Trade-off between real activities earnings management and accrual-based manipulation-evidence from China

Trade-off between real activities earnings management and accrual-based manipulation-evidence from China

Accepted Manuscript Title: Trade-off between Real Activities Earnings Management and Accrual-based Manipulation-Evidence from China Authors: Gao Jie, ...

593KB Sizes 0 Downloads 70 Views

Accepted Manuscript Title: Trade-off between Real Activities Earnings Management and Accrual-based Manipulation-Evidence from China Authors: Gao Jie, Gao Baichao, Wang Xiao PII: DOI: Reference:

S1061-9518(17)30016-2 http://dx.doi.org/doi:10.1016/j.intaccaudtax.2017.08.001 ACCAUD 222

To appear in:

Journal of International Accounting, Auditing and Taxation

Received date: Revised date: Accepted date:

16-7-2014 10-7-2017 2-8-2017

Please cite this article as: Jie, Gao., Baichao, Gao., & Xiao, Wang., Trade-off between Real Activities Earnings Management and Accrual-based ManipulationEvidence from China.Journal of the Chinese Institute of Chemical Engineers http://dx.doi.org/10.1016/j.intaccaudtax.2017.08.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Trade-off between Real Activities Earnings Management and Accrual-based Manipulation-Evidence from China

GAO Jie1* Harbin Institute of Technology (Shenzhen),Shenzhen University Town, Xili, Nanshan District, Shenzhen, P.R. China (518055) Email: [email protected] Phone:+8618680666996 GAO Baichao1 Harbin Institute of Technology (Shenzhen),Shenzhen University Town, Xili, Nanshan District, Shenzhen, P.R. China (518055) WANG Xiao1 Harbin Institute of Technology (Shenzhen),Shenzhen University Town, Xili, Nanshan District, Shenzhen, P.R. China (518055)

Abstract: This study investigates the trade-off between real activities earnings management and accrual-based earnings management of Chinese listed companies. We first establish a model in which the manager could choose between accrual-based or real activities earnings management, and solve for the optimal level of each kind of earnings management. Based on the results of the theoretical model, we empirically investigate how managers choose between the two kinds of earnings management. Our empirical tests use China's institutional background with data of listed companies from 2008 to 2012. Results indicate that the level of real activities earnings management is higher for firms under lower government intervention, firms with higher financial leverage and lower corporate governance. Firms in a less stringent legal environment, double-listed firms and firms with higher growth prospects are more likely to engage in accrual-based earnings management. The results provide

-1-

implications about the accounting choice of listed companies under China’s unique institutional characteristics.

Keywords: real activities earnings management, accrual-based earnings management, trade-off, accounting choice 1 Introduction

Earnings management is a classic research topic in the accounting field. Extant studies have generally classified earnings management into accrual-based and real activities earnings management. Accrual-based earnings management is achieved by changing the accounting policies or estimates adopted when recognizing a given transaction in the financial statements. Examples include changing the depreciation method for fixed assets or the provision for doubtful accounts. Accrual-based earnings management can bias reported earnings in a particular direction without changing underlying transactions and it does not alter cash flows. Real activities earnings management, however, is achieved by changing the actual operational activities, such as providing excess cash discounts, reducing R&D expenditures, and overproduction (Roychowdhury 2006). This kind of earnings management usually changes firm cash flows. Given the difference between accrual-based and real activities earnings management, it would be meaningful to investigate the trade-off between these two types of earnings management by corporate managers. Previous studies provide empirical evidence in the US setting (Cohen et al. 2008; Cohen and Zarowin 2010; Badertscher 2011; Zang 2012). Due to different institutional characteristics, the incentives of Chinese listed companies to choose between these two types of earnings management should be different from those in the US. For example, unlike in the US, many companies in China are under frequent intervention by the government, and the level of government intervention is likely to affect the incentives for managers to engage in different types of earnings management. In addition, firms located in different administrative areas in China face different levels of legal environment

-2-

stringency, which is likely to impact the accounting choice associated with earnings management. In addition, firm-specific factors, such as corporate governance level and double-listing status are also likely to affect the relative cost of the two types of earnings management, and thus impact managers' accounting choice incentives. Identifying factors unique to China's institutional characteristics and examining their impact on the trade-off between accrual-based and real activities earnings management would generate meaningful implications for regulators, investors, listed companies and other capital market participants. Toward this goal, we first build an agency model to study the trade-off of these two kinds of earnings management by managers. We then develop and test empirical hypotheses based on the implications of the theoretical model under China’s unique institutional background. Our theoretical model shows that, in equilibrium, the optimal level of accrual-based earnings management increases in the relative cost of real activities earnings management versus accrual-based earnings management. Conversely, the optimal level of real activities earnings management decreases in the relative cost. Based on the results of the theoretical model, we predict and find empirical evidence that in China, higher leveraged firms and firms under lower government intervention and weaker corporate governance are more likely to engage in real activities earnings management. Double-listed firms, firms in a less stringent legal environment and firms with a higher growth rate are more likely to engage in accrual-based earnings management. This study contributes to the extant literature in several aspects. First, it contributes to the theoretical literature on earnings management. Most previous analytical studies examine only one type of earnings management (Fischer and Stocken 2004; Goldman and Slezak 2006; Crocker and Slemrod 2007; Chan and Gao 2014), while the current study analyzes the trade-off between two types of earnings management. Second, it identifies a set of variables that affect managers’ trade-off between accrual-based and real activities earnings management unique to China's institutional background, as distinct from the factors examined in previous studies in -3-

the US setting (Roychowdhury 2006, Zang 2012).The findings of the current study indicate that it is important to consider the effect of institutional characteristics when examining the trade-off between different types of earnings management in different economies. Previous studies, such as Ball et al. (2003) and Leuz et al. (2003) find that financial reporting quality is influenced by the legal enforcement regime at the country level. The current study, however, examines the effect of legal environment and government intervention on earnings quality at the province level within one country. Findings indicate that province-level differences do impact the choice of earnings management type. This study also adds to the literature on the economic effects of firm-specific factors such as corporate governance, indicating that a stronger corporate governance system could effectively deter real activities manipulation for listed firms in China. Finally, while previous empirical studies mainly examine the effect of institutional characteristics on accrual-based earnings management, this study extends extant literature by investigating their effect on both accrual-based and real activities manipulation. The findings indicate that only examining one kind of earnings management could generate biased results. The paper proceeds as follows. Section 2 reviews the related literature. Section 3 presents the model and develops empirical hypotheses. Section 4 describes the data and econometric specifications. Section 5 presents empirical results and the last section concludes.

