Killing two birds with one stone? Auditor choice in merger & acquisitions and subsequent auditor assurance quality

ABSTRACT Listed companies commonly appoint their incumbent annual auditors to audit a target company during a merger and acquisition (M&A). However, some companies choose an alternative auditor to perform the M&A audit. To understand the consequences of this apparently unconventional auditor choice, we argue that the acquiring company introduces competition between the M&A auditor and the annual auditor, which makes the M&A auditor incentivised to compete for the annual audit engagement, whereas the annual auditor is motivated to protect her business from rivalry. This game facilitates corporate management to gain favourable treatment from auditors both in the annual audit and in the performance commitment attestation for post-M&A periods. Using data from Chinese M&A market from 2008 to 2017, we find evidence consistent with our hypotheses. Our findings suggest that an audit client is capable of compromising external auditors’ monitoring by employing multiple auditors on various assurance services.


Introduction
Merger and acquisitions (M&As) are considered an important expansion strategy for listed companies. As an acquirer, a listed company employs an external auditor to provide assurance service on target company's historical financial statements and earnings forecast in an M&A deal. In terms of the auditor selection, we find that although the acquiring companies commonly choose their incumbent annual auditors to provide assurance services for the M&A deal, a considerable number of acquiring companies hire another audit firm (other than the incumbent annual auditor) to audit target companies. 1 This study aims to understand the apparent anomaly in auditor choice in the M&A assurance market, along with its impact upon assurance service quality.
Conducting an M&A audit by the acquirer's annual auditor has obvious benefits of trust, cost saving, and efficiency. First, the acquirer can attach greater reliance upon her own annual auditor (relative to another auditor with whom the acquirer has no previous relationship) when evaluating the target company. Second, an M&A assurance service may also involve an audit of the acquirer's accounting information (Lennox et al., 2018), along with continuing post-M&A assurance services. This implies that choosing the acquirer's annual auditor as the M&A auditor has advantages in cost saving and efficiency as well (Cai et al., 2016;Dhaliwal et al., 2016).
Given these benefits, we find that there are still more than 30% of the acquiring companies employing another accounting firm to provide the M&A assurance service. This indicates that, in addition to the apparent benefits, there could be other considerations in M&A auditor selection. Prior studies find that, through company's M&A activities, management boosts their own interests, in terms of a significant increase in executive compensation (Han & Chen, 2007), and the chances of political promotion for management in state-owned enterprises (Chen et al., 2015). Therefore, we conjecture that a potential motivation for the unconventional M&A auditor choice also comes from management opportunism.
We argue that the management of the acquiring company obtains two potential benefits when they choose a non-annual auditor to provide the M&A audit service. On one hand, the annual auditor is likely to sense a stronger threat to annual audit engagement from a new peer audit firm that provides the M&A assurance service, as well as a higher dismissal threat from the management. Therefore, the annual auditor may become more inclined to meet management preference both on the annual audit in the year of M&A, and on post-M&A performance commitment attestation. On the other hand, if the management hires a non-annual auditor to provide the M&A audit service, the M&A auditor not only has a good opportunity to continue to provide the post-M&A performance commitment attestation service, but also the chances to be charged with the annual audit engagement of the acquiring company. Therefore, the M&A auditor is more likely to cater to management preferences on performance commitment attestation services. To sum up, the acquiring company's management is capable of compromising the independence of the M&A auditor and the annual auditor by assigning the M&A assurance service to a non-annual auditor, which helps explain why some companies make an unconventional M&A auditor choice.
To test our theory, we manually collected M&A auditor choices in the Chinese capital market between 2008 and 2017. Empirical evidence shows that, for listed companies who choose a non-annual auditor to conduct the M&A audit, the probability of audit opinion improvement in the year of M&A is significantly higher than that of listed companies who choose the annual auditor to conduct the M&A audit (after controlling for other factors that may have an impact on annual audit opinion improvement), which is consistent with the unconventional M&A auditor choice facilitating management to obtain a more favourable annual audit outcome. We also find that, for listed companies who choose an alternative auditor to conduct the M&A audit, the reported degree of performance commitment fulfilment is significantly higher than that of listed companies who choose the incumbent annual auditor as the M&A auditor (after controlling for other factors that may have an impact on performance commitment fulfilment), which is consistent with non-annual auditors adopting a more lenient standard when providing attestation services on M&A performance commitment.
Our study contributes to the literature in three ways. First, we use the setting of M&A auditor choice to demonstrate the capability of corporate management to weaken the effects of external governance through creating conflict of interests between multiple peers who are expected to play a governance role but who also aim to earn service fee revenues from clients. Specifically, in our setting, both the assurance quality of the annual audit for the M&A year and that of post-M&A performance commitment attestation services are compromised when the acquiring company assigns two assurance services to two different audit firms. Prior corporate governance literature focuses on whether and how various governance forces constrain or motivate the management (Jiang & Kim, 2015;Wong, 2016), whereas our paper shows how corporate management can implement countermeasures on the external monitoring.
Second, it is difficult for traditional literature on audit opinion shopping to directly observe the potential or private connections between client management and the backup auditor (who are not the annual auditor) for favourable audit outcomes (Chow & Rice, 1982;Lennox, 2000;Liu and Liu, 2007). Our setting extends this line of research by illustrating the negative effect on annual auditor's independence when the client management maintains business relationships with the annual auditor and another accounting firm in the same reporting period, which helps better understand the channel of audit opinion shopping.
Third, this paper also contributes to the literature on M&A performance commitment. Prior literature finds that performance commitment is useful to motivate the target company for higher performance (Pan et al., 2017), while companies with performance commitments are more likely to engage in earnings management to meet performance commitment goals (Hou et al., 2015). Our study finds that the management of the acquiring company can affect M&A assurance service quality and the reported level of fulfilment of performance commitments via M&A auditor choices.
The remainder of this study is organised as follows: Section 2 develops research hypotheses; Section 3 describes and empirically models M&A auditor choices; Sections 4 and 5 test research hypotheses H1 and H2, respectively; Section 6 concludes.

