Major asset restructuring performance commitments and classification shifting through non-recurring items

ABSTRACT We examine whether firms engage in classification shifting to meet performance targets during mergers and restructuring. Using a sample of listed firms that complete major asset restructuring and sign performance commitment agreements from 2008 to 2019, we find that during the commitment period, nearly 39% of firms ‘step on the line’ to achieve net income before non-recurring items, i.e., the realised performance slightly exceeds the promised performance target. Compared to control firms and non-commitment years, firms that ‘step on the line’ to meet the target are more likely to achieve this by misclassifying recurring expenses as non-operating losses. Furthermore, this effect is more pronounced in firms with larger committed amounts, firms using stock to compensate for non-performance, and firms audited by non-Big 4 auditors. Overall, our paper extends the research on incentives for classification shifting and has implications for regulators to strengthen the regulation of accounting treatment in performance commitments.


Introduction
Mergers and acquisitions (M&As) are an important way for a firm to achieve industrial upgrading and operating synergies.To facilitate the completion of M&A deals, more and more listed firms are introducing performance commitments in M&As.Zhai et al. (2019) estimate that the proportion of M&As that include performance commitments in all M&As has increased from less than 1% in 2011 to 40% in 2015.Theoretically, the performance commitment is a useful protection mechanism for acquirers when there is information asymmetry between the acquirer and the target firm.Specifically, by agreeing the future earnings of the target asset and the specific compensation method in the event of nonperformance ex ante, performance commitments can, to a certain extent, reduce the risk of inaccurately valuing targets and improve the fairness of M&A transactions, thus protecting the interests of acquirers and promoting the healthy development of capital markets.However, with the arrival of the performance commitment period, the fulfilment of promised performance has presented more and more problems: the accurate realisation of the promised performance during the commitment period and the reversal of performance immediately after the expiration of the commitment period occur frequently.According to a report in Shanghai Securities News,1 among 360 major asset restructurings completed in the Shanghai Stock Exchange from 2015 to 2019, nearly 60% of firms had performance completion ratios (measured as the realised performance divided by promised performance) between 100% and 110%.The phenomenon of 'stepping on the line' to meet the performance target was evident.Some firms performed well during the commitment period, but profits fell sharply once the commitment period expired.For example, a pharmaceutical firm completed a major restructuring in 2016.During the three-year performance commitment period, the firm 'stepped on the line' to meet its performance targets.However, after the commitment period expired, its performance dropped by more than 70%.Its financial statements were also issued with an unqualified audit opinion.This raises the question: Is the target firm's accurate realisation of the promised performance a coincidence or the result of management manipulation?Does management have the incentive to meet the performance target through accounting manipulation?
To this end, in this paper, we examine whether listed firms engage in earnings manipulation to meet the performance target.The study of this issue relates to whether the performance commitment in M&As has ex-post credibility, and thus to the effectiveness of M&A market operation, and thus to the effectiveness of resource allocation in China's capital market.In particular, we examine whether listed firms engage in classification shifting to meet the performance target.Classification shifting is a kind of earnings management tool through which managers adjust the classification of items within the income statement to change the amounts of different items and mislead related stakeholders.For example, a firm can misclassify core expenses as non-recurring items to inflate core earnings.The performance target in M&A performance commitments is usually linked to net income before non-recurring items.To increase net income before non-recurring items, firms can take two approaches.First, they can manipulate nonrecurring items.By changing the classification of non-recurring items, for example, reclassifying recurring expenses as non-recurring losses, net income before nonrecurring items will increase.Second, they can manipulate net income.By increasing net income through accruals management and real activity earnings management to increase net income, net income before non-recurring items will also increase.Since the first approach corresponds directly to net income before non-recurring items, managers are more likely to choose the first approach, i.e. classification shifting, to meet the performance target.
In addition, compared with traditional earnings management tools (accruals management and real activity earnings management), classification shifting is endowed with the following advantages: (1) Low cost.Accruals management mainly increases current earnings by accelerating revenue recognition or decelerating expense recognition.Since it involves 'settling up' between different accounting periods, an increase in current earnings generally implies a reversal of future earnings (Baber et al., 2011).On the other hand, real activity earnings management increases current earnings mainly through price discounts, overproduction, and cuts in discretionary expenses.Because these real activities deviate from a firm's normal business practices, an increase in current earnings is usually accompanied by a decline in future operating performance (Cohen & Zarowin, 2010).In contrast, classification shifting does not affect future earnings.Managers can inflate current core earnings by simply changing the classification of recurring and nonrecurring items within the income statement, greatly reducing the cost of earnings management.(2) Concealment.Since classification shifting does not change bottomline earnings (net income), it is less likely to be scrutinised by regulators and auditors (Nelson et al., 2002).Moreover, in China, the definition of non-recurring items is mainly enumerative and the distinction between recurring and non-recurring items is ambiguous (X.Fan & Zheng, 2009), which leaves room for management to engage in classification shifting.As a result, we expect that listed firms with performance targets linked to net income before non-recurring items have a strong incentive to engage in classification shifting to meet the target.
To test the impact of 'using net income before non-recurring items' as the performance metric in M&As on firms' classification shifting through non-recurring items, we identify 667 major asset restructurings that include performance commitments for A-share firms listed in the Shanghai and Shenzhen Stock Exchanges from 2008 to 2018.Then, we use the 'firm-year' observations composed of three years before the start and after the end of performance commitment periods for these 667 major asset reorganisation events as the research sample.Based on the sample, we find: (1) during the performance commitment period, nearly 39% of firms 'step on the line' to achieve the promised net income before non-recurring items, i.e. the realised net income before non-recurring items slightly exceeds the promised net income before non-recurring items.(2) The positive relation between abnormal net income before non-recurring items and non-operating losses is more pronounced in firms that 'step on the line' to meet the target, and this positive relation is mainly caused by more hidden other non-operating losses, suggesting that firms 'stepping on the line' to meet the target are more likely to boost their net-income before non-recurring items by misclassifying recurring expenses as non-recurring losses, especially other non-operating losses.(3) After controlling for accruals management and real activity earnings management, we still find the classification shifting effect, suggesting that classification shifting has incremental 'contribution' to performance achievement even after taking into account traditional earnings management.Besides, our findings still hold after a series of robustness tests including changing the sample period, changing the 'stepping on the line' interval, and replacing non-operating losses with non-recurring losses.(4) Cross-sectional analyses suggest that the effect is more pronounced in firms with larger committed amounts and firms that use stock to compensate for nonperformance, and Big 4 auditors play a monitoring role in curbing classification shifting.
(5) Economic consequence analysis shows that, after the expiration of the commitment period, the performance of firms that 'step on the line' to meet the target declines significantly.This provides further support for the view that firms engage in earnings manipulation during the commitment period.
Our paper contributes to the literature in several ways.First, our paper adds to the research on earnings management.The existing literature generally focuses on how firms can increase net income through accruals management and real activity earnings management.In contrast, research on how firms increase core earnings through classification shifting is much less.As scholars and regulators have paid more attention to traditional earnings management in recent years, firms have less room to manipulate earnings through these tools.In this case, whether firms will manipulate earnings through other less investigated and more hidden tools, such as classification shifting, becomes an important issue to examine.Some studies have examined classification shifting in the circumstances of catering to capital market expectations, equity offerings, and management equity incentives (Lu et al., 2019;McVay, 2006;Xie et al., 2019), but there is no literature examining classification shifting in the context of M&A performance commitments.Considering the important role of M&A performance commitments in influencing the efficiency of resource allocation in capital markets, examining whether firms will engage in classification shifting in this context becomes an important issue worth studying.Our study based on this context also helps us gain a better understanding of the incentive and means of classification shifting.
Second, our paper extends the research on the economic consequences of performance commitments.Existing studies have examined the impact of performance commitments on M&A performance (Lv & Han, 2014;Pan et al., 2017;Yang et al., 2018), market valuation (Zhai et al., 2019) and investor protection (Dou & Zhai, 2020, Li et al., 2020;Gui et al., 2011).However, little attention has been paid to the fulfilment of performance commitments and the resulting earnings management, which is of great concern to investors.Although G. Zhang et al. (2020) and X. Liu and Wu (2021) find that performance commitments in M&A transactions induce more earnings management, they ignore a more hidden form of earnings management -classification shifting.They also do not conduct an in-depth analysis of specific performance metrics and the fulfilment of promised performance.By focusing on the unique perspective of classification shifting, our paper examines firms' earnings manipulation during the commitment period, thus enriching the literature on the economic consequences of performance commitments.
Finally, our findings are informative to regulators and investors.Our paper finds that listed firms tend to engage in classification shifting to meet their performance targets.This finding has implications for both regulators and investors.For regulators, they should consider the reasonableness of performance commitments and earnings management behaviour caused by such commitments, and improve the regulation of non-recurring item disclosure.For investors, they should be concerned about the true value of target assets and the reliability of performance, and beware of earnings manipulation behaviour conducted by firms to meet performance targets.
The remainder of the paper is organised as follows: Section 2 reviews the related literature.Section 3 proposes our main hypotheses.Section 4 introduces the research design.Section 5 examines the relationship between performance commitments and classification shifting.Section 6 conducts robustness checks.Section 7 presents further analysis, including cross-sectional analyses and economic consequence analyses.Section 8 concludes and makes policy suggestions.