2 Literature review Numerous studies provide empirical evidence about the existence and determinants of accrual-based earnings management (Healey 1985; Burgstahler and Dichev 1997; DeFond and Jiambalvo 1994; DuCharme et al. 2004; Lakshmanan 2000; Rangan 1998; Sweeney 1994; Teoh et al. 1998). More recent studies investigate the possibility that managers manipulate real transactions to distort earnings. For example, Baber et al. (1991), Dechow and Sloan (1991), Bushee (1998) and Cheng (2004) find that managers reduce R&D expenditures to increase the annual earnings. Jackson and

-4-

Wilcox (2000) find that managers employ price discounts to temporarily increase sales. Other types of real activities earnings management that have been explored include cutting advertising expenditures, stock repurchases, sales of profitable assets, derivative hedging, debt-equity swaps and securitization. Roychowdhury (2006) first developed an empirical model to systematically measure the extent of real activities earnings management, and provided evidence suggesting that managers use overproduction to report lower cost of goods sold, adopt price discounts to temporarily increase sales, and reduce discretionary expenditures to improve reported margins. After that, several studies investigate the trade-off between accrual-based earnings management and real activities earnings management in the US setting. Cohen et al.(2008) document that accrual-based earnings management increased steadily from 1987 until the passage of the Sarbanes-Oxley Act (SOX) in 2002, followed by a significant decline after the passage of SOX. The level of real activities earnings management declined prior to Sarbanes-Oxley Act (SOX) and increased significantly after the passage of SOX, suggesting that firms switched from accrual-based to real earnings management methods after the passage of SOX. Cohen and Zarowin (2010) investigate real activities earnings management for SEO companies, and provide evidence that the decline in post-SEO performance due to real activities management is more severe than that due to accrual management. Zang (2012) identifies a series of factors that affect the trade-off between accrual-based and real activities earnings management in the US setting, providing large-sample evidence consistent with the argument that managers use real activities earnings management and accrual-based earnings management as substitutes. Research on accrual-based earnings management in China mostly focuses on capital market incentives (Lin and Wei 2000; Lu 1999). Other factors that affect managers’ choice of accrual-based earnings management include the level of corporate governance, property rights, tax rates and credit rating. (Gao and Zhang 2008; Wang et al. 2009). Only in recent years has real activities earnings management in China been studied. Zhang et al. (2008) find firms reporting small increases in earnings engage in -5-

real activities earnings management. Fang and Jin (2011) find that the extent of real activities and accrual-based earnings management are negatively related with the quality of internal control. Li et al. (2011) find that firms engage in more real activities earnings management if their expected tax rate increases and engage in more accrual-based earnings management if their expected tax rate decreases. Cai et al.(2012) find that firms reporting losses in two or three consecutive years employ both real activities and accrual-based earnings management to improve the distressed situation. In all, there is lack of comprehensive study of the trade-off between accrual-based and real activities earnings management in China, especially after the adoption of new accounting standards in 2007. In terms of theoretical research on earnings management, besides the studies that focus on identifying the conditions that lead to earnings management (Bagnoli and Watts 2000; Demski et al. 2004; Dye 1988; Evans and Sridhar 1996; Ronen et al. 2006), many studies investigate how manager’s earnings management incentives affect investors’ perception about firm value (Fischer and Stocken 2004; Fischer and Verrecchia 2000; Narayanan 1985; Stein 1989). More recently, Goldman and Slezak (2006), Crocker and Slemrod (2007) and Chan and Gao (2014) investigate how the equilibrium compensation contract is derived in the presence of earnings management incentives. However, the extant theoretical studies generally do not distinguish between real activities earnings management and accrual-based earnings management. In this study, we build an agency model in which the agent (i.e. manager) can choose between real activities earnings managements and accrual-based earnings management and solve for the equilibrium level of each kind of earnings management.

3. Theoretical model and hypotheses development 3.1 Theoretical model The study builds a one-period model. The model consists of a publicly traded firm with a principal and a manager. The manager is risk-averse while the principal is

-6-

risk-neutral. At the first stage, the principal designs a compensation contract to motivate the manager to run the firm. At the second stage, the manager exerts unobservable effort that stochastically affects the firm’s end-of-period gross cash flow and, at the same time, covertly creates opportunities to engage in real activities and accrual-based earnings management. At the third stage, the manager publicly issues an earnings report that reflects the firm’s end-of-period gross cash flow with noise and bias (i.e., earnings management). At the fourth stage, the manager gets paid by the principal, and at the last stage, the firm’s end-of-period cash flow is realized. ~

~

The firm’s stochastic production process is represented by v  e  1 , where v is the firm’s gross cash flow, e is the unobservable effort taken by the manager at stage two and  1 is the noise term that is normally distributed with mean 0 and precision h . The distribution of the production noise term is independent of the manager’s effort. At stage one, the principal designs a compensation contract based on accounting earnings to efficiently motivate the manager to exert effort. By doing so, the principal takes into account that the manager is able to manipulate the earnings report to be issued at stage three. We assume that the managerial compensation w is linear: w     r , where r is the earnings report publicly issued by the manager, and  and  are the linear compensation coefficients. At stage two, the manager exerts unobservable effort and covertly creates opportunities to engage in costly real activities and accrual-based earnings management. We assume the costs of managerial effort and earnings management are C0 (e)  e2 / 2 , C1 (a)  c1a 2 / 2 , C2 (b)  c2b2 / 2 , where a , b are the extent of accrual-based earnings management and real activities earnings management respectively. Moreover, c1 , c2  0 , and they are the marginal cost of the two kinds of earnings management. The cost function of earnings management might reflect, for instance, the manager’s litigation, reputation and psychic cost of manipulating the firm’s earnings report. At stage three, the manager publicly issues an earnings report r  e  1  a  b   2 , where  2 represents noise in the financial

-7-

reporting system and is assumed to be normally distributed with mean 0 and precision m. The manager’s preferences are represented by a constant absolute risk aversion (CARA) utility function defined over his compensation w , minus the cost of effort,

C0 (e) , and earnings management C1 (a) , C2 (b) : U m (e, a, b)   exp{ [w  (e2 / 2  c1a 2/ 2  c2b 2/ 2)]} ,

(1)

Where  is the manager’s coefficient of constant absolute risk aversion. The reservation utility of the manager is assumed to be -1. We use backward induction to solve for the equilibrium. At stage 2, given a compensation contract w, the manager’s objective is to choose e , a and b to maximize his expected utility:

Max E[U m (e, a, b)] e , a ,b

(2)

Substituting (1) into (2) and taking the first order condition generates the following results: a*   / c1

(3)

b*   / c2

(4)

e*  

(5)

At stage 1, the principal would like to maximize his/her own interests: ~

Max E (v  w)  ,

subject to the agent’s incentive compatibility constraint:

e , a , b   arg max E[U *

*

*

m

(e, a, b)],

and the agent’s participation constraint:

E[U m (e, a, b)]  1 Solving problem (6) yields the optimal solution for the linear compensation -8-

(6)

coefficient:

* 

1  (1/ h  1/ m)  1/ c1  1/ c2  1

(7)

Substituting (7) back to (3) and (4) generates the following results: a* 

1  c1 (1/ h  1/ m)  c1  c1 / c2  1

b* 

(8)

1  c2 (1/ h  1/ m)  c2  c2 / c1  1

(9) From equation (8),

a / c1  0 and a / c2  0 , which indicates that the higher the cost of accrual-based earnings management, the lower the optimal level of accrual-based earnings management chosen by the manager. The higher the cost of real activities earnings management, the higher the level of accrual-based earnings management the manager will adopt. Equation (9) shows that b / c1  0 and b / c2  0 , which indicates that the higher the cost of accrual-based earnings management, the higher the level of real activities earnings management. The higher the cost of real activities manipulation, the lower the level of real activities earnings management the manager will adopt.

3.2 Hypotheses development Both real activities earnings management and accrual-based earnings management are costly activities. The model developed in the previous section indicates that a manager’s trade-off decision between the two types of earnings management depends on their relative cost, which, in turn, is determined by the firm’s operational and accounting environment.