The acquiring company's demand for external assurance
When a listed company is the acquiring party in an M&A deal, it has three types of demands for assurance services.
First, there is a demand for auditing the target company during the acquisition. To reduce the information asymmetry between both parties in an M&A deal, the acquiring company and the target company's shareholders employ an external auditor to provide auditing and attestation services on the target company's historical financial statements and earnings forecasts. On this basis, both parties can agree upon a reasonable valuation of the target company and arrive at a mutually acceptable performance commitment.
Second, there is a demand for an attestation service based on the reported degree of performance commitment fulfilment after the M&A deal. According to Chinese regulatory requirements, the acquiring company should employ an audit firm to verify the reported degree of performance commitment fulfilment on a regular basis in post M&A periods, and disclose the auditor's attestation report to the public.
Third, there is a routine demand for the audit of annual financial statements after M&A.
The abovementioned 2 three types of assurance services are closely interrelated. Specifically, the main contract terms contained in an M&A deal and the M&A audit outcomes potentially affect the setting of subsequent performance commitment, thereby being associated with post-M&A performance commitment attestation service. On the other hand, as a major transaction, M&A activity apparently affects the acquiring company's annual audit in the year of M&A.

Relationship between external auditors and the acquirer's M&A auditor choice
Although the acquiring company could assign different assurance services to various audit firms, the management has an incentive to minimise the transaction cost of M&A by avoiding employing multiple external auditors given the start-up cost for any external assurance agency (DeAngelo, 1981). 3 Figure 1 exhibits a regular pattern of the acquiring company's auditor choice when faced with multiple demands for external assurance. In Figure 1, the acquiring company employs its annual auditor A to provide various external assurance services for the M&A deal. In other words, annual auditor A is responsible for the M&A assurance service during year t, the acquiring company's annual audit for year t, and the performance commitment attestation service in post-M&A years.
Extant literature has documented that, when the M&A auditor is the same as the acquiring company's annual auditor, the M&A deal is more likely to be completed, along with lower transaction costs (Cai et al., 2016;Dhaliwal et al., 2016). As mentioned above, in addition to the M&A assurance services, the acquirer also needs an external auditor to provide continuous attestation services on the degree of the performance commitment fulfilment in post M&A periods. Taken together, demands for currentperiod and long-term assurance services suggest that it is a reasonable decision for an acquiring company to choose its annual auditor to provide M&A-related assurance services.
In contrast to Figure 1, Figure 2 exhibits a less common M&A auditor choice, where the acquiring company employs another audit firm (other than the incumbent annual auditor) to conduct the M&A audit. Such a choice makes the selection of annual auditors for the M&A year complicated, which means both the annual auditor A by the end of period t-1 and the M&A auditor B in period t have a chance to become the annual auditor for period t (i.e. the annual auditor for the M&A year). This scenario leads us to our first research question: does such a less common M&A auditor choice affect the audit quality for the M&A year? We develop our hypothesis H1 regarding this research question in section 2.3.
In addition, the less common choice of M&A auditor also makes the selection of external auditor for the target company's performance commitment attestation in post M&A periods complicated, which means both the annual auditor A by the end of period t-1 and the M&A auditor B in period t have a chance to provide subsequent attestation services. This situation leads us to a second research question: does such a less common M&A auditor choice affect the performance commitment attestation quality in post M&A periods? We develop our hypothesis H2 regarding the second research question in section 2.4.

