff | 0.143 | 0.198 | 0.027 | 0.072 | 0.168 |
| CFO_Min | −0.027 | 0.225 | −0.015 | 0.052 | 0.081 |
| LossProb_Diff | 0.299 | 0.289 | 0.063 | 0.188 | 0.500 |
| LossProb_Min | 0.189 | 0.283 | 0.000 | 0.063 | 0.250 |
| Std_Sales_Diff | 0.061 | 0.053 | 0.011 | 0.044 | 0.114 |
| Std_Sales_Min | 0.015 | 0.013 | 0.003 | 0.013 | 0.029 |
| Std_CFO_Diff | 0.064 | 0.059 | 0.007 | 0.047 | 0.131 |
| Std_CFO_Min | 0.032 | 0.024 | 0.010 | 0.027 | 0.056 |
| Std_SalesGr_Diff | 0.461 | 1.238 | 0.033 | 0.088 | 0.256 |
| Std_SalesGr_Min | 0.147 | 0.273 | 0.047 | 0.080 | 0.142 |
| RET_COV (138,265) | 0.217 | 0.304 | 0.000 | 0.009 | 0.408 |
| CFO_COMP_COV (138,265) | 0.177 | 0.242 | 0.017 | 0.078 | 0.231 |
| ProductSimilarity (23,844) | 0.070 | 0.059 | 0.023 | 0.053 | 0.106 |
| AuditFee_Diff | 0.147 | 0.216 | 0.017 | 0.063 | 0.173 |
| AuditFee_Min | −0.154 | 0.493 | −0.153 | 0.000 | 0.000 |
| ReportLag_Diff (108,397) | 0.121 | 0.105 | 0.039 | 0.095 | 0.182 |
| ReportLag_Min (108,397) | 0.602 | 0.071 | 0.560 | 0.615 | 0.651 |
| Panel B: Correlations | ||||||
|---|---|---|---|---|---|---|
| Variables | (1) | (2) | (3) | (4) | (5) | (6) |
| (1) SameB4Partner | 1.00 | |||||
| (2) Sameoffice_DiffPartner | −0.01* | 1.00 | ||||
| (3) Samefirm_Diffoff_DiffPart | −0.03* | −0.08* | 1.00 | |||
| (4) Tacc_Diff | −0.01* | 0.03* | −0.02* | 1.00 | ||
| (5) Dacc_Diff | −0.02* | 0.01* | −0.02* | 0.55* | 1.00 | |
| (6) ECOMP_COV | 0.03* | 0.00 | 0.02* | −0.01 | −0.01* | 1.00 |
- * Pearson correlation significant at the 0.01 level.
Panel B reports the correlations between the variables of interest and other key variables. The correlations between SameB4Partner and both of the accruals-difference metrics (Tacc_Diff and Dacc_Diff) are significantly negative. This result corroborates our prediction that firm-pairs audited by the same Big 4 partner have more similar accruals, implying that the earnings are more comparable to each other as compared with a firm-pair not audited by the same Big 4 partner. Panel B also reports a significantly positive correlation between SameB4Partner and ECOMP_COV, providing univariate evidence that firm-pairs audited by the same Big 4 partner have a higher degree of earnings comovement than firm-pairs audited by different partners.
4 EMPIRICAL RESULTS
4.1 Audit partner style and financial statement comparability
4.1.1 Accrual differences
Table 3 reports the estimation of firm-pair differences for total and discretionary accruals. Columns 1 and 2 present the estimation results for the test of H1 using Tacc_Diff and Dacc_Diff as the dependent variables. We find a significantly negative coefficient on SameB4Partner for Tacc_Diff (β1 = −0.007, p value < 0.01) and Dacc_Diff (β1 = −0.017, p value < 0.01), consistent with greater similarity in the accruals structure for firm-pairs audited by the same Big 4 audit partner. The signs of the control variable coefficients are generally as expected. Accr_Min, CFO_Min and Size_Min have negative coefficients and are comparable to those reported in Francis et al. (2014). These results suggest that firm pairs are more likely to be similar when they are at the minimum level of accruals, cash flow and firm size.
| (1) | (2) | (3) | |
|---|---|---|---|
| Variables | Tacc_Diff | Dacc_Diff | ECOMP_COV |
| SameB4Partner (β1) | −0.007*** | −0.017*** | 0.107*** |
| (2.83) | (5.16) | (6.82) | |
| Sameoffice_DiffPartner (β2) | 0.001 | −0.007*** | 0.008* |
| (0.71) | (3.65) | (1.65) | |
| Samefirm_Diffoff_DiffPart (β3) | −0.001* | −0.001 | 0.007*** |
| (1.65) | (0.93) | (4.65) | |
| Tenure_Diff | −0.000 | −0.000 | −0.007*** |
| (1.38) | (0.05) | (10.38) | |
| Tenure_Min | −0.001*** | −0.001*** | −0.000 |
| (6.65) | (3.29) | (0.29) | |
| IndustrySpec | −0.000 | −0.002** | −0.002 |
| (1.27) | (2.49) | (1.11) | |
| Accr_Min | −0.744*** | −0.459*** | |
| (287.16) | (123.47) | ||
| Size_Diff | 0.001*** | −0.001*** | −0.004*** |
| (7.26) | (3.47) | (6.15) | |
| Size_Min | −0.001*** | −0.003*** | −0.002*** |
| (6.02) | (9.62) | (3.14) | |
| Lev_Diff | −0.002** | 0.024*** | 0.005* |
| (2.53) | (20.85) | (1.83) | |
| Lev_Min | −0.011*** | 0.039*** | 0.008 |
| (8.53) | (18.46) | (1.53) | |
| MB_Diff | −0.000*** | −0.000*** | −0.000** |
| (4.99) | (14.95) | (2.31) | |
| MB_Min | −0.001*** | −0.002*** | 0.001*** |
| (18.15) | (19.58) | (3.61) | |
| CFO_Diff | 0.004 | 0.078*** | |
| (1.61) | (24.54) | ||
| CFO_Min | −0.078*** | 0.026*** | |
| (30.60) | (7.90) | ||
| LossProb_Diff | −0.008*** | 0.014*** | −0.017*** |
| (12.62) | (13.57) | (6.02) | |
| LossProb_Min | −0.049*** | −0.037*** | 0.008** |
| (45.47) | (22.01) | (2.25) | |
| Std_Sales_Diff | 0.011*** | 0.044*** | −0.006 |
| (3.77) | (8.38) | (0.43) | |
| Std_Sales_Min | 0.174*** | 0.070*** | 0.044 |
| (15.75) | (3.22) | (0.70) | |
| Std_CFO_Diff | −0.010*** | −0.004 | −0.017 |
| (4.30) | (1.10) | (1.46) | |
| Std_CFO_Min | −0.029*** | 0.054*** | −0.004 |
| (5.17) | (5.21) | (0.12) | |
| Std_SalesGrth_Diff | 0.002*** | 0.001*** | −0.003*** |
| (9.51) | (4.72) | (4.89) | |
| Std_SalesGrth_Min | 0.018*** | −0.004** | −0.000 |
| (14.17) | (2.56) | (0.10) | |
| RET_COV | −0.004 | ||
| (0.94) | |||
| CFO_COMP_COV | 0.024*** | ||
| (4.68) | |||
| Constant | 0.013*** | 0.088*** | 0.231*** |
| (8.04) | (34.76) | (32.50) | |
| Industry fixed effect | Yes | Yes | Yes |
| Year fixed effect | Yes | Yes | Yes |
| Adjusted R2 | 0.551 | 0.293 | 0.095 |
| Num. of firm-pairs | 233,812 | 233,812 | 138,265 |
| Test of coefficient | |||
| β1 − β2 | −0.008** | −0.023** | 0.104*** |
| β1 − β3 | −0.006* | −0.017*** | 0.103*** |
- This table reports the estimation of firm-pair differences for total and discretionary accruals (columns 1 and 2) and for earnings comovement (column 3). Industry and year fixed effects are included. Robust t-statistics are reported in parentheses. ***, ** and * indicate significance levels at 1, 5 and 10%, respectively (two-tailed). All variables are defined in the appendix.
