External auditor types and the cost stickiness of listed companies

Abstract

External auditing is an important method of corporate governance that can effectively monitor and restrict a manager’s opportunistic behaviour and reduce cost stickiness. We focus on Chinese A-shares listed companies from 2002 to 2010 to explore the governance role of external auditors in cost stickiness. We classify audit firms into three types: the Big Four International, the Big Ten Domestic, and the non-Big Ten Domestic. The results show that the cost stickiness of firms audited by the Big Four International is significantly lower, while the cost stickiness difference between firms audited by the latter two types is not significant. Further tests show that the restriction on cost stickiness for the Big Four International is more significant when the largest shareholder’s shareholding ratio is lower or when the local market is less developed. These results indicate that the restriction placed by audit quality on cost stickiness varies with other corporate governance mechanisms. Our results are qualitatively unchanged under different robustness tests, including propensity score matching. This study shows that Big Four International audit firms can better reduce cost stickiness, which provides evidence of the higher audit quality of these firms and extends the study of restrictive factors on cost stickiness to external corporate governance mechanisms.

Keywords:

• Big Four International
• cost stickiness
• shareholding ratio
• marketisation level
• propensity score matching

View correction statement:

Erratum

1. Introduction

External auditing is an indispensable part of internal corporate governance (Cohen, Krishnamoorthy, & Wright, 2002; Fan & Wong, 2005). The modern enterprise system features the separation of ownership and management. This mechanism might improve the efficiency of capital and human talents. However, it also inevitably leads to conflict between principals and agents (Jensen & Meckling, 1976). It is possible that agents (managers) with information superiority would harm the interest of principals (shareholders) to maximise their own interests. External audits can mitigate information asymmetry between contracting parties and reduce internal agency cost through independent audit work (Watts & Zimmerman, 1983). Many Chinese studies focus on the choice and impact of external auditors, including constraints imposed by external auditors on earnings management and financial restatement (Cai, Huang, & Zhao, 2005; Li, Wang, & Wang, 2004; Liu, 2009; Liu & Yan, 2006; Wang & Zhang, 2005), and on related-party transactions (Gao & Wu, 2007; Hong, Xu, & Li, 2011; Yang, Lin, & Wang, 2009; Yue, 2006; Zeng & Ye, 2005). There is also a constraint through the governance role of the external auditors in the turnover of chief executive officers (CEOs) (Jiang, Liu, & Liu, 2007; Wang, 2011). However, there are fewer studies on the constraints imposed by external auditing on managerial behaviour (Jiang et al., 2007) in this strand of literature. Such studies as have been carried out focus mainly on the quality of information disclosure, which is the final output of the information disclosure process. This study argues that external audit governance is effective not only in the final part of information disclosure, but also during operational and production procedures. To explore this possibility, we examine the influence of external audit on cost stickiness. Cost stickiness refers to the phenomenon where costs decrease less when business activity falls than they increase when business activity rises (Anderson, Banker, & Janakiraman, 2003). Anderson et al. (2003) tested this phenomenon for the first time in the field of finance and accounting, using US data.

We look at the association between external audit and cost stickiness in China for the following three reasons. First, this study allows us to explore the mechanism of formation of cost stickiness from an external perspective. The literature on cost stickiness is mainly based on an internal perspective. Anderson et al. (2003), Calleja, Steliaros, and Thomas (2006), Chen, Lu, and Sougiannis (2012), and Sun and Liu (2004) point out that managerial opportunistic behaviour is one of the primary causes of cost stickiness and they mainly focus on internal factors such as the structure of the board of directors.¹ However, our study extends the investigation of the causes of cost stickiness to external governance mechanisms and looks at the role of external auditors for the first time, which contributes to the current literature. Second, it explores and shows the incentive for the demand for high-quality external audits, given China’s institutional background. Ball, Robin, and Wu (2003) argue that the large number of related-party transactions in East Asian countries leads to a lack of motivation for listed firms to provide high-quality accounting information. Defond, Wong, and Li (2000) also point out that, under government regulation, Chinese listed companies lack demand for high-quality audit services. However, the environment of Chinese listed firms has changed with the launch of non-tradable share reform, the development of marketisation,² and the move to public listing of many non-state-owned firms. Since the conflict between shareholders and managers is becoming increasingly severe, the need to monitor managers can become one source of demand for high-quality audits. Third, it provides evidence for the debate on the audit quality of the Big Four International firms. Most international studies agree that the audit services of the Big Four International firms are of higher quality (Becker, Defond, Jiambalvo, & Subramanyam, 1998; Francis, Maydew, & Sparks, 1999; Lennox, 1999).³ However, although the same opinion is dominant in Chinese studies, controversy still exists (Guo, 2011; Liu & Xu, 2002; Liu & Zhou, 2007; Qi, Chen, & Zhang, 2004; Wang & Wang, 2006; Yu, 2007).⁴ We focus on the governance role of the Big Four International firms and the Big Ten Domestic audit firms in China in order to provide evidence for this debate from a new perspective.

When there is a business boom, optimistic managers are likely to increase expenses (especially unproductive expenditures) drastically to enhance personal utility or to control more resources (Chen, Lu, & Sougiannis, 2012). When there is a business recession, however, managers may not be willing to decrease such expenses or may just decrease them slightly. Such behaviour results in high cost stickiness; that is, expenses decrease more slowly when revenue falls than expenses increase when revenue rises by the same amount. To the extent that high-quality external audits can effectively regulate the corporate financial process and constrain managers’ opportunistic behaviour (Watts & Zimmerman, 1983), we suggest high-quality external audits may help to reduce cost stickiness. Using Chinese A-shares listed companies from 2002 to 2010, we test the relationship between high quality external audit and cost stickiness. We classify audit firms into three types: the Big Four International, the Big Ten Domestic, and the non-Big Ten Domestic. The results show that the cost stickiness is significantly lower in listed companies audited by the Big Four International, while there is no significant cost stickiness difference between firms audited by the latter two types. In addition, the restriction on cost stickiness associated with the Big Four International is much more significant when the largest shareholder’s shareholding ratio is lower or when the local market is less developed, which indicates that the audit quality’s restriction on cost stickiness varies with other corporate governance mechanisms. This study reveals that the Big Four International are associated with cost stickiness and extends the study on cost stickiness factors to external governance mechanisms.

This paper contributes to the literature in the following three ways. First, it enriches the literature that explores factors associated with cost stickiness and sheds light on how external audit quality interacts with factors identified in prior literature. This paper investigates empirically for the first time how auditor supervision, an external governance method, impacts cost stickiness and finds that the Big Four International considerably inhibit cost stickiness. In addition, we find that the impact of audit quality on cost stickiness varies with changes in other corporate governance mechanisms, such as ownership structure and external environment, which helps to understand cost stickiness factors and how these other factors affect cost stickiness. Second, it shows motivation for the demand for high-quality auditing in China. We find the Big Four International firms are associated significantly with cost stickiness (e.g. when the shareholding ratio of major shareholders is lower). Such a restrictive effect on cost stickiness could effectively increase the benefits of hiring audit firms with higher audit quality and add to the demand for high-quality audit services. Third, we provide evidence of the quality differences between different audit firms from a new perspective. We show that while the Big Four International are significantly associated with cost stickiness, the Big Ten Domestic firms do not have a similar impact. Our results are consistent with the argument that the Big Four International produce higher-quality audit work.

The rest of the paper is structured as follows. Section 2 presents the institutional background, literature review and hypothesis development. Section 3 describes the research design. Section 4 shows the empirical test results. Section 5 concludes and discusses potential limitations.