3.2.1 Real activities earnings management Based on the comparative static results derived from problem (6) in the previous

-9-

section regarding the optimal level of real activities earnings management:

and

b 0 c1

b  0 , we predict that the optimal level of real activities earnings management c2

increases in the relative cost of accrual-based earnings management. Although China adopted market-oriented economic policies a few decades ago, government still plays a very important role in many aspects of the operations of Chinese companies. Different government sectors interfere with enterprises through examining and authorizing business activities from time to time. In China, the central government controls the overall regulations over various administrative authorization issues, however, it gives the government in different administrative areas a lot of freedom in setting their specific examination and approval procedures. The complexity of the approval process is largely affected by the efficiency of individual governments, and it differs vastly from one province to another. Generally, the degree of intervention is lower in provinces with more developed economies, located mainly in the southeast of China, such as Zhejiang, Jiangsu, Guangdong province (Li et al., 2012). Generally, there are two reasons for government intervention toward enterprises in China. First, to help the local government achieve its public goals, such as maintaining social welfare, facilitating employment and local economic growth, reducing fiscal deficit, etc. Government tends to transfer part of the public goals to firms through intervention. Second, to facilitate the political career of government officials (Lin & Li, 2004). The main evaluation criteria for government officials in promotion include local GDP, fiscal revenue, etc. Thus, local government tends to interfere with corporate operations to induce over-investment (to boost GDP) (Cheng, et al,2008),or to collect more tax (to increase fiscal income). Difference in the development stage of different provinces, as well as varied political incentives of local government officials also contribute to varied levels of government intervention in different provinces. In order to maintain the normal operations of business, firms must spend a lot of - 10 -

time dealing with government officials. For example, the establishment and annual inspection of an enterprise would involve a series of authorization and approval processes under various government agencies in the province it is located in, such as the administration bureau of industry & commerce and the bureau of taxation. The use of foreign venture capital would first involve the approval of the state administration of foreign exchange, which takes at least three to four months. Firms also need the approval of bureau of taxation to enjoy certain tax benefits or tax return policies. Moreover, examination and authorization by the government is required for firms to be eligible for certain fiscal subsidies. A lot of government intervention activities are based on accounting data. In order to avoid exploitation from government intervention (such as excessive investment), firms have incentives to manipulate accounting numbers. Compared with accrual-based earnings management, which only involves changing the accounting records, real activities earnings management cannot be achieved without affecting daily operation activities. Thus, it is less convenient, and more time-consuming (and thus more costly) to engage in real activities earnings management. Hence, to avoid/attenuate the negative impact brought by government intervention, firms are less likely to engage in real activities manipulation. Compared with firms in the US, Chinese firms spend much more time coping with the local government. Frequent government intervention reduces operational efficiency and allows government officials to grab more personal benefits, which adversely impacts long-term firm development. Extant studies have found that firms under more frequent government intervention generally have worse financial performance, and this effect is more distinguished for state-owned enterprises (Zhang et al 2015). To alleviate the negative effect on firm value arising from frequent government intervention, managers are likely to engage in less real activities earnings management, which is believed to more severely impact the long-term value of the firm than accrual-based earnings management. Based on the above analysis, we posit the following hypothesis: - 11 -

H1a: Other things being equal, firms under lower government intervention engage in a higher level of real activities earnings management. To get enough funding, sometimes firms need debt financing. To meet the criteria set by creditors, firms may engage in earnings management to avoid losses, which might lead to more binding debt covenants (Roychowdhury 2006).The higher the level of debt outstanding, the more likely firms will engage in earnings management. However, if earnings management is detected by creditors, they might call the loans before their maturity date, stop providing future loans to the borrower and terminate the debt contract, which is likely to increase the cost of future debt financing. This effect becomes more severe for firms with a larger proportion of debt financing. Since accrual-based earnings management is more easily detected than real activities earnings management, the relative cost for higher leveraged firms to engage in accrual-based earnings management is higher, thus, they are more likely to participate in real activities manipulation to reduce the potential risk of being detected by current and potential creditors. Moreover, accrual-based earnings management usually takes place at the end of the accounting period, while real activities earnings management could be adopted throughout year. Thus, it would be more convenient for firms to engage in real activities earnings management to satisfy the uncertain demand of funds during the year. Consequently, managers in higher leveraged firms tend to substitute real activities manipulation for accrual-based earnings management. Hence, we can posit the following hypothesis: H1b: Other things being equal, firms with higher leverage engage in a higher level of real activities earnings management. Corporate governance mechanisms are designed to alleviate the conflicts of interest between shareholders and managers. Characteristics of strong corporate governance include separation of CEO and chairman of the board, higher percentage of independent directors and higher shareholding of top management. Core et al. (1999) show that the level of corporate governance is positively associated with future - 12 -

operating and stock return performance in the US setting. Bai et al. (2005) further find that firms with higher governance levels have higher market valuation, and investors are willing to pay a considerable premium for well-governed firms under the Chinese setting. Firms under stronger corporate governance are more concerned with long-term value. Compared with accrual-based earnings management, real activities earnings management is achieved through a departure from optimal operational decisions, and is likely to have a stronger negative impact on long-term firm value. In addition, efficient corporate governance provides oversight for the management throughout the year, making it more difficult to conceal real activities earnings management. Thus, the relative cost of real activities earnings management is higher for well-governed firms. Conversely, the relative cost of accrual-based earnings management is higher for poorly-governed firms, and these firms are more likely to engage in real activities manipulation. Based on the above argument, we posit the following hypothesis: H1c: Other things being equal, firms under lower corporate governance levels engage in a higher level of real activities earnings management.

3.2.2 Accrual-based earnings management Based on the comparative static results derived from equation (8) in the previous section regarding the optimal level of accrual-based earnings management:

a 0 c1

and a  0 , we propose that the optimal level of accrual-based earnings management c2

increases in the relative cost of real activities earnings management. Accrual-based earnings management is more likely to be detected in a stringent legal environment compared with real activities manipulation (Wongsunwai 2013). Once detected, companies face the risk of lawsuit and loss of reputation. Thus, the cost of accrual-based earnings management relative to real activities earnings management is higher when the legal environment becomes more stringent. Cohen et al. (2008) have shown that the level of real earnings management activities declined prior to SOX (Sarbanes-Oxley Act) and increased significantly after the passage of SOX, suggesting - 13 -

that firms switched from accrual-based to real activities earnings management in the post-SOX period when the legal environment became more stringent. Zang (2012) uses Big 8 (whether the firm’s auditor is one of the international Big 8 audit firms), auditor tenure and SOX (whether the observation is from the post-SOX period) as proxies for the level of scrutiny, and finds that firms facing greater scrutiny from auditors and regulators have a higher level of real activities earnings management. In China, due to market segregation and local protectionism in different administrative areas, different provinces are characterized by a varied level of property rights protection for enterprises. The development level of market intermediaries (such as law firms and accounting firms) across provinces also varies. A diversified level of economic and social development also results in a diversified level of law enforcement efficiency in varied provinces. All these factors contribute to a varied level of legal environment stringency across provinces. Che and Qian (1998) show that in an insecure property rights protection environment, firms have more incentive to hide their profit, and this incentive becomes higher in areas with less stringent law enforcement where profit hiding behavior has lower cost. The difference in legal environment stringency across provinces leads to a varied cost of accrual-based earnings management relative to real activities manipulation. In a less stringent legal environment, the cost of accrual-based earnings management would be lower relative to that of real activities earnings management, thus, higher level of accrual-based earnings management will be adopted. Hence, we formulate the following hypothesis: H2a: Other things being equal, firms in a less stringent legal environment engage in a higher level of accrual-based earnings management. A-shares are issued by the firms listed in mainland China and are traded by RMB (The Chinese currency). H-shares refer to the shares registered in mainland China and listed on the Hong Kong stock exchange. Institutional investors can invest in H-shares, while individuals in mainland China cannot invest in H-shares directly. Thus, firms listed in both A-shares and H-shares markets could attract more institutional investors,