The results also provide evidence that audit partners have an influence on financial statement comparability over and above that of audit firms and audit offices. While we find significantly negative coefficients on SameB4Partner in columns 1 and 2, the coefficient of Sameoffice_DiffPartner is only significantly negative in column 2 (β2 = −0.007, p value < 0.01). Likewise, the coefficient of Samefirm_Diffoff_DiffPart is significantly negative only in column 1 (β3 = −0.001, p value < 0.1). Moreover, the average effect of audit partner style on the similarity of total accruals is more than twice as large as the average effect of office style or firm style. In column 2, the average effect of audit partner style on Dacc_Diff is 1.7% of lagged total assets, whereas the average effect of audit office style on Dacc_Diff is 0.7%. The average effect of Samefirm_Diffoff_DiffPart is not statistically significant. Results for Tacc_Diff are similar. Finally, when we perform a test of coefficient differences between SameB4Partner, Sameoffice_DiffPartner and Samefirm_Diffoff_DiffPart, we find that β1 is significantly smaller than β2 and β3. These results provide evidence consistent with H1 and indicate that audit partner style dominates the audit office and audit firm style effects.
Our results are also economically significant. For instance, having the same Big 4 audit partner leads to a decrease of 12.59% (0.017/0.135) from the mean value of difference in total accruals for firms with the same audit partner. Performing the test of economic significance suggested by Chen et al. (2020), we find that the mean firm in our study has profitability ((sales revenues − COGS)/lagged total assets) of 16.45%. Therefore, the estimated effect of the audit partner on change in gross profitability due to a 1.7% shift in abnormal accruals is a 10.3% change in profitability for the average firm in our sample (1.7/16.45). That we find a significant partner style effect in the highly regulated and litigious US setting suggests that strong institutions may play a role in audit partner style, particularly because some evidence in non-US settings casts doubt on a strong engagement signing auditor effect in weaker institutional environments. For example, Chen et al. (2020) do not find a significant difference between the coefficient of the same engagement signing auditor and the coefficient of the same audit office but different signing auditor for total accruals and only marginal significance for abnormal accruals.19 One explanation for the differences we observe is that greater levels of experience and education for audit partners in the United States lead to the development of stronger audit style. Our results suggest that rigorous education requirements and experience associated with practicing in the United States lead to a greater style effect, consistent with the idea that US partners have education and experience that allows them to align closely with industry norms. These results also corroborate the PCAOB's observation that audit partners individually contribute to the outcome of the audit apart from their offices and firms.
4.1.2 Earnings covariation
Table 3, column 3 reports the test of audit partner style effects using earnings covariation (ECOMP_COV). Again, we control for Sameoffice_DiffPartner and Samefirm_Diffoff_DiffPart. The coefficient of SameB4Partner is positive and statistically significant (0.107, p value < 0.01). In addition, while ECOMP_COV is significantly related to Sameoffice_DiffPartner (0.008, p value < 0.1) and Samefirm_Diffoff_DiffPart (0.007, p value < 0.01), the magnitudes of the coefficients are far smaller than the coefficient of SameB4Partner. These results are consistent with our evidence in the accruals model that the individual audit partner effect is greater than the unique effect of the specific audit firm or audit office. In a test of difference of coefficients, both the difference between SameB4Partner and Sameoffice_DiffPartner and Samefirm_DiffOff_DiffPart are statistically significant at the 1% level. These results are consistent with the reasoning presented that in regulatory environments that foster a rigorous accounting education and high-quality experience, this training leads to a more defined audit style. Thus, firms audited by the same Big 4 partner have a higher degree of earnings comovement than firm-pairs audited by different partners.
4.1.3 Auditor fixed effect estimation
Table 4 presents the results for the audit partner fixed effects test. The R2 of the individual auditor fixed-effect models (columns 2 and 4), are significantly higher than the R2s of the base models in which we include audit firm and audit office fixed effects but not audit partner fixed effects. For example, when comparing column 3 results with column 4 results, the base model explains 24.3% of the variation of Dacc; and our audit partner fixed effect model explains 28.9% of the variation of Dacc. The increase in explanatory power is 18.9%. Similarly, the F-tests on each of the partner fixed effect regressions are large (i.e., F-values substantially exceed the relevant critical values) and allow us to reject the null hypothesis that the partner fixed effects are zero. Overall, we find that financial statement comparability is most influenced by the audit partner's decision, consistent with Cameran et al.’s (2020) evidence in the United Kingdom.20
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| Variables | Tacc | Tacc | Dacc | Dacc |
| SIZE | 0.068*** | 0.076*** | 0.009 | −0.001 |
| (11.90) | (8.98) | (1.45) | (−0.13) | |
| MTB | 0.003*** | 0.003*** | 0.001** | 0.001* |
| (4.43) | (3.81) | (2.17) | (1.65) | |
| LEV | −0.166*** | −0.176*** | −0.103** | −0.122** |
| (−7.84) | (−6.46) | (−2.56) | (−2.29) | |
| CFO | 1.084*** | 0.976*** | 0.308*** | 0.265*** |
| (80.29) | (60.37) | (8.95) | (6.75) | |
| LossProb | 0.324*** | 0.310*** | −0.018 | −0.052* |
| (12.09) | (7.91) | (−0.91) | (−1.78) | |
| StdSales | −0.000 | −0.000 | −0.000 | −0.000 |
| (−1.06) | (−1.34) | (−0.60) | (−0.42) | |
| StdCFO | −0.000*** | −0.000* | −0.000 | −0.000 |
| (−3.67) | (−1.82) | (−1.47) | (−0.62) | |
| StdSalesGrth | −0.000*** | −0.000*** | −0.000** | −0.000* |
| (−4.71) | (−3.31) | (−2.14) | (−1.91) | |
| Constant | −1.138*** | −0.824 | 0.149 | −0.296 |
| (−4.42) | (−0.86) | (0.90) | (−0.73) | |
| Industry and year fixed effect | Yes | Yes | Yes | Yes |
| Audit firm fixed effect | Yes | Yes | Yes | Yes |
| Audit office fixed effect | Yes | Yes | Yes | Yes |
| Audit partner fixed effect | No | Yes | No | Yes |
| Adjusted R2 | 0.493 | 0.517 | 0.243 | 0.289 |
| F-statistics | 61.14 | 29.67 | 22.19 | 9.52 |
| Observations | 12,488 | 12,488 | 12,488 | 12,488 |
- This table indicates that the results for the audit partner fixed effects test. We compare the R2 of specifications with audit partner fixed effects (columns 2, 4 and 6) and without audit partner fixed effects (columns 1, 3 and 5). The sample size represents the sample of 25,485 partner-firm-years for our primary tests. We drop 12,997 partner-firm-years with missing observations to calculate variables and controls. Industry and year fixed effects are included. Robust t-statistics are reported in parentheses. ***, ** and * indicate significance levels at 1, 5 and 10%, respectively (two-tailed). All variables are defined in the appendix.