2. Institutional background, literature review, and hypothesis development

2.1. Institutional background

The development of external auditing in China is closely related to politics and the economy. After China’s reform and opening up (after 1978), the CPA (Certified Public Accountants) system was restored and the audit business redeveloped (Liu & Xu, 2002). According to the Temporary Regulations on the Examination and Approval Procedures of Limited Liability Accounting Firms passed by the Ministry of Finance in 1993, early accounting firms were mostly established relying on government agencies, such as the local Finance Departments of the different provinces. This situation remained the same until April 1998, when the Ministry of Finance released The Notice about Decoupling Accounting Firms Which Have Securities or Futures Related Business from Their Linked Units, which launched the decoupling reform. By the end of 1999, the decoupling reform was mostly accomplished. However, domestic accounting firms that originally relied on government agencies still had many inherent deficiencies. Along with several serious incidents of fraud in China’s capital market, market attention to audit quality gradually grew, initiating the demand for the audit work of the original Big Five International firms. In 2001, the China Securities Regulatory Commission together with the Ministry of Finance released The Regulation on the Temporary License for Foreign Accounting Firms to Conduct Business in Financial Listed Firms. In the same year, the two regulatory departments authorised temporary licences to the Big Four (originally the Big Five) international accounting firms and required that listed firms in the banking, security, and insurance industry be audited by both domestic and international firms. Similar administrative regulations requiring certain types of firms to be audited by international audit firms were also implemented afterwards (Liu & Xu, 2002; Liu & Zhou, 2007). In 2001 China signed an agreement to open up the CPA service market when it joined the World Trade Organization (Qi et al., 2004), which led to the complete opening of the auditing market to the Big Four International. Over the past ten years, the Big Four International firms have developed rapidly in the audit market in China. Their total income accounted for more than 55% of the total income of China’s top 100 accounting firms at the end of 2007. After 2007, as a large number of small companies went public in the SME (small and medium sized enterprises) and GEM (growth enterprises market) boards,⁵ the income share of the Big Four International declined but still surpassed 30% of the total income of the top 100 accounting firms (according to data from The Comprehensive Evaluation of Top 100 Accounting Firms in China published by the Chinese Association of Certified Public Accountants from 2003 to 2013). In addition, the Big Four International audit firms are widely recognised as having professional advantages in human capital, accumulative knowledge, and non-audit services (Qi et al., 2004; Wang & Wang, 2006).

2.2. Literature review

Anderson et al. (2003) were the first in the field of finance and accounting to propose the concept of cost stickiness, which refers to the phenomenon where costs decrease less when business activity falls than they increase when business activity rises. The authors use a sample of US-listed companies to test this theory empirically. They find that selling, general, and administrative (SG&A) costs increase by 0.55% when sales revenue increases by 1% but only decrease by 0.35% when sales revenue decreases by 1%, which provides preliminary evidence of cost stickiness.

Numerous studies further confirm the existence of cost stickiness (Calleja et al., 2006; Chen, Lu, & Sougiannia, 2012; Dierynck, Landsman, & Renders, 2012) and they focus more on the causes of cost stickiness. Among these causes, managers’ opportunistic behaviour is considered to be crucial (Banker, Byzalov, & Plehn-Dujowich, 2011; Sun & Liu, 2004)⁶ and empirical evidence is relatively abundant. Calleja et al. (2006) compare the degree of cost stickiness in four countries. They point out that weaker corporate governance accounts for the higher cost stickiness in France and Germany than in the United States and the United Kingdom. Chen, Lu, and Sougiannis (2012) further examine the effect of agency costs between stockholders and managers on cost stickiness. They find that cost stickiness is highly positively related to the incentive for managerial empire building (such as free cash flow, executive tenure, expectation to leave, and compensation structure). They also find that strong corporate governance can mitigate this positive association. Dierynck et al. (2012) provide evidence that executives’ incentive to conduct earnings management affects labour cost stickiness. They find that, to meet the goal of profit, management tends to manipulate labour costs. Apart from managers’ opportunistic behaviour, adjustment cost and management optimistic expectations can also cause cost stickiness. Regarding adjustment cost, Anderson et al. (2003) propose that firms with large proportions of physical assets and human capital tend to have greater cost stickiness, since the adjustment costs of these two assets are higher. Banker, Byzalor, and Chen (2013) find that the power of a country’s labour union has a significant influence on the adjustment costs of human resource. If a county’s labour union is more powerful, the adjustment cost of human resources will be greater and labour cost stickiness will be correspondingly higher. Regarding management optimistic expectations, Anderson et al. (2003) suggest that when the macro-economy grows rapidly, management will be optimistic about the future, even after a short-term decline in revenue, so cost stickiness will be greater. Similarly, Banker et al. (2011) find that companies in growing industries have greater cost stickiness.

In terms of Chinese domestic studies, Sun and Liu (2004) find that cost stickiness also exists in China’s listed companies, but macroeconomic growth and capital intensive cannot explain it well. Therefore, managers’ opportunistic behaviour may be a better explanation. Wan and Wang (2011) find that CEO non-duality and a higher proportion of independent directors can effectively reduce cost stickiness.⁷ Chen, Song, and Shi (2012) propose that controlling for managers’ incentives to manage earnings, as in a ‘big bath’, can drastically decrease the estimated cost stickiness. Liang (2013) finds that cost stickiness is negatively associated with the level of ownership concentration.

Given the findings of the foreign and domestic literature, there have been many studies on the impact of managers’ opportunistic behaviour on cost stickiness. However, these studies mainly focus on the effect of internal corporate governance mechanisms rather than external governance mechanisms.⁸ Based on the studies above, we further include audit supervision, an important external governance factor, in this line of study and examine the effect of external audit quality on the cost stickiness of listed companies.

2.3. Hypothesis development

In this paper, we seek to develop the argument that that high-quality external audits can reduce listed companies’ cost stickiness. Managers’ opportunistic behaviour is a main cause of cost stickiness (Banker et al., 2011). In modern corporations, with the separation of management and ownership, minority shareholders have to pay large amounts to obtain comprehensive information about companies’ operation and development, while managers do not (Jensen & Meckling, 1976). In addition, since executive incentive contracts cannot completely reconcile shareholders’ and managers’ interests, managers pervasively have an ‘empire building’ incentive. Managers tend to over-expend firm size and spend a great amount of resources on personal gains instead of improving company value (Jensen, 1986; Stulz, 1990).⁹ When business performance declines, managers’ optimal behaviour should be to cut down costs and keep them in accordance with activities. However, managers with strong opportunistic incentives will not adjust or only slowly adjust costs when sales decline (Chen, Lu, & Sougiannis, 2012). For example, to pursue personal prestige and rights, managers hire large numbers of employees, frequently update office equipment, and incur many other expenses. In such a situation, costs decrease less when business activity falls than costs increase when business activity rises and we regard this asymmetric cost change as cost stickiness.

We propose that high-quality external audits can decrease managers’ opportunistic behaviour and further help decrease cost stickiness. Becker et al. (1998) and Francis et al. (1999) point out that high-quality external audits have a great influence on corporate behaviour, such as standardising the financial procedures of listed companies, improving information transparency, and reducing financial fraud and earnings management. High-quality external audits can identify and correct management fraud and earnings management (Qi et al., 2004). They can also alert shareholders by providing non-standard audit opinions (Jiang et al., 2007), which indirectly limits managers’ opportunistic behaviour. Under weak audit supervision, managers can package costs or conduct earnings management to hide the negative impact of such costs on company performance. However, under strong audit supervision, it is difficult to hide the negative impact. Considering the difficulties of conducting financial fraud or earnings management, managers will decrease their opportunistic behaviour under strong audit supervision and cost stickiness will decrease.¹⁰ In addition, through standardising financial procedures, high-level external audits can also improve the quality of financial report information to help shareholders compare the information with other companies in the same industry accurately and in a timely manner and identify the relation of abnormal revenues and abnormal costs. Shareholders can thus make better decisions about executive compensation and replacement and strengthen control on management.

