Is integrated auditing superior to separate auditing? Evidence from China

ABSTRACT The purpose of this study is to compare audit effectiveness and audit efficiency between companies that have integrated auditing of internal control over financial reporting (ICFR) and financial statements and companies that have separate auditing. We analyse a sample of Chinese public companies that disclosed their ICFR audit reports from 2011 to 2015. Using the full sample, sub-samples, and a propensity score matching (PSM) sample, we consistently find that companies having integrated auditing exhibit higher financial reporting quality measured by excess non-operating income. The findings regarding audit efficiency are mixed, with limited evidence indicating that an integrated auditor is potentially able to complete two audits without further delay. Overall, our results seem to suggest that integrated auditing is superior to separate auditing in that it enhances audit effectiveness and might improve audit efficiency. Such superiority might be due to knowledge spillover when two related audit services are jointly provided.


Introduction
Standard setters seem to favour integrated auditing, i.e. the audit of internal control over financial reporting (ICFR) integrated with the financial statements (FS) audit. Since 2004, ICFR audits have been required for U.S. accelerated filers. According to the Public Company Accounting Oversight Board (Public Company Accounting Oversight Board [PCAOB], 2007), 'The audit of internal control over financial reporting should be integrated with the audit of the financial statements . . . In an integrated audit of internal control over financial reporting and the financial statements, the auditor should design his or her testing of controls to accomplish the objectives of both audits simultaneously . . .' Unlike the U.S., in China, ICFR audit can be integrated with the FS audit or performed separately. Chinese companies have the option to choose integrated auditing or separate auditing. While both options are available, the vast majority of Chinese companies choose integrated auditing. Is integrated auditing superior to separate auditing? More specifically, Given that the institutional background has changed significantly since the study period in Zhang and Han (2016), new empirical evidence on the consequences of integrated auditing vs. separate auditing is needed. Gunn et al. (in press) address this issue by comparing outcomes (i.e. audit quality and audit efficiency) of integrated auditing vs. separate auditing using a sample period of 2011-2016. Different from Zhang and Han, Gunn et al. (in press) examine both ICFR and FS audit quality with a focus on ICFR audit. Specifically, they find that separate ICFR audits are associated with higher quality compared to integrated ICFR audits. The audit cost analysis indicates that FS audits are more efficient when conducted in the separate auditing setting. In their additional analysis, Gunn et al. (in press) provide limited evidence that separate FS audits are associated with higher quality compared to integrated FS audits, which is consistent with the finding in Zhang and Han (2016). 3 Overall, Gunn et al. (in press) challenge the assumption that integrated audits are more effective and efficient than separate audits. As the first study that tests this assumption, it has significant implications to policy makers. Perhaps, one size does not fit all, and it is beneficial to allow some companies to adopt alternative ICFR audit arrangements.
Although our study addresses similar issues, such as audit quality and audit efficiency, it differs from Gunn et al. (in press) in the following two primary aspects. First, while Gunn et al. (in press) provide a thorough coverage of how separate audits might impact both ICFR and FS audit quality and audit efficiency, our study exclusively focuses on the FS audit quality and efficiency. Such a focus stems from the theory of knowledge spillover. Assuming there is knowledge spillover between FS audit and ICFR audit, we expect that integrated auditor is more effective and efficient compared to the separate FS auditor in performing the FS audit service. Second, our measures for FS audit quality and FS audit efficiency are different from the ones used in Gunn et al. (2022). 4 Additionally, our sample period is one year shorter.
Analysing a sample of Chinese companies that disclosed their ICFR audit reports 5 during the period 2011 to 2015, we find that companies having integrated auditing show a lower level of excess non-operating income (ENOI) than companies having separate auditing. Since ENOI reversely measures financial reporting quality, this result indicates that integrated auditing yields a higher level of financial reporting quality, which is an outcome of enhanced audit effectiveness. We interpret this result as the existence of knowledge spillover between ICFR and FS audits. The results are consistent and robust when we run the analysis using (1) a sub-sample of companies that mandatorily disclose their ICFR audit reports, (2) a sub-sample of companies that voluntarily disclose their ICFR audit reports, and (3) a propensity score matching (PSM) sample.
We analyse audit report lags 6 to explore whether integrated auditing improves audit efficiency. The full sample analysis indicates that audit report lag for integrated auditing is not different from that for separated auditing. Note that the integrated auditor not only performs the FS audit but also the ICFR audit, an additional audit service. Improved efficiency and increased workload have opposite effects on audit report lag. Because the integrated auditor has a higher workload compared to the external auditor who only performs the FS audit, such a result indicates that integrated auditing not only improves audit efficiency, but such improved audit efficiency fully offsets the effect of increased workload on audit report lag. The result using a sub-sample of companies that are required to disclose their ICFR audit reports is consistent with the full sample result. That is, we find no significant difference in audit report lag between integrated auditing and separate auditing. That means, the integrated auditors are able to complete both FS and ICFR audits as timely as the auditors that only perform the FS audits.
