Does client size contribute to audit quality? : empirical evidence from China

(1)

Does Client Size Contribute to Audit Quality?

–Empirical Evidence from China

Student name: Shilun Wu Student number: 10605509 Date: 13th August, 2014 MSc Accountancy & Control, Control Supervisor: Dr. Georgios Georgakopoulos Second reader: Dr. Rui Vieira Amsterdam Business School Faculty of Economics and Business, University of Amsterdam

(2)

1

Abstract

The purpose of this paper is to examine the relationship between client size and audit quality, since it is still a controversial topic. In order to identify the relationship, an empirical research is conducted using the data from Chinese stock market. Linear regression model is applied to testify the hypotheses afterwards. The absolute value of discretionary accruals is employed as the proxy for audit quality and two types of proxies for client size are employed. The statistical results show that there is a significant correlation between client size and audit quality for big 4 accounting firms. However for non-big 4 accounting firms, the correlation is not significant. This result suggests that even big 4 accounting firms can compromise to their clients issuing lower quality auditing reports. This will contribute to future studies since big 4 accounting firms are always standing with excellent reputation on audit quality. Besides, audit quality is the main concentration for investors as well. The empirical evidence from China is valuable for its own growing stock market.

Key words: China, client size, audit quality, big 4 accounting firm, non-big 4 accounting firm, stock market

(3)

2

Introduction

Companies provide stakeholders in capital market with financial information for various purposes and one of the most vital purposes is to attract the investors. Under such circumstances, audited entities need independent external auditors to issue audit reports for their financial statements, proving the financial statements are reliable and fair. Auditor firms normally deliver such auditing services to their clients by collecting auditing fee charges. Thus such external audit reports would eventually increase the value of public financial statements by providing insurance and decreasing risk for investors, because investors in stock markets generally prefer investees with proper auditing insurance. Particularly, in terms of “increased value”, it is difficult to identify the exact incremental value amount. In fact, audit quality can be implemented as one of the proxies for such “increased value” (A.A. Al-Thuneibat, 2011). In other words, there is a positively linear relation between audit quality and “increased value”. The higher audit quality is, the more increased value will be.

As a popular research topic, audit quality is absolutely a complex conception and it has been researched for years among audit-related studies. In terms of auditing result, audit quality can be simply conceptualized as a dichotomy of audit failure or no audit failure (J.R. Francis, 2011). Although it is feasible and reliable to measure audit quality by the dichotomous view (audit failure & no audit failure), such method is being questioned by the fact audit failures are rarely recognized in real life. According to J.R. Francis (2004), the frequency for audit failure is lower than 1% at an annual basis, which is obviously very low in the huge auditing market context. Furthermore, Z.U. Palmrose (1997) argues that for those audit failures associate with lawsuits, the successfully defending rate is only 50%. Therefore, a far more accepted concept refers to a theoretical continuum distributed from low to high audit quality (J.R. Francis, 2004).

It is certain that auditor independence is the key factor influencing the audit quality. Additionally, in terms of auditor independence, economic incentive plays a vital role along with the auditing process. Companies stand a larger size normally offer larger economic incentives to auditing firms. Consequently, auditing firms would always encounter a trade-off between economic incentive and firm reputation in practice. According to J.K. Reynolds & J.R. Francis (2001), a concept economic dependence was introduced to describe the indirect

(5)

4

influence of economic incentive to audit quality. The game-relationship is that auditing firms with fewer clients generally commits more economic dependence. On the contrary, holding more clients would mitigate the economic dependence relatively. From auditing firms’ side, larger auditing firms generally have more clients than smaller ones. Moreover, larger auditing firms generally benefit more from clients than smaller auditing firms in term of auditing fee. Thus, larger auditing firms have more “choices” on the audit market, which will bring more “bargain power” as a result. Take big 4 audit companies as an example, J.K. Reynolds & J.R. Francis (2001) pointed out that they do not give more favourable report to larger clients comparing to smaller ones.

Regarding to the discussion that whether accounting firms would issue more favourable auditing reports for larger clients, this study will use the empirical data from Chinese stock market to conduct further correlation research. One of the motivations for this research is that this is still a controversial topic in terms of “economic incentive”. Although a number of prior literatures state that larger clients normally create larger economic incentive, audit fee mostly, it seems not always the case in real auditing market. According to the research results of E. Carson et al (2004), they summarized that there is no evidently linear relation between audit fee and client size in Australia. Furthermore, in Hong Kong audit market, F. Liu et al (2009)’s research results show that clients hold larger size with more audit fee did not make their auditors compromise to negatively influencing the audit quality. Thus this research will enrich the knowledge in the relation between client size and audit quality.