Previous studies show that Big Four International firms, with high-level specialisation and independence, can represent high-quality external auditing. Dopuch and Simunic (1980) find that big accounting firms tend to undergo more specialised training, stricter certification, and peer review. O’Keefe and Westort (1992) find that big accounting firms invest more in knowledge. When the size of auditor grows, the quasi-rent for each client will decrease. Then the probability of an audit firm behaving opportunistically will decrease, so the audit firm is more independent (DeAngelo, 1981). Dye (1993), Moore and Scott (1989) and Watts and Zimmerman (1981, 1986) demonstrate that big accounting firms are more independent from the perspective of the brand reputation mechanism and deep pocket theory. Greater ability helps audit firms identify and correct financial procedure weakness and restrict managers’ earnings management. If the accounting firms are more independent, they are more likely to provide objective and fair audit reports and can increase deterrence to managers. Compared with domestic audit firms, Big Four International firms have obvious advantages due to their history, size, specialisation, market share, and so forth, which ensure their higher audit quality (Qi et al., 2004). In addition, Wang, Wong, and Xia (2008) find that, in China, government intervention can reduce audit quality, especially the quality of small local audit firms. We can infer from the previous analysis that, compared with domestic firms, the Big Four International provide better audit services and companies audited by them tend to have lower degrees of cost stickiness. Similarly, according to the quasi-rent and deep pocket theories, the Big Ten Domestic audit firms should have higher audit quality than other domestic firms and their clients’ cost stickiness should therefore be lower than that of the clients of other domestic firms. The above discussion leads to our first hypothesis.

H1a: Compared with listed companies audited by domestic audit firms, companies audited by the Big Four International have a lower level of cost stickiness.

H1b: Compared with listed companies audited by non-Big Ten Domestic audit firms, companies audited by the Big Ten Domestic have a lower level of cost stickiness.

Our prior analysis suggests high-quality external audit can mitigate cost stickiness by reducing agency costs between stockholders and managers. However, external audit is only part of corporate governance and other governance mechanisms can also reduce agency costs and control managers’ opportunistic behaviour. If other mechanisms can successfully reduce agency costs, external audit may not be the dominant factor in the supervision, but if other mechanisms fail, external audit can be more important.

Specifically, we analyse the interaction between different forms of external governance regarding two aspects: ownership structure and the external environment. The largest shareholder’s holding ratio and marketisation level of the province where a firm’s headquarter is located represent two important dimensions of external governance mechanisms and are relatively exogenous to the Big Four International indicator.¹¹ Since these two factors can vary cross-sectionally, the degree of corporate agency problem can vary accordingly. If the largest shareholder possesses a higher proportion of shares, it will monitor the managers more actively (Shleifer & Vishny, 1986). The proportion of shares held by the largest shareholder is negatively associated with agency costs. Chen and Liang (2010) find that non-pecuniary compensation is attenuated when the largest shareholder holds a higher proportion of shares. Dou, Liu, and An (2011) find that large shareholders’ complete control of a firm can prevent inefficient investments. In addition, when the market is highly developed, listed companies will face more competition and less government intervention and investors can identify managers’ efforts accurately and in a timely fashion. As a result, a high level of marketisation leads to low agency costs. Li (2007) finds that firms’ SG&A rate and efficiency loss both decrease when market competition increases. Jiang, Huang, and Zhang (2009) find that administrative expense rates decrease and asset turnover increases in highly competitive product markets.

When the largest shareholder holds a lower proportion of shares and the market level is lower, agency costs are relatively higher and an external audit can play a more important role in reducing cost stickiness. The above discussion leads to our second hypothesis.

H2a: When the largest shareholder holds a lower proportion of shares, the negative association between being audited by the Big Four International or Big Ten Domestic and cost stickiness is more pronounced.

H2b: When the market level is lower, the negative association between being audited by the Big Four International or Big Ten Domestic and cost stickiness is more pronounced.

3. Research design

3.1. Sample selection

Our initial sample starts with all A-shares listed firms in the Shanghai and Shenzhen Stock Exchanges from 2002 to 2010.¹² We then delete 1670 observations that were listed for less than two years and 133 observations in the financial industry. We also delete observations whose change in operating revenue or expense from the previous year is lower than 0.5% or higher than 99.5%, which includes 723 observations (Chen, Lu, & Sougiannis, 2012). Finally, we delete 534 observations with missing controlling variables. Our final sample includes 10,494 observations. Table 1 shows the sample selection procedures.

Table 1. Sample selection process.

Year	Initial observations	Listed less than 2 years	Financial companies	Firms with extreme changes in income or expense (0.5%)	Missing data	Remaining observations
2002	1,211	150	7	64	56	934
2003	1,278	138	7	71	43	1,019
2004	1,377	167	8	88	34	1,080
2005	1,366	115	10	88	55	1,098
2006	1,421	81	10	98	65	1,167
2007	1,537	192	14	107	52	1,172
2008	1,611	203	17	47	72	1,272
2009	1,703	176	30	80	82	1,335
2010	2,050	448	30	80	75	1,417
Total	13,554	1,670	133	723	534	10,494

3.2. Research design and variable definitions

Following Anderson et al. (2003), Chen, Lu, and Sougiannis (2012) and Sun and Liu (2004), we establish the following model to test our hypothesis:(1)

where ∆Cost and ∆Income represent changes in expense and in operating income, respectively. We first calculate the ratio of expenses in year t to expense in year t – 1 and the ratio of the operating income in year t to that in year t – 1 and then take the natural logarithm of these two ratios. Specifically, expense is the sum of operating cost and general cost; Income_decrease is a dummy variable for income decline, which equals one if the operating income in year t was lower than that in year t – 1 and zero otherwise; Big4 is an indicator variable for Big Four International audit firms that equals one if the listed company was audited by Ernst & Young, KPMG, Deloitte, or PwC in year t and zero otherwise; and Domestic_big10 is a dummy variable of Big Ten Domestic audit firms that equals 1 if the firm was audited by a Big Ten Domestic audit firm and zero otherwise.¹³

Following Anderson et al. (2003), Chen, Lu, and Sougiannis (2012) and Sun and Liu (2004), our control variables include an indicator variable for Income decline for two consecutive years (Two_year_decrease), economic growth (GDP_growth), human capital intensity (Employee_intensity), fixed capital intensity (Fixed_asset_intensity), financial leverage (Leverage), the market index (Marketisation_index), proportion of independent directors (Independent_director), CEO duality (Duality), and proportion of managerial shareholdings (Managerial_share).

Table 2 shows the definitions and explanations of the main variables. To reduce the effect of extreme values, all continuous variables are winsorized at the 0.5% level. We control for year fixed effects and clustered observations by firms (Petersen, 2009) to mitigate the effect of serial correlation. We also report robust t-values to adjust for heteroscedasticity (White, 1980). These methods improve the robustness of our empirical results.

Table 2. Variable definitions.

Type	Variables name	Variables symbol	Variables definition
Dependent variable	Changes in the costs	∆Cost	Natural logarithm of the ratio of the expense in year t to the expense in year t–1
Independent variables	Changes in operating income	∆Income	Natural logarithm of the ratio of the operating income in year t to the operating income in year t–1
	Income declining	Income_decrease	Dummy variable that equals 1 if operating income in year t is lower than that in year t–1, and equals 0 otherwise
	The Big Four International	Big4	Dummy variable that equals 1 if the firm is audited by Big Four International, 0 otherwise
	The Big Ten Domestic	Domestic_big10	Dummy variable that equals 1 if the firm is audited by Big Ten Domestic, 0 otherwise
Control variables	Income declining for two consecutive years	Two_year_decrease	Dummy variable that equals 1 if the operating income declines for two consecutive years, and equals 0 otherwise
	Economic growth	GDP_growth	GDP growth for year t
	Human capital intensity	Employee_intensity	Number of employees/operating revenue (in millions)
	Fixed capital intensity	Fixed_asset_intensity	Total assets/operating revenue
	Financial leverage	Leverage	Debts/assets
	Marketisation index	Marketisation_index	Marketisation index (Fan et al., 2010)
	Proportion of independent directors	Independent_director	Proportion of independent directors
	CEO duality	Duality	Dummy variable that equals 1 if the CEO and Chairman are the same one, otherwise 0
	Proportion of management shareholdings	Managerial_share	Proportion of shares held by managers

If cost stickiness exists, the regression coefficient a₂ in model (1) should be significantly negative. If H1a holds, we would observe that a₃ in model (1) is significantly positive. If H1b holds, a₅ should be significantly positive. If H2 holds, we expect that a₃ and a₅ in model (1) exhibit significant differences between groups with different ownership structures or local marketisation levels.