4.2 Additional tests
4.2.1 Influence of partner experience and expertise
We perform two cross-sectional tests to examine whether the stronger influence of audit partners compared with audit offices varies with audit office and audit partner characteristics. Professional experience is an important factor in consistently implementing accounting standards (Bonner & Lewis, 1990), and prior research finds that audit partner (Chi et al., 2017) and audit office experience (Anantharaman & Wans, 2019) influence audit quality. Therefore, we examine the influence of audit partner experience relative to audit office experience on financial statement comparability. We estimate Equation 4 after (1) replacing SameB4Partner with SameB4Partner_HighExp and SameB4Partner_LowExp and (2) replacing Sameoffice_ DiffPartner with Sameoffice_DiffPartner_HighExp and Sameoffice_DiffPartner_LowExp. We measure partner experience as the number of years since the partner became a partner (Chi et al., 2017) and office experience as the number of years that office has performed the client's audit (Anantharaman & Wans, 2019).21 A smaller (larger) coefficient on SameB4Partner_HighExp relative to Sameoffice_DiffPartner_HighExp is consistent with partner experience having a stronger influence on accruals comparability (earnings covariance) than office experience. Table 5, Panel A presents the results. The coefficient of SameB4Partner_HighExp is significantly negative (−0.006, p value < 0.05) in column 1, whereas the coefficient of Sameoffice_DiffPartner_HighExp is not statistically significant. The test of differences in coefficients suggests that the coefficient of SameB4Partner_HighExp is significantly smaller than the coefficient of Sameoffice_DiffPartner_HighExp and indicates that audit partner experience dominates audit office experience for accruals similarity. In column 2, the coefficient of SameB4Partner_HighExp is significantly negative (−0.015, p value < 0.01) but larger than the coefficient of Sameoffice_DiffPartner_HighExp (−0.006, p value < 0.05), and the test of differences in coefficients is significant. In column 3, the coefficient of SameB4Partner_HighExp is significantly positive (0.101, p value < 0.01) while the coefficient of Sameoffice_DiffPartner_HighExp is not statistically significant. Moreover, the test of differences in coefficients suggests that the clients of experienced partners exhibit higher earnings comovement than the clients of experienced audit offices. Overall, the evidence partially suggests that experience at the partner level is more influential than experience at the office level on client financial statement comparability.
| Panel A: Cross-Sectional Test: Auditor Experience | |||
|---|---|---|---|
| (1) | (2) | (3) | |
| Variables | Tacc_Diff | Dacc_Diff | ECOMP_COV |
| SameB4Partner_HighExp (β1) | −0.006** | −0.015*** | 0.101*** |
| (2.32) | (3.99) | (5.83) | |
| SameB4Partner_LowExp | −0.008 | −0.024*** | 0.132*** |
| (1.61) | (3.67) | (3.57) | |
| Sameoffice_DiffPartner_HighExp (β2) | −0.001 | −0.006*** | 0.010 |
| (0.06) | (2.70) | (1.49) | |
| Sameoffice_DiffPartner__LowExp | 0.002 | −0.006** | 0.005 |
| (0.90) | (2.32) | (0.89) | |
| Controls | Yes | Yes | Yes |
| Industry FE | Yes | Yes | Yes |
| Year FE | Yes | Yes | Yes |
| Adjusted R2 | 0.055 | 0.291 | 0.023 |
| Num. of firm-pairs | 233,812 | 233,812 | 138,265 |
| Test of coefficient | |||
| β1 − β2 | 0.005*** | 0.009*** | 0.091*** |
| Panel B: Cross-Sectional Test: Auditor Expertise | |||
|---|---|---|---|
| (1) | (2) | (3) | |
| Variables | Tacc_Diff | Dacc_Diff | ECOMP_COV |
| SameB4Partner_Expert (β1) | −0.004* | −0.016*** | 0.125*** |
| (1.74) | (4.20) | (5.63) | |
| SameB4Partne_NonExpert | −0.010*** | −0.018*** | 0.088*** |
| (2.29) | (3.19) | (3.96) | |
| Sameoffice_DiffPartner_Expert (β2) | −0.005 | −0.022 | −0.028 |
| (0.94) | (1.09) | (0.63) | |
| Samefirm_DiffPartner_NonExpert | −0.003* | −0.005*** | 0.011** |
| (1.92) | (2.25) | (2.14) | |
| Controls | Yes | Yes | Yes |
| Industry FE | Yes | Yes | Yes |
| Year FE | Yes | Yes | Yes |
| Adjusted R2 | 0.055 | 0.291 | 0.023 |
| Num. of firm-pairs | 233,812 | 233,812 | 138,265 |
| Test of coefficient | |||
| β1 − β2 | 0.009 | 0.038** | 0.153*** |
- This table reports cross-section tests for partner and office experience and expertise. Dependent variables are total and discretionary accruals (columns 1 and 2) and earnings comovement (column 3). Industry and year fixed effects are included. Robust t-statistics are reported in parentheses. ***, ** and * indicate significance levels at 1, 5 and 10%, respectively (two-tailed). Variables are defined in the appendix.