- 14 -

who are more concerned about the firms’ long-term value compared with individual investors. Institutional investors are also believed to generate more impact on firms’ operational and financial activities compared with individual investors, making it more difficult for listed firms to engage in real activities manipulation. While real activities earnings management is more likely to be harmful to firms’ long-term value compared with accrual-based earnings management (Cohen and Zarowin, 2010), the relative cost of real activities earnings management is thus higher for double listed companies. When managers pay more attention to the long-term development of firms, accrual-based earnings management is less costly compared with real activities earnings management. Hence, we posit the following hypothesis: H2b: Other things being equal, firms listed in both A-shares and H-shares markets engage in a higher level of accrual-based earnings management. Firms with higher growth prospects reveal to the market better future financial and operational performance. Capital markets react more favorably toward growth firms. However, growth firms are penalized more by the stock market if they fall short of expectations (Skinner and Sloan 2002). Thus, compared with low growth firms, high growth firms are more concerned about long-term value. Real activities earnings management is achieved through a deviation from optimal operational decisions; thus, it is likely to generate a more severe impact on the long-term firm value. Firms with higher growth prospects also attract more financial analysts; higher analyst coverage increases the relatively cost to firms of engaging in real activities manipulation. Consequently, the relative cost of engaging in accrual-based earnings management is lower for companies with higher growth rates, so managers tend to adopt more accrual-based manipulation instead of real activities earnings management. Hence, we posit the following hypothesis: H2c: Other things being equal, firms with higher growth prospects engage in a higher level of accrual-based manipulation.

- 15 -

4 Data and research design 4.1 Sample selection Most of the data for this study are obtained from GENIUS FINANCE and CCER databases in China. Data about the level of legal stringency and government intervention are obtained from NERI INDEX of Marketization of China’s Provinces Report. To avoid the influence of new accounting standards adopted in China since 2007, we use all firms listed in Shenzhen and Shanghai Stock Exchange between 2008 and 2012 as our sample. We exclude firm-years without sufficient data available to estimate dependent and independent variables. The industry classification is defined following the standards set by China Securities Regulatory Commission (CSRC). Banks and financial institutions are eliminated. ST firms (firms with loss for two consecutive years) and *ST firms (firms with loss for three consecutive years) are also eliminated. The final sample includes 6766 firm-year observations.

4.2 Variable measurement 4.2.1 Dependent variables We use discretionary accruals to proxy for accrual-based earnings management. Discretionary accruals are the difference between firms’ actual accruals and the normal level of accruals. We estimate the latter using the following adjusted Jones model (1991): Accrualsit / Ait 1  0  1 1/ Ait 1   2  REVit  RECit  / Ait 1 

 3  IAit / Ait 1   4 ( FAit / Ait 1 )   it Where Accrualsit is net profit minus net operating cash flows reported in the statement of cash flows in year t ; Ait 1 is total assets in year t  1 ; REVit is the change in operating revenue from year t  1 to t ; RECit is the change in accounts

- 16 -

receivable from year t  1 to t ; IAit is intangible assets in year t ; and FAit is gross property, plant and equipment. The estimated residual ( AM it ) by year and industry captures discretionary accruals. We use the absolute value of the residual as the proxy for accrual-based earnings management. Following Roychowdhury (2006), we examine three types of real activities earnings management: accelerating the timing of sales and/or generating additional unsustainable sales through increased price discounts or more lenient credit terms; increasing earnings by reducing cost of goods sold through overproduction; and cutting discretionary expenditures, including R&D, advertising, and selling, general, and administrative (SG&A) expenditures. The first type is measured by the abnormal level of cash flow from operations. The second type is measured by the abnormal level of production cost, and the third type is measured by the abnormal level of discretionary expenditure. Following Roychowdhury (2006) and Li and Zhang (2010), we estimate the normal level of cash flow from operation with the following equation: CFOit / Ait 1  0  1 1/ Ait 1   2  Sit / Ait 1   3  Sit / Ait 1   4  Sit 1 / Ait 1   5  ECit / Ait 1   6 TCit / Ait 1   7  OCit / Ait 1  + it

Where CFOit is net operating cash flows reported in the statement of cash flows in year t ; Sit is sales in year t ; Sit is the change in sales from year t to t  1 ;

Sit 1 is the change in sales from year t  1 to t  2 ; ECit is cash paid to employees in year t ; TCit is tax payments in year t ; and OCit is cash received relating to other operating activities in year t . The abnormal level of cash flow from operation is measured as the estimated residual from the above equation. We multiply the residuals by 1 (denoted as ACFOit ) such that higher values indicate greater abnormal amounts of cash flow from operations.

- 17 -

We estimate the normal level of production costs following Roychowdhury (2006):

PRODit / Ait 1  0  1 1/ Ait 1   2  Sit / Ait 1   3  Sit / Ait 1   4  Sit 1 / Ait 1    it Where PRODit is the sum of cost of goods sold in year t and the change in inventory from year t  1 to t . The abnormal level of production costs ( APRODit ) is measured as the estimated residual from the above equation. The higher the residual, the larger the amount of inventory overproduction, and the greater the increase in reported earnings through reducing cost of goods sold. Also following Roychowdhury (2006), we estimate the normal level of discretionary expenditures using the following equation:

DISX it / Ait 1  0  1 1/ Ait 1   2  Sit 1 / Ait 1    it Where DISX it denotes discretionary expenditures, the sum of operating costs and administration costs; Sit 1 is sales in year t  1 . The abnormal level of discretionary expenditures is measured as the estimated residual from the above regression. We multiply the residuals by 1 (denoted as ADISX it ) such that higher values indicate greater amounts of discretionary expenditures cut by firms to increase reported earnings. We aggregate the three real activities earnings management measures into one proxy,

RM it , by taking their sum. RM it  ACFOit  APRODit  ADISX it

4.2.2 Independent variables For the six hypotheses developed in Section 3, there are six independent variables.

LAWit indicates the level of legal environment stringency, obtained from NERI INDEX of Marketization of China’s Provinces Report (Fan et al., 2011). The law index is composed from four aspects. The first one is the extent of development of market intermediaries, such as law firms, CPA firms, as well as the degree of the assistance - 18 -

which industry associations provide to local enterprises; the second one is the degree of protection of producers’ legitimate rights, measured by the justice and effectiveness of law enforcement agencies; the third one is the degree of intellectual property rights protection; the last one is the protection level of consumers' rights and interests. A higher index indicates a more stringent legal environment. The index is obtained from surveys conducted of firms operating within different administrative areas (provinces) in China. Detailed indices for each province are presented in Appendix A. DOU it is a dummy variable indicating whether the firm is listed in both China and Hong Kong stock markets, 1 for double-listed companies, and 0 otherwise. We use GROWit as the proxy for growth level, which is measured as the market-to-book value of assets.

GOVit indicates the extent of local government intervention, also obtained from NERI INDEX of Marketization of China’s Provinces Report (Fan et al., 2011). It is measured by the degree of complexity of the administrative approval/authorization process of government agencies over the establishment and various operation stages of companies. The index is also obtained from surveys conducted of representative firms in each province. Detailed indices for each province are also presented in Appendix A. The higher the index, the lower the level of government intervention. As Appendix A shows the raw index, in this study, we use 90 minus the raw index to measure government intervention, so that higher values indicate higher levels of government intervention.