The economic consequences of performance commitments
The Performance commitment originated from the Share Split Structure Reform (SSSR, for short).At that time, in order to ensure the smooth implementation of the reform and enhance the liquidity of the stock market, the regulator encouraged group firms to inject high-quality assets into listed firms through acquisitions and major restructurings.At the same time, to protect the interests of small and medium shareholders, the regulator required listed firms to make specific commitments on future business goals.Thus, the commitments in the SSSR period were mainly designed to alleviate the agency problem between dominant non-tradable shareholders and tradable minorities.Based on the SSSR setting, early scholars examine the impact of performance commitments on the protection of minority shareholders' interests.Gui et al. (2011) find that commitments included in the SSSR reduce the reform costs of non-tradable shareholders and gain the support of tradable shareholders, leading to Pareto improvements.Xu et al. (2008) find that commitments play an important signalling role in helping non-tradable shareholders signal firm quality to tradable shareholders.However, some scholars find negative effects of performance commitments.For example, H. Liu et al. (2011) find that firms making commitments in the SSSR have an incentive to engage in earnings management through nonrecurring items. 2 Hou et al. (2015) find that large shareholders may engage in accruals management to meet performance targets.
The regulator's preference for performance commitments has continued into M&A restructuring.In 2008, the China Securities Regulatory Commission (CSRC) issued the Measures for Administration of Major Asset Restructuring of Listed Firms (the Measures, for short), stipulating that the bidder and the target should sign a performance commitment compensation agreement if the target asset is valued using the future-earnings-based method.Since then, performance commitments have been widely used in M&A transactions.The reason for the regulator to introduce performance commitments are (1) to reduce information asymmetry between the acquirer and the target, thus reducing the risk of inaccurately valuing targets; (2) to alleviate the post-merger adverse selection problem of the target firm and motivate the management of the target firm to work harder to enhance firm value; and (3) to protect the interests of the acquirer and investors.However, empirical evidence is mixed as to whether performance commitments have achieved the desired effect.
Some scholars find that performance commitments have positive effects.For example, Lv and Han (2014) find that performance commitments have a signalling effect and generate synergy gains.With the introduction of performance commitments, both the takeover premium and acquirers' announcement returns increase.Pan et al. (2017) use change in return on assets to measure merger performance and find that performance commitments have an incentive effect and can motivate the target firm to improve its performance.However, because speculation is prevalent in China's capital market and accounting earnings is susceptible to manipulation, it is doubtful whether performance commitments really generate synergistic and incentive effects.
Recently, some scholars have found that performance commitments have negative effects.In the context of M&As, the main agency problem has shifted from the conflict between 'large shareholders' and 'minority shareholders' during the SSSR period to the conflict between 'large shareholders + the committed party (usually shareholders of the target firm)' and 'minority shareholders'.Driven by interests, large shareholders and the committed party have an incentive to set inflated promises, thus pushing up the valuation of target assets and the share price of listed firms.This creates a 'high promise, high valuation, and high share price' interest chain (J.Wang & Fan, 2017;Zhai et al., 2019).Benefiting from the high valuation and high share price, large shareholders with information advantages can take advantage of performance commitments and sell shares at the right time to gain benefits.In contrast, minority shareholders suffer significant losses due to stock purchases (Dou & Zhai, 2020).J. Li et al. (2020) also find that the opportunistic behaviour of stakeholders resulting from performance commitments further leads to stock price crash risk.In addition, X. Liu et al. (2018) and Z. Wang et al. (2021) explore the effect of performance commitments on auditor behaviour.They find that managers may engage in accruals management and real activity management to meet performance targets, which increases audit risk and leads to more audit work and higher audit fees.
Taken together, the existing literature has studied performance commitments in the contexts of SSSR and M&As, respectively.Although some studies find that performance commitments facilitate M&A completion, reduce transaction costs, and protect investors' interests, more evidence suggests that performance commitments push up the valuation of target assets and the share price of listed firms, imposing significant losses on minority investors.In addition, the existing literature pays little attention to the design of performance metrics and the fulfilment of promised performance.Although J. Wang and Fan (2017) study the realisation of performance and suggest the possibility of performance manipulation, they do not provide direct evidence.Several studies provide evidence of earnings management in target firms (X.Liu & Wu, 2021;X. Liu et al., 2018;G. Zhang et al., 2020), but they ignore a more hidden form of earnings management -classification shifting, and do not conduct an in-depth analysis of specific performance metrics.By examining firms' classification shifting behaviour during the performance commitment period, our paper not only enriches the literature on the economic consequences of performance commitments, but also provides empirical evidence for regulators to improve the system related to M&A performance commitments.