We follow a similar process to evaluate whether the influence of partner industry expertise is stronger than the influence of audit office industry expertise on clients’ financial statement comparability. We create indicator variables for firm pairs whose partners are industry experts (SameB4Partner_Expert, SameB4Partner_NonExpert) to replace SameB4Partner in Equation 4. We also create indicator variables for audit offices that are industry experts (Sameoffice_DiffPartner_Expert, Sameoffice_DiffPartner_NonExpert). We measure audit partner (office) expertise as the partner (office) with the highest number of clients (e.g., Chin et al., 2014; Chin & Chi, 2009) in each industry-year. A smaller (larger) coefficient on SameB4Partner_Expert relative to Sameoffice_DiffPartner_Expert is consistent with partner industry expertise having a stronger influence on accruals comparability (earnings covariance) than office industry experience. Table 5, Panel B presents the results. Tests of coefficient differences between SameB4Partner_Expert and Sameoffice_DiffPartner_Expert are significant in columns 2 and 3. Moreover, the coefficient of SameB4Partner_Expert (−0.016, p value < 0.01) is smaller in column 2 than the coefficient of Sameoffice_DiffPartner_Expert (−0.022, p value > 0.1), indicating that partner expertise has a stronger influence on accruals comparability than office expertise. Similarly, in column 3, the coefficient of SameB4Partner_Expert (0.125, p value < 0.01) is greater than the coefficient of Sameoffice_DiffPartner_Expert (−0.028, p value > 0.1). The results provide additional evidence of an audit partner style effect on financial statement comparability over and above that of the audit office or firm effect, consistent with the PCAOB's decision to require partner identification.
4.2.2 Partner education and experience
To this point, our results indicate a strong audit partner style effect in the highly regulated US market where accounting education standards are rigorous and signing partners have significant experience. We have suggested that the partner style effect in the United States may stem from US audit engagement partners’ high levels of experience and education. To provide more direct evidence, we collect data on audit partners’ experience and education from LinkedIn and other publicly available sources. We consider a partner to be experienced if the partner has industry expertise (Expert). A partner has higher levels of education if the partner has a master's degree (Master). We replace SameB4Partner in Equation 4 with two indicator variables: SameB4Partner_With and SameB4Partner_Without. SameB4Partner_With is an indicator variable equal to one if the firm-pairs are audited by the same Big 4 partner who has higher industry or educational experience, and zero otherwise. SameB4Partner_Without is an indicator variable equal to one if the firm-pairs are audited by same Big 4 partners that do not have higher levels of experience or education, and zero otherwise.
We expect more experienced and educated partners to have higher accruals similarity and earnings covariance. Consistent with these expectations, we find that partners that are industry experts and partners with master's degrees have significantly higher accruals similarity and earnings covariation than their less experienced and less educated peers. The results are presented in Table 6. In columns 1–2 when Tacc_Diff is the dependent variable, we find that partners that are industry experts (−0.004, p value < 0.1) and hold master's degrees (−0.003, p value < 0.1) have higher accruals similarity than auditors that do not. We find similar results in columns 3–4 when Dacc_Diff is the dependent variable. Turning to earnings covariation in columns 5–6, our results suggest that partners who are industry experts (0.125, p value < 0.01) and hold master's degrees (0.127, p value < 0.01) have higher earnings covariance than auditors who are not industry experts. In a test of difference of coefficients, the difference between SameB4Partner_With and SameB4Partner_Without are statistically significant in five of the six columns. Overall, these results provide evidence consistent with the argument that higher education and experience increase financial statement comparability, helping to explain why we observe a partner style effect in the US market.
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Variables | Tacc_Diff | Dacc_Diff | ECOMP_COV | |||
| Expert | Master | Expert | Master | Expert | Master | |
| SameB4Partner_With (β1) | −0.004* | −0.003* | −0.017*** | −0.005* | 0.125*** | 0.127*** |
| (1.72) | (1.68) | (4.23) | (1.72) | (5.63) | (4.76) | |
| SameB4Partner_Without (β2) | −0.010*** | −0.012** | −0.018*** | −0.028*** | 0.057** | 0.055*** |
| (3.27) | (2.49) | (3.21) | (4.96) | (2.25) | (2.65) | |
| Sameoffice_DiffPartner | 0.001 | 0.001 | −0.006*** | −0.007*** | 0.007 | 0.007 |
| (0.70) | (0.69) | (3.64) | (3.64) | (1.63) | (1.61) | |
| Samefirm_Diffoff_DiffPart | −0.001 | −0.001 | −0.001 | −0.000 | 0.007*** | 0.007*** |
| (1.64) | (1.63) | (0.92) | (0.88) | (4.6) | (4.53) | |
| Test of coefficient | ||||||
| β1 − β2 | 0.006* | 0.009* | 0.001 | 0.023*** | 0.068*** | 0.072** |
| Controls | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry FE | Yes | Yes | Yes | Yes | Yes | Yes |
| Year FE | Yes | Yes | Yes | Yes | Yes | Yes |
| Adjusted R2 | 0.550 | 0.550 | 0.291 | 0.291 | 0.023 | 0.023 |
| Num. of firm-pairs | 233,812 | 233,812 | 233,812 | 233,812 | 138,265 | 138,265 |
- This table reports the results for the cross-sectional analyses of the influence of audit partner experience and education on financial statement comparability. Dependent variables are total and discretionary accruals (columns 1–2, 3–4) and earnings comovement (columns 5–6). Industry and year fixed effects are included. Robust t-statistics are reported in parentheses. ***, ** and * indicate significance levels at 1, 5 and 10%, respectively (two-tailed). All variables are defined in the appendix.
4.2.3 Non-Big 4 auditors
We also examine whether clients with the same non-Big 4 audit partner experience similar levels of comparability to clients with Big 4 audit partners. Che et al. (2020) provide evidence that Big 4 firms are better able to identify and recruit high-quality partners than non-Big 4 firms, and that Big 4 firms emphasize and provide better opportunities for learning to their partners compared with their non-Big 4 counterparts. These results are consistent with higher levels of experience and education for partners in Big 4 firms compared with partners of non-Big 4 firms. We expect that firms whose financial statements are audited by the same non-Big 4 partner will have lower levels of comparability than we have found in our Big 4 sample. In the sample of firms with non-Big 4 audit partners, we find that when two firms share the same non-Big 4 audit partner, there is no association with accruals comparability (untabulated). In addition, we do not find significant evidence of earnings comovement in the non-Big 4 sample. Therefore, in our sample of firms audited by partners with arguably lower levels of experience and education, we do not find evidence of a strong audit partner style effect. The results from these tests complement our primary tests and provide more evidence that education and quality/length of experience are factors influencing audit partner style.