LEVit is financial leverage. G  Indexit is the proxy for corporate governance level, obtained through principal component analysis (PCA) with seven variables reflecting internal and external corporate governance practices (Bai et al. 2005): duality of CEO and chairman of the board, the percentage of independent directors on board, the ownership of top managers, the ownership of the largest shareholder, the concentration of shareholding from second to tenth largest shareholders, whether or not the firm has a parent firm, and whether or not the firm is a state-owned enterprise (SOE). A description of these seven variables can be found in Appendix B. For LAWit (legal

- 19 -

environment) and GOVit (local government intervention), we identify the province in which the offices of listed firms are located, and then match the corresponding index with the listed companies.

4.2.3 Control variables To test the hypotheses, we include six control variables that are likely to affect the incentive of managers to engage in earnings management based on previous literature. LnAit is the industry-adjusted log value of total assets used to control for relative firm size in the industry. ROAit , computed using net income divided by total assets in year t , is used to control for firm performance. EU it is environmental uncertainty, measured by the adjusted sales revenue variation coefficient for five years, i.e., the standard deviation of sales revenue for five years divided by the mean of sales revenue. BIG10it indicates whether the firm’s auditor is one of the Big 10 audit firms in China (PwC Zhongtian, Deloitte Huayong, Ruihua, EY Huaming, Lixin, KPMG Huazhen, Daxin, Tianjian, ShineWing, Dahua, from China CPA Association, 2013).

SEOit is an indicator of seasoned equity offering, equal to 1 if the firm issues equity in the next year, and 0 otherwise. We also include industry and year dummies. LOSSit is a dummy variable, taking the value of 1 if the net profit of the firm is less than 0, and 0 otherwise. Table 1 describes the main variables for regressions. Insert Table 1 here

4.3 Regression model To investigate how managers choose between accrual-based earnings management versus real activities earnings management, we use the following regression models: Regression model for H1:

- 20 -

RM it   0  1 LAWit   2 DOU it   3GROWTH it   4GOVit   5 DEBTit  6G  Indexit   7 LnAit  8 ROAit   9 EU it  10 Big10it  11SEOit

(10)

12 LOSSit  13YEAR  14 INDUSTRY  it Regression model for H2:

AM it  0  1LAWit   2 DOU it  3GROWTH it   4GOVit  5 DEBTit  6G  Indexit   7 LnAit  8 ROAit  9 EU it  10 Big10it  11SEOit

(11)

 12 LOSSit  13UNEXPRM it  14YEAR  15 INDUSTRY   it According to Hypothesis 1, Firms under low government intervention are more likely to engage in real activities earnings management,  4 is expected to be negative. The higher the debt ratio, the more likely firms will engage in real activities earnings management, so the coefficient for DEBTit ,  5 is expected to be positive. Meanwhile, firms under stronger corporate governance mechanism are expected to engage in a lower level of real activities earnings management,  6 is expected to be negative. According to Hypothesis 2, firms in a more stringent legal environment are expected to engage in less accrual-based earnings management, so the coefficient for

LAWit , 1 is expected to be negative. Double-listed companies and growth firms are expected to engage in a higher level of accrual-based earnings management, so the coefficients for DOU it and GROWit ,  2 and  3 are expected to be positive. Finally, according to Zang (2012), accrual-based earnings management takes place at the end of each period, while real activities earnings management is incurred during the whole accounting period, thus the level of accrual-based earnings is affected by the unexpected level of real activities manipulation. Thus, we include UNEXPRM it in the regression on accrual-based earnings management. UNEXPRM it is the estimated residual from equation 10.

- 21 -

5 Empirical results 5.1 Descriptive statistics Panel A of Table 2 presents the regression results for the estimation of normal level of accruals, Panel B of Table 2 presents regression results for the estimation of normal levels of cash flow from operation, production cost and discretionary expenditure respectively. The coefficients are consistent with extant studies on accrual-based and real activities earnings management in China. Table 3 reports descriptive statistics for the independent variables and control variables. From table 3, we observe that around 3% of the firms are listed in both A-Share and H-Share stock market; the average debt ratio is 0.46; the average ROA is 6.6%; 26% of the firms are audited by Big 10 auditors in China, and around 9% of the firms in the sample period have a loss. Table 4 presents the Pearson correlation results among the variables in the empirical tests. From the correlation results, we observe that there is a significant and negative correlation between AM it and RM it . The correlation between LAWit and

GOVit is significantly negative, consistent with the fact that government intervenes less in a more stringent legal environment. Insert Tables 2, 3, and 4 here

5.2 Empirical results Table 5 presents regression results regarding the factors that affect real activities earnings management (t-statistics are presented in brackets). To better understand the effect of independent variables on accrual-based earnings management, we present the results with and without control variables respectively. Consistent with H1a, for firms under higher level of government intervention, the cost of real activities manipulation is higher than that of accrual-based earnings management. Thus firms tend to engage in a lower level of real activities manipulation, and the coefficient on GOVit is negative and significant at the 0.05 level. Firms with higher debt ratios face higher costs of - 22 -

accrual-based earnings management, and are more likely to engage in real activities earnings management, as demonstrated by a positive coefficient on DEBTit , consistent with H1b1. As predicted by H1c, firms with stronger corporate governance are more concerned with long-term firm value, resulting in higher relative cost of real activities earnings management. The coefficient on G  Indexit is negative and significant at the 0.01 level, showing that a strong corporate governance mechanism is effective in deterring real activities earnings management for Chinese listed firms. Firms facing a seasoned equity offering are more likely to engage in real activities earnings management, with the coefficient on SEOit positive and significant, consistent with the capital market incentive of earnings management observed in the extant literature. The regression results without control variables are consistent with those in the fully specified model. Insert Table 5 here Table 6 presents results regarding the factors that affect managers’ choice of accrual-based earnings management2 (t-statistics are presented in brackets). To better understand the effect of independent variables on accrual-based earnings management, we present the results with and without control variables respectively. Consistent with H1a, the coefficient on LAWit is negative and significant at the 0.05 level, indicating that firms in a less stringent legal environment engage in a higher level of accrual-based earnings management due to lower cost associated with this kind of earnings management. Consistent with H1b, compared with firms only listed in A-share market, double-listed firms are more likely to participate in accrual-based manipulation, as indicated by the positive coefficient on DOU it (significant at the 0.01 level). This prediction is further supported by the negative and significant 1

Following Roychowdhury (2006), we also employ the dummy variable of existence of debt as the alternative proxy for debt incentive. The results are qualitatively unchanged both for the regression of real activities earnings management and accrual-based earnings management. 2 Following Kothari et al. (2005), we also adopt the performance-based discretionary accrual measure when examining the factors that affect accrual-based earnings management, i.e., we adjust the discretionary accrual with matched firms based on performance (ROA) and industry, and the empirical results are qualitatively the same.

- 23 -

coefficient on DOU it about real activities manipulation in Table 5. Finally, the coefficient on GROWTHit is significantly positive at the 0.01 level, suggesting that firms with higher growth prospects are more likely to engage in accrual-based earnings management, supporting H1c. The negative coefficient on GROWTHit in Table 5 on real activities earnings management further supports the hypothesis, indicating that the relative cost related to real activities earnings management is higher for firms with greater growth prospects. Differing from extant studies in the US setting (Zang 2012), the coefficient on BIG10it is positive rather than negative. One explanation for this result might be that in economies with relatively weak legal system and investor protection such as China, auditors with higher reputation are not necessarily more stringent scrutinizer of earnings management. In fact, some larger audit firms even face lower audit risk (Liu and Zhou 2007)3. Firms with a higher level of discretionary accrual might hire more reputable auditors to deliver a positive signal about the credibility of their financial report to the market. This finding is also consistent with several studies in the Chinese setting (Liu et al. 2011;Wang et al. 2011). SEOit is also positively and significantly associated with accrual-based earnings management. Consistent with Zang (2012), the coefficient on UNEXPRM it is negative and significant at 0.01 level, showing that accrual-based earnings management is negatively associated with the unexpected part of real activities manipulation. This result suggests that managers adjust the amount of accrual-based earnings management after real activities manipulation is realized. The regression results without control variables are consistent with those of the fully specified model. Insert Table 6 here

3

We also use BIG4 rather than BIG10 in the regressions, and the empirical results are qualitative the same, with the coefficient on BIG4 positive and significant at 5% level in the accrual-based earnings management model, and the coefficient in the real activities earnings management model insignificant.