From audit quality perspective, this paper contributes to the literature by adding evidence from a developing country with booming audit business and capital market. In China, audit quality improvement is still in progress thus it needs more relevant academic research support.

The following structure will begin with literature review on several relevant topics, researched variables that will contribute to this study, and brief introduction to Chinese stock market. After that, some prior literature’ arguments on the relation between client size and audit quality are listed and hypotheses are established consequently. On the basis of Jones Model, the regression test is applied to identify the relation between client size and audit quality. The results will be followed afterwards. Finally conclusions and further discussion about the study will be stated as a summary with both contribution and limitation analysis.

(6)

5

Prior Literature & Hypotheses Establishment

In China, the merits for a particular company to choose big-four audit companies have reached to consensus. Big-four audit firms hold more integrated, mature and standard auditing procedures, which will consequently decrease the audit failure rate. Although J.R. Francis (2004) points out that audit failure rates are extremely low in real life, which is much lower than 1% every year. Even if an audit failure comes to the case of big-four companies, they can still afford clients’ loss from financial perspective. Thus, the accounting information audited by big-four audit companies is more reliable and can earn more trust from investors who will generally more accepted financial reports audited by them. In contrast, a number of non-big four companies stand as smaller firms size usually hold poorer quality control and lower risk-resisting capabilities comparing to big-fours. As a consequence, lower audit quality will be and it would bring enormous loss to investors, even the whole capital market. Currently many researches focus on audit quality related topics with much empirical evidence. In fact, based on prior literature, the majority of financial fraudulent behaviours are quite simple and easily to detect. Then it will logically come to the question that why such fraudulent behaviours are not always disclosed in audit reports. This question definitely cannot be simply explained as the poor professionalism of auditors. Instead, impact directly come from auditees is one of the key contributor to low audit quality or even audit failures. Chinese Capital Market

Stock market in China has witnessed many audit failures since 1990s, audit quality consequently are becoming the focus of all stakeholders in stock market. GUANG XIA (YIN CHUAN) INDUSTRY LIMITED LIABILITY COMPANY and MAIKETE GUANGDIAN FOODSTUFF CO., LTD are two Chinese firms involved in audit scandals. Thus more entities are entering capital market will lead to increasing demand for audit services delivery. On the other hand, more growing companies, especially for those who start to involve in overseas business, need big-four audit firms’ services to gain trust and legitimacy from both domestic and foreign capital market. From audit firm perspective, the prosperous Chinese market can benefit both big-four audit companies and non-big-fours. Cited from SSE (Shanghai Stock Exchange) official annual report, “since China's market economy is growing

(7)

6

astonishingly, there is an increasing demand for more market-oriented resource allocation, leading to the gradual establishment and development of China's capital market”.

Auditor Firm Size

In fact, in term of contributors to audit quality, “auditor size” has been more widely researched in previous literature than “client size”. H. Arnett & P. Danos (1979) state that on the premises professional audit standards are accurately settled during auditing process, there is no difference between largest 8 audit firms and others in term of audit quality. However, L.E. DeAngelo (1981) argues that auditor firm size can eventually influence audit quality when auditors earn “quasi-rents” from specific customers. Furthermore, L.E. DeAngelo (1981) concludes that audit quality can be affected by accounting firm size due to auditor independence level. The discussion on whether auditor firm size can significantly influence the audit quality is still continuing. In this study, big-four auditor firms stand a role as larger auditor firms and some selected non-big-four auditor firms stand as smaller auditor firms. Client Size

Simultaneously, regard to audit quality, L.E. DeAngelo (1981) does not ignore the fact that client size might be another contributor. On the one hand, he argues that no large accounting firms are willing to issue a misreport for a single client at the cost of losing their reputation. On the other hand, accounting firms providing services to only one client may consequently involve in misreporting since the economic incentive is crucial to their survivals. While F. Liu et al (2009) argues that clients with larger size normally are monitors by more stakeholders in stock market. As a result, auditor firms will issue averagely high quality auditing reports under the supervision. Thus the impacts of client size to audit quality differ from larger auditor firms to smaller ones. Currently client size assumes to be a contributor to audit quality is still a controversial topic with insufficient empirical research support. Audit quality

Audit quality is an abstract concept and cannot be measured and observed directly. In addition, there is no universally accepted standard for measuring audit quality. The most three popular proxies for audit quality is auditing fees, discretionary accruals and types of audit opinions.