4.3 Audit fees and report lag
To explore our prediction that audit partner style influences financial statement comparability, we also examine whether audit partner style is associated with similarities in audit effort among clients audited by the same partner. Unique risk tolerances can influence partners’ implementation of standards and internal working rules and affect the amount of effort they exert in testing. In examining comparability of financial reporting between audit firms, Johnston and Zhang (2021) report that a higher level of audit fee comovement is associated with a higher likelihood of financial statement similarity between audit clients. They also report that audit delay is positively associated with financial reporting similarity. Therefore, we consider whether audit fees and audit delay are a function of audit partners’ unique working style.
We calculate AuditFee_Diff and ReportLag_Diff as the absolute value of the difference of the natural logarithm of audit fees and audit report lag between two firms. We then estimate Equation 4 replacing Accrual_Diffijt with AuditFee_Diff and ReportLag_Diff. As expected, audit fees and audit report lag are more comparable between clients served by the same audit partner than by clients served by the same audit office or audit firm, but different partners. Our results (untabulated) are consistent with H1, suggesting that audit characteristics of clients with the same audit partner are more comparable than those for clients served by different audit partners, even if they are served by the same audit firm and/or audit office.22
4.4 Endogeneity concerns
We argue that firms with the same partner will have higher financial statement comparability because of their audit partner's individual auditing style. A plausible alternative explanation is that similarities in firms’ economic fundamentals influence both their choice of auditor and financial statement comparability. While we cannot completely rule out this possibility, we conduct several tests to mitigate this concern, including audit partner switching, controlling for economic similarity between clients, placebo tests, and entropy balancing tests.23 As more data become available—especially data related to audit partners switching firms, offices and clients—it will be possible to construct additional tests with greater identification power.
4.4.1 Audit partner switches
We perform tests on the effects of changes in audit partners on financial statement comparability over the 4 years in our sample. If our results accurately reflect practice, and individual audit partners imprint their own audit style on the financial statements of their clients, we should observe that two firms with different audit partners who switch to the same audit partner have more comparable financial statements following the switch. We would also expect the opposite to be true: where two firms share the same audit partner but one of the firms switches to a different audit partner, those firm-pairs’ financial statements should not be comparable following the switch.
We use a change model to test the effects of changes in audit partners on financial statement comparability. The change model setting provides more robust and reliable evidence on the influence of individual audit partner style on financial statement comparability. We identify 35 firms that switch to the same audit partner from a different partner in our sample, and 30 firms that switch to a different partner from the same partner during our sample period. We create an indicator variable that captures the two possible outcomes: (1) a change from the same partner to a different partner (SwitchToDiffBig4) and (2) a change from different partners to the same partner (SwitchToSameBig4). These two indicator variables replace SameB4Partner in Equation 4.24
Table 7 corroborates our previous evidence that audit partners actively influence comparability of financial reporting. The coefficient on SwitchToSameBig4 is significantly negative for ∆Dacc_Diff (−0.037, p value < 0.05), indicating that the accruals of the switching firm become more comparable with the new partner's other clients when a pair of firms that had different audit partners in the previous year switch to having the same audit partner in the current year. The coefficient on SwitchToDiffBig4 is not significant in either model. Thus, we do not find evidence that switching to a different engagement auditor decreases the accrual comparability of firm-pairs that had the same audit partner in the previous year.
| (1) | (2) | |
|---|---|---|
| Variables | ∆Tacc_Diff | ∆Dacc_Diff |
| SwitchToSameBig4 (β1) | 0.015 | −0.037** |
| (1.11) | (2.12) | |
| SwitchToDiffBig4 (β2) | −0.017 | −0.004 |
| (1.24) | (0.15) | |
| SwitchToSameOffice | −0.008 | −0.018 |
| (1.09) | (1.64) | |
| SwitchToDiffOffice | 0.028*** | −0.002 |
| (2.97) | (0.18) | |
| SwitchToSameFirm | −0.008 | −0.035*** |
| (1.54) | (4.73) | |
| SwicthToDiffFirm | −0.004 | −0.021*** |
| (0.62) | (2.67) | |
| ∆Tenure_Diff | 0.002 | 0.009*** |
| (0.89) | (2.84) | |
| ∆Tenure_Min | 0.001 | 0.008*** |
| (1.00) | (4.05) | |
| ∆IndustrySpec | −0.003*** | −0.006*** |
| (2.96) | (4.92) | |
| ∆Accr_Min | −0.351*** | −0.466*** |
| (39.12) | (45.90) | |
| ∆Size_Diff | 0.018*** | 0.025*** |
| (6.93) | (7.34) | |
| ∆Size_Min | 0.049*** | 0.053*** |
| (14.12) | (11.65) | |
| ∆Lev_Diff | −0.012* | −0.000 |
| (1.93) | (0.05) | |
| ∆Lev_Min | 0.021** | −0.019 |
| (2.27) | (1.64) | |
| ∆MB_Diff | 0.000*** | 0.000 |
| (3.46) | (0.83) | |
| ∆MB_Min | −0.001*** | −0.001** |
| (3.60) | (2.11) | |
| ∆CFO_Diff | 0.109*** | 0.053*** |
| (15.88) | (6.21) | |
| ∆CFO_Min | −0.028*** | −0.120*** |
| (3.21) | (11.87) | |
| ∆LossProb_Diff | 0.004 | −0.018** |
| (0.71) | (2.16) | |
| ∆LossProb_Min | −0.054*** | −0.061*** |
| (4.83) | (4.42) | |
| ∆Std_Sales_Diff | −0.000*** | 0.000** |
| (6.75) | (2.14) | |
| ∆Std_Sales_Min | −0.000*** | 0.000*** |
| (4.29) | (4.49) | |
| ∆Std_CFO_Diff | 0.000*** | 0.000 |
| (3.74) | (0.86) | |
| ∆Std_CFO_Min | −0.000 | 0.000 |
| (0.12) | (0.44) | |
| ∆Std_SalesGrth_Diff | −0.005*** | −0.004*** |
| (4.44) | (3.60) | |
| ∆Std_SalesGrth_Min | 0.015*** | 0.005 |
| (2.90) | (0.84) | |
| Constant | −0.007*** | 0.013*** |
| (5.55) | (7.44) | |
| Industry Fixed Effect | Yes | Yes |
| Year Fixed Effect | Yes | Yes |
| Adjusted R2 | 0.145 | 0.119 |
| Num. of firm-pairs | 44,570 | 44,570 |
- This table reports the estimation of firm-pair audit partner switching tests. Dependent variables are the change of total and discretionary accruals. The variables of interest are SwitchToSameBig4 and SwitchToDiffBig4. The sample represents all firms with more than 1 year of observations to calculate audit partner changes between years. Industry and year fixed effects are included. Robust t-statistics are reported in parentheses. ***, ** and * indicate significance levels at 1, 5 and 10%, respectively (two-tailed). All variables are defined in the appendix.
Overall, the partner switching test provides additional evidence that the individual audit partner has a significant influence on the comparability of financial statements controlling for the influence of the audit firm and audit office.