Hypothesis H1
As shown in Figure 2, if the acquiring company chooses auditor B (i.e. an auditor other than the incumbent annual auditor A) to be M&A auditor, we argue that such a choice may affect the quality of the annual audit for the M&A year in two channels.
In the first scenario, the acquiring company chooses the M&A auditor B to be its annual auditor for the year of M&A. This arrangement offers auditor B with not only the fee revenues for performing the M&A audit but also those for the annual audit (which usually has a larger amount and greater persistence). Because auditor B gains incremental financial benefits from the acquiring company's management by replacing the incumbent auditor A, we expect auditor B would be more inclined to please the management.
In the second scenario, the acquiring company continues to engage the incumbent annual auditor A as the annual auditor for the M&A year. Since the management also hires auditor B to provide the M&A assurance services, auditor A is likely to sense an imminent competition from auditor B on the annual audit engagement, as well as a greater dismissal threat from the management when compared with the scenario in Figure 1. 4 Therefore, auditor A may have an incentive not to challenge management during the annual audit for the M&A year.
To sum up, for either mechanism discussed above, we predict that the acquiring company is more likely to obtain a more favourable audit outcome for the year of M&A compared to the conventional scenario in Figure 1, where the incumbent annual auditor is engaged as the M&A auditor. Accordingly, we propose our first hypothesis as follows: H1. An acquiring company is more likely to obtain an improvement in the annual audit opinion for the year of M&A when it chooses an alternative audit firm (other than its incumbent annual auditor) to perform the M&A audit than when the company engages its incumbent annual auditor to perform the M&A audit.
On the other hand, in the scenario where the acquiring company chooses its incumbent annual auditor A to perform the M&A audit, auditor A's independence could be compromised because the incumbent annual auditor not only has a continued annual audit engagement for year t, but also an additional M&A audit, which means more financial benefits granted by the client management. 5 Therefore, it remains an empirical question whether auditor independence is impaired to a greater extent due to management introducing peer auditor competition (as in Figure 2) or management offering incremental business revenue to the incumbent annual auditor (as in Figure 1). If the net effect is dominated by the latter, it would be difficult to obtain evidence supporting our hypothesis H1.

Hypothesis H2
As shown in Figure 2, if the acquiring company chooses auditor B (i.e. an auditor other than the incumbent annual auditor A) to be M&A auditor, we argue that such a choice may also affect the quality of attestation services on the reported degree of target company's performance commitment fulfilment in post M&A periods.
After the M&A transaction is completed, the target company begins to deliver the performance commitment made in the deal. In post-M&A periods, an acquiring company is required to report the degree of performance commitment fulfilment to the public on a yearly basis, and engages an audit firm to verify the relevant performance with an attestation report. 6 Extant literature shows that if the target company fails to meet the performance commitment goal in post-M&A periods, the stock market reaction is negative for the acquiring company (Wang & Fan, 2017). Several reasons may account for this finding. First, the failure to fulfil performance commitment implies that the target company's profitability is lower than the initial expectation in the M&A deal, which sends a signal of a low-quality M&A and triggers inventors' concerns about management capability to make acquisition decisions. Second, if the target company fails to fulfil the performance commitment, uncertainties arise whether the acquiring company could be compensated. For example, if the shares of the target company are pledged by their pre-M&A shareholders, it would be difficult for the acquiring company to be compensated with the target company's shares (Zhang, 2017). In addition, a failure to fulfil performance commitment may lead to the acquiring company's goodwill impairment, which has a negative impact on the acquiring company's financial performance (Gao & Yuan, 2017). Such an impact goes beyond the cash or equity compensation in case of fulfilment failure of performance commitment. Therefore, the acquirer, once bonded with the target company, has an incentive to make the target company meet performance commitment goals, to avoid the adverse impacts on the acquiring company itself.
In a scenario where the acquiring company appoints auditor B (other than the incumbent annual auditor A) to conduct performance commitment attestation in post M&A periods, auditor B may well recognise the fact that the acquiring company did not engage its incumbent annual auditor for the M&A audit and subsequent performance commitment attestations, and likely perceive to be granted a favour that usually is unexpected. To return the favour, auditor B may tend to cater to the acquiring company's preference when conducting performance commitment attestation. In addition, auditor B may also realise that he has an opportunity to be engaged with annual audits by replacing the acquiring company's incumbent annual auditor A, which may further reinforce auditor B's tendency to meet management's preferred outcome in reported degree of performance commitment fulfilment.
In an alternative scenario where the acquiring company chose auditor B to conduct M&A audit, but employs the incumbent annual auditor A to provide an attestation service on performance commitment fulfilment in post-M&A periods. Compared with the scenario in Figure 1, the annual auditor A in Figure 2 plausibly sense a more imminent competition from a peer auditor (i.e. auditor B) who just provided the M&A assurance service, as well as a higher dismissal threat from the management. Therefore, once the acquiring company appoints auditor A to conduct performance commitment attestation, auditor A may be more inclined to please the management by opining on a favourable attestation outcome.
Taken together, we predict that the acquiring company is more likely to obtain favourable outcomes regarding the post-M&A performance commitment attestations in a scenario where an alternative auditor is introduced (as in Figure 2) than in a scenario where the acquiring company uses its incumbent annual auditor for the M&A assurance services (as in Figure 1). We propose our second hypothesis as follows:

H2:
The acquiring company is more likely to report a higher degree of the target company's performance commitment fulfilment when it chooses an alternative audit firm (other than its incumbent annual auditor) to perform the M&A audit than when the company engages its incumbent annual auditor to perform the M&A audit.