- 24 -

6. Conclusion In this study, we theoretically and empirically investigate the trade-off between real activities earnings management and accrual-based earnings management of Chinese listed companies. We provide theoretical evidence that managers choose earnings management methods based on the relative costs of accrual-based and real activities earnings management. Our empirical results show that firms in a less stringent legal environment, double-listed firms, and firms with higher growth prospects are more likely to engage in accrual-based earnings management due to lower relative costs of this kind of earnings management. Meanwhile, higher leveraged firms, firms under lower government intervention and with weaker corporate governance mechanisms, which are faced with lower relative cost of real activities earnings management, are more likely to engage in real activities manipulation. This study contributes to the literature about the trade-off between accrual-based and real activities earnings management. It shows that internal factors (leverage, growth level and corporate governance mechanism) and external factors (legal environment, government intervention and dual-listing status) specific to China's unique institutional characteristics play an important role in managers' incentive to choose between the two kinds of earnings management, suggesting that it is important to consider institutional influence when studying the accounting choice in different economies. Moreover, while previous empirical studies on the effect of institutional characteristics on financial reporting quality mainly examine the country level impact (Ball et al. 2003, Leuz et al. 2003), the current study shows that province/state level features also play a role in affecting the accounting choice on earnings management. Moreover, while previous empirical studies mainly examine the effect of firm specific factors (such as corporate governance) and institutional characteristics (such as government intervention and legal environment) on accrual-based earnings management, this study extends the extant literature by investigating their effect on both accrual-based and real activities manipulation, suggesting that considering only

- 25 -

one kind of earnings management could generate biased results for investors, regulatory bodies and other market participants.

APPENDIX A Rule of Law Index and Government Intervention Index (NERI INDEX of Marketization of China’s Provinces 2011 Report) Index Province Peking Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hubei Hunan Guangdong Guangxi Hainan Chongqing Sichuan Guizhou Yunnan Xizang Shanxi Gansu Qinghai

Rule of Law Index 16.27 11.57 5.60 5.55 5.32 8.46 6.00 5.96 19.89 18.72 19.85 7.32 8.30 5.90 8.18 6.07 7.15 6.02 13.99 4.88 5.25 7.60 7.39 4.47 5.44 0.18 5.88 4.86 3.51 - 26 -

Government Intervention Index 7.20 8.02 3.43 2.16 1.08 6.38 4.72 5.68 10.00 9.89 9.32 7.59 5.41 6.48 6.85 5.51 5.38 4.91 8.12 3.76 0.84 5.44 6.91 0.24 7.02 -12.95 0.66 3.93 1.30

Ningxia Xinjiang

4.66 4.98

1.30 0.00

APPENDIX B Description of G-Index G-Index is the corporate governance index estimated with the principal component analysis (PCA). Following Bai et al.(2005), we set up an evaluation index system which includes seven variables indicating both the internal and external corporate governance practice of Chinese listed firms. These indicators reflect the board governance mechanism, managerial compensation and ownership structure effect. Description of the variables is tabulated as below. Variable Name

Definition

Top1

Ownership of the largest shareholder

Equil

Ownership concentration of the second to the tenth largest shareholders

Parent

Dummy variable indicating whether the firm has a parent, 1 if there is a parent company, 0 otherwise

Indratio

The proportion of the number of independent directors on board

Duality

Dummy variable for duality, 1 if CEO is also board director, 0 otherwise

Top5

Ownership of the five top management members

State

Dummy variable. 1 if the firm is a state-controlled enterprise, and 0 otherwise.

Applying the principal component analysis method (PCA) on these seven variables, the first principal component is the G-Index, specifically, G-Index= -0.540*Top1+0.815*Equil-0.778*Parent+0.337*Indratio+0.722*Duality +0.960*Top5+0.668*State

- 27 -

References Baber W.R., Fairfield P.M., and Haggard J.A. (1991). The Effect of Concern About Reported Income on Discretionary Spending Decisions: The Case of Research and Development. The Accounting Review, 66(4), 818. Bagnoli M., and Watts S.G. (2000). The Effect of Relative Performance Evaluation On Earnings management: A Game-Theoretic Approach. Journal of Accounting and Public Policy, 19(4,5), 377-397. Bai C. E., Liu Q, Lu J., Song M. F. and Zhang J. X (2005). An Empirical Study on Chinese Listed Firms'Corporate Governance. Economic Research Journal, 02, 81-91. Ball, R., A. Robin, and J. Wu. 2003. Incentives versus standards: Properties of accounting income in four East Asian countries. Journal of Accounting and Economics 36 (1/3): 235–270. Burgstahler D., and Dichev I. (1997). Earnings management to Avoid Earnings Decreases and Losses. Journal of Accounting and Economics, 24(1), 99-126. Bushee B.J. (1998). The Influence of Institutional Investors On Myopic R&D Investment Behavior. The Accounting Review, 73(3), 305-333. Cai C., Zhu R., He H., and Xie L. (2012). The Choice of Earnings management Methods,Hidden Behavior and the Improvement of Condition for the Dying Firms:Empirical Evidence from ST Listed Companies in a-stock Market [in Chinese]. Accounting Research, 09, 31-39. Chan D. and Gao J. (2014). Earnings management, Incentive Contracts and Private Information Acquisition. Journal of Accounting and Public Policy 33(6), 529-550. Che J and Qian Y. (1998). Insecure Property Rights and Government Ownership of Firms. Quarterly Journal of Economics, ,113(2), 467-496. Chen, H., J. Chen, G. J. Lobo, and Y. Wang. (2011). Effects of audit quality on earnings management and cost of equity capital: Evidence from China. Contemporary Accounting Research 28 (3), 892-925. Cheng S. (2004). R&D Expenditures and CEO Compensation. The Accounting Review, 79(2), 305-328. Cheng Z.M., Xia X.P. and Yu M.G.(2008). Government Intervention, Pyramidal Structure and Investment of Local SOEs.[in Chinese] Management World, 2008(9), 37-47. Cohen D.A., Dey A., and Lys T.Z. (2008). Real and Accrual-Based Earnings management in the Pre- and Post-Sarbanes-Oxley Periods. The Accounting Review, 83(3), 757-787. Cohen D.A., and Zarowin P. (2010). Accrual-Based and Real Earnings management Activities Around Seasoned Equity Offerings. Journal of Accounting and Economics, 50(1), 2. Crocker K.J., and Slemrod J. (2007). The Economics of Earnings Earnings management and Managerial Compensation. RAND Journal of Economics, 38(3), 698-713. Dechow P.M., and Sloan R.G. (1991). Executive Incentives and the Horizon Problem: An Empirical Investigation. Journal of Accounting and Economics, 14(1), 51. DeFond M.L., and Jiambalvo J. (1994). Debt Covenant Violation and Earnings management of Accruals. Journal of Accounting and Economics, 17(1-2), 145-176. Demski J.S., Frimor H., and Sappington D.E.M. (2004). Efficient Earnings management in a Repeated Setting. Journal of Accounting Research, 42(1), 31-49. DuCharme L.L., Malatesta P.H., and Sefcik S.E. (2004). Earnings management, Stock Issues, and Shareholder Lawsuits. Journal of Financial Economics, 71(1), 27-49. Dye R.A. (1988). Earnings management in an Overlapping Generations Model. Journal of Accounting Research, 26(2), 195-235.