The analysis using a sub-sample of voluntary companies indicates that more time is required for integrated auditors to complete both FS and ICFR audits than for FS auditors to complete the FS audits. Such a result does not warrant a conclusion. There are two possible explanations for this result: (1) integrated auditing might not improve audit efficiency; or (2) integrated auditing improves audit efficiency, but the magnitude of the improved audit efficiency is not large enough to offset the effect of increased workload on audit report lag.
Although the above samples yield different results, the analysis using the PSM sample shows a significantly shorter audit report lag for integrated auditing. This suggests that integrated auditing not only improves audit efficiency, but the magnitude of improved efficiency is significant enough to outweigh the effect of increased workload on audit report lag. Such improved audit efficiency might be attributable to knowledge spillover when two audits are performed jointly. After all, the integrated auditor's client-specific knowledge is likely to help improve audit efficiency by avoiding duplication and overlapping of audit procedures.
In summary, the results of analysing audit report lag are inconsistent using different samples. Notice that, of the total 7,027 observations in our full sample, 6,869 (97.8%) and 158 (2.2%) represent cases of integrated auditing and separate auditing, respectively. Despite the limitations of propensity score matching, it provides a more balanced sample (integrated auditing vs. separate auditing) and addresses potential issues caused by severely imbalanced samples. Therefore, we consider the results using the PSM sample more reliable.
Collectively speaking, our results seem to suggest that integrated auditing is superior to separate auditing because it enhances FS audit effectiveness, and it potentially improves FS audit efficiency. This study makes several important contributions. First of all, it adds new evidence on the consequences of integrated auditing vs. separate auditing in a mandatory regime. Contrary to Gunn et al. (in press) and Zhang and Han (2016), we find that integrated auditing enhances audit effectiveness, and therefore, improves financial reporting quality. We attribute the increased audit effectiveness to knowledge spillover between ICFR and FS audits. Secondly, in addition to furthering the empirical literature on China's ICFR audit requirement, our study adds to the existing literature on knowledge spillover by identifying a mechanism through which the ICFR audit improves financial reporting quality. Furthermore, we document that such knowledge spillover not only helps improve audit effectiveness but might also enhance FS audit efficiency. Finally, to a certain degree, this study empirically confirms the superiority of integrated auditing and provides a theoretical foundation for the regulatory requirement of integrated auditing. Specifically, our results suggest that integrated auditing improves audit effectiveness and might improve audit efficiency, thereby supporting the choice of integrated auditing.
The remainder of the paper is organised as follows. Section 2 provides institutional background, reviews related literature, and develops hypotheses. Section 3 describes the data and research design, and Section 4 presents the results. Finally, section 5 concludes with insights for researchers and regulators.

Institutional background
Although no accounting scandals have drawn a high level of global attention as Enron and WorldCom, fraud is not uncommon among Chinese companies. According to Lisic et al. (2015), the frequency of accounting frauds ranged between 2.2% to 3.94% during 1999 to 2005 in China. The consequences of accounting fraud are costly and disruptive. To discourage accounting fraud, among other regulations to improve transparency of capital markets, China started its formal development of internal control regulations at the beginning of this century. At around the same time in the U.S., a series of high-profile accounting scandals triggered the passage of Sarbanes-Oxley Act (SOX). Since SOX's passage, numerous academic studies and anecdotal evidence have provided implications for Chinese researchers and regulators. A few years after the enactment of SOX, China started developing its own internal control standards using the COSO's framework. In 2008, five Chinese governmental departments 7 jointly issued the Enterprise Internal Control Standard (EICS). With improving financial reporting quality being one of the primary goals, EICS provides regulatory standards for Chinese companies to develop and implement effective internal controls. EICS places an emphasis on internal control over financial reporting (ICFR) and requires management disclosure on the effectiveness of ICFR. The scope of EICS is broader but similar to that of SOX 302 and SOX 404.