(8)

7

Z.U. Palmrose (1986) and L. Zhang (2005) use auditing fee to measure audit quality. The nature of auditing services is paid products and services. Hence, it must be compromised to pricing rules which means value determine price and price fluctuates around value. In other words, the level of auditing fees is correspondingly changing with the level auditing quality. Nevertheless, auditing service is a special product. It is intangible and difficult to measure, even along with certain firm problems. Specifically, clients are like investors and auditors are like management. Both sides cannot be easily reached to consensus on audit quality. Under such circumstances, purchasing audit opinions can be occurred logically. This is exactly the disadvantages of using audit fee to measure audit quality, because high auditing fees can be explained as the result of high audit quality or audit opinion purchase.

C. Becher et al, (1998), J.K. Reynolds & J.R. Francis, (2001) and R.M. Frankel et al, (2002) use discretionary accruals to measure audit quality. Auditors may discover accounting improprieties during auditing process, hence auditors can notice clients make adjustments to all those inappropriate accounting treatments. If auditors can eventually stand for their fair opinion on dubious accountings, the audit reports quality will be higher. If auditors comprise to clients and failing to report significant audit adjustments, the audit reports will be lower. Correspondingly the discretionary accruals will change as the results of audit adjustments. Using audited financial statements as an observable auditing outcome is normally associated with abnormal accruals paradigm (J.R. Francis, 2004). This is based on prior literature that clients can exploit accruals for more aggressive earnings management (J.J. Jones 1991; M. DeFond & C. Park, 1997). After that, many studies start employing Jones Model to measure abnormal accruals. For instance, by using cross-sectional Jones Model, higher abnormal accruals show higher level of earnings management (D. F. Prawitt et al, 2009).

As for auditor audit opinions, A. Graswell et al (2002) argue that audit opinions can be a proxy for audit quality. Furthermore, they consider that modified auditor opinions as the result of inconsistency between auditors and clients. Such inconsistency can reflect high independency of auditors. In other words, more modified audit opinions are consistent with higher audit quality. Additionally, under the circumstances that modified audit opinions account for higher percentage of the whole audit opinions, audit quality will be higher consistently. J.V. Carcello & T.L. Neal (2000) argues that the types of auditor reporting can reflect the level of audit quality as well. Incumbent management of clients may increase audit fee pressure or decrease the non-audit services as a tool to aggressively push auditors issuing

(9)

8

non-going concern audit reports which is “satisfying” for the clients. Thus auditors with higher independency can avoid such pressure as much as possible from clients. As a result, higher audit quality will be. The evidence that clients encounter a going-concern report with modified audit opinions is more willing to change their auditors (C. W. Chow & S. J. Rice. 1982; J.F. Muchler, 1984; M. Geiger et al. 1998). This can be regarded as examples to confirm the inconsistency between clients and auditors, which will eventually influence the audit quality.

Hypotheses development

In sum, when it comes to the relation between client size and audit quality, relevant prior literature generally hold following 3 views:

There is a negative correlation between client size and audit quality. Auditor firms averagely issue audit report with lower quality for larger clients.

L.E. DeAngelo et al. (1981) summarize the relationship between auditors and auditees in terms of economic dependency. Auditor firms are not non-profit organization. Instead they need to survive by charging auditing service fees. Thus basically auditor firms need to satisfy their customers which are auditee entities by issuing client-preferred audit opinions. From clients’ perspective, it always comes to the case that they prefer “satisfying” audit reports to “fair” audit reports. Thus auditor firms are always encountering the dilemma trading-off between high revenue and high audit reports quality. M.C. Knapp (1985) points out that the importance of audit clients is significantly correlated with audit quality. In addition, comparing to smaller clients, larger clients are more easily influence auditors’ judgement during auditing process. Eventually lower audit quality and more “satisfying audit reports” will be as a consequence. In fact, such correlation between client size and audit quality might vary along with the change of legal and political background.

There is a positive correlation between client size and audit quality. Auditor firms averagely issue audit report with higher quality for larger clients.