Data source, sample selection and sample distribution
We select listed companies who are acquirers in M&A deals during the year from 2008 to 2017 as our sample observations. In May 2008, the CSRC issued Rules on Material Asset Restructuring, which greatly changed the regulatory environment for M&A transactions. Therefore, we begin our sample period from 2008. We manually collect M&A auditor information by checking M&A transaction reports and other relevant announcements disclosed by the listed companies, and compare it with the companies' annual auditor information.
We start from 1,707 M&A transactions disclosed between 2008 and 2017 on www. cninfo.com.cn, a website where the CSRC requires Chinese listed companies to disclose official information. We drop 102 observations where M&A auditor information is not available. We also drop 179 observations where the listed company is not an acquiring party. To ensure that our empirical tests are consistent with the scenarios discussed in our hypotheses development, we require the annual auditor for year t to either be the same as the annual auditor for year t-1 (a scenario in Figure 1), or switch to the M&A auditor in year t (a scenario in Figure 2). This requirement excludes 63 observations where other auditor switching situations are involved. We further drop 150 observations with missing values. The final sample for modelling M&A auditor choices consists of 1,213 observations. The annual report information and other relevant data are obtained from the CSMAR database.
Among the 1,213 sample observations, 830 (68.4%) acquiring companies choose their incumbent annual auditor to conduct the M&A audit, whereas 383 (31.6%) choose an alternative auditor to perform the M&A audit. The distribution suggests that it is a common practice for acquiring companies to choose an auditor with whom the acquirer is familiar to conduct the M&A audit. On the other hand, some acquiring companies do make a less common choice of the M&A auditor. Panels A and B of Table 1 present the sample distribution by year and industry, respectively.

Modelling the M&A auditor choice
To understand the acquiring company's auditor choice in an M&A deal, we construct the following binary logistic model: In model (1), the dependent variable DIFFAUD t is an indicator variable equal to one if the acquiring company employs an alternative auditor (other than its incumbent annual auditor) to conduct at least one M&A audit in period t, and zero if the acquiring company employs its incumbent annual auditor to provide all M&A audits in period t. 7 Firstly, attributes of M&A transactions may explain the acquirer's M&A auditor choice. As discussed earlier, both parties in an M&A deal have an incentive to appoint their own auditor to perform the M&A audit, thereby reducing information asymmetry and maximising the acquisition benefits. Given this incentive, the larger the size of a target company is relative to the acquiring company, the chances that the target company's annual auditor is appointed to perform the M&A audit are higher (assuming that the target company's annual auditor is usually different from the acquiring company's annual auditor). Therefore, we include RELSIZE t (which equals the total assets of the target company at the time when the M&A deal is disclosed in period t, scaled by the total assets of the acquiring company at the end of period t-1). We expect the coefficient on RELSIZE t to be positive. If a listed company is engaged in multiple M&A transactions in a year, the chances for the company to choose a non-annual auditor are mechanically higher. Therefore, we include MULTIDEAL t (an indicator variable equals to 1 if the listed company is engaged in more than one M&A transaction in period t, and zero otherwise), and expect the coefficient of MULTIDEAL t to be positive. 8 Second, factors documented in prior literature on annual auditor switch may also explain why the acquiring company chooses its incumbent annual auditor to perform an M&A audit. For example, management turnover in the current period may break the business relationship between the company and its incumbent annual auditor (Chow & Rice, 1982). We therefore include MGMTCH t (an indicator variable equal to one if the chairperson or the Chief Executive Officer (CEO) of the acquiring company is changed in period t, and zero otherwise). We expect the coefficient of MGMTCH t to be positive. It is possible that the acquiring company had a disagreement with its incumbent annual auditor in the prior period, which makes the acquiring company choose an alternative auditor. We control for MOD t-1 (an indicator variable equal to one if the incumbent auditor issued a modified audit opinion, and zero otherwise), and expect its coefficient to be  positive. If the acquiring company has employed a larger audit firm as its incumbent annual auditor, it indicates the acquirer to have a higher demand for assurance quality (Francis & Wilson, 1988), which suggests a lower likelihood for the acquiring company to appoint a new auditor in the M&A transaction. Therefore, we include BIG10 t-1 (an indicator variable equal to one if the acquiring company's annual auditor for period t-1 ranks top 10 according to the CICPA national ranking for that year, and zero otherwise), and expect the coefficient on BIG10 t-1 to be negative. Growing companies tend to adjust their demands for external assurance and realign with auditors more often (Johnson & Lys, 1990). We therefore include GROWTH t-1 (i.e. change in revenues from period t-2 to t-1, divided by revenues for period t-2), and expect its coefficient to be positive. Finally, we follow prior literature to include a number of control variables (Lennox, 2000): ROA t-1 for the acquiring company's operating performance (which equals the operating profit for period t-1 scaled by total assets at the end of period t-1), LEV t-1 for the financial leverage (i.e. total liabilities at the end of period t-1 scaled by total assets at the end of period t-1), and SIZE t-1 (i.e. the natural log of total assets at the end of period t-1), along with industry and year fixed effects. To mitigate the impact of outliers, we winsorise all continuous variables at the top and the bottom 1 percentiles. Table 2 reports descriptive statistics of the main explanatory variables in model (1) for the full sample, the regular M&A auditor choice subsample (i.e. DIFFAUD t = 0), and the unconventional auditor choice subsample (i.e. DIFFAUD t = 1), respectively. Compared with the regular auditor choice group, the unconventional auditor choice group is featured with a larger size of the target company relative to the acquiring company (RELSIZE t ), a higher likelihood of modified opinion from the acquiring company's incumbent annual auditor (MOD t-1 ), a higher likelihood of management turnover (MGMTCH t = 1), and the incumbent annual auditor being more likely a small audit firm (BIG10 t-1 = 0). In addition, the acquiring companies that made a less common M&A auditor choice are associated with lower profitability (ROA t-1 ), higher financial leverage (LEV t-1 ) and smaller size (SIZE t-1 ).  (1) has a reasonable explanation power. The coefficient on RELSIZE t is significantly positive (z-stat. = 3.29), which indicates that when the relative size of the target company (to the acquiring company) is larger, the acquiring company is less likely to engage its incumbent annual auditor to perform the M&A audit. The coefficient of MOD t-1 is significantly positive (z-stat. = 2.31), which suggests that the acquiring company is less likely to employ its incumbent annual auditor to perform the M&A audit when the prior-year annual audit opinion is modified. The coefficient of BIG10 t-1 is significantly negative (z-stat. = −2.50), indicating that the acquirer is more likely to use its incumbent annual auditor to perform the M&A audit when the incumbent annual auditor is a large accounting firm. The coefficient of ROA t-1 is significantly negative (z-stat. = −2.81), suggesting that an acquiring company with better performance is more likely to engage its incumbent annual auditor to conduct the M&A audit. Moreover, the coefficients of MGMTCH t and GROWTH t-1 are positive and marginally significant, which suggests that an acquirer with management turnover or with higher growth is more likely to engage a new audit firm. We do not find evidence that a higher frequency of M&A activity in a year (MULTIDEAL t ) significantly increases the incidence of unconventional M&A audit choice.