Unlike the requirement in the U.S., the audit of ICFR in China does not have to be integrated with the FS audit. Chinese companies are allowed to appoint a different external auditor to perform the ICFR audit separately. While there is a lack of documentation on the rationality of allowing both integrated auditing and separate auditing, such a unique regulation offers an opportunity to empirically examine the consequences of these two options and to provide evidence testing the 'assumed' superiority of integrated auditing. In the following, we attempt to provide possible explanations for the establishment of a standard that allows both integrated auditing and separate auditing. There has been a debate in China on whether the ICFR audit should be integrated with the FS audit or performed separately by a different auditor. Proponents of integrated auditing argue that the ICFR audit should be integrated with the FS audit because the two audits require similar audit procedures regarding the client's ICFR. Having these two related audit services integrated could help achieve a higher level of audit efficiency via reducing duplicated audit procedures. Proponents of separate auditing believe that, with two audits performed separately, each auditor is able to stay focused on the specific underlying audit objectives. Using such an objective-driven approach is important in maintaining audit quality for each specific type of audit. Although the above reasoning is not datadriven, it might have had an impact on the development of EICS. Additionally, prior to the enactment of EICS, some companies that voluntarily disclosed their ICFR audit reports chose separate auditing. Thus, allowing both integrated auditing and separate auditing in the mandatory regime might have been due to practical reasons (i.e. allowing the continuation of existing separate auditing choices).
Although the vast majority of companies opt for integrated auditing, some choose separate auditing. 8 In 2011, the first year of the mandatory implementation of EICS, three of the 67 cross-listed companies chose separate auditing. In separate auditing, the FS audit and the ICFR audit are conducted independently by two different auditors. Therefore, there are duplication and overlapping of audit procedures between the two auditors. Whereas in integrated auditing, the one external auditor is able to avoid these redundant audit procedures. It is reasonable to expect increased audit efficiency when the FS and ICFR audits are integrated. The phenomenon of separate auditing seems counterintuitive assuming possible benefits of integrated auditing. To understand why some companies chose separate auditing, we identified two public companies that had chosen separate auditing: Company ABC (hereafter ABC) and Company XYZ (hereafter XYZ). 9 Since 2011, ABC has had a national accounting firm for its ICFR audit and an international Big 4 for its FS audit. One author conducted interviews with ABC's financial officer and a senior audit manager at the international Big 4 who was on ABC's FS audit engagement team during our sampled period. ABC's financial officer explained that the decision for separate auditing was made based on the following considerations: (1) to reduce the total audit fees as the international Big 4 charges a higher audit fee; (2) to maintain the existing well-established 'guanxi' with the national accounting firm; both ABC's and the accounting firm's headquarters are located in Shandong province and this accounting firm provides various audit services to many of ABC's subsidiaries; and (3) to seek ICFR 'audit independence' as the international Big 4 was also the accounting firm that helped ABC to establish its internal control system.
The senior audit manager at the international Big 4 confirmed some of the reasons listed above, such as the local 'guanxi' and audit fee considerations. Additionally, the audit manager believed that having a different ICFR auditor might have been a strategy for ABC to gain leverage on the audit fee negotiation.
XYZ had separate auditing from 2012, the first year of its ICFR audit, to 2017. Unlike ABC, XYZ's FS and ICFR auditors were both Chinese national accounting firms. Therefore, 'guanxi' consideration (audit fee consideration) is more (less) likely to be the main driver for choosing separate auditing. According to a securities practitioner who served as an IPO sponsor for XYZ, besides the 'guanxi' consideration, a company is more likely to choose a different external auditor for the ICFR audit when its FS auditor's existing workload is high. In summary, such qualitative evidence suggests that audit fee, guanxi, ICFR audit independence, and FS auditor workload are potential factors that drive the ICFR auditor choice. All three interviewees suggested that there were a wide variety of reasons for making the ICFR auditor choices. There is also empirical evidence that confirms the 'guanxi' consideration. For example, Zhang and Han (2016) examine the antecedents of separate auditing using a 'social network' (guanxi) perspective. They find that a company that is more concerned about its external relationships with other companies and organisations tends to choose separate auditing. Although it is interesting to explore why some companies choose separate auditing, the determinants of ICFR auditor choice are not addressed in this study. Instead, we focus on two aspects of the consequences of such choice, i.e. audit effectiveness and audit efficiency.