3.3. Descriptive statistics and correlation coefficients

Table 3 reports the descriptive statistics of the main variables. Panel A shows the descriptive statistics for the full sample. The means of ∆Cost and ∆Income are 0.133 and 0.145, respectively, and the ratio of observations with an income decrease (mean of Income_decrease) is 25.6%. These statistics are close to those reported by Gong, Liu, and Shen (2010). The proportion of firm-years audited by Big Four International (mean of Big4) is 5.65%, which is close to the result of Guo (2011), Wang and Wang (2006), Yu (2007), and others. A total of 33.5% of observations are audited by Big Ten Domestic (mean of Domestic_big10). All the statistics of the control variables are within reasonable ranges. Panel B reports the descriptive statistics for subsamples partitioned by the types of audit firms. The results show that the overall condition is better in listed companies audited by Big Four International firms. Financial leverage (Leverage), of the likelihood of having a decreased income, the likelihood of two consecutive years of decreased income, and the likelihood of CEO duality are lower for such firms and, on average, the local marketisation level is higher. These results are consistent with previous studies (e.g. Qi et al., 2004; Wang & Wang, 2006).¹⁴ Table 4 shows the correlation coefficients of the main variables. The correlation matrix shows no strong correlation between the independent variables and the control variables.

Table 3. Descriptive statistics.

Panel A: Full sample
Variables symbol	Mean	Lower quartile	Median	Upper quartile	Standard deviation
∆Cost	0.133	–1.499	0.132	1.752	0.375
∆Income	0.145	–1.525	0.141	2.151	0.391
Income_decrease	0.256	0.000	0.000	1.000	0.436
Big4	0.057	0.000	0.000	1.000	0.231
Domestic_big10	0.335	0.000	0.000	1.000	0.472
Two_year_decrease	0.096	0.000	0.000	1.000	0.295
GDP_growth	10.726	9.100	10.100	14.200	1.618
Employee_intensity	2.717	0.023	1.700	28.418	3.503
Fixed_asset_intensity	2.717	0.304	1.794	36.022	3.630
Leverage	0.506	0.062	0.515	0.957	0.186
Marketisation_index	8.060	0.630	7.920	11.710	2.202
Independent_director	0.346	0.000	0.333	0.600	0.073
Duality	0.096	0.000	0.000	1.000	0.295
Managerial_share	0.005	0.000	0.000	0.273	0.028

Panel B: Sub-sample
	Big Four International		Big Ten Domestic		Non-Big Ten Domestic		Big Four International–Non-Big Ten Domestic
Variables symbol	Mean	Median	Mean	Median	Mean	Median	Mean	Median
∆Cost	0.145	0.127	0.138	0.133	0.128	0.132	0.98	1.44
∆Income	0.165	0.154	0.145	0.144	0.145	0.138	0.76	0.08
Income_decrease	0.189	0.000	0.260	0.000	0.259	0.000	–3.84***	–3.83***
Two_year_decrease	0.057	0.000	0.093	0.000	0.101	0.000	–3.30***	–3.29***
GDP_growth	10.647	10.100	10.630	10.100	10.786	10.100	–1.22	–1.47
Employee_intensity	1.657	0.994	2.239	1.508	3.079	1.886	–7.60***	–10.87***
Fixed_asset_intensity	2.261	1.605	2.503	1.670	2.877	1.893	–3.15***	–4.37***
Leverage	0.471	0.478	0.517	0.531	0.503	0.512	–4.74***	–4.87***
Marketisation_index	8.826	8.960	8.580	8.660	7.703	7.420	8.74***	8.88***
Independent_director	0.341	0.333	0.355	0.333	0.342	0.333	–1.88*	–1.24
Duality	0.057	0.000	0.103	0.000	0.096	0.000	–3.29***	–3.29***
Managerial_share	0.001	0.000	0.006	0.000	0.004	0.000	–3.39***	–0.75

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively

Table 4. Correlation coefficients.

	A	B	C	D	E	F	G	H	I	J	K	L	M
(A) ∆Cost	1
(B) ∆Income	0.294^***	1
(C) Income_decrease	–0.200^***	–0.536^***	1
(D) Big4	0.007	0.010	–0.037^***	1
(E) Domestic_big10	0.009	–0.001	0.008	–0.174^***	1
(F) Two_year_decrease	–0.142^***	–0.344^***	0.556^***	–0.032^***	–0.008	1
(G) GDP_growth	–0.073^***	0.045^***	–0.088^***	–0.012	–0.042^***	–0.040^***	1
(H) Employee_intensity	–0.042^***	–0.246^***	0.178^***	–0.074^***	–0.098^***	0.220^***	–0.057^***	1
(I) Fixed_asset_intensity	–0.058^***	–0.353^***	0.244^***	–0.031^**	–0.042^***	0.267^***	–0.042^***	0.410^***	1
(J) Leverage	0.048^***	0.047^***	0.005	–0.046^***	0.043^***	–0.005	0.043^***	–0.026^**	–0.006	1
(K) Marketisation_index	–0.047^***	–0.019	0.029^**	0.085^***	0.168^***	0.003	0.124^***	–0.261^***	–0.066^***	–0.012	1
(L) Independent_director	–0.031^**	0.016	–0.005	–0.018	0.087^***	–0.025^*	0.117^***	–0.129^***	–0.027^**	0.021^*	0.223^***	1
(M) Duality	0.005	–0.000	–0.009	–0.032^**	0.017	–0.005	0.001	0.011	0.007	–0.031^**	0.038^***	0.020^*	1
(N) Managerial_share	0.026^**	0.012	–0.031^**	–0.033^***	0.036^***	–0.033^***	–0.015	–0.024^*	–0.034^***	–0.081^***	0.144^***	0.071^***	0.161^***

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4. Empirical analysis

4.1. Testing H1: types of external auditors and cost stickiness

Table 5 shows the regression results of model (1). Column (1) contains only ∆Income and ∆Income*Income_decrease, while columns (2) to (6) further include control variables and our variables of interest. The regression result shows that the F-values of the regression model are mostly above 20 and the adjusted R² values are about 0.10, close to the results of Sun and Liu (2004). The coefficient of ∆Income*Income_decrease is negative and significant at the 1% level in column (1), suggesting that cost stickiness exists in China’s listed companies, which is consistent with the results of Gong et al. (2010) and Sun and Liu (2004). The sign and significance of the coefficient of ∆Income*Income_decrease remain unchanged in column (2) when control variables are included. Columns (3) to (5) present the regression results for the subsamples by type of external auditors. The coefficient of ∆Income*Income_decrease is positive and not significant in column (3) for companies audited by Big Four International firms, negative but not significant in column (4) for companies audited by Big Ten Domestic firms, and negative and significant at the 5% level in column (5) for companies audited by other audit firms. Columns (6) and (7) include the interaction term of three variables in the model to test for H1a and H1b. The coefficient of ∆Income*Income_decrease*Big4 is positive and significant at the 10% level in column (6), meaning that Big Four International audited firms have significantly less cost stickiness compared with other firms. Column (7) further includes the interaction of Domebig10 and ∆Income*Income_decrease. The coefficient of ∆Income*Income_decrease*Big4 remains positive and significant at the 10% level, but the coefficient of ∆Income*Income_decrease*Big10 is positive but not significant, which means that there is no significant difference in cost stickiness between firms audited by Big Ten Domestic and by the non-Big Ten Domestic. The results in columns (3) to (7) support H1a but not H1b. We think that the non-significant result for H1b is not surprising. For different reasons, such as a less diversified business, brain drain, and government intervention, most previous studies have failed to find significant audit quality differences between domestic audit firms of different sizes (Fang, Hong, & Li, 2004; Wu, 2006; Yuan & Li, 2003). In addition (untabled), the joint test shows that the sum of the coefficients of ∆Income*Income_decrease and ∆Income*Income_decrease*Big4 is not significantly different from zero.¹⁵

Table 5. Types of external auditors and the cost stickiness.