Classification shifting
In recent years, classification shifting, as a new earnings management tool, has received more and more attention from academics.Unlike accruals management and real activity management which aim to increase net income, classification shifting overstates core earnings by changing the classification of items within the income statement.Specifically, firms can misclassify recurring expenses as non-recurring losses, or misclassify nonrecurring gains as recurring revenue, to overstate core earnings (McVay, 2006).
Current foreign research on classification shifting focuses on the specific means by which firms conduct classification shifting and the motivations for classification shifting.McVay (2006) is the first to study the specific means by which firms conduct classification shifting.She finds that listed firms opportunistically classify recurring expenses (costs of goods sold (COGS) and selling, general, and administrative expenses (SGA)) as income-decreasing special items to inflate core earnings.Y. Fan et al. (2010) reached similar conclusions based on quarterly data.Y. Fan and Liu (2017) further decompose recurring expenses into COGS and SGA components, and find that managers misclassify both COGS and SGA as special items to inflate earnings.In addition to using special items, Barua et al. (2010) find that firms also misclassify recurring expenses as discontinued operations to inflate earnings.As for the motivations for classification shifting, foreign studies mainly focus on stock market-based incentives.They find that firms engage in classification shifting to meet analysts' earnings expectations (Y.Fan et al., 2010;McVay, 2006) and to obtain high equity valuations prior to IPOs (J.Liu et al., 2020).Recently, several scholars have examined other incentives for classification shifting.For example, Y. Fan et al. (2019) extend the stock-market-based motivation into debt-market-motivation by showing that firms with debt covenants based on earnings before interest, taxes, depreciation, and amortisation (EBITDA) are more likely to engage in classification shifting to avoid debt covenant violations.Based on a sample of Korean listed firms, Chung et al. (2020) find that controlling shareholders are less likely to engage in classification shifting to reduce core earnings and obtain tax benefits.
Compared with foreign studies, research on classification shifting in China started relatively late.Unlike foreign studies, there is no widespread classification shifting behaviour among Chinese listed firms (Xie et al., 2019;Z. Zhang & Zhang, 2012).Domestic scholars have studied classification shifting mainly based on particular scenarios.Some scholars examine the impact of regulatory changes on firms' classification shifting behaviours.For example, X. Li et al. (2015) find that after the performance indicator for seasoned equity offerings changed from 'net income' to 'net income before nonrecurring items', listed firms are more likely to engage in classification shifting to increase core earnings before public offerings.Lu et al. (2019) reach similar conclusions based on the IPO setting.Ye and Zang (2016) find that after the introduction of the 'Eight-point Policy', state-owned enterprises (SOEs, hereafter) are more likely to misclassify 'other cash paid for operating activities' items cash as more hidden 'inventory' items to avoid regulatory supervision.Other scholars examine firms' classification shifting behaviours in the context of equity incentives.In the design of equity incentives, the performance assessment metric should exclude non-recurring gains and losses if it is accounting earnings.B. Liu et al. (2016) find that this design induces firms to use classification shifting to meet the conditions for exercising options and increase stock prices.Xie et al. ( 209) also find that managers will 'step on the line' to meet the conditions for exercising options by misclassifying recurring expenses as non-recurring losses.
In summary, previous studies have examined firms' classification shifting behaviour from capital market incentives, contract incentives, and regulatory incentives.However, few studies have examined the impact of M&A performance commitments on firms' incentives to engage in classification shifting.By exploring firms' classification shifting behaviour in the context of M&As, our paper not only enriches the understanding of the motivations for classification shifting, but also has important implications for regulators, investors, and auditors in identifying the true performance of firms.