Literature review and hypothesis development
According to the Public Company Accounting Oversight Board (PCAOB), effective internal control over financial reporting provides reasonable assurance regarding the reliability of financial reporting and the preparation of financial statements for external purposes. Assessing the effectiveness of the ICFR is an important part of FS auditing. Therefore, an auditor's intricate knowledge about a client's ICFR might improve both effectiveness and efficiency of the FS audit. Previous studies have investigated whether mandatory ICFR audits indeed improve financial reporting quality (e.g. Iliev, 2010;Nagy, 2010;Bhaskar et al., 2019;Lennox & Wu, 2021;McCallen et al., 2022). While Iliev (2010) and Nagy (2010) provide evidence of improved financial reporting quality (i.e. more conservative earnings; reduced material misstatements), Bhaskar et al. (2019) find that the FS-only audits (i.e. without the presence of ICFR audits) are associated with lower likelihood of material misstatement compared to integrated audits (with the presence of ICFR audits). Specifically, Bhaskar et al. (2019) suggest that mandatory ICFR audits harm financial reporting quality. This finding is confirmed by Lennox and Wu (2021). Using a sample of Chinese public companies, Lennox and Wu (2021) suggest that such a negative impact on financial reporting quality is caused by a reduction in adjustments to clients' pre-audit financial statements when ICFR audits are required The studies cited above address an important research issue -whether mandatory ICFR audits improve financial reporting quality. Although more research is needed to address this issue, our study focuses on a different but related question: when ICFR audits are mandated, is integrated auditing superior to separate auditing? Specifically, in this study, we examine the superiority by comparing FS audit effectiveness and FS audit efficiency between integrated auditing and separate auditing. Because separate auditing only exists in China, there is a lack literature on comparisons of integrated auditing and separate auditing. Up to this date, only two studies (i.e. Zhang & Han, 2016;Gunn et al., in press) have empirically examined the consequences of integrated and separate auditing. Gunn et al. (in press) extended Bhaskar et al. (2019)'s research on integrated FS audit quality, which suggests that increased opportunities for judgement-based integration issues in integrated auditing negatively affect FS audit quality. From a legislative perspective, Zhang and Han (2016) hypothesise that 'separate auditing' improves FS audit quality because the separation of powers can strengthen the government's oversight. Different from the above studies, our hypotheses are based on the theory of knowledge spillover. When ICFR and FS audits are integrated, we expect knowledge spillover to improve both audit effectiveness and audit efficiency.

Audit effectiveness
Prior literature has documented knowledge spillover due to synergies and knowledge transfers between NAS and audit services. Although the ICFR audit is not a NAS, in China, it is treated as a separate service from the FS audit. The literature on joint provision of NAS and audit services provides insight for our study. Simunic (1984) suggests that some NAS improve audit effectiveness through knowledge spillover. Knowledge of a client's tax accounting might spill over to the audit and improve audit quality, and that in turn would increase financial reporting quality (Gleason & Mills, 2011;Kinney et al., 2004;Robinson, 2008). By performing NAS, auditors can enrich their client-specific knowledge, which in turn improves audit effectiveness (Beck & Wu, 2006). IS-related consulting services improve the auditor's knowledgebase and therefore lead to improved audit/financial reporting quality (Koh et al., 2013). Since effective ICFR can help ensure reliable financial reporting, it is plausible to expect the knowledge spillover effect between FS and ICFR audits. If such knowledge spillover exists, then audit effectiveness would be improved when ICFR and FS audits are integrated and provided by the same external auditor. An auditor's possessing deep knowledge about its client's internal control could improve the FS audit effectiveness, which results in less earnings management and higher financial reporting quality. Consistent with studies cited above, we assume that audit effectiveness has a positive relationship with financial reporting quality. Therefore, we state our first hypothesis as follows: Bamber et al. (1993) define efficiency as 'the use of fewer inputs to obtain a given output' (page 2). Besides improved audit effectiveness, improved audit efficiency can also be achieved through knowledge spillover. For example, Knechel and Sharma (2012) suggest that joint provision of NAS and audit services yield audit efficiency because the auditor accumulates client-specific knowledge and expertise. Using the same logic, the FS auditor gains a greater level of familiarity with its client's ICFR if it also performs the ICFR audit. Such client-specific knowledge would help improve audit efficiency by avoiding duplication and overlapping of audit procedures. Therefore, based on the theory of knowledge spillover, our second hypothesis is stated as follows:

Audit efficiency
H2: Compared to Chinese companies that choose separate auditing, companies that choose integrated auditing exhibit higher audit efficiency.

Data and sample
We collect data regarding accounting, financial statement audits, and ICFR audits from the China Stock Market & Accounting Research (CSMAR) database. The internal control index data is collected from DIB Internal Control and Risk Management database. Because the staged mandatory implementation of EICS began in 2011, our sample covers year 2011 to 2015. Sample selection, sample distributions by year, and sample distributions by industry are provided in Panel A, B, C of Table 1, respectively. Our initial sample consists of 10,942 firm-year observations from all public companies that disclosed ICFR audit reports during fiscal/calendar years 2011-2015. After deleting 266 observations in the financial industry and 3,649 observations that do not have required data, our sample reduces to 7,027 firm-year observations, of which 158 observations represent separate auditing. Note that the number of observations in 2011 is significantly lower because that year was the first year of mandatory implementation of EICS. After 2011, more and more companies were required to implement EICS. The final sample covers 16 major industry sectors, of which more than half of the sampled firms are manufacturers.