	(1) Full sample	(2) Full sample	(3) The Big Four	(4) The Big Ten Domestic	(5) Others domestic	(6) Full sample	(7) Full sample
	0.309***	0.294***	0.351***	0.332***	0.274***	0.294***	0.294***
	(10.68)	(10.10)	(3.51)	(6.69)	(7.40)	(10.10)	(10.08)
∆Income*Income_decrease	–0.151***	–0.145**	0.076	–0.116	–0.153**	–0.148***	–0.169***
	(–3.01)	(–2.57)	(0.48)	(–1.45)	(–2.14)	(–2.61)	(–2.71)
∆Income*Income_decrease*Big4						0.214*	0.238*
						(1.70)	(1.84)
						0.024**	0.031***
						(2.20)	(2.71)
∆Income*Income_decrease *Domestic_big10							0.086 (1.27)
Domestic_big10							0.020***
							(2.68)
Two_year_decrease		–0.085***	–0.041	–0.078***	–0.086***	–0.085***	–0.083***
		(–4.86)	(–0.57)	(–3.08)	(–3.77)	(–4.83)	(–4.82)
GDP_growth		–0.022***	0.014	–0.022***	–0.025***	–0.022***	–0.022***
		(–5.43)	(1.26)	(–2.65)	(–4.95)	(–5.35)	(–5.40)
Employee_intensity		0.001	–0.008	0.001	0.001	0.001	0.001
		(0.39)	(–1.50)	(0.34)	(0.28)	(0.42)	(0.38)
Fixed_asset_intensity		0.002	0.017***	0.003	0.002	0.002	0.002
		(1.04)	(2.84)	(1.03)	(0.52)	(1.03)	(1.08)
Leverage		0.075***	0.114	0.046	0.082***	0.076***	0.074***
		(3.81)	(1.56)	(1.55)	(2.86)	(3.85)	(3.73)
Marketisation_index		–0.005***	–0.001	–0.007***	–0.005**	–0.005***	–0.005***
		(–3.09)	(–0.19)	(–2.89)	(–2.13)	(–3.20)	(–3.39)
Independent_director		–0.099*	–0.159	–0.108	–0.095	–0.099*	–0.100*
		(–1.70)	(–1.12)	(–1.01)	(–1.21)	(–1.70)	(–1.72)
Duality		0.001	0.044*	–0.004	0.003	0.001	0.001
		(0.12)	(1.83)	(–0.26)	(0.19)	(0.14)	(0.13)
Managerial_share		0.436***	1.131***	0.497***	0.417***	0.442***	0.445***
		(5.07)	(3.61)	(4.18)	(3.59)	(5.12)	(5.22)
Constant	0.139***	0.137***	0.057	0.165***	0.140***	0.137***	0.134***
	(8.74)	(4.93)	(0.84)	(3.31)	(3.71)	(4.90)	(4.82)
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	10,494	10,494	593	3,520	6,381	10,494	10,494
Adjusted R^2	0.101	0.107	0.130	0.157	0.088	0.107	0.108
F value (model)	34.15	27.62	7.737	14.67	14.71	26.14	25.08
F value (∆Income* Income_decrease)			(3) vs (5)=1.48

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4.2. Testing H2: types of external auditors, agency costs, and cost stickiness

For H2, we separate the entire sample into two subsamples according to the proportion of shares held by the largest shareholders. If the proportion is above the median, the observation is classified as a high stake of major shareholders; otherwise the observation is classified as a low stake of major shareholders. We want to use subsample regressions to explore the difference in the restrictive roles of Big Four International and Big Ten Domestic audit firms on cost stickiness in different environments.¹⁶

Table 6 shows the regression results. The coefficient of ∆Income*Income_decrease*Big4 is not significant in column (1) for firms with high stakes of major shareholders, suggesting that Big Four International audit firms do not significantly reduce cost stickiness when the largest shareholder’s holding ratio is high. However, the coefficient of ∆Income*Income_decrease*Big4 is positive and significant at the 1% level in column (2) for firms with low stakes of major shareholders, which means that Big Four International audit firms can reduce cost stickiness significantly when the largest shareholder’s holding ratio is low. In addition, the F-value of the Chow test of the coefficients of ∆Income*Income_decrease*Big4 between the two groups is 6.46 and significant at the 1% level. All of these results show that Big Four International audit firms play an active role when major shareholders hold fewer stakes and when agency costs are high. Similarly, the coefficient of ∆Income*Income_decrease*Big4 is not significant in column (3) for firms in areas with higher levels of marketisation, indicating that Big Four International audit firms do not suppress cost stickiness significantly when firms are located in areas with well-developed markets. On the contrary, the coefficient of ∆Income*Income_decrease*Big4 is positive and significant at the 5% level in column (4) for firms in areas with lower levels of marketisation, which means that Big Four International firms can significantly restrict cost stickiness when firms are located in areas with less developed markets. In addition, the F-value of the Chow test of the coefficients of ∆Income*Income_decrease*Big4 is 4.01 and significant at the 5% level. All these results suggest that Big Four International audit firms play a positive role in reducing cost stickiness when the market is not well developed and when agency cost is high. We do not find evidence that the restrictive roles on cost stickiness are significantly different between Big Ten Domestic and non–Big Ten Domestic audit firms. The above results together show in the role of Big Four International firms in mitigating agency costs and reducing cost stickiness changes under different circumstances.¹⁷

Table 6. Types of external auditors, agency costs and the expense stickiness.

	(1) High stake of major shareholders	(2) Low stake of major shareholders	(3) High level of marketisation	(4) Low level of marketisation
∆Income	0.332***	0.241***	0.271***	0.317***
	(7.67)	(5.88)	(7.37)	(6.98)
∆Income*Income_decrease	–0.161	–0.137*	–0.109	–0.217**
	(–1.27)	(–1.83)	(–1.38)	(–2.38)
∆Income*Income_decrease*Big4	0.011	0.690***	0.079	0.717**
	(0.09)	(2.98)	(0.77)	(2.36)
Big4	0.011	0.064***	0.030**	0.028
	(0.77)	(3.43)	(2.49)	(1.14)
∆Income*Income_decrease *Domestic_big10	0.172 (1.47)	0.059 (0.71)	0.035 (0.40)	0.150 (1.20)
Domestic_big10	0.018*	0.021*	0.010	0.032***
	(1.77)	(1.81)	(1.04)	(2.71)
Two_year_decrease	–0.048	–0.103***	–0.084***	–0.077***
	(–1.54)	(–4.64)	(–3.97)	(–2.84)
GDP_growth	–0.017***	–0.027***	–0.014**	–0.027***
	(–3.21)	(–3.89)	(–2.07)	(–4.53)
Employee_intensity	–0.001	0.002	0.000	0.001
	(–0.50)	(0.65)	(0.05)	(0.34)
Fixed_asset_intensity	0.004	0.002	0.003	0.002
	(1.09)	(0.72)	(1.02)	(0.54)
Leverage	0.040	0.101***	0.045*	0.103***
	(1.42)	(3.51)	(1.82)	(3.21)
Marketisation_index	–0.009***	–0.002	–0.010***	–0.005
	(–4.09)	(–0.93)	(–2.75)	(–1.04)
Independent_director	–0.102	–0.110	–0.068	–0.125
	(–1.15)	(–1.28)	(–0.95)	(–1.36)
Duality	0.011	–0.004	0.008	–0.006
	(0.68)	(–0.30)	(0.54)	(–0.33)
Managerial_share	0.490***	0.510***	0.516***	0.151
	(2.89)	(5.16)	(6.70)	(0.68)
Constant	0.174***	0.114**	0.180***	0.123**
	(4.38)	(2.56)	(3.59)	(2.52)
Year	Yes	Yes	Yes	Yes
Observations	5,249	5,245	5,263	5,231
Adjusted R^2	0.150	0.078	0.126	0.095
F value (model)	15.24	12.79	18.28	11.89
F value (∆Income* Income_decrease)	(1) vs (2) = 6.46**		(3) vs (4) = 4.01**

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4.3. Robustness tests

4.3.1. Test using propensity score matching (PSM)

So far, we have not ruled out the impact of self-selection on our results. For instance, companies with strict internal control may tend to choose audit services provided by Big Four International audit firms, whose services are of higher quality (e.g. Thornton & Moore, 1993). At the same time, these companies may exhibit lower cost stickiness due to strict internal control (Chen, Lu, & Sougiannis, 2012). If this is the case, our result that companies audited by Big Four International have lower cost stickiness is just a natural result of the companies’ choices. To address such concerns, we use the PSM method to further test our results. The PSM method can eliminate the effect of self-selection by matching the treated group (Big Four International-audited firms in this article) with the control group (non-Big Four International-audited firms) and is widely employed in related accounting and auditing research (e.g. Carcello, Vanstraelen, & Willenborg, 2009; Lawrence, Minutti-Meza, & Zhang, 2011).¹⁸

Specifically, the PSM test contains two stages. The first stage is to grade based on choice tendency and the second stage is to match samples based on the grades generated. The grading is based on a logit model whose dependent variable is an indicator variable of Big Four International audited firms (Big4). We choose independent variables and control variables according to Wang et al. (2008). The grading model is (2)

where SOE is a dummy variable that equals one for state-owned companies and zero otherwise, Growth is the growth rate of operating revenue, Size is the natural logarithm of total assets, CROA is the profit from operations divided by total assets, Lev is financial leverage, CurrentR is the ratio of current assets to current liability, InventoryR is the ratio of inventory to total assets, and RecR is the ratio of accounts receivable to total assets. We also control for year fixed effects. The first-stage regression test shows that Size, Leverage, and other factors have significant influence on whether a company hires a Big Four International audit firm. Matching by grades, we generate a total sample of 1,084 observations, including both the treated and control groups, and we use this sample to conduct the second-stage regression.