Theoretical analysis and hypothesis development
To reduce the information asymmetry between the acquirer and the target in M&A, reduce the risk of inaccurately valuing target assets, and protect the interests of investors, the CSRC issued the Measures in 2008.This policy requires both the M&A parties to sign a performance commitment agreement if the target asset is valued using a future-earnings-based method.In a typical performance commitment, the target firm promises to achieve certain performance targets in future years, and agrees to compensate the acquirer in cash or equity if the realised performance falls below the target.Since the release of the policy, performance commitments have been widely used in M&A transactions.Although in 2014, the CSRC relaxed the regulation on performance commitments and only required certain types of M&As to include performance commitments, performance commitments are still common in M&A transactions and are on the rise year by year (Zhai et al., 2019).This is because performance commitments have a signalling effect.Not signing a performance commitment can easily raise the acquirer's doubts regarding the future profitability of the target firm and thus leads to M&A failure.
After signing a M&A performance commitment, a firm has a strong incentive to take various measures to achieve the performance target.First, from the perspective of performance compensation, failure to meet the performance target will trigger penalties.Under the performance commitment arrangement, the target firm is required to compensate the acquirer in cash or stock when the realised earnings fall short of the performance target.Secondly, from the perspective of regulatory pressure, failure to meet the performance target will bring a loss of reputation and penalties from regulators.According to Article 54 of the Measures, if the performance achieved by the target firm does not reach 80% of the promised performance, public firm managers and related intermediaries shall make a public explanation and apologise to the investors; if the achieved performance does not reach 50% of the promised performance, public firms and relevant personnel shall be subject to supervisory talks, warning letters, and periodic reports.Prior literature shows that the SEC/CSRC sanctions have a series of negative effects on the penalised firms, such as a fall in share price (G.Chen & Gao, 2005), a decline in reputation (Cu & Xia, 2012;Karpoff et al., 2008) and a reduction in financing (X.Liu & Chen, 2018).Therefore, failure to meet the performance target tends to have significant negative effects on the target firm.Thirdly, from the perspective of goodwill impairment, failure to meet the performance target will lead to significant impairment of goodwill, which negatively affects the listed firm's performance and increases stock price crash risk (Yuan et al., 2020;H. Zhang et al., 2020;Li et al., 2020).J. Zhang (2017) cites a case of significant goodwill impairment due to failure to meet the promised performance: in the last year of the performance commitment, Ganfeng Lithium recorded a goodwill impairment loss of more than 200 million yuan due to a significant decline in the performance of the target asset.The impairment amount accounted for 85.34% of the original value of goodwill and 46.77% of the net profit in that year, which had a significant impact on the listed firm's performance and stock price.This, in turn, reduced the value of shares held by the target shareholders.Taken together, to avoid triggering compensation, reduce the possibility of being regulated, and reduce welfare losses, the target firm has a strong incentive to meet its performance target by various means.
However, if the target firm is weak in financial performance, it may resort to earnings management tools to meet the performance target.In most M&A deals, the performance commitment agreement is tied to net income and net profit before non-recurring items.Moreover, since net income before non-recurring items is more representative of the firm's sustainable profitability, net income before non-recurring items accounts for a large proportion of all performance metrics.In our sample, M&A deals with performance metrics tied to net income before non-recurring items account for 94.5% of all M&A deals.In such cases, manipulating earnings using classification shifting through non-recurring items is a more likely tool for management than traditional earnings management tools.
First, classification shifting through non-recurring items is more directly related to net income before non-recurring items than traditional earnings management methods.To reach the net income before non-recurring items, firms can take two approaches: First, they can start from non-recurring items, increasing net income before non-recurring items by classifying recurring expenses as non-operating expenses; secondly, they can start from net profit, increasing the overall net profit through accruals management and real activity management.Since the first approach corresponds directly to net income before non-recurring items, it is easier and more direct to adopt the first approach to achieve the earnings target.Second, classification shifting through non-recurring items does not reduce future earnings.Accruals management increases earnings by advancing or postponing the recognition of revenues and expenses, which can lead to performance reversals in the future (Baber et al., 2011).Real activity management increases earnings through price discounts, overproduction and cutting discretionary expenses, which sacrifices the firm's future benefits (Cohen & Zarowin, 2010).For firms that enter into performance commitments, the performance commitment period is usually three years.Although a firm can achieve the promised performance in a given year through accruals management or real activity management, it cannot address the negative effects of accrual reversals or declining performance in the remaining years.In contrast, classification shifting does not change future earnings.A firm can directly boost its net income before nonrecurring items by classifying recurring expenses as non-recurring losses.This method is not only easy to implement, but also does not have the 'settling up' issue across different accounting periods.Third, classification shifting through non-recurring items is more hidden and harder to detect by auditors and regulators than traditional manipulation methods.The existing literature finds that classification shifting is subject to less scrutiny of auditors and regulators because it does not change the bottomline earnings (Nelson et al., 2002).In addition, the current classification of nonrecurring gains and losses in China uses the enumeration method, and the division between recurring and non-recurring gains and losses is ambiguous (X.Fan & Zheng, 2009).This makes it more difficult for auditors and regulators to detect classification shifting.
Overall, compared with other earnings management methods, classification shifting through non-recurring items is the most beneficial and least costly earnings management method.Thus, we expect that managers have a strong incentive to engage in classification shifting to meet the target when performance metrics specified in the M&A agreement are tied to net income before non-recurring items.We propose the following hypothesis.
H1: Other things being equal, firms are more likely to engage in classification shifting through non-recurring items to meet the performance target when the performance target specified in M&A performance commitments is tied to net income before nonrecurring items.

Sample selection and data source
Our initial sample includes all completed major asset restructurings of Chinese listed firms from 2008 to 2018.The M&A event selection period starts in 2008 because, in April 2008, the CSRC released the Measures, stipulating that listed firms should sign performance commitment agreements in major asset restructuring deals, as described earlier.Therefore, 2008 became the origin year of M&A performance commitments.The M&A event selection period ends in 2018 because the performance commitment period is generally 3 years and we want to ensure that there is enough data to conduct empirical analyses.We focus on major asset restructurings because we want to investigate whether target firms have an incentive to engage in classification shifting during the commitment period.However, the post-merger financial statement data of target firms are not publicly available and are difficult to obtain.As a result, we adopt an alternative approach, i.e. using the financial data of listed firms (acquirers) to examine the classification shifting behaviour of target firms.Considering that in a major asset restructuring, the total asset or sales of the target firm usually account for more than 50% of the pre-merger asset or sales of the listed firm, which has a significant impact on the performance of the listed firm, we believe that it is reasonable to use the overall performance of the listed firm to measure the classification shifting behaviour of the target firm.We are also aware of the measurement error associated with this alternative approach.Thus, in the subsequent robustness tests, we conduct a series of analyses to mitigate measurement errors.
Then, we select M&A events according to the following criteria: (1) keeping only the earliest restructuring if a firm undergoes multiple major asset restructurings in the M&A event selection period.We do so to reduce the impact of different M&A events.(2) excluding M&A events with performance commitment periods beyond 2020.(3) excluding M&A events with missing performance commitment data.Finally, these requirements yield 667 M&A events for 667 listed firms.Table 1 details the selection process of M&A events.Table 2 further presents the distributions of M&A events by the year of M&A completion.Most M&As are completed in 2015 and 2016, which is in line with the merger wave in China.Data on M&A performance commitments are obtained from the China Security Market and Accounting Research (CSMAR) database.3Among 667 M&A events, M&A events with performance metrics tied to net income before non-recurring items account for about 94.45% of all M&A events, indicating that net income before nonrecurring items is a common metric in performance commitment agreements.
Based on the selected 667 M&A events, we construct the research sample in this paper, which consists of 'firm-year' observations from three years before the start of performance commitment periods to three years after the expiration of performance commitment periods.We choose three years before and after the commitment period as the control period because the commitment period is usually three years and firms have no incentive to engage in classification shifting to meet performance targets before the start and after the end of the commitment period.To facilitate understanding, we draw Figure 1 to explain the composition of the sample period.Taking Firm A as an example, if its performance commitment period is 2011, 2012 and 2013, its observation years consist of three years before the start of the performance commitment (2008)(2009)(2010), years during the performance commitment period (2011-2013), and three years after the expiration of the performance commitment period (2014)(2015)(2016).After excluding observations in the financial industry and observations with missing variables, we obtain the final sample consisting of 4379 'firm-year' observations.Table 3 details the sample selection process.Financial data are obtained from the CSMAR database.Industry classification is based on