Research design
To examine whether the unconventional M&A auditor choice affects the annual audit quality for the M&A year (H1), we follow prior studies on audit opinion improvement (Fang & Hong, 2008;Chen et al., 2005) and estimate the model below: In model (2), the dependent variable OPIMPROVE t is an indicator variable equal to one if the severity of the annual audit opinion is lessened in period t from period t-1, and zero otherwise. Consistent with prior literature, the audit opinion ranges from the most severe types (adverse opinion or disclaimer of opinion), less severe types (qualified opinion, followed by unqualified opinion with emphasis-of-matter paragraph, and the least severe type (standard unqualified opinion). The variable of our main interest in model (2) is DIFFAUD t , which is the same as that defined in model (1). We expect the coefficient of DIFFAUD t to be significantly positive if H1 is supported. Because we measure OPIMPROVE as a change variable, we design control variables similarly. If the target company's relative size is larger, the M&A deal may bring the acquiring company more growth and improvement opportunities. We therefore control for RELSIZE t . The management turnover suggests a potential managerial improvement, thus we control for MGMTCH t . A change in audit opinion may also be related to a change in financial position or operating performance (Johnson & Lys, 1990). Therefore, we control for ΔROA t (i.e. ROA t -ROA t-1 ), ΔLEV t (i.e. LEV t -LEV t-1 ), and ΔSIZE t (i.e. SIZE t -SIZE t-1 ), where the definitions of ROA, LEV and SIZE are the same as those defined in model (1). In addition, we control for year fixed effect, whereas the industry fixed effect is differenced out in a change model.
Besides the conventional binary logistic regression, we also report the results based on rare event logistic regression for model (2) (King & Zeng, 2001), considering the low frequency of dependent variable events. Moreover, taking into account the systematic differences between the regular and less conventional M&A auditor choices (i.e. DIFFAUD t = 0 vs. 1), we employ Heckman's two-stage approach by calculating the inverse Mills' ratio (IMR) based on the M&A auditor choice model (i.e. model (1)) and then including IMR in model (2). 9

Regression results of model (2)
In the full sample, the severity of annual audit opinion is lessened in 1.4% observations. Among the observations where the annual audit opinion is improved (OPIMPROVE t = 1), 76.5% choose an alternative audit firm (who is not the incumbent annual auditor) to conduct the M&A audit (DIFFAUD t = 1). In contrast, among observations where the annual audit opinion is not improved (OPIMPROVE t = 0), 30.9% choose an alternative auditor to perform the M&A audit (DIFFAUD t = 1). Univariate tests suggest that the difference (30.9% vs. 76.5%) is significant (t-stat. = −4.03), which offers preliminary support for the hypothesis H1. Table 4 reports the results of conventional binary logistic regression of model (2) (Col. 1), rare-event logistic regression (Col. 2), and Heckman's two-stage approach (Col. 3). Col. 1 shows that the coefficient of DIFFAUD t is significantly positive (z-stat. = 2.42), which indicates that the annual audit opinion for the M&A year is more likely to be improved if the acquiring company employs a non-annual auditor to conduct the M&A audit.
Col. 2 of Table 4 shows that, after adjusting the potential measurement bias caused by the low incidence of the dependent variable event (King & Zeng, 2001), the coefficient of DIFFAUD t is still significantly positive (z-stat. = 2.21). Col. 3 shows that, after considering the potential self-selection bias of DIFFAUD t , the coefficient of DIFFAUD t is still significantly positive (z-stat. = 2.27). At the same time, the coefficient on the inverse Mills' ratio IMR is significant, which justifies our controlling for the potential self-selection bias of the M&A auditor choice. 10