4.4.2 Product similarity
Our results suggest that audit partners’ unique audit style contributes to comparability of financial statements between firms that share the same audit partner. An alternative view is that our evidence is attributable to economic similarities between the two companies rather than the active influence of the partner on the audit. To address this concern, we control for Hoberg and Phillips’ (2016) product similarity score in Equations 4 and 5. The results (untabulated) corroborate our previous evidence that audit partners contribute to financial statement comparability, over and above economic similarities related to product offerings.
4.4.3 Placebo test
Although the partner switching and product similarity tests provide convincing evidence that our results are not attributable to similarities between the audit firms for which an auditor partner works rather than due to partners’ active influence on audit engagements, we also conduct placebo tests to provide additional evidence consistent with our results. Our results, untabulated, reveal that none of the placebo partner coefficients are statistically significant. While the placebo test does not resolve the concern that the auditor-client pairs we observe in the data are nonrandom, this test paired with the product similarity test, provides comfort that our results are not attributable to similarities in the audit firms where partners work.
4.4.4 Entropy balancing
We attempt to alleviate endogeneity concerns related to observable differences between firms with and without the same Big 4 partner by using entropy balancing (Hainmueller, 2012). We find that in the entropy balanced sample, firms audited by the SameB4Partner have significantly greater similarity in accruals and greater covariance of earnings. On the other hand, the coefficients of Sameoffice_DiffPartner and Samefirm_Diffoff_DiffPart are not statistically significant. The results (untabulated) are consistent with H1.25
5 CONCLUSION
Using a sample of US audit partners at Big 4 audit firms, we find that two firms in the same industry-year who share the same audit partner have more comparable financial statements than two firms in the same industry-year who have different audit partners, even after considering the effects of audit firm and audit firm office. Specifically, we find that firm-pairs in the same industry-year with the same Big 4 audit partner exhibit greater cross-sectional similarity in accruals, while individual audit partner fixed effects explain a higher proportion of the variance in audited accruals than audit firm and office fixed effects alone. We also find that earnings comovement is more similar for firm-pairs audited by the same Big 4 audit partner. We interpret these results as evidence of audit partner style, which is likely a result of more rigorous educational requirements and greater experience at the audit partner level for firms in the United States. We provide cross-sectional tests consistent with this argument.
Interestingly, the results across our models indicate that audit partners are more important than audit firms and offices in the determination of financial statement comparability. Our tests of economic significance indicate that having the same Big 4 audit partner leads to a 12.59% decrease in abnormal accruals, compared with the mean, while the corresponding decrease for firm pairs audited by the same audit office is only 4.4%. We also provide evidence that partners’ unique working styles are associated with similarities in audit effort among their own clients. Our results are robust to industry and year fixed effects, as well as to auditor switching, placebo and self-selection tests, and to the inclusion of product similarity measures.
We corroborate the PCAOB's arguments for partner identification by providing evidence suggesting that individual audit partners are significantly related to the degree of comparability between firms’ financial statements, suggesting that audit partners are indeed “of singular importance” to the audit. Thus, we contribute to the literature examining the role of audit partners on financial statement comparability in a highly developed regulatory setting and contribute to the developing literature investigating the role of audit partners in a US setting. We find that newly available partner identification information is useful in assessing how partners affect the comparability of audited financial statements. In a departure from prior research, which finds only weak results for the influence of engagement auditors apart from senior signing auditors on financial statement comparability, we find a strong style effect for audit partners in the US setting. Our results provide evidence consistent with suggestions in prior research that greater education and experience influence a stronger individual audit style.
ACKNOWLEDGMENTS
We thank Thomas Bourveau, John Campbell, Ted Christensen (editor), Peter Chen, Lorien Stice-Lawrence, two anonymous reviewers and workshop participants at the 2021 American Accounting Association annual meeting for helpful comments and suggestions.
Appendix
Variable Definitions
| Main variables | |
|---|---|
| Tacc_Diff | The absolute value of the difference in total accruals (Tacc) between two firms in a pair. |
| Dacc_Diff | The absolute value of the difference in discretionary accruals (Dacc) between two firms in a pair. |
| ECOMP_COV | Earnings comovement of a firm-pair, measured as the R2 of the regression of Equation 3 over 16 consecutive quarters. |
| SameB4Partner | An indicator variable equal to one for firm-pairs with the same Big 4 partner, and zero otherwise. |
| Sameoffice_DiffPartner | An indicator variable equal to one for firm-pairs with the same audit office but different audit partners, and zero otherwise. |
| Samefirm_Diffoff_DiffPart | An indicator variable equal to one for firm-pairs with the same audit firm but different audit offices and different audit partners, and zero otherwise. |
| Tacc | Signed total accruals measured as the difference between income before extraordinary items and operating cash flow excluding cash flow from extraordinary items, scaled by lagged total assets. |
| Dacc | Signed discretionary accruals calculated using Jones (1991) model of discretionary accruals as modified by Kothari et al. (2005). |
| Control variables | |
|---|---|
| Tenure_Diff | The absolute value of the difference in audit partner tenure between two firms in a pair. |
| Tenure_Min | Minimum value of audit partner tenure in a firm-pair. |
| IndustrySpec | An indicator variable equal to one if an audit partner is an industry specialist, and zero otherwise. |
| Accr_Min | Minimum value of total, discretionary or absolute discretionary accruals in a firm-pair. |
| Size_Diff | The absolute value of the difference in natural logarithm of total assets in a firm-pair. |
| Size_Min | Minimum value of natural logarithm of total assets in a firm-pair. |
| Lev_Diff | The absolute value of the difference in debt-to-equity ratio in a firm-pair. |
| Lev_Min | Minimum value of debt-to-equity ratio in a firm-pair. |
| MB_Diff | The absolute value of the difference in market-to-book ratio in a firm-pair. |
| MB_Min | Minimum value of market-to-book ratio in a firm-pair. |
| CFO_Diff | The absolute value of the difference between operating cash flow scaled by lagged total assets in a firm-pair. |
| CFO_Min | Minimum value of operating cash flow scaled by lagged total assets in a firm-pair. |
| LossProb_Diff | The absolute value of the difference in loss probability in a firm-pair, measured as the proportion of fiscal quarters with negative income before extraordinary items in the previous 16 quarters. |
| LossProb_Min | Minimum value of loss probability in a firm-pair in the previous 16 quarters. |
| Std_Sales_Diff | The absolute value of the difference in standard deviation of sales over the past 16 quarters in a firm-pair. |
| Std_Sales_Min | Minimum value of standard deviation of sales over the past 16 quarters in a firm-pair. |
| Std_CFO_Diff | The absolute value of the difference in standard deviation of operating cash flow over the past 16 quarters in a firm-pair. |