"Stepping on the line" to meet the performance target: primary evidence
To test whether listed firms will engage in classification shifting to meet the performance target, we depict the distribution of the performance completion ratio of listed firms.We calculate the performance completion ratio as the realised net income before nonrecurring items divided by the promised net income before non-recurring items.The performance completion ratio is equal to 100% if the realised net income before nonrecurring items is exactly equal to the promised net income before non-recurring items.If the sample size increases suddenly when the performance completion ratio is just above 100%, it indicates that the firm is likely to engage in classification shifting.Figure 2 shows the distribution of listed firms' performance completion ratio during the performance commitment period.The horizontal coordinate is the performance completion ratio and the interval width is 10%.The vertical coordinate is the proportion of observations in the interval to the total sample.As shown in Figure 2, the observations exhibit a significant jump on the right side of 100%, i.e. observations with performance completion ratios between 100% and 110% increase suddenly.This suggests that firms are likely to engage in classification shifting to achieve the performance target.As a robustness check, we also depict the sample distribution using 5% as the interval width, and the graph is similar to Figure 2. Based on the above analysis and following Xie et al. (2019), we focus on firms that happen to 'step on the line' to meet the performance target when examining the classification shifting behaviour of listed firms.In fact, not all firms with performance metrics linked to net income before non-recurring items have the incentives to engage in classification shifting.On the one hand, managers have no need to engage in classification shifting when performance targets are easy to meet; on the other hand, managers have a lower incentive to engage in classification shifting when the performance targets are difficult to meet.Therefore, firms that 'step on the line' to meet the target are more likely to engage in classification shifting than other firms.Following J. Wang and Fan (2017), we define the dummy variable of 'stepping on the line' to meet the target (MEET): if the firm's cumulative net income before non-recurring items during the performance commitment period is 100% to 110% of the promised cumulative net income before non- recurring items (i.e. the performance completion ratio is between 100% and 110%), MEET takes the value of one.Otherwise, MEET takes the value of zero.In other words, we treat firms that 'step on the line' to meet the net income before non-recurring items target as the treatment group, while treating other firms (firms with performance metrics not tied to net income before non-recurring items and firms with performance metrics tied to netincome before non-recurring items but do not 'step on the line' to achieve them) as the control group.

Discussion on the specific means of classification shifting
In  This figure shows the distribution of firms' performance completion ratio, defined as the realised net income before non-recurring items divided by the promised net income before non-recurring items.The horizontal coordinate is the completion ratio.The vertical coordinate is the proportion of observations to the total sample.The column width is 10%.We winsorise the completion ratio at the 1% and 99% quartiles.
classifying non-recurring gains (such as government subsidies, fair value gains and losses, and gains from the sale of fixed assets) as recurring revenues is easily detected by the auditor.A firm is also less likely to adopt the second means to engage in classification shifting.In contrast, the accounting treatment of classifying recurring expenses as nonrecurring losses is more hidden and therefore more likely to be used by the firm.As such, we analyse firms' classification shifting behaviour by focusing on the first means.Existing studies also analyse classification shifting mainly from this perspective (Y.Fan et al., 2010;McVay, 2006;Xie et al., 2019;Z. Zhang & Zhang, 2012).

Abnormal net income before non-recurring items (UE_CE)
We follow McVay (2006) and Xie et al. (2019) to measure classification shifting through non-recurring items.Specifically, we first estimate model ( 1) by industry and year to obtain the expected net income before non-recurring items.Then, we subtract the expected net income before non-recurring items from the actual net income before nonrecurring items to obtain the abnormal net income before non-recurring items (UE_CE).
In model (1), CE is net income before non-recurring items scaled by sales.ATO is the asset turnover ratio.ACCRUALS is operating accruals.ΔSALES is sales growth.NEG_ΔSALES equals ΔSALES when ΔSALES is negative, and zero otherwise.Table 4 presents the detailed definitions of these variables.

Research design
Following Xie et al. (2019), we construct the following model to test listed firms' classification shifting behaviour during the performance commitment period.
where UE_CE is abnormal net income before non-recurring items.MEET is a dummy variable for 'stepping on the line' to meet the performance target (treatment indicator).MEET takes the value of one for the entire observation period (before, during, and after the commitment period) if a firm's performance completion ratio is 100%~110%, and zero otherwise, as defined earlier.Take Figure 1 as an example.Assuming that Firm A 'steps on the line' to meet the promised net income before non-recurring items, then MEET is equal to one for the entire observation period (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016).Assuming that Firm B does not 'step on the line' to meet the promised net income before non-recurring items during the commitment period (2011)(2012)(2013) or the performance target is not net income before non-recurring items, then MEET is equal to zero for the entire observation period (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016).COMMIT is a dummy variable for the performance commitment period.COMMIT is equal to one if the observation year falls within the performance commitment period, and zero otherwise.Take Figure 1 as an example.The performance commitment period is 2011-2013.Thus, COMMIT is equal to one for 2011-2013, and zero for other years.DBL is non-operating losses (the main account of non-recurring losses) scaled by sales.The coefficient on MEET*DBL, β 2 , represents the extent to which recurring expenses are classified as non-operating losses for firms that 'step on the line' to meet the target relative to those that do not.The coefficient on COMMIT*DBL, β 3 , represents the extent of classification shifting during the performance commitment period relative to the non-commitment period.The coefficient of interest is the coefficient on MEET*COMMIT*DBL, i.e. β 1 .It represents the extent to which recurring expenses are classified as non-operating losses for treatment firms relative to control firms during the commitment period relative to the non-commitment period.If treatment firms are more likely to engage in classification shifting during the commitment period, then the positive relationship between abnormal net income before non-recurring items (UE_CE) and non-operating losses (DBL) should be stronger for treatment firms than control firms during the commitment period relative to non-commitment period.That is, β 1 is expected to be positive.
Following Xie et al. (2019) and Chung et al. (2020), we control for a series of variables that may affect a firm's classification shifting behaviour: firm size (SIZE), firm leverage (LEV), growth opportunities (MTB), profitability (ROA), loss indicator (LOSS), firm age (AGE), accruals management (DA), and real activity earnings management (RM).Table 4 details the variable definitions.To mitigate endogeneity concerns, we control for firm fixed effects and year fixed effects.Standard errors are clustered at the firm level.
In addition, following Xie et al. (2019), we examine the relationship between UE_CE and DBL based on all A-share listed firms from 2008 to 2020.The regression results show that UE_CE is significantly and negatively associated with DBL at the 1% level, suggesting that listed firms generally do not engage in classification shifting.Thus, if we can find a positive  et al., 1995).The model is regressed cross-sectionally by industry and year.