Mechanism analysis for H1
We propose two potential mechanisms when developing hypothesis H1. The first mechanism is that the M&A auditor replaces the incumbent annual auditor and returns the favour to the acquiring company by pleasing the management during the audit for the M&A year (i.e. a real dismissal effect). The second mechanism is the acquiring company engages an alternative auditor to perform the M&A audit but still uses the incumbent annual auditor to perform the annual audit for the M&A year, which makes the incumbent annual auditor sense a stronger dismissal threat and compromise her independence (i.e. a dismissal threat effect). To test the above two mechanisms, we differentiate DIFFAUD t into DIFFAUD_M&AAUD t and DIFFAUD_INCUMAUD t in model (2), with the rest of the model specification unchanged. DIFFAUD_M&AAUD t is an indicator variable equal to one if DIFFAUD t = 1 and the M&A auditor becomes the annual auditor for period t, and zero otherwise. DIFFAUD_INCUMAUD t is an indicator variable equal to one if DIFFAUD t = 1 and the incumbent annual auditor for period t-1 is retained for period t, and zero otherwise. Then, the coefficient on DIFFAUD_M&AAUD t captures the real dismissal effect of the M&A auditor choice, and the coefficient on DIFFAUD_INCUMAUD t captures the dismissal threat effect of the M&A auditor choice.
Statistics show that, among 383 observations where DIFFAUD t = 1, 62 acquiring companies replace the incumbent annual auditor with the M&A auditor to perform an annual audit for period t (i.e. DIFFAUD_M&AAUD t = 1), whereas 321 acquiring companies retain their incumbent annual auditor for period t (i.e. DIFFAUD_INCUMAUD t = 1). The likelihood of the incumbent annual auditor being replaced by the M&A auditor is 16.2% (= 62/383), which supports the notion that the incumbent annual auditor senses competition pressure from the M&A auditor and a stronger dismissal threat from the management. Table 5 presents the regression results of the modified model (2). The coefficients on DIFFAUD_M&AAUD t and DIFFAUD_INCUMAUD t are both significantly positive, which indicates that the unconventional M&A auditor choice not only has the dismissal effect but also the dismissal threat effect on the annual audit quality for the M&A year. Further tests suggest that there is no significant difference between the two coefficients.

Research design
To examine whether the unconventional M&A auditor choice affects the attestation quality of the target company's post-M&A fulfilment of performance commitment (H2), we construct the following models: In model (3), the dependent variable COMPRTO t+n measures the completion degree of the target company's performance commitment, which is specified in the M&A contract in period t for post-M&A period t + n (where n ranges from 1 to 5 according to performance commitment agreements we have collected). We observe three ways to evaluate the target company's performance in M&A contracts: (a) annual assessment, (b) annual cumulative assessment, and (c) cumulative assessment. Below we discuss how we measure COMPRTO t+n based on each evaluation approach. The annual assessment is to compare the actual performance realised in a post-M&A year with the committed performance in the same year. If the realised performance is no less than the committed performance, the performance commitment is deemed as completed for this post-M&A year; otherwise, the target company's previous shareholders need to make compensation to the acquirer. Moreover, the amount of performance in excess of the committed performance in a specific year cannot be carried forward to a subsequent year. In the annual assessment approach, COMPRTO t+n equals the target company's realised performance in post-M&A period t + n divided by the committed performance for post-M&A period t + n stipulated in the M&A contract in period t.
Annual cumulative assessment is to compare the cumulative realised performance up to a post-M&A year with the cumulative committed performance up to the same year, which means that the realised performance in excess of the committed performance in a single prior year can be carried forward into a subsequent year for performance evaluation. However, excessive realised performance in a specific year is not eligible to undo compensation already made in a prior period. In the annual cumulative assessment approach, COMPRTO t+n equals the cumulative realised performance up to post-M&A period t + n divided by the cumulative committed performance up to post-M&A period t + n stipulated in the M&A contract in period t. 11 Cumulative assessment is used to compare the aggregate-realised performance within the whole commitment years with the aggregate-committed performance in the M&A contract. If the aggregate-realised performance is no less than the aggregate-committed performance, the performance commitment is deemed as completed; otherwise, the target company's previous shareholders need to make compensation to the acquirer. In the cumulative assessment approach, COMPRTO t+n equals the aggregate realised performance during the whole commitment years divided by aggregate committed performance for the whole commitment years stipulated in the M&A contract. For cases under this approach, we include in the sample only observations in the final year (but not earlier years) of the whole commitment period.
In model (4), the dependent variable COMPDUM t+n indicates whether the performance commitment for period t + n is fulfiled, which is an indicator variable equal to one if COMPRTO t+n is no less than one, and zero otherwise. DIFFAUD t is defined in the same way as in model (1).
We include a number of control variables in models (3) and (4) that capture contract attributes of M&A transactions in period t. First, considering that different performance assessment approaches 12 are accompanied with various levels of difficulty in fulfiling performance commitment goals, we control for the performance assessment approaches by including ANNCUMUL t (an indicator variable equal to one if the performance assessment approach is the annual cumulative assessment, and zero otherwise) and CUMUL t (an indicator variable equal to one if the performance assessment approach is cumulative assessment, and zero otherwise).
Second, how the key performance indicator is measured also affects the fulfilment of performance commitment. Compared with indicators such as net income after deducting non-recurring items (or the lower of net income after deducting non-recurring items and net income), net income is more likely to be subject to manipulation. Therefore, we control for TARGMEASURE t (an indicator variable equal to one if the performance indicator involves deduction of non-recurring items, and zero otherwise).
Third, the monetary amount of committed performance also affects the difficulty to fulfil the performance commitment. The larger the amount of committed performance, the more difficult it is to fulfil the performance commitment. Therefore, we control for TARGPERF t , which equals the committed performance amount for post-M&A period t + n scaled by the target company's total asset at the time of M&A transaction.
Fourth, we observe that an M&A contract is more likely to include a bonus term to reward excessive performance when the performance commitment goal is more difficult to achieve. We control for REWARD t , which is an indicator variable equal to one if the M&A contract contains a bonus term for excessive performance, and zero otherwise.
Fifth, the information asymmetry is relatively low when an M&A transaction involves related parties, which makes the performance commitment goal easier to achieve. Therefore, we control for RPT t which is an indicator variable equal to one if the acquirer and the target company in an M&A deal are related parties, and zero otherwise.
In addition to M&A contract attributes, we control for the type of auditors who perform attestation services in each post-M&A year. Specifically, we control for BIG10 t+n , which is an indicator variable equal to one if the auditor in charge of performance commitment attestation for period t + n ranks top 10 per the CICPA national ranking, and zero otherwise.
Finally, we control for year fixed effects, industry fixed effects, and post-M&A period fixed effects in models (3) and (4).