| Std_CFO_Min | Minimum value of standard deviation of operating cash flow over the past 16 quarters in a firm-pair. |
| Std_Sales_Grth_Diff | The absolute value of the difference in standard deviation of sales growth over the past 16 quarters in a firm-pair. Sales growth is the change in total sales in year t divided by total sales in year t−1. |
| Std_Sales_Grth_Min | Minimum value of standard deviation of sales growth over the past 16 quarters in a firm-pair. |
| RET_COV | Return comovement between two firms in a pair, measured as the R2 of the following regression for firm i and j from the same industry over 16 consecutive quarters: |
| RETiq = α0+α1RETjq+εijq, where RET is the quarterly stock return. | |
| CFO_COMP_COV | Operating cash flow comovement between two firms in a pair, measured as the R2 of the following regression for firm i and j from the same industry over 16 consecutive quarters: CFOiq = α0+α1CFOjq+εijq |
| Control variables in fixed effect models | |
|---|---|
| Size | The natural logarithm of total assets. |
| MTB | Market to book ratio. |
| Lev | Total debt divided by total equity. |
| CFO | Operating cash flow before extraordinary items, scaled by lagged total assets. |
| LossProb | The proportion of fiscal quarters with negative income before extraordinary items in the previous 16 quarters. |
| StdSales | Standard deviation of sales over the previous 16 quarters. |
| StdCFO | Standard deviation of operating cash flow (CFO) over the previous 16 quarters. |
| StdSalesGrth | Standard deviation of growth of sales over the previous 16 quarters. |
| Variables in additional tests | |
|---|---|
| SwitchToSameBig4 | An indicator variable equal to one if a pair of firms had different Big 4 audit partners in the previous fiscal year and switched to the same Big 4 audit partner in the current year, and zero otherwise. |
| SwitchToDiffBig4 | An indicator variable equal to one if a pair of firms shared the same Big 4 audit partner in the previous fiscal year but switched to different Big 4 audit partners in the current year, and zero otherwise. |
| SwitchToSameOffice | An indicator variable equal to one if a pair of firms had different audit office and audit partners in the same firm in the previous fiscal year and switched to the same office in the firm in the current year, and zero otherwise. |
| SwitchToDiffOffice | An indicator variable equal to one if a pair of firms had the same audit office and different audit partners in the same firm in the previous fiscal year and switched to a different office in the same firm in the current year, and zero otherwise. |
| SwitchToSameFirm | An indicator variable equal to one if a pair of firms had different audit firms in the previous fiscal year and switched to the same audit firm in the current year, and zero otherwise. |
| SwitchToDiffFirm | An indicator variable equal to one if a pair of firms had the same audit firm in the previous fiscal year and switched to a different audit firm in the current year, and zero otherwise. |
| ProductSimilarity | Product similarity score for the company pair (Hoberg & Phillips, 2016). |
| AuditFee_Diff | The absolute value of the difference in log of audit fees in a firm-pair. |
| ReportLag_Diff | The absolute value of the difference in report lag in a firm-pair. |
| AuditFee_min | Minimum value of log of audit fees in a firm-pair. |
| ReportLag_min | Minimum value of report lag in a firm-pair. |
| SameNB4Partner | An indicator variable equal to one for firm-pairs with the same non-Big 4 partner, and zero otherwise. |
| SameB4Partner_With | An indicator variable equal to one if the firm-pair is audited by the same Big 4 partner who has higher industry or educational experience, and zero otherwise. |
| SameB4Partner_Without | An indicator variable equal to one if the firm-pair is audited by same Big 4 partner who does not have higher levels of experience or education, and zero otherwise. |
| SameB4Partner_Expert | An indicator variable equal to one if the firm-pair is audited by the same Big 4 partner who is an industry expert, and zero otherwise. Partner industry expert is defined the partner with the highest number of clients in the industry (SIC2). |
| SameB4Partner_NonExpert | An indicator variable equal to one if the firm-pair is audited by the same Big 4 partner that who is not an industry expert, and zero otherwise. |
| Sameoffice_DiffPartner_Expert | An indicator variable equal to one if the firm-pair is audited by the same industry-expert office but different partners after excluding the partner in question, and zero otherwise. Office industry expert is defined as the office with the highest number of clients in the industry by SIC2. |
| Sameoffice_DiffPartner_NonExpert | An indicator variable equal to one if the firm-pair is audited by the same office but different partners in an office that is not considered an industry expert after excluding the partner in question, and zero otherwise. |
| SameB4Partner_HighExp | An indicator variable equal to one if the firm-pair is audited by the same Big 4 partner with high experience, and zero otherwise. High experience is measured as an above-median length of time since the partner became a partner at the firm. |
| SameB4Partner_LowExp | An indicator variable equal to one if the firm-pair is audited by the same Big 4 partner with low experience, and zero otherwise. Low experience is measured as a below-median length of time since the partner became a partner at the firm. |
| Sameoffice_DiffPartner_HighExp | An indicator variable equal to one if the firm-pair is audited by the same office with high experience but different partners, and zero otherwise. High experience is measured as an above-median length of time that the client has been audited by that audit office. |
| Sameoffice_DiffPartner_LowExp | An indicator variable equal to one if the firm-pairs are audited by the same office with low experience but different partners, and zero otherwise. Low experience is measured as a below-median length of time that the client has been audited by that audit office. |
Open Research
DATA AVAILABILITY STATEMENT
All the data used in this study are available from commercial datasets commonly used in this literature. We describe our sample selection process in detail in the paper.
REFERENCES
- 1 We use the term “engagement signing auditor style” to refer to style effects found in prior research using Chinese data (rather than “partner style”) because the signing engagement auditor may not be a partner under the Chinese system. In China, two engagement auditors sign the financial statements—a review auditor and an engagement auditor. The review auditor is a partner and the engagement auditor is typically a senior manager. The role of the engagement auditor is analogous to the role of the audit partner in a US audit. See Li et al. (2016) for additional information about the differences between Chinese and US audit settings.
- 2 Chen et al. (2020) test the influence of the review and engagement signing auditors on financial statement comparability separately. However, in most of their tests, they do not find that the regression coefficients on the same engagement signing auditor are significantly larger than the coefficients on the same audit office but different signing auditor.
- 3 Li et al. (2021) argue that individual auditor style exists because each auditor “has her/his own unique knowledge structure, education background, working experience, and cognitive style, which can influence her/his understanding of auditing and accounting standards. In addition, each individual auditor has her/his own unique diligence, risk preference, and ethics, which can affect her/his professional judgment and independence during the auditing process. These personal attributes lead individual auditors to develop her/his own testing approaches for implementing audit standards along with personal working rules for interpreting and implementing accounting standards” (p. 76).
- 4 We discuss these measures in more detail in Section 3, and we provide variable definitions in the appendix.