Table 7 shows the regression results. Column (1) is based on full sample regression, while columns (2) to (5) report the results of subsamples grouped by largest shareholder’s holding ratio (columns (2) and (3)) or local marketisation levels (columns (3) and (4)). The results shows that the coefficient of ∆Income*Income_decrease*Big4 is positive and significant at the 10% level in column (1), suggesting the restrictive role of Big Four International audit firms in cost stickiness. The coefficient of ∆Income*Income_decrease*Big4 is not significant in column (2) for firms with a high stake of major shareholders, but is positive and significant at the 1% level in column (3) for firms with a low stake of major shareholders. Meanwhile, the F-value of the Chow test of the coefficients of ∆Income*Income_decrease*Big4 in the two columns is 3.86 and significant at the 5% level. The coefficient of ∆Income*Income_decrease*Big4 is not significant in column (4) for firms in areas with a well-developed market and is positive and significant at the 10% level in column (5) for firms in areas with lower marketisation levels. In addition, the F-value of the Chow test of the two coefficients is 4.17 and significant at the 5% level. These results are qualitatively the same as the results in previous sections. The results in Table 7 show that using PSM does not change the outcome we found earlier.

Table 7. Test using the PSM method.

	(1) Full sample	(2) High stake of major shareholders	(3) Low stake of major shareholders	(4) High level of marketisation	(5) Low level of marketisation
∆Income	0.469***	0.479***	0.485***	0.453***	0.487***
	(8.94)	(5.27)	(11.54)	(7.62)	(4.35)
∆Income*Income_decrease	–0.431***	–0.420**	–0.629***	–0.163	–0.531**
	(–3.21)	(–2.44)	(–3.18)	(–0.88)	(–2.45)
∆Income*Income_decrease*Big4	0.306*	0.144	0.753***	–0.035	0.848*
	(1.81)	(0.92)	(2.87)	(–0.19)	(1.96)
Big4	–0.014	–0.038*	0.047	–0.010	–0.018
	(–0.81)	(–1.84)	(1.63)	(–0.51)	(–0.56)
Two_year_decrease	0.028	0.130*	–0.181**	0.006	0.087
	(0.50)	(1.75)	(–2.57)	(0.12)	(0.64)
GDP_growth	0.001	–0.006	0.020	0.016	–0.015
	(0.10)	(–0.56)	(1.42)	(1.29)	(–0.99)
Employee_intensity	–0.005	–0.004	–0.005	–0.001	–0.005
	(–0.99)	(–0.71)	(–0.90)	(–0.13)	(–0.77)
Fixed_asset_intensity	0.005	0.005	0.009	0.012**	–0.001
	(1.04)	(0.92)	(0.89)	(2.05)	(–0.12)
Leverage	0.010	–0.003	–0.006	0.060	–0.017
	(0.16)	(–0.04)	(–0.06)	(0.89)	(–0.13)
Marketisation_index	–0.006	–0.010*	0.004	–0.016	0.000
	(–1.25)	(–1.66)	(0.51)	(–1.63)	(0.00)
Independent_director	–0.139	–0.015	–0.473**	–0.160	–0.090
	(–1.02)	(–0.09)	(–2.08)	(–0.90)	(–0.41)
Duality	0.065*	0.051	0.103**	0.060	0.068
	(1.77)	(1.04)	(2.03)	(1.19)	(1.60)
Managerial_share	–0.539	–3.056***	–0.391	–0.414	–0.817***
	(–1.27)	(–3.90)	(–0.95)	(–0.58)	(–2.65)
Constant	0.140**	0.158**	0.081	0.171	0.125
	(2.22)	(2.05)	(0.73)	(1.56)	(0.98)
Year	Yes	Yes	Yes	Yes	Yes
Observations	1,084	725	359	641	443
Adjusted R^2	0.179	0.143	0.298	0.254	0.101
F value (model)	6.908	4.905	13.89	8.414	45.14
F value (∆Income* Income_decrease)		(2) vs (3) = 3.86**		(4) vs (5) = 4.17**

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4.3.2. Test on industrial listed companies

Weiss (2010) uses only industrial listed companies to examine cost stickiness for the following two reasons: first, the cost and profit model of companies from a single industry allows for greater comparability. Second, industrial listed companies face more severe competition and using such a sample can better avoid the influence of monopolistic prices. Therefore, we choose a subsample of industrial firms according to the industry classification of the China Securities Market and Accounting Research (CSMAR) and conduct robust tests.

Table 8 shows the regression results. Column (1) presents the full sample regression, while columns (2) to (5) report the results of subsamples grouped by the largest shareholder’s holding ratio (columns (2) and (3)) or local marketisation levels (columns (3) and (4)). The results shows that the coefficients of ∆Income*Income_decrease*Big4 are positive and significant at the 5% level in columns (1), (3), and (5), but are not significant in column (2) for firms with a high stake of major shareholders or in column (4) for firms in areas with lower marketisation levels. These results support the positive role of the Big Four International audit firms in reducing cost stickiness and also the inference that the restrictive role varies with the largest shareholder’s holding ratio and local marketisation level. The results in Table 8 show that using a subsample of industrial firms does not change the outcome we found earlier.

Table 8. Test on industrial listed companies.

	(1) Full sample	(2) High stake of major shareholders	(3) Low stake of major shareholders	(4) High level of marketisation	(5) Low level of marketisation
∆Income	0.328***	0.334***	0.305***	0.324***	0.329***
	(6.82)	(5.54)	(3.56)	(5.13)	(4.83)
∆Income*Income_decrease	–0.121	–0.007	–0.122	–0.140	–0.107
	(–1.22)	(–0.06)	(–0.91)	(–1.17)	(–0.73)
∆Income*Income_decrease*Big4	0.561**	0.301	0.700**	0.387	0.750**
	(2.47)	(1.04)	(2.33)	(1.62)	(2.48)
Big4	0.018	–0.006	0.042**	0.026*	0.001
	(1.26)	(–0.35)	(2.06)	(1.73)	(0.05)
∆Income*Income_decrease *Domestic_big10	0.052 (0.47)	–0.011 (–0.09)	0.063 (0.45)	0.054 (0.42)	0.041 (0.19)
Domestic_big10	0.017*	0.006	0.024*	0.015	0.021
	(1.93)	(0.54)	(1.75)	(1.31)	(1.51)
Two_year_decrease	–0.062***	–0.020	–0.083***	–0.070**	–0.054
	(–2.81)	(–0.57)	(–2.85)	(–2.40)	(–1.60)
GDP_growth	–0.023***	–0.021***	–0.024**	–0.016*	–0.028***
	(–4.53)	(–2.94)	(–2.57)	(–1.77)	(–4.01)
Employee_intensity	–0.001	–0.004	0.001	–0.001	–0.002
	(–0.34)	(–1.05)	(0.29)	(–0.13)	(–0.60)
Fixed_asset_intensity	0.007	0.008	0.007	0.003	0.010
	(1.41)	(1.43)	(1.07)	(0.69)	(1.28)
Leverage	0.057**	0.019	0.089**	0.024	0.081**
	(2.32)	(0.57)	(2.40)	(0.72)	(2.19)
Marketisation_index	–0.004**	–0.008***	–0.001	–0.013***	0.000
	(–1.97)	(–2.80)	(–0.18)	(–2.83)	(0.05)
Independent_director	–0.063	–0.089	–0.051	–0.100	–0.028
	(–0.78)	(–0.72)	(–0.47)	(–0.97)	(–0.23)
Duality	–0.010	–0.001	–0.015	–0.003	–0.017
	(–0.66)	(–0.04)	(–0.75)	(–0.14)	(–0.79)
Managerial_share	0.329***	0.305	0.382***	0.441***	–0.060
	(3.30)	(1.43)	(3.29)	(4.65)	(–0.38)
Constant	0.132***	0.198***	0.072	0.239***	0.082
	(3.46)	(3.93)	(1.18)	(3.50)	(1.40)
Year	Yes	Yes	Yes	Yes	Yes
Observations	6,236	3,295	2,941	2,830	3,406
Adjusted R^2	0.106	0.128	0.086	0.138	0.089
F value (model)	14.33	8.998	7.523	11.65	7.891
F value (∆Income* Income_decrease)		(2) vs (3) = 0.88		(4) vs (5) = 1.00