- 28 -

Evans J.H.I., and Sridhar S.S. (1996). Multiple Control Systems, Accrual Accounting, and Earnings management. Journal of Accounting Research, 34(1), 45-65. Fan G., Wang X., and Zhu H. (2011). NERI INDEX of Marketization of China's Provinces 2011 Report. (1st ed). Beijing: Economic Science Press. Fang H., and Jin Y. (2011). Can High Quality Internal Control Reduce Earnings management? An Empirical Research Based on Voluntary Internal Control Audits Reporting [in Chinese]. Accounting Research, 08, 53-60. Fischer P.E., and Stocken P.C. (2004). Effect of Investor Speculation On Earnings management. Journal of Accounting Research, 42(5), 843-870. Fischer P.E., and Verrecchia R.E. (2000). Reporting Bias. Accounting Review, 75(2), 229-245. Gao L., and Zhang J. (2008). Corporate Governance, Institutional Investors, and Earnings management [in Chinese]. Accounting Research, 09, 64-72. Goldman E., and Slezak S.L. (2006). An Equilibrium Model of Incentive Contracts in the Presence of Information Earnings management. Journal of Financial Economics, 80(3), 603-626. Jackson S.B., and Wilcox W.E. (2000). Do Managers Grant Sales Price Reductions to Avoid Losses and Declines in Earnings and Sales? Quarterly Journal of Business and Economics, 39(4), 3-20. Jones J.J. (1991). Earnings management During Import Relief Investigations. Journal of Accounting Research, 29(2), 193. Kothari, S.P., Leone J. A. and Wasley. C. 2005. Performance matched discretionary accrual measures. Journal of Accounting and Economics 39: 163–197. Lakshmanan S. (2000). Do Firms Mislead Investors by Overstating Earnings Before Seasoned Equity Offerings? Journal of Accounting and Economics, 29(3), 339-371. Lennox, C. S., X. Wu, and T. Zhang. 2014. Does mandatory rotation of audit partners improve audit quality? The Accounting Review 89 (5): 1775-1803. Leuz, C., D. Nanda, and P. D. Wysocki. 2003. Earnings management and investor protection: An international comparison. Journal of Financial Economics 69 (3): 505–527. Li B., and Zhang J. (2010). Research On Economic Consequences of Real Earnings management: Experimental Evidence From Sale Earnings management [in Chinese]. Management Review, 22(09), 84-92. Li Y. X., Chen K. J., Yao H. and Liu L. (2012). Research on the Relation between External Governance Environment and Earnings Management Based on Regional Differences. [in Chinese]. Nankai Business Review, 15 (4), 89-100. Li Z., Dong Z., and Lian Y. (2011). Accrual-Based Earnings management or Real Activities Earnings management? A Study Based On China's Income Tax Reform in 2007 [in Chinese]. Management World, 01, 121-134. Lin S., and Wei M. (2000). The Earnings management by Chinese A-Share Firms in the IPO Process [in Chinese]. China Accounting and Finance Review, 2(2), 87-130. Lin Y.F. and Li. Z.Y.(2004). Policy Burden, Moral Hazard and Soft Budget Constraint. [in Chinese].Economic Research, 2004(2),17-27. Liu F. and Zhou F. Y. (2007). Do Big 4 Mean Higher Audit Quality? [In Chinese]. Accounting Research, 3,79-94. Liu Q. L., He W. F. and Luo L. (2011). Mandatory Adoption of IFRS, Implementation of New Laws, and Accrual and Real Earnings Management. China Accounting and Finance Review,13 (1), 57-121.

- 29 -

Lu J. (1999). An Empirical Study On Earnings management of Chinese ST Listed Companies [in Chinese]. Accounting Research, 09, 25-35. Narayanan M.P. (1985). Managerial Incentives for Short-Term Results. Journal of Finance, 40(5), 1469-1484. Rangan S. (1998). Earnings management and the Performance of Seasoned Equity Offerings. Journal of Financial Economics, 50(1), 101-122. Ronen J., Tzur J., and Yaari V.L. (2006). The Effect of Directors' Equity Incentives On Earnings management. Journal of Accounting and Public Policy, 25(4), 359-389. Ross S.A. (1973). The Economic Theory of Agency: The Principal's Problem. American Economic Review, 63(2), 134-139. Roychowdhury S. (2006). Earnings management through Real Activities Earnings management. Journal of Accounting and Economics, 42(3), 335. Skinner D. and Sloan R. (2002). Earnings Surprises, Growth Expectations and Stock Returns. Review of Accounting Studies, 7 (2), 289-312. Stein J.C. (1989). Efficient Capital Markets, Inefficient Firms: A Model of Myopic Corporate Behavior. The Quarterly Journal of Economics, 104(4), 655. Sweeney A.P. (1994). Debt-Covenant Violations and Managers' Accounting Responses. Journal of Accounting and Economics, 17(3), 281-308. Teoh S.H., Wong T.J., and Rao G.R. (1998). Are Accruals During Initial Public Offerings Opportunistic? Review of Accounting Studies, 3(1-2), 175-208. Wang B., Su, W. B and He M. Z. (2011). Is There Difference in Audit Quality among Big 4 Audit Firms in China? Auditing Research, 6, 89-97. Wang Y., Wang L., and Gong C. (2009). Reform of Enterprise Income Tax, Earnings management and its Economic Consequences [in Chinese]. Economic Research Journal, 03, 86-98. Wongsunwai W. (2013). The Effect of External Monitoring On Accrual-Based and Real Earnings management: Evidence From Venture-Backed Initial Public Offerings. Contemporary Accounting Research, 30(1), 296. Zang A.Y. (2012). Evidence On the Trade-Off Between Real Activities Earnings management and Accrual-Based Earnings management. The Accounting Review, 87(2), 675-703. Zhang J., Li B., and Liu D. (2008). Real Activities Earnings management on Earnings management: A Loss-Avoidance Based Empirical Evidece. Application of Statistics and Management, 05, 918-927. Zhang W. D., Zhang J. K., Cheng Z. C. and Xu X. (2015). Asset Injection under the Government Intervention, Pyramid Ownership Structure and Corporate Governance, [In Chinese] Accounting Research, 3, 42-49.

- 30 -

TABLE 1 Variable definitions Dependent variable

Definition

AM

Accrual-based manipulation, estimated from modified Jones (1991) model

RM

Real activities manipulation, estimated from Roychowdhury (2006) model

Independent variable LAW

Definition

The level of legal environment stringency, obtained from NERI INDEX of Marketization of China’s Provinces Report

DOU

A dummy variable indicating whether the firm is listed in both China and Hong Kong stock market, 1 for double-listed companies, and 0 otherwise

GROWTH

Market-to-book value

DEBT

Debt ratio

GOV

The extent of local government intervention, obtained from NERI INDEX of Marketization of China’s Provinces Report

LOSS

A dummy variable indicating whether the firm is under loss, 1 if the net profit of the firm is less than zero, and 0 otherwise

LnA

Firm size, measured by industry-adjusted log value of total assets

ROA

Net income divided by total assets, multiplied by 100%

EU

Environmental uncertainty, measured by the standard deviation of sales revenue for the past five years divided by the mean of sales revenue

BIG10

A dummy variable which equals 1 if the firm’s auditor is one of the Big 10 audit firms in China, and 0 otherwise

SEO

A dummy variable for seasoned equity offering, which equals 1 if the firm issues equity in the next year, and 0 otherwise

G-Index

Corporate governance level, obtained through principal component analysis

- 31 -

(PCA) with seven variables reflecting the corporate governance practice

- 32 -

Table 2 Measurement of Accrual-based Earnings management and Real Activities Manipulation Panel A: Estimation of the Normal Levels of Accruals TAit/Ait-1 -0.125***

Intercept

(-6.644) 0.000***

1/Ait-1

(138.806) (ΔREVit-ΔRECit)/Ait-1

0.359*** (94.608) -7.802***

IAit/Ait-1

(-98.757) -0.713***

FAit/Ait-1

(-26.660) Obs.