RM
Real earnings management estimated from the model of Roychowdhury (2006).First, we calculate abnormal operating cash flow, abnormal production costs, and abnormal discretionary expenses, respectively.Then, we calculate RM as abnormal production costs minus abnormal cash flow from operations minus abnormal discretionary expenses.
correlation between UE_CE and MEET*COMMIT*DBL, it suggests that our results are robust, not just a reflection of the average classification shifting behaviour of listed firms.

Descriptive statistics
Table 5 Panel A reports descriptive statistics for the main variables.To avoid the effect of outliers, we winsorise all continuous variables at the top and bottom 1% levels.As shown, the mean value of MEET is 0.390, suggesting that 39% of firms 'step on the line' to meet the net income before non-recurring items target, i.e. the realised net income before nonrecurring items slightly exceeds the promised net income before non-recurring items.The mean value of COMMIT is 0.393, suggesting that 39.3% of the observations fall within the performance commitment period.On average, the value of UE_CE is −0.066.Nonoperating losses account for 1.1% of sales.The values of the remaining variables are comparable to the existing literature (B.Liu et al., 2016).Table 5 Panel B presents correlations of variables.UE_CE is significantly and positively associated with COMMIT, suggesting that abnormal net income before non-recurring items is higher during the commitment period.UE_CE is significantly and negatively associated with DBL, suggesting that listed firms generally do not engage in classification shifting.We also conduct the multicollinearity test.Specifically, we find that the variance inflation factors (VIF) of the model are all less than 10, suggesting that multicollinearity is not a concern (Neter et al., 1996).

"Stepping on the line" to meet the target and classification shifting
Table 6 reports the estimation results of model (2).In column (1), we include industry fixed effects and year fixed effects.In column (2), we include firm fixed effects and year fixed effects. 4As the table shows, in both columns, the coefficient on MEET*COMMIT*DBL is positive and significant at the 1% levels (β 1 = 2.815, t-statistic = 3.542; β 1 = 2.916, t-statistic = 3.861).This suggests that firms that 'step on the line' to meet the target are more likely to classify recurring expenses as non-operating losses to inflate core earnings relative to other firms during the commitment period relative to the non-commitment period.H1 is supported.In addition, UE_CE is negatively and significantly associated with DBL, suggesting that treatment firms do not engage in classification shifting during the sample period.

Decomposing non-operating losses
We further decompose non-operating losses into 'losses on fixed asset disposals' (DBL_PPE) and 'other non-operating losses' (DBL_Others) to examine which item is more likely to be used by firms to engage in classification shifting.Xie et al. (2019) argue that firms are less likely to classify recurring expenses as 'losses on fixed asset disposals' because the fixed asset disposal is usually accompanied by a large number of accounting 4 MEET is absorbed due to firm fixed effects.documents and thus more likely to attract auditors' attention.In contrast, firms are more likely to classify recurring expenses as 'other non-operating losses' because 1) their types are diverse; 2) their classification is ambiguous; 3) they are difficult to identify; and 4) they are easily overlooked by auditors.As such, we expect that firms that 'step on the line' to meet the target are more likely to classify recurring expenses as 'other non-operating losses' to inflate core earnings.
To test this prediction, we interact DBL_PPE and DBL_Others with MEET and COMMIT to model (2) to examine which items firms are more likely to be used by firms to inflate earnings.Table 6 column (3) reports the estimation results.The coefficient on MEET*COMMIT*DBL_Others is significantly negative at the 1% level (β 1 = 3.372, t-statistic = 3.834), while the coefficient on MEET*COMMIT*DBL_PPE is non-significant (β 1 = 11.370,t-statistic = 1.320).This suggests that firms are more likely to engage in classification shifting through 'other non-operating losses'.

Robustness checks
To strengthen the main conclusions, we conduct a series of robustness tests.

Mitigating measurement errors
So far, we have used the data of listed firms (acquiring firms) to measure the classification shifting behaviour of target firms.However, this approach suffers from measurement error.To mitigate this problem, we adopt two approaches.One is to add a cross-sectional test conditional on the target firm's size and the other is to use financial statement data of parent firms to conduct indirect tests.
First, we perform a cross-sectional test based on the ratio of the target firm's premerger sales to the acquiring firm's pre-merger sales (sales ratio, for short).We expect that the classification shifting effect is more pronounced for firms with higher sales ratios, as these firms have a greater impact on the acquiring firm's post-merger financial performance.Following prior studies (S.Chen & Ma, 2018;Du & Sui, 2021), we split the sample into two sub-samples according to the upper quartile and lower quartile of sales ratio and conduct the estimation of model (2) again.Columns (1) and (2) of Table 7 present the estimation results for firms with higher sales ratios and firms with lower sales ratios, respectively.As shown, the coefficient of MEET*COMMIT*DBL is significantly positive in firms with higher sales ratios (β 1 = 8.202, t-statistic = 2.747), while the coefficient is insignificant in firms with lower sales ratios (β 1 = 1.019, t-statistic = 1.339).The difference between the two estimated coefficients is statistically significant (p-value of 0.000).The results show that the observed classification shifting effect is more pronounced in firms in which the merger has a larger effect on acquiring firms.
Second, we use the financial statement data of parent firms to examine whether the classification shifting effect originates from the parent firm or the target firm.Specifically, we recalculate abnormal core earnings (UE_CE_parent) and nonoperating losses (DBL_parent) using the financial data of parent firms and then rerun the model (2).Table 7 column (3) presents the regression results.The coefficient on MEET*COMMIT*DBL_parent is insignificant (β 1 = 55.011,t-statistic = 1.189), indicating that there is no significant evidence showing that parent firms engage in classification shifting during the commitment period.In conclusion, we find evidence of classification shifting using the consolidated financial data, while we do not find evidence of classification shifting using the data of parent firms.This difference suggests that the classification shifting effect observed in our paper originates from target firms rather than parent firms.