Financial reporting quality
FS auditing enhances reliability of information contained in financial statements for intended users and ensures financial reporting quality. Effective FS auditing helps constrain earnings management. In this study, we use the level of earnings management as an indicator of financial reporting quality. Following previous studies that examine the earnings management using Chinese data (Chen & Yuan, 2004;Gul et al., 2009;Cang et al., 2014;Fan et al., 2015), we measure earnings management by 'excess non-operating income' (ENOI). While accrual measures are commonly used in the U.S., as suggested in Chen and Yuan (2004), excess nonoperating income might be a better alternative to measure earnings management among Chinese companies.
We select our control variables based on prior research on earnings management phenomenon in China (e.g. Cheng et al. 2015;Fan et al., 2015;Gul et al., 2017). We include company size (Size), financial risk (Lev), financial distress (Loss), growth in sales (Growth), book-to-market ratio (BTM), operating cash flows (OCF), internal control effectiveness (ICIndex), auditor quality (Bigauditor and Specialist), a dummy variable indicating whether a company is state-owned (SOE), complexity of a company (Complexity), and whether a company reports its ICFR report mandatorily or voluntarily (Mandatory). Industry effects and year effects are also controlled. To test H1, we estimate the following regression model: where ENOI is the absolute value of the before-tax non-operating income (or loss) divided by equity and adjusted by the industry median; Integrate is our interest variable, and is equal to 1 for integrated auditing, and 0 otherwise; Lev is measured by total liabilities divided by total assets; Size is measured by the natural logarithm of total assets measured in RMB; Loss is equal to 1 if the company reports a loss, and 0 otherwise; the growth rate, Growth, is measured by change in sales divided by sales; the book to market ratio, BTM, is calculated by the book value of total assets scaled by the market value of equity at the end of fiscal year; operating cash flows, OCF, is calculated as net operating cash flows divided by total assets; internal control quality, ICIndex, is measured by the square root of the Bigauditor is equal to 1 if the company's financial statements auditor is an international Big 4 or a Chinese national Big 10 accounting firm, and 0 otherwise; Specialist measures an auditor's industry specialisation, which is equal to 1 if the company's auditor is an industry specialist, and 0 otherwise; SOE is equal to 1 if the company is a state-owned enterprise, and 0 otherwise; Complexity is measured by a 'complexity score' following the procedure described in Gul et al. (2017). Mandatary is equal to 1 when a company is mandated to disclose its audited ICFR report, and 0 otherwise. We expect α 1 to be negative and significant. Bamber et al. (1993) define efficiency as 'the use of fewer inputs to obtain a given output'.

Audit efficiency
In this research context, the given output refers to completed ICFR and FS audits. There are two options to obtain such an output: (1) one auditor performs both ICFR and FS audits and achieves the objectives of the two audits simultaneously, and (2) two auditors share the workload and achieve the objectives of the two audits separately. Since we are not able to directly measure the inputs required to complete an audit, we follow prior literature (e.g. Bailey et al., 2018;Bamber et al., 1993;Knechel & Payne, 2001;Knechel & Sharma, 2012;Masli et al., 2010) and use audit report lag to measure audit efficiency. In this study, the audit report lag in integrated auditing measures the audit efficiency for both FS and ICFR audits. In separate auditing, it is not possible to use a single audit report lag measure for both audits because FS and ICFR audits are conducted separately by two different auditors. The audit report lag in separate auditing refers to the number of calendar days from fiscal year end to the date when the FS audit report is filed. Therefore, the observed audit report lag in integrated auditing captures time used to complete both FS and ICFR audits, while the observed audit report lag in separate auditing only captures the time used to complete the FS audit. Although the integrated auditor's knowledge about a client's ICFR would improve audit efficiency and therefore reduce the total time needed to complete the audit work compared to the separate FS auditor, the integrated auditor has a higher workload. Increased workload and improved audit efficiency have opposite effects on audit report lag, which makes it more complex to use audit report lag to test the audit efficiency hypothesis. When the observed audit report lag in integrated auditing is longer than that of the FS audit in separate auditing, a conclusion is not warranted. That is because even if integrated auditing improves audit efficiency, such an improved audit efficiency might not fully offset the effect of the increased workload on audit report lag. However, if the observed audit report lag in integrated auditing is shorter than or not different from that of the FS audit in separate auditing, there is evidence to conclude that integrated auditing improves audit efficiency. This is because the