6766 2

Adj. R

0.891

Panel B: Estimation of the Normal Levels of Cash Flow from Operation, Production Costs and Discretionary Expenditures CFOit/Ait-1 Intercept

PRODit/Ait-1 ***

-0.055

DISXit/Ait-1

Intercept

0.087

(-5.266) 1/Ait-1

0.000

0.000

(1.533) Sit/Ait-1

0.266

ΔSit/Ait-1

-0.312

Sit/Ait-1

0.343

ΔSit-1/Ait-1

0.195

ΔSit/At-1

0.259

3.367

0.498*** (12.078)

1/Ait-1

0.000*** (-251.759)

Sit/Ait-1

0.958*** (311.683)

***

(47.583) ΔSit-1/Ait-1

-0.094***

(11.489) ECit/Ait-1

***

(55.051)

(-141.774) **

***

(11.545)

(69.666) ***

Intercept

(1.994) 1/Ait-1

***

***

(-1.256)

***

(104.911) TCit/Ait-1

-4.660*** (-130.672)

OCit/Ait-1

0.000 (-0.445)

Obs. 2

Adj. R

6766

6766

6766

0.988

0.960

0.943

***, **, * represent significance at the level of 1 percent, 5 percent, and 10 percent, respectively.

- 33 -

TABLE 3 Descriptive Statistics Variable

Obs

Min

Max

Mean

Median

Std.Dev.

LAW

6766

0.180

19.890

11.680

8.640

5.780

DOU

6766

0.000

1.000

0.030

0.000

0.170

GROWTH

6766

-7.714

49.548

1.716

1.126

3.200

GOV

6766

80.000

102.950

83.277

82.800

2.971

DEBT

6766

0.000

1.240

0.464

0.474

0.207

G-Index

6766

-4.944

29.714

-1.042

-2.258

4.704

LnA

6766

18.833

28.405

21.922

21.754

1.244

ROA

6766

-25.163

47.829

6.597

5.834

6.398

EU

6766

1.001

1.517

1.151

1.128

0.087

BIG10

6766

0.000

1.000

0.260

0.000

0.438

SEO

6766

0.000

1.000

0.040

0.000

0.194

LOSS

6766

0.000

1.000

0.090

0.000

0.280

AM

1.000

RM

-.052*** 1.000

LAW

-.027** .000

1.000

DOU

.067*** .023

.023*

1.000

GROWTH .047*** .012

-.022*

-.010

GOV

.016

-.799*** -.036*** .031** 1.000

DEBT

-.051*** .098*** -.055*** .070*** .080*** .039*** 1.000

G-Index

-.012

LnA

-.092*** .119*** -.006

.349*** .124*** .016

.393*** .005

1.000

ROA

.020*

.000

-.241*** -.003

-.274*** -.004

.081*** 1.000

EU

-.033*** -.036*** -.012

.012

-.016

BIG10

.029**

SEO

.033*** .048*** .023*

.005

.027** -.022

-.007

LOSS

.015

.011

-.114*** .007

.167*** .006

-.016

-.034*** .062*** -.021* -.031** .009 .017 -.003

LOSS

SEO

BIG10

1.000

.015

.068*** .061*** -.011 -.022*

EU

ROA

LnA

G-Index

DEBT

GOV

GROWTH

DOU

LAW

RM

AM

TABLE 4 Pearson Correlations

-.029** .024*

.041*** .009

1.000

.034*** -.017

-.038*** .060*** -.067*** .008 .012

.002

-.007

.027** -.019 1.000 .016

.001 -.025** 1.000

-.087*** -.504*** .015 -.006

***, **, * represent significance at the level of 1 percent, 5 percent, and 10 percent, respectively.

- 34 -

1.000

-.013 1.000

TABLE 5 Regression Result of Real Activities Earnings Management

RM it   0  1 LAWit   2 DOU it   3GROWTH it   4GOVit   5 DEBTit   6G  Indexit  7 LnAit  8 ROAit   9 EU it  10 Big10it  11 SEOit  12 LOSSit  13YEAR 14 INDUSTRY  it Pred.Sign

Coefficients

Intercept -

LAW DOU

-

GROWTH

-

GOV

1.681*

-1.553

(1.65)

(-1.40)

-0.007

-0.004

(-1.14)

(-0.74)

0.100

-0.254**

(0.82)

(-2.01)

-0.081***

-0.087***

(-4.45)

(-4.02)

-0.021

*

(-1.84) DEBT

+

0.691

***

(6.66) G-Index

(-2.08) 0.205* (1.79)

-0.013

-0.013***

(-2.96)

(-3.11)

***

-

-0.023**

0.163***

LnA

-7.67 -0.011***

ROA

(-2.81) -0.584**

EU

(-2.53) -0.039

Big10

(-0.76) 0.346***

SEO

-3.38 -0.076

LOSS

(-0.90) YEAR

controlled

INDUSTRY

controlled

Obs. Adj.R

2

6766

6766

0.014

0.086

Notes: 1. Dependent variable is RM, real activities earnings management based on Roychowdhury (2006) model. The other variables are defined in Table 1. 2. ***, ** and * represent significance at the level of 1 percent, 5 percent, and 10 percent, respectively.

- 35 -

TABLE 6 Regression Result of Accrual-based Earnings Management

AM it   0  1LAWit   2 DOU it  3GROWTH it   4GOVit  5 DEBTit   6G  Indexit  7 LnAit  8 ROAit  9 EU it  10 Big10it  11SEOit  12 LOSSit  13UNEXPRM it  14YEAR  15 INDUSTRY   it Pred.Sign

Coefficients

Intercept LAW

-

DOU

+

GROWTH

+

GOV

-

DEBT

-

G-Index

-

0.806

3.049***

(1.55)

(5.32)

-0.006**

-0.008**

(-2.06)

(-2.48)

0.408***

0.549***

(6.57)

(8.40)

0.082***

0.061***

(8.78)

(5.48)

-0.005

-0.005

(-0.77)

(-0.94)

-0.124**

0.039

(-2.35)

(0.66)

-0.001

-0.002

(-0.47)

(-1.00) -0.081***

LnA

(-7.41) 0.002

ROA

(0.80) -0.363***

EU

(-3.05) 0.098***

Big10

(3.65) 0.150***

SEO

(2.84) 0.038

LOSS

(0.86) -0.097***

UNEXPRM

(-13.24) YEAR

controlled

INDUSTRY

controlled

Obs. Adj.R

2

6766

6766

0.019

0.062

Notes: 1. Dependent variable is AM, accrual-based earnings management based on modified Jones model (1991). UNEXPRM is the estimated residuals from Eq.(10). Description of all the other variables is tabulated in Table 1.

- 36 -

2. ***, ** and * represent significance at the level of 1 percent, 5 percent, and 10 percent, respectively.

- 37 -