Deleting sample with re-occurrence of M&A during the commitment period
If a firm completes other M&As during the performance commitment period, the conclusions drawn from our may be confounded.Therefore, to reduce the influence of the re-occurrence of M&A events, we exclude the sample with reoccurrences of M&A during the performance commitment period.Table 7 column (4) presents the estimation results.The coefficient on MEET*COMMIT*DBL remains positive (β 1 = 2.737, t-statistic = 3.602), suggesting that our main conclusions are robust.

Changing the sample period
In the preceding analysis, the sample period in our paper consists of three years before and after the performance commitment period, i.e. three years before the start of the performance commitment, years during the performance commitment period, and three years after the end of the performance commitment.To alleviate the concern that the financial data of listed firms are not comparable before and after the merger, we only retain the observations during the performance commitment period and three years after the expiration of the performance commitment period.Table 8 column (1) presents the regression results.The coefficient on MEET*COMMIT*DBL is significantly positive (β 1 = 2.332, t-statistic = 3.832), indicating that our main conclusions remain unchanged after changing the sample period.
In addition, we also shorten the sample period from three years before and after the performance commitment period to two years before and after the commitment period.Table 8 column (2) presents the regression results.Again, the coefficient on MEET*COMMIT*DBL remains significantly positive (β 1 = 2.908, t-statistic = 3.167).

Changing the "stepping on the line" interval
In our main analysis, we define the 'stepping on the line' interval as [100,110].That is, the realised net income before non-recurring items is between 100% and 110% of the promised net income before non-recurring items.As a robustness check, we consider two alternative intervals, i.e. [100, 105] and [90, 110].As shown in Table 8 columns ( 3) and (4), the results based on alternative intervals are similar.

Replacing "non-operating losses" with "non-recurring losses"
Further, following Lu and Bu (2020), we also use an alternative measure of DBL, i.e.EXLOSS.EXLOSS is calculated as non-recurring losses scaled by sales.Table 8 column (5) presents the estimation results.Consistent with our prediction, the coefficient on (MEET*EXLOSS) is positive and significant at the 1% level (β 1 = 2.888, t-statistic = 4.195).

Cross-sectional tests
To reinforce the theoretical analysis of our paper, we conduct a battery of cross-sectional tests.

Committed performance amount
First, we examine whether the classification effect is conditioned on the committed performance amounts.Our theoretical analysis suggests that firms have a strong incentive to engage in classification shifting to meet the performance target and thus avoid triggering compensation.This indicates that higher committed performance may result in

Compensation method
Second, we examine whether the classification shifting effect is conditioned on the compensation method of performance commitments.The compensation methods of performance commitments can be classified into three types: all-cash, all-stock, and a mix of cash and stock.Since the mixed compensation method can be further classified as cash-dominant compensation method or stock-dominant compensation method, we simplify all compensation methods into two types: cash and stock.In the case of the cash compensation method, a target firm is required to compensate the acquirer a certain amount of cash in the event that the firm misses the promised target.The payment amount is usually the difference between the reported performance and the promised performance.In the case of the stock compensation method, the target firm is required to compensate the acquirer in a certain number of shares in the event that the firm misses the promised target.Generally, the acquirer buys back a certain number of shares from the target firm at book value and then cancels them.Because stock compensation due to non-performance reduces the shares held by the target firm, which has a negative impact on its shareholder status, the stock compensation method causes greater losses to the target firm than the cash compensation method (Pan et al., 2017;G. Zhang et al., 2020;Zhao & Yao, 2014).Therefore, we expect that firms using stock to compensate for nonperformance are more likely to engage in classification shifting.To test this hypothesis, we split the sample into two groups according to compensation methods.Columns (3) and (4) of Table 9 report the results for firms using stock to compensate for non-performance and firms using cash to compensate for non-performance, respectively.The results show that the coefficient on MEET*COMMIT*DBL in column (3) (β 1 = 3.618, t-statistic = 4.635) while insignificant in column (4) (β 1 = 1.125, t-statistic = 1.035), indicating that firms that use stock to compensate for non-performance have a greater incentive to engage in classification shifting.

The monitoring role of auditors
Finally, we examine whether the classification shifting effect is conditioned on the type of auditor.Although firms can shift recurring expenses to hard-to-detect non-recurring losses to boost core earnings, such manipulation may be constrained by auditors.
According to the 'Explanatory Bulletin No. 1 on Disclosure of Information by Firms Issuing Public Securities -Non-recurring Gains and Losses' issued by the CSRC, when issuing audit reports for firm prospectuses, periodic financial reports, and financial reports for securities issuance, CPAs should pay sufficient attention to the items, amounts and notes of non-recurring items, and verify the truthfulness, accuracy, completeness, and reasonableness of non-recurring items disclosed in financial reports.Therefore, auditors with higher independence or competence are more likely to give sufficient attention to the classification of non-recurring items when reviewing financial statements, thereby restraining management's opportunistic behaviour and reducing the possibility of classification shifting.Previous studies have shown that Big 4 auditors are more effective in constraining managers' earnings management behaviours than non-Big 4 auditors (Lawrence et al., 2011).Therefore, we expect the classification shifting effect to be less pronounced among firms audited by Big 4 auditors.We split the sample into firms audited by Big 4 auditors and firms not audited by Big 4 auditors, and then conduct the main regression.Columns ( 5) and ( 6) of