integrated auditor is able to complete two audits in more or as timely manner as the FS auditor (in separate auditing) to complete one audit. Therefore, H2 will be supported if the observed audit report lag in integrated auditing is not longer than that for the FS audit in separate auditing. Our dependent variable is audit report lag (AudLag). Following Sharma et al. (2017) and Bailey et al. (2018), we include Size, Lev, ARINV, Complexity, Loss, ROA, Bigauditor, Specialist, ICIndex, CleanOpn, and Mandatory as control variables. We also control year and industry effects. Therefore, to test H2, we estimate the following regression: AudLag ¼ γ 0 þ γ 1 Integrate þ γ 2 Size þ γ 3 Levþγ 4 ARINV þ γ 5 Complexity þγ 6 Loss þ γ 7 ROA þ γ 8 Bigauditor þ γ 9 Specialist þ γ 10 ICIndex þγ 11 CleanOpn þ γ 12 Mandatory þ Industry effects þ Year effects þ ε The dependent variable, AudLag is the natural log of the number of calendar days from fiscal year-end to the date when the auditor files the audit report (FS audit report in separate auditing). Integrate is our interest variable and is equal to 1 for integrated auditing, and 0 otherwise. ARINV is the percentage of accounts receivable and inventory in total assets. ROA is measured by operating earnings before tax divided by average total assets. CleanOpn is a dummy variable which is equal to 1 if a company received an unqualified audit opinion. All other variables are as defined in Equation (1). As discussed earlier, based on the theory of knowledge spillover, we expect that integrated auditing helps improve audit efficiency. Instead of measuring audit efficiency directly, we observe audit report lag. If the observed audit report lag in integrated auditing the same as or shorter than that of the FS audit in separate auditing, there is evidence that integrated auditing improves audit efficiency. Therefore, we expect γ 1 to be either insignificant (improved audit efficiency effect fully offsets the increased workload effect) or significantly negative (improved audit efficiency effect outweighs the increased workload effect). Table 2 reports descriptive statistics of the variables used in this study. Of the 7,027 observations, 6,869(97.8%)/158(2.2%) observations represent integrated/separate auditing (Integrate). The mean excess non-operating income (ENOI) is about 1.1% of total stockholders' equity. On average, it takes about 92 days to complete a FS audit (the mean AudLag = 4.503) Approximately 57.2% of total sample represent mandatory ICFR auditing (Mandatory). The mean Size is 22.212. The average total accounts receivable and inventory to total assets (ARINV) is 26.5%. The mean debt ratio (Lev) is about 44.7%. 8.8% of observations show a net loss (Loss). The average ROA is 4.0%. The average operating cash flow to total assets (OCF) is 4.7%. Book-tomarket ratio (BTM) has a mean of 0.545. About half of the observations (50.3%) represent state-owned enterprises (SOE). The mean complexity score is 2. About 73.1% of observations have either an international Big 4 or a Chinese national Big 10 as the FS auditor (BigAuditor). 10 About 35.8% of the observations have an industry specialist as their FS auditor (Specialist). The average ICindex score is 25.47. 99.7% of the observations received an unqualified audit opinion (CleanOpn).

Propensity score matching sample
As described earlier, of the 7,027 firm-year observations, only 158 represent separate auditing. This imbalanced and small sample size of companies that choose separate auditing may bias our inferences. To address this concern as well as the selection bias, we use a propensity score matching (PSM) sample. Additionally, we exclude companies that are listed on SME and GEM boards. Therefore, the PSM sample only include 137 pairs that are traded on the main board to ensure that all observations represent ICFR audits. For each separate auditing case (Integrate = 0), we select one case from the integrated auditing sample (Integrate = 1) using the nearestneighbour approach without replacement. 11 Table 4 presents the results of bivariate analysis using the full sample and the PSM sample in Panel A and Panel B, respectively.   Variable definitions appear in Appendix A. †, *, **, *** denotes significance level at 10%, 5%, 1%, 0.1% respectively (two-tailed).
The bi-variate analysis using the full sample shows that ENOI is lower in integrated auditing than in separate auditing but there is no significant difference in AugLag between integrated auditing and separate auditing. As also shown in Panel A, the full sample bi-variate analysis shows significant differences in Size, Lev, Loss, Growth, BTM, ICIndex, Bigauditor, Specialist, Mandatory, SOE, Complexity, ARINV, and ROA between integrated auditing and separate auditing.
The bivariate-analysis of ENOI using the PSM sample is consistent with that using the full sample. Additionally, as shown in Panel B in Table 4, AudLag of integrated auditing is significantly lower than that of the FS audit in separate auditing. That means, on average, companies choosing integrated auditing experience a shorter audit report lag. The bivariate analysis result using the PSM sample is consistent with our expectation. Variable definitions appear in Appendix A. †, *, **, *** denotes significance level at 10%, 5%, 1%, 0.1% respectively (two-tailed).