Note:

For definitions of the variables, please refer to Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4.3.3. Control for discretionary accruals

Chen, Song, and Shi (2012) find that companies may use a big bath to manage earnings to avoid future losses and such behaviour can lead to the overestimation of stickiness. Thus, we follow Dechow, Sloan, and Sweeney (1995) and use the adjusted Jones model to calculate discretionary accruals (DA). We then include DA and an interaction term ∆Income*Income_decrease*DA in our model and re-estimate the model.¹⁹The results are presented in Table 9. Column (1) presents the full sample regression, while columns (2) to (5) report the results of subsamples grouped by the largest shareholder’s holding ratio (columns (2) and (3)) or local marketisation levels (columns (3) and (4)). The results show that, after Da and the interaction term ∆Income*Income_decrease*Da are controlled for, the coefficient of ∆Income*Income_decrease*Big4 is positive and significant in column (1), while the coefficient of ∆Income*Income_decrease*Da is positive and significant at the 1% level. The combined results indicate that Big Four International audit firms reduce cost stickiness mainly by influencing managers’ earnings management behaviour, since the introduction of discretionary accruals lowers the explanatory power of Big Four International audit firms for cost stickiness. Further subsample tests suggest that the coefficients of ∆Income*Income_decrease*Big4 are positive and significant at the 1% and 5% levels, respectively, in columns (3) and (5), but are not significant in columns (2) and (4). The results of Table 9 suggest that controlling discretionary accruals will not change our earlier outcomes.²⁰

Table 9. Control for discretionary accruals.

	(1) Full sample	(2) High stake of major shareholders	(3) Low stake of major shareholders	(4) High level of marketisation	(5) Low level of marketisation
∆Income	0.304***	0.358***	0.240***	0.270***	0.348***
	(10.63)	(8.56)	(5.85)	(7.36)	(7.99)
∆Income*Income_decrease	–0.170***	–0.177	–0.130**	–0.104	–0.237***
	(–3.00)	(–1.42)	(–1.97)	(–1.36)	(–3.11)
∆Income*Income_decrease*DA	0.773***	0.488	0.846***	0.350	1.322***
	(3.84)	(1.47)	(3.37)	(1.58)	(3.84)
DA	–0.021	–0.091	0.016	0.019	–0.091
	(–0.38)	(–1.27)	(0.18)	(0.31)	(–0.97)
∆Income*Income_decrease*Big4	0.207	–0.020	0.645***	0.070	0.651**
	(1.63)	(–0.16)	(2.96)	(0.68)	(2.44)
Big4	0.029**	0.009	0.062***	0.029**	0.022
	(2.56)	(0.62)	(3.38)	(2.44)	(0.92)
∆Income*Income_decrease *Domestic_big10	0.063 (0.98)	0.105 (0.85)	0.066 (0.82)	0.030 (0.36)	0.077 (0.63)
Domestic_big10	0.020***	0.016	0.022*	0.010	0.029**
	(2.62)	(1.58)	(1.89)	(1.06)	(2.43)
Two_year_decrease	–0.085***	–0.050	–0.107***	–0.087***	–0.076***
	(–5.01)	(–1.62)	(–4.80)	(–4.06)	(–2.86)
GDP_growth	–0.022***	–0.017***	–0.026***	–0.014**	–0.027***
	(–5.32)	(–3.28)	(–3.81)	(–2.06)	(–4.58)
Employee_intensity	–0.001	–0.002	0.000	–0.000	–0.001
	(–0.31)	(–0.90)	(0.18)	(–0.13)	(–0.49)
Fixed_asset_intensity	0.002	0.004	0.002	0.003	0.000
	(1.12)	(1.11)	(0.80)	(1.17)	(0.09)
Leverage	0.058***	0.022	0.086***	0.042*	0.065**
	(2.93)	(0.77)	(3.07)	(1.70)	(1.99)
Marketisation_index	–0.005***	–0.009***	–0.002	–0.010***	–0.005
	(–3.58)	(–4.16)	(–1.05)	(–2.78)	(–1.06)
Independent_director	–0.098*	–0.089	–0.119	–0.068	–0.120
	(–1.70)	(–1.00)	(–1.43)	(–0.95)	(–1.34)
Duality	0.003	0.011	–0.002	0.008	–0.000
	(0.27)	(0.65)	(–0.12)	(0.55)	(–0.02)
Managerial_share	0.449***	0.530***	0.497***	0.517***	0.137
	(5.27)	(3.26)	(4.95)	(6.77)	(0.60)
Constant	0.143***	0.175***	0.129***	0.183***	0.141***
	(5.29)	(4.57)	(2.99)	(3.63)	(3.07)
Year	Yes	Yes	Yes	Yes	Yes
Observations	10,481	5,244	5,237	5,262	5,219
Adjusted R^2	0.117	0.162	0.086	0.128	0.114
F value (model)	25.48	15.74	13.02	17.16	13.65
F value (∆Income* Income_decrease)		(2) vs (3) = 6.71***		(4) vs (5) = 4.28**

Note:

DA is discretionary accruals, which is calculated following Dechow et al. (1995). The other variables are defined in Table 2.

^*** ,

^** ,

^* indicate significance levels of 1%, 5% and 10% respectively.

4.3.4. Other robustness tests

In addition, we carry out the following unreported tests.²¹ (1) Following Anderson et al. (2003), Chen, Lu, and Sougiannis (2012), and Sun and Liu (2004), we incorporate the interactions of the four economic factors Two_year_decrease, GDP_growth, Employee_intensity, and Fixed_asset_intensity and stickiness (∆Income*Income_decrease) or the interaction terms of all control variables and stickiness (∆Income*Income_decrease) into the model. The results are qualitatively the same for both methods, with coefficients of ∆Income*Income_decrease*Big4 that are positive but not significant in the full sample regressions and a significant difference in the coefficients of ∆Income*Income_decrease*Big4 between different subsamples. (2) Following Chen, Song, and Shi (2012), we measure fixed capital intensity by the ratio of fixed assets to operating income. (3) We substitute Da with its absolute value. (4) We use yearly total income ranking to identify the Big Ten Domestic audit firms. (5) We choose the Big Five Domestic or the Big 20 Domestic instead of the Big Ten Domestic. None of the robustness tests changes the previous results qualitatively.

5. Conclusions and limitations

Cost stickiness is an important problem for the issue of cost behaviour and an important measure of enterprise production and managerial efficiency. However, for a long time, related studies on cost stickiness have been limited and mainly focused on the impact of internal corporate governance mechanisms (Calleja et al., 2006; Chen, Lu, & Sougiannis, 2012). We introduce an important external governance mechanism, namely external audit, into this topic and propose that an external audit, which can mitigate information asymmetry and reduce internal agency costs through independent audit work, can play a restrictive role in a firm’s cost stickiness.