The analysis of traditional earnings management
The preceding analysis focuses on classification shifting using non-recurring items.In addition to this earnings management tool, firms can also increase their overall net income through traditional earnings management tools, such as accruals management and real activity earnings management, which also increase net income before nonrecurring items.In this subsection, we test these two earnings management tools.In particular, we construct the following model: DA/RM it =α 0 +α 1 ×MEET it ×COMMIT it +α 2 ×MEET it +α 3 ×COMMIT it +α i ×Controls it +ε it (3) where DA and RM denote accruals management and real activity earnings management, respectively.The remaining variables are defined in the same way as in model ( 2).The coefficient of interest is the coefficient on MEET*COMMIT, α 1 .We expect that firms that 'step on the line' to meet the target are more likely to engage in earnings management during the performance commitment period.Thus, α 1 is expected to be greater than zero.Table 10 column (1) reports the estimation results using DA as the dependent variable.Columns (2) and (3) report the regression results when DA is greater than zero and when DA is less than zero, respectively.Column (4) reports the regression results using RM as the dependent variable.Although the coefficients on MEET*COMMIT in columns (1) and ( 4) are insignificant, the coefficient on MEET*COMMIT in column (2) is significant.This suggests that firms that 'step on the line' to meet the target are more likely to engage in upward accruals management and less likely to engage in real activity earnings management.This may be due to the fact that real-activity earnings management can harm a firm's future performance and thus the manipulation cost is relatively large.

Economic consequence analysis
Our main analysis shows that firms 'stepping on the line' to meet the target tend to engage in classification shifting activities to inflate earnings.Since the manipulated earnings are not sustainable and the incentive to manipulate core earnings decreases after the expiration of the performance commitment period, firms that 'step on the line' to meet the target should experience declining performance after the expiration of the performance commitment period.We construct the following model to test this conjecture: ROA_NexEx it =α 0 +α 1 ×MEET it ×POST it +α 2 ×MEET it +α 3 ×POST it +α i ×Controls it +ε it (4) where ROA_NexEx is net income before non-recurring items scaled by total assets.MEET is defined as earlier.POST is a dummy variable that equals one if the observations fall within three years after the expiration of the performance commitment period, and zero otherwise.The coefficient of interest is the coefficient on MEET*POST, α 1 .We expect that firms that engage in classification shifting during the commitment period will experience declining performance after the commitment.Thus, α 1 is expected to be less than zero.
Following prior research (Luo & Liu, 2009;Z. Wang & Hu, 2016;Xu et al., 2019;Yu & Chi, 2004), we control for a series of variables that may affect financial performance: firm size (SIZE), firm leverage (LEV), firm age (AGE), sales growth (GROWTH), state ownership (STATE), the proportion of shares held by the largest shareholder (BIGHOLD), board size (BOARDSIZE), and the share of independent directors on the board (INDEP).Table 11 reports the estimation results of model ( 4).As the table shows, the coefficient on MEET*POST is negative at the 5% level (α 1 =-0.018,tstatistic = −2.191),suggesting that firms 'stepping on the line' to meet the target indeed experience a decline in performance after the expiration of the performance commitment period.This evidence of declining performance in the post-commitment period also corroborates the earnings manipulation behaviour during the commitment period.

Conclusions
Based on a sample of listed firms that complete major asset restructurings and sign performance commitment agreements from 2008 to 2019, we examine the effect of performance commitments on firms' classification shifting behaviours.Empirical results show that, relative to other firms, firms that 'step on the line' to meet the promised net income before non-recurring items target are more likely to engage in classification shifting through non-recurring items.Specifically, these firms are more likely to shift recurring expenses to non-operating losses, especially hard-to-identify other non-operating losses, to meet performance targets.The results are robust to a series of robustness tests.Cross-sectional tests show that the classification shifting effect is more pronounced in firms with larger committed amounts and firms that use stock to compensate for non-performance, and Big 4 auditors play a monitoring role in curbing this form of earnings management behaviour.Economic consequence analysis shows that, after the expiration of the commitment period, the performance of firms that 'step on the line' to meet the target declines significantly, providing further evidence of earnings manipulation during the commitment period.
Our findings have implications for regulators, investors, and auditors.First, regulators should strengthen the supervision of post-merger commitment compliance, paying particular attention to the phenomenon of 'stepping on the line' to meet performance targets and the resulting classification shifting behaviours.When designing standards, regulators should standardise and refine the classification shifting criteria of accounting accounts, clarifying which business activities can be classified as non-recurring losses (gains) or recurring expenses (revenues).Secondly, investors should treat the performance commitment in M&A rationally, make an objective evaluation of the firm's ability to meet the performance, and be careful of firms' earnings manipulation through classification shifting.Third, as the gatekeepers of capital markets, external auditors should pay more attention to whether firms misclassify recurring expenses as non-recurring losses when auditing accounting accounts, so as to better play the role of accounting information in the capital market.

Figure 1 .
Figure 1.Example of year observations of performance commitments.

Figure 2 .
Figure 2. Distribution of firms' performance completion ratio.This figure shows the distribution of firms' performance completion ratio, defined as the realised net income before non-recurring items divided by the promised net income before non-recurring items.The horizontal coordinate is the completion ratio.The vertical coordinate is the proportion of observations to the total sample.The column width is 10%.We winsorise the completion ratio at the 1% and 99% quartiles.

Table 2 .
The distribution of M&A events.

Table 3 .
Sample selection process.

Table 4 .
Variable definitions.MEET A dummy variable for "stepping on the line" to meet the performance target.MEET equals one if a firm's M&A performance target is tied to net income before non-recurring items and the firm "steps on the line" to meet the target, and zero otherwise.NEG_ΔSALES Percent change in sales (ΔSALES) if ΔSALES<0, and zero otherwise.SIZE Firm size, calculated as the natural logarithm of total assets.LEV Firm leverage, calculated as total liabilities scaled by total assets.MTB Market to book ratio, calculated as market value divided by book value.ROA Return on assets, calculated as net income scaled by total assets.LOSS Loss indicator.Loss is equal to one if net income is less than zero, and zero otherwise.AGE Firm age, calculated as the natural logarithm of the number of years since the IPO.DA Abnormal accruals estimated from the modified Jones Model (Dechow

Table 6 .
'Stepping on the line' to meet the target and classification shifting.

Table 10 .
'Stepping on the line' to meet the target and traditional earnings management.

Table 11 .
'Stepping on the line' to meet the target and future performance.