Multiple regression results: audit effectiveness (financial reporting quality)
Following Chen and Yuan (2004), Gul et al. (2009), Cang et al. (2014, and Fan et al. (2015), we use the absolute value of excess non-operating income (ENOI) to proxy for financial reporting quality. We run the regression model specified in Equation (1) for the full sample, two sub-samples that represent mandatory ICFR auditing and voluntary ICFR auditing, and the PSM sample. Regression results to test H1 for the full sample, mandatory auditing sample, voluntary auditing sample, and PSM sample are presented in Table 5 Panel A. The coefficient on Integrate is negative and statistically significant for all the four samples. That is, after controlling for other factors, companies choosing integrated auditing report lower excess non-operating income (earnings management). Thus, integrated auditing seems to improve financial reporting quality through enhanced audit Variable definitions appear in Appendix A. †, *, **, *** denotes significance level at 10%, 5%, 1%, 0.1% respectively (two-tailed).
effectiveness. This result supports H1 and differs from that documented in Gunn et al. (in press) and Zhang and Han (2016). It is not to our surprise that our result is different from Zhang and Han (2016) due to significant changes in the institutional background. With more guidance and more practice on implementing EICS, ICFR audits have become more normalised. Additionally, external auditors have gained more experience and knowledge in conducting the ICFR audit. Therefore, it is reasonable to expect an improved audit effectiveness in more recent years. Although the sample period in our study overlaps with that in Gunn et al. (in press), the results regarding FS audit quality are different. Such difference is likely due to different FS audit quality measures used in the two studies. While we capture audit effectiveness indirectly via measuring 'earnings management', Gunn et al. (in press) use an ex-post indicator to measure FS audit quality directly. Finally, the coefficients on Lev and Complexity, are statistically significant across all four samples. Coefficients on Growth, BTM, ICIndex, and SOE are significant for full, mandatory, and voluntary samples. These results indicate that while state-owned companies (SOE) with a higher debt ratio (Lev) and a higher level of complexity (Complexity) report higher excess non-operating income, undervalued companies (BTM) with a higher sales growth rate (Growth) and higher internal control quality (ICIndex) report lower excess non-operating income.

Multiple regression results: audit efficiency
As stated earlier, we compare audit report lag (AudLag) between integrated auditing and FS audit in separate auditing to test H2. H2 will be supported if AudLag for integrated auditing is not longer than (i.e. either not different from or shorter than) that for the FS audit in separate auditing. We run the regression model specified in Equation (2) for the full, mandatory, voluntary, and PSM samples. Regression results are reported in Table 5 Panel B. The coefficient on Integrate is insignificant for the full sample. Because improved efficiency and increased workload have opposite effects on audit report lag, such a result potentially suggests increased audit efficiency in integrated auditing.
While the coefficient on Integrate for the mandatory sample is insignificant (p-value = 0.371), it is significantly positive (p-value = 0.036) for the voluntary sample. These results suggest that, for Chinese companies that are required to have their ICFR audited, it does not take more time for the integrated auditors to complete two audits than the FS auditors in separate auditing to complete one audit. In other words, the integrated auditors of the mandatory companies are able to complete the audit work more efficiently. For companies that voluntarily have their ICFR audited, the integrated auditors spend more time to complete two audits than the FS auditors in separate auditing to complete one audit. Such a result does not warrant a conclusion. It is possible that integrated auditing improves audit efficiency, but the increased efficiency is not large enough to offset the effect of the increased workload on audit report lag. It is also possible that integrated auditing does not improve audit efficiency. Therefore, the integrated auditors of voluntary companies may or may not complete the audit work more efficiently. The testing results using mandatory and voluntary sub-samples are inconsistent.
The coefficient on Integrate for the PSM sample is negative and significant. Specifically, the audit report lag in integrated auditing is significantly shorter than that of the FS audit in integrated auditing. That means, even with a higher workload, the integrated auditor is able to complete two audits in a more or as timely manner as FS auditor in separate auditing to complete one audit. Therefore, the result using the PSM sample supports H2 and suggests that integrated auditing improves audit efficiency. In the propensity score matching process, we identify Mandatory as one of the predictors for the choice of integrated vs. separate auditing. Whether a company is mandated to implement EICS seems to impact the consequences of integrated auditing vs. separate auditing. Once we address this endogeneity issue using a matched sample, the coefficient on Integrate becomes negative and significant. Furthermore, the bivariate analysis and multiple regression analysis results using the PMS sample are consistent. The propensity score matching addresses the issues caused by severely imbalanced samples and potentially address the endogeneity. Therefore, we consider the results using the PSM sample more reliable.
Note that our result regarding FS audit efficiency is different from that in Gunn et al. (in press). While we find that the FS audit is more efficient in the integrated auditing setting than in the separate auditing setting, Gunn et al. (in press) suggest the opposite. We believe that such a difference is likely due to different audit efficiency measures used in these two studies. As mentioned earlier, audit reporting lag might be a 'noisy' measure for audit efficiency.