Using China A-shares listed firms from 2002 to 2010, we test this hypothesis empirically. The results show that cost stickiness is less pronounced for firms audited by the Big Four International audit firms than for firms audited by non-Big Four International audit firms, but such a difference in cost stickiness is not found between firms audited by the Big Ten Domestic audit firms and by non-Big Ten Domestic audit firms. In addition, we find that the restrictive role of the Big Four International is more pronounced when a firm’s largest shareholder’s holding ratio is lower or when a firm is located in an area with a lower marketisation level. The results are unchanged when we use PSM or industrial listed company subsamples or when we control for discretionary accruals. Our study enriches the literature on factors influencing cost stickiness and indicates new incentives for the demand for high-quality audit services and implies the higher audit quality of Big Four International audit firms.

Our paper also has the following limitations. First, there can be model misspecification. (1) Our basic model follows Anderson et al. (2003), Chen, Lu, and Sougiannis (2012) and Sun and Liu (2004), but the missing quadratic term in the model can lead to econometric errors.²² (2) The selection of economic factors can cause problems. There are no conclusive standards for this issue and the domestic literature discusses this even less. This paper controls for the four most frequently selected factors, which are the indicator of two consecutive years of a decline in operating income, economic growth, human capital density, and fixed capital intensity. However, this does not mean that the influence from the contract perspective and efficiency perspective is perfectly controlled for in our model. More variables reflecting these two perspectives may be gradually identified in the future. (3) There are other measures of cost stickiness. Weiss (2010) measures stickiness using quarterly operating fluctuations, but seasonal sales will naturally affect the estimation of stickiness (e.g. the seasonal sales fluctuation of electric appliances may not suggest a decline of sales). In addition, quarterly financial statements are not mandatorily audited and may suffer more from earnings management. As a result, we do not use this method for robustness checks. Second, the self-selection problem may still affect our results. Although we use PSM to match firms audited by Big Four International firms with control firms, we have to admit that such methods can only mitigate instead of rule out the impact of the self-selection problem and we should interpret the results with caution. These limitations can be addressed in future research.

Acknowledgements

This research is supported by grants from the National Natural Science Foundation of China (No. 71402198), the Beijing Municipal Commission of Education ‘Joint Construction Project’, the Beijing Municipal Commission of Education ‘Pilot Reform of Accounting Discipline Clustering’, ‘2011 Synergetic Innovation’ Key Project on ‘Development of Public Accounting Profession’ and ‘121 Talent Project for Young Doctoral Graduates Development 2014’ of the Central University of Finance and Economics.

Notes

1. Chen, Lu, and Sougiannis (2012) provide direct evidence for this argument. They find that cost stickiness has a positive correlation with corporate free cash flow and this positive correlation can be weakened when the corporate governance level is higher. Calleja et al. (2006) find that expense stickiness in France and Germany is higher than in the United States or Britain in their international study and they conjecture that the difference in corporate governance is the main reason.

2. China is a large country where the levels of market development and government intervention vary significantly across the 30 areas (provinces, autonomous regions or municipalities) (Jian & Wong, 2010). Fan, Wang, and Zhu (2010) offer an index that measures the development of marketisation of different areas in China.

3. In different years, it can be the Big Five International or the Big Eight International instead of Big Four International. For the sake of brevity, we use the term Big Four International for all conditions without differentiation. All these Big N international audit firms refer to the well-recognised top international audit firms.

4. The findings of these studies conflict for the following reasons, in addition to the difference in research focus (discretionary accruals, audit opinion types, ERC (earnings response coefficient), and conservatism): first, the sample period is short. Most sample periods are one year to three years and the longest are five years. The shorter the sample period, the more the variables of interest can fluctuate, especially discretionary accruals. Second, the difference in the sample size of firms audited by Big Four International and non-Big Four International is not well addressed. Third, the self-selection problem is paid little attention. Our paper, instead, chooses a nine-year sample period (from 2002 to 2010) to conduct the empirical tests and methods such as propensity score matching (PSM) to mitigate the potential problem.

5. The SME started in 2004 and GEM started in 2009. But the government stopped the IPO approval during the non-tradable share reform (2005–2006), so companies listed before 2007 are mainly on the main board.

6. Other reasons include the consideration of cost adjustment and managers’ optimism (Banker et al., 2011). Sun and Liu (2004) attribute these reasons to an opportunistic behaviour perspective, a contract perspective, and an efficiency perspective.

7. However, their research design does not control for common variables, economic factor variables, or year or industry fixed effects.

8. Calleja et al. (2006) mention external supervision theoretically, but do not conduct empirical research.

9. Such as overextending the number of employees (Williamson, 1963) and building luxurious offices.

10. Controlling earnings management is not the only way external governance works. If a company’s performance worsens and managers still spend 1 million RMB to build luxurious offices, they can cover the adverse influence of such expenses on performance through earnings management. A capable and independent auditor can either correct the earnings management during the auditing procedure, which reduces the degree of earnings management, or alert shareholders by signing non-standard audit opinions without correcting for earnings managements (in such cases, the degree of earnings management is unchanged).

11. Although many variables have been used in prior studies to measure agency costs, such as the administrative expense rate and free cash flow, most of these variables might present the consequence of the whole corporate governance including auditing (Gul & Tsui, 1998; Li, 2007).

12. Our sample starts in 2002 because the concept of the Big Four International tends to have become stable with the voluntary surrender by Arthur Andersen of its licence to practice in the US. Also, the Chinese CPA Association started issuing its Comprehensive Evaluation of the Top 100 Accounting Firms in 2002, which provides a good reference for defining the Big Ten Domestic audit firms.

13. The firm ranking is derived from yearly Comprehensive Evaluation of the Top 100 Accounting Firms released by China CPA Association (http://www.cicpa.org.cn/Column/swszhpm).

14. These differences further support the necessity of using PSM as a robustness test.

15. Some later tests show that the sum of the coefficients of ∆Income*Income_decrease and ∆Income*Income_decrease*Big4 is significantly positive, meaning that there is an anti-stickiness phenomenon. This kind of result is mainly found in groups with lower largest shareholder’s shareholding ratio and groups with a lower market index. We conjecture that one possible reason is that the agency cost in this kind of company is higher. When sales decrease, managers trying to achieve performance goals may have to adopt real earnings management if their audit firms (Big Four International) effectively restrict them from conducting earnings management. In such a situation, excessive cost cutting can lead to an anti-stickiness phenomenon. Cohen and Zarowin (2010) and Zang (2012) both point out that the degree of real earnings management is higher when firms are audited by the Big Eight International. We do not find an anti-stickiness phenomenon when PSM methods are used.

16. The medians of the largest shareholder’s holding ratio (Sh) and the market index (Marketisation_index) are 0.357 and 7.92, respectively.

17. We also estimate the model using 2×2 subsamples based on whether the firm is audited by Big Four International firms and whether the largest shareholder’s holding ratio (or market index) is higher than the median. The results show that the cost stickiness in groups when firms are audited by non-Big Four International firms and when the largest shareholder’s holding ratio (or when the market index is lower than the median) is the strongest.

18. The result is more significant if we compare the Big Four International sample with the non-Big Ten Domestic sample (e.g. the control group does not include the Big Ten Domestic sample).

19. The idea is to mitigate the influence of earnings management in the form of a big bath, instead of the influence of the absolute value of earnings management. In accordance with this idea, we control for DA instead of its absolute value. In robustness tests, we control for the absolute value of discretionary accruals (AbsDA) and the interaction term ∆Income*Income_decrease*AbsDA.

20. We think the restrictions on cost stickiness and earnings management are two parallel consequences. Big Four International firms can use different methods, such as restricting earnings management, signing non-standard audit opinions, and providing comparable financial information of higher quality, to help shareholders make executive compensation and turnover decisions to reduce managers’ opportunistic behaviour. Reducing earnings management is only one of these methods and, when the latter two methods are used, the degree of earnings management cannot completely explain the change in cost stickiness.

21. Unreported results are available from the authors upon request.

22. To address this problem, we also conduct subsample tests, but this method is convincing only in the test of the basic hypothesis. Further tests still rely on the basic model and may still be affected by potential model misspecification (the comparison between subsamples with different ownership structure or different external environments is a Chow test of the cubic item).