Sample selection
We first hand-collected 4236 observations where listed companies report the completion degree of the target company's performance commitment during the period 2008-2017. We take the following steps in sample selection: (i) We drop 405 observations where the listed company is not an acquirer, but a target company; (ii) We drop 231 observations where the identity of the M&A auditor is not available; (iii) We drop 196 observations where the identity of the post-M&A auditor who performs the attestation is not available; (iv) We drop 408 observations where the attestation service is neither performed by the incumbent annual auditor nor by the M&A auditor 13 ; (v) To ensure the comparability among attestation opinions, we drop 12 observations receiving a non-standard attestation opinion; (vi) As nearly all performance commitment agreements use earnings as the key performance indicator, we drop one observation where a non-earnings indicator is used (i.e. the target company's book value of total assets); (vii) We drop 131 observations with missing values of control variables, which is mainly due to the missing value for the target-company's total assets at the time of the M&A deal (used for calculation of TARGPERF t ).   Table 6 presents the descriptive statistics of the main variables in models (3) and (4). In the full sample, the average reported completion degree of performance commitment (COMPRTO t+n ) is 1.080, and the likelihood of fulfiling performance commitment (COMPDUM t+n ) is 85.1%. 458 observations (16%) adopt the annual assessment approach, 2,321 observations (81.4%) adopt the annual cumulative assessment approach (ANNCUMUL t = 1), and 73 observations (2.6%) use the cumulative assessment approach (CUMUL t = 1). 2,382 observations (83.5%) measure their key performance indicator by considering the deduction of non-recurring items. On average, the monetary amount of performance commitment (TARGPERF t ) is set as 72.5% of the target company's total assets at the time of the M&A deal. 961 observations (33.7%) contain a bonus term for excessive performance in the M&A contract (REWARD t ). 1,717 observations (60.2%) are related-party M&A transactions (RPT t ). The mean DIFFAUD t is 0.302, which indicates that 30.2% of the observations chose an alternative auditor (other than the incumbent annual auditor) to conduct the M&A audit. About 55.8% observations have a top-10 auditor perform the attestation service for a post-M&A period (BIG10 t+n ).

Regression results of models (3) and (4)
Untabulated univariate analysis shows that in the regular auditor choice group (i.e. DIFFAUD t = 0), the mean (median) completion degree of performance commitment (COMPRTO t+n ) is 1.065 (1.045), and the likelihood of fulfiling performance commitment (COMPDUM t+n ) is 83.7%; in contrast, in unconventional auditor choice group (i.e. DIFFAUD t = 1), the mean (median) COMPRTO t+n is 1.114 (1.057), and the mean COMPDUM t+n is 88.3%, which are both significantly higher than those in the regular auditor choice group at 1% level. Table 7 reports the regression results of the models (3) and (4). For both dependent variables COMPRTO t+n and COMPDUM t+n , the coefficients in DIFFAUD t are significantly positive (t-stat. = 2.34, z-stat. = 2.67). These findings suggest that if the acquiring company chose an alternative auditor (other than its incumbent annual auditor) to conduct the M&A audit, the acquirer can report a higher completion degree of performance commitment, and is more likely to fulfil the performance commitment for a post-M&A year. Note that these outcomes are accompanied by a clean attestation opinion from the auditor, we conclude that auditors who perform the attestation services offer more favourable opinions to the management when the M&A auditor is not the acquirer's incumbent annual auditor. Therefore, the results in Table 7 lend support to hypothesis H2.
In terms of control variables, the coefficients of ANNCUMUL t are significantly positive (t-stat. = 4.59, z-stat. = 6.12), which indicate that the performance commitment is easier to fulfil when the target company's performance is evaluated on an annual cumulative basis. The coefficients on RPT t is significantly positive (t-stat. = 4.21, z-stat. = 2.90), which suggests that if the M&A deal is a related-party transaction, the performance commitment is more likely to be fulfiled.