- 5 In contrast, having the same audit office but different audit partners results in a decrease of only 4.4% from the mean value of differences in abnormal accruals for all firm-pairs in our sample. The decrease in mean value of difference in accruals for firms with the same audit firm but different audit partners is not statistically significant.
- 6 We find that adding audit partner fixed effects to the model containing audit firm and office fixed effects increases the explanatory power of the main model for discretionary accruals by 18.9% relative to the base model's adjusted R2.
- 7 We limit our sample to Big 4 audit firms in our main tests because prior studies find that financial statement comparability is attributable to Big 4 firms (Francis et al., 2014), thus making it the most powerful setting to test our predictions.
- 8 Nearly all individual auditor style studies are carried out using Chinese data. One notable exception of which we are aware is Ahn and Sonu (2021) which uses Korean data to examine audit partner style. We also acknowledge two concurrent working papers that investigate audit partner style in a US setting—Baugh et al. (2022) who focus on non-Big 4 audit firms and Kim et al. (2022) who focus on audit partner “individualism.”
- 9 Francis and Michas (2013) examine the “contagion” effects within offices and find that offices with audit failures also have clients with higher discretionary accruals, suggesting that audit offices have systematic and persistent audit styles.
- 10 Lennox and Wu (2018) find that most existing partner studies originate in China where an audit partner signatory law has existed since 1995 (Ministry of Finance, 1995). Following these China-based studies, subsequent studies have used data from Taiwan and Australia, with a smaller number of studies also using European data.
- 11 See Brown et al. (2014) and Eierle et al. (2021) who find that the US scores higher than any other country in enforcement of accounting standards and measures of auditor effort and audit quality.
- 12 Certainly, one can find similarities between US and non-US audit partners and institutional settings. However, here we highlight that differences do exist even in similar country settings and that these differences make clear the importance of studying the United States separately.
- 13 For example, examining audit partners in the United Kingdom, Carcello and Li (2013) find a significant decline in abnormal accruals and the propensity to meet earnings benchmarks, and a significant increase in informativeness of earnings when partner signatures are required. In contrast to their results, Blay et al. (2014) do not find a difference in audit quality following partner identification in the European Union. Using US partner data, Burke et al. (2018) and Dao et al. (2019) find that audit quality increases and that audit delays decrease when individual audit partners are publicly known. However, using a difference-in-differences design, Cunningham et al. (2019) do not find similar results. Likewise, Doxey et al. (2019) do not find evidence that the financial markets value audit partner disclosure in the United States.
- 14 Educational standards are fundamental to auditors’ ability to serve the public interest. These types of standards vary by country and are stronger in environments with more sophisticated regulatory and institutional frameworks. Accounting standards for education and related high-level experience are notably lower in environments with weaker regulatory and institutional systems. For example, adoption of International Education Standards lags on average for countries with weaker institutions (International Accounting Education Standards Board, 2014). For a more complete treatment of accounting education standards and development of accounting institutions see Enthoven (1976, 1981).
- 15 The analysis extends from q − 15 to q.
- 16 While it is possible that a partner switch occurs during the 16-quarter period, there are limitations involved in keeping only samples in which the partners are the same over the 16 consecutive quarters. Principally, keeping observations with only 16 full quarters of partner data leaves only one fiscal year in which to perform our analyses.
- 17 We evaluate our control variables under the framework of Swanquist and Whited (2021) and justify their use based on their relation both with dependent variables (comparability of earnings and audit outcome) and with the variable of interest (choice of audit partner). We determine that our control variables have the characteristics of “good control variables,” as suggested by Swanquist and Whited (2021). We control for the comparability and scale of other firm-specific characteristics which are likely determinants of accounting earnings or which may affect audit outcomes. We control for auditor firm tenure (Tenure_Diff and Tenure_Min), auditor specialization (IndustrySpec), minimum value of accruals (Accr_Min), the similarity and scale of firm size (Size_Diff and Size_Min), financial leverage (Lev_Diff and Lev_Min), growth (MB_Diff and MB_Min), probability of having a loss (LossProb_Diff and LossProb_Min) and volatility in sales, sales growth, and operating cash flow (CFO_Diff, CFO_Min, Std_Sales_Diff, Std_Sales_Min, Std_CFO_Diff, Std_CFO_Min, Std_SalesGrth_Diff and Std_SalesGrth_Min). For Equation 5, we follow Francis et al. (2014) by adding returns comovement (RET_COV) and operating cash flows comovement (CFO_COMP_COV) as controls and removing CFO_Diff, CFO_Min and Accr_Min.
- 18 Big 4 firms audit the vast majority of the largest listed firms in the US (99% of firms on the S&P 500 are audited by Big 4 firms [McKeon, 2020]); therefore, it follows that results in our setting are attributable to Big 4 audit partners. In untabulated tests for non-Big 4 firms, we find that our results are attributable to partners in Big 4 firms. We leave to future research a thorough investigation of the factors that lead to these differences in audit partner style effects on financial statement comparability between audit firm types.
- 19 In China, two engagement auditors sign the financial statements—a review (senior) auditor and an engagement (junior) auditor. The review auditor is a partner and the engagement auditor is typically a senior manager. The role of the engagement auditor is analogous to the role of the audit partner in a US audit (Li et al., 2016).
- 20 In untabulated tests, we also find that the incremental explanatory power of audit partners exists over audit firms and audit offices alone. In addition, Chen et al. (2018) question the reliability of the two-stage process to calculate measures of accruals. When we use the one-stage process they suggest, our results are qualitatively similar.
- 21 We measure partner experience as the number of years as a partner rather than the number of years that the partner has been engaged on the audit because that information is not available.
- 22 We also perform the partner fixed effects test for audit fees and audit report lag. In untabulated results, we find that the base model explains 25.3% of the variation of audit fees while the audit partner fixed effect model explains 60.4% of the variation audit fees. For audit report lag, the base model explains 8.0% of the variation. However, when we add audit partner fixed effects, the R2 increases to 16.6%.
- 23 Because firms are not randomly assigned an audit partner, we also perform two untabulated tests to address self-selection bias. First, we remove from the sample firm-pairs that have an audit partner who meets the definition of an auditor specialist in prior research. Second, self-selection bias may also arise when industry competitors choose different audit partners to protect proprietary information, but the two industry competitors nevertheless share a similar financial statement production process. We address this issue by using the Herfindahl Index to proxy for proprietary costs of disclosure (e.g., Francis et al., 2014). We remove firm-pairs in the highest decile of Hirfindahl Index to control for the possibility that a firm-pair's similar earnings are induced by competition. We then re-estimate the reduced sample regression models. The results of both tests are consistent with our primary results.
- 24 We do not conduct the partner switching test for the earnings comovement model because that measure requires 16 consecutive quarters of data, and the audit partner could change during the 16-quarter window.
- 25 We also estimate the model using a propensity-score matched sample, with no change to our inferences (untabulated).