Finally, the coefficients on Specialist are statistically significant across all four samples. Coefficients on Size, ROA, ICIndex, and CleanOpn are statistically significant for the full, mandatory, and voluntary samples. Interestingly, industry specialists (Specialist) seem to use more time to conduct FS audits. This result is consistent with Abidin and Ahmad-Zaluki (2012). Although an industry specialist's industry expert knowledge would result in higher audit efficiency in auditing clients in a given industry, it might also have a higher audit workload due to the high market share in that industry. Such higher workload might cause a longer audit delay. Consistent with our expectations, larger companies (Size) experience a longer audit delay while companies with higher internal control quality (ICIndex) experience a shorter audit delay. Audit report lag is shorter for companies that receive an unqualified audit opinion (CleanOpn).

Sensitivity analysis: alternative measure for earnings management
In summary, results using the PSM sample suggest that integrated auditing yields higher audit effectiveness (financial reporting quality) and higher audit efficiency. We interpret such results as the effects of knowledge spillover associated with integrated auditing.
In addition to using a PSM sample as a sensitivity analysis, we use performance matched discretionary accruals (DA) to measure earnings management. Specifically, DA is measured as the absolute value of residuals from cross-sectional estimation of the industry-year matched on performance (i.e. ROA) in year t-1. The results (not tabulated) are similar to, but somewhat weaker than, those reported in Table 5 Panel A when ENOI is used as a proxy for earnings management. This could be due to high levels of error in estimates of abnormal accruals as documented in Chen and Yuan (2004).

Conclusion, limitations, and future research
In this study, we attempt to compare audit effectiveness and audit efficiency between companies that have integrated auditing of internal control over financial reporting (ICFR) and financial statements and companies that have separate auditing. Compared to companies that choose separate auditing, companies that have the same auditor for both FS and ICFR audits exhibit a lower level of excess non-operating income, an indicator of financial reporting quality. Such a result seems to suggest that integrated auditing improves audit effectiveness, which is an outcome of knowledge spillover. We also document limited evidence suggesting that integrated auditing is likely to improve audit efficiency. Collectively speaking, integrated auditing seems to be superior to separate auditing. Perhaps, that could explain the pervasiveness of integrated auditing and that the number of companies that choose separate auditing has been decreasing since 2015. Our study contributes to the literature by providing additional empirical evidence on the consequences of integrated auditing vs. separate auditing. According to our results, integrated auditing seems to enhance audit effectiveness and might improve audit efficiency. Such findings confirm the effect of knowledge spillover when ICFR and FS audits are integrated.
While providing important insights for researchers and regulators, the current study is subject to the following major limitations. First of all, to the best of our knowledge, China is the only country in the world that allows separate auditing. Thus, while our study provides opportunity to test theory/practice, it has limited generalisability. Secondly, in this study, we focus on the consequences of integrated auditing vs. separate auditing. We only provide limited qualitative evidence to explain why some companies choose separate auditing. Future research can expand the understanding of determinants of ICFR auditor choice. Thirdly, the sample period covers five years (i.e. 2011-2015) from the beginning year of ICFR mandatory implementation. The primary reason to choose 2011-2015 as our sample period is that, although never a popular choice, separate auditing was the most 'pervasive' during that time. After 2015, the number of companies that choose separate auditing begins to decrease. Additionally, since 2015, all main board companies have implemented EICS, and the institutional background has changed as well. Nonetheless, future research could examine a period after all main board companies have implemented EICS. Finally, the research methodology is subject to multiple flaws. For example, earnings management (measured by excess nonoperating income) is an indirect measure for the FS audit quality. Excess non-operating income is likely to have noises and biases. Audit report lag might be a 'noisy' measure for audit efficiency. Given our results regarding FS audit effectiveness and efficiency are different from Gunn et al. (in press), caution should be exercised when interpreting these results. Additionally, while the difference in non-operating income is statistically significant, the effect size might be too small (Cohen's d = 0.17) to demonstrate the importance of this finding. In analysing audit report lag, the results of using different samples are inconsistent. Specifically, H2 is supported when using the full sample, mandatory sub-sample, and the PSM sample but is not supported using the voluntary sample. Although we consider the results using the PSM sample more reliable, once again, caution should be exercised when interpreting these results. Such a matching approach has its limitations. As suggested in Shipman et al. (2017), 'PSM does not address most concerns relating to self-selection or endogeneity' (page 216). The findings using the PSM sample might be sensitive to PSM design choices.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
Qiaoling Fang would like to acknowledge financial support from National Natural Science Foundation of China (72172144).