Mechanism analysis for H2
For the 861 observations who make an unconventional M&A auditor choice, there are three scenarios of auditor attestation to post-M&A fulfilment of performance commitment. First, there are 84 observations (9.8%) where the M&A auditor performs the post-M&A attestation and the acquiring company retains its incumbent annual auditor. Second, there are 222 observations (25.8%) where the M&A auditor performs the post-M&A attestation and the acquiring company replaces the incumbent annual auditor with the M&A auditor. Third, there are 555 observations (64.4%) where the acquiring company's incumbent annual auditor performs the post-M&A attestation.
In the first scenario, the M&A auditor is hired to perform the attestation and may be aware of the opportunity to further obtain the annual audit by replacing the incumbent annual auditor. Therefore, the M&A auditor may have the incentive to cater to managerial preferences. In the second scenario, the M&A auditor replaces the incumbent annual auditor, which may also motivate the M&A auditor to return the favour from the management by catering to managerial preference. In the third scenario, the incumbent annual auditor is more likely to sense the dismissal threat from management as compared to the scenario where the incumbent annual auditor has been engaged to perform the M&A audit in the first place, which makes the incumbent annual auditor incentivised to offer a favourable attestation outcome to management.
To test the above mechanisms, we differentiate DIFFAUD t in models (3) and (4) into AUD1 t+n , AUD2 t+n , and AUD3 t+n (with each variable equal to one for the first, second, and third scenario, respectively, and zero otherwise), and rerun the regressions. Table 8 reports the regression results. Although all three variables (AUD1 t+n , AUD2 t+n , and AUD3 t+n ) show a positive sign, only the coefficient of AUD3 t+n is significant (t-stat. = 2.42, z-stat. = 2.49). These findings suggest that the dismissal threat effect dominates the results in Table 7. That is, if the acquiring company chooses an alternative audit firm to perform the M&A audit, such an arrangement has a continuously intimidating effect on the incumbent annual auditor.

Findings and implications
We examine the M&A auditor choices in the Chinese capital market from 2008 to 2017, and find that if the acquiring company employs an alternative audit firm (other than the incumbent annual auditor) to conduct the M&A audit, both the annual audit quality for the M&A year and the performance commitment attestation quality in post-M&A periods are impaired. We attribute the impairment of external auditor's independence to business competition between two peer audit firms, which is introduced by the acquiring company. The competition makes the incumbent annual auditor perceive a credible dismissal threat from the management, and incentivises the newly introduced M&A auditor to compromise independence for incremental fee revenues.
It is worth pointing out that our findings do not require or assume that the acquiring company's management is necessarily intentional to introduce competition by choosing an alternative auditor to perform the M&A audit. In other words, regardless of whether an acquiring company intentionally makes an M&A auditor choice, as long as the acquirer chooses an alternative auditor to provide the M&A assurance service, competition will be introduced and weaken the external monitoring from auditors.
Our study has implications for information users, including regulators, to pay attention to a listed company's auditor choice in an M&A transaction, and to subsequent annual audit quality and performance commitment attestation quality when an alternative auditor other than the incumbent annual auditor is appointed to conduct the M&A audit. Broadly speaking, our study suggests that, when faced with multiple demands for external assurance services, the acquiring company's management apparently has discretion in choosing external auditors, and is capable of exerting a negative impact on the quality of external assurance services.

Limitations and further research opportunities
We are motivated by relatively scant prior research on the acquiring company's M&A auditor choice. Therefore, our exploration of the topic is preliminary, which is subject to future research for a better understanding of more determinants. When testing hypothesis H1, we are aware that the proportion of observations with audit opinion improvement in the full sample is relatively low (1.4%). One potential concern is that the dependent variable examined is trivial. However, given the low incidence of non-standard audit opinion for Chinese listed companies (2%-3%), it is understandable for audit opinion improvement percentage to be small. In addition, the descriptive statistics of OPIMPROVE show that the incidence of annual audit opinion improvement is 0.5% (3.4%) in the regular (unconventional) M&A auditor choice group, and the difference between these two groups is significant both statistically and economically. Third, as shown in Table 1, a majority of the unconventional M&A auditor choice took place in more recent years (70.6% over the period 2015-2017 vs. 29.4% during 2008-2014), which suggests that the phenomenon deserves attention. Finally, besides annual audit opinion, we investigate other aspects such as the performance commitment attestation in post-M&A periods to triangulate annual audit-based evidence. Nevertheless, we may not identify all the consequences following the (unconventional) M&A auditor choice in a single study, which warrants future research.