From auditor firms’ perspective, J.K. Reynolds & J.R. Francis (2001) conclude client size and audit quality are significantly correlated. Specifically, accounting firms are normally more careful issuing audit report for larger companies. Auditor independence will not be influenced by “economic incentives” consequently. Moreover, larger companies generally

(10)

9

show more impacts to capital market and they would be more monitored by legal authorities, regulators and public institution. Any auditing profession violation can be revealed much more easily. A. Craswell et al (2002) confirm such positive correlation, using empirical evidence that the level of audit fee does not show significant impact toward auditor independence.

Prior research suggests that, in stock market larger companies normally issue more public information for different stakeholders (M.J. Brennan et al., 1996; W. Gebhardt et al., 2001). As a result, the more publicly available information of a firm issue, the lower earnings management risk is due to more public monitor from different stakeholders (J. Boone et al., 2008).

There is no significant correlation between client size and audit quality.

Moreover, a number of empirical researches conceive auditor size do influence audit quality while the relation is notably insignificant. Z. Guo (2011) confirmed this insignificant relation between auditor size and audit quality by using empirical evidence from mainland China. Consistent with this finding, F. Liu et al (2009) confirmed the relation by using empirical evidence from Hong Kong China. These 2 papers are valuable to this study since the both employ empirical evidence from China.

The two hypotheses are logically generated base on the assumptions above:

Hypotheses H1: Under the circumstances clients are audited by big 4 audit firms, client size is correlated with audit quality.

H1a: There is a significant correlation between total assets and discretionary accruals in big 4 audited clients.

H1b: There is a significant correlation between firm values and discretionary accruals in big 4 audited clients.

Hypotheses H2: Under the circumstances companies are audited by non-big 4 audit firms, client size is correlated with audit quality.

H2a: There is a significant correlation between total assets and discretionary accruals in non-big 4 audited clients.

(11)

10

H2b: There is a significant correlation between firm values and discretionary accruals in non-big 4 audited clients.

Study Design and Methodology

According to J. K. Reynolds & J.R. Francis (2001), there are two types of accounting accruals can be applied for hypotheses H1 and H2: total accruals and discretionary accruals. Prior literatures argue that the magnitude of discretionary accruals is consistent with the magnitude of earnings management (R.M. Frankel et al, 2002; J.R. Francis, 2004; W. Chi & H. Huang, 2004; D.F. Prawitt et al, 2009). The measure for this study is the absolute value of discretionary accruals.

In Jones (1991),

TA

_ijt denotes total accruals,

DA

_ijt denotes discretionary accruals and

NA

ijt denotes non-discretionary accruals:

1jt

*

[PPE

ijt

/A

ijt-1

] +

ijt

(12)

11

TA

ijt: Total accruals of company i (of industry j in year t) which is calculated as the

difference of net income generated from on-going operations and cash flow generated from ordinary operations (C. Becker et al., 1998).

A

_ijt-1

:

Total assets of company i in industry j for year (t - 1), which is the previous period.

∆REV

ijt: Net change in revenue of company i of industry j in year t and previous period

ijt

/A

ijt-1

]

Where :

∆AR

ijt: Net change in accounts receivable of company i of industry j in year t and

previous period (t - 1)

Other variables are same as in

TA

_{ijt equation definition above.}

After that, discretionary accruals are calculated as the difference between total accruals’ absolute value and value of non-discretionary accruals, scaled by lagged total assets (A. A. Al-Thuneibat et al, 2011):

(13)

12

When it comes to identify the curvilinear relation between a particular variable and discretionary accruals, W. Chi & H. Huang (2004) use an approach in quadratic form. This study will use similar regression model for the purpose to test hypotheses H1 and H2. Logically I will discuss about the control variables that are added in the regression model in term of reason and importance. Moreover most of those control variables have been identified in prior research that can affect the level of discretionary accruals.

Control variables

According to (C. Becker et al., 1998), the most vital factors would be operating cash flow, debt level of a company and the absolute client size. I will identify the control variables that would be used in the regression model, because appropriately controlling variables would eventually contribute to the reliability of results by mitigating control variables’ systematic influences.

According to P. Dechow et al (1995), there is an inverse relation between discretionary accruals and operating cash flows (scaled by lagged total assets) within a company. As for the magnitude of debt in a client, M. Defond & J. Jiambalvo (1994) point out that companies show higher possibility involving in aggressive earnings management due to the incentives caused by higher debt level. Accordingly, higher financial leverage generally lead to higher risk and increase in equity risk premium of clients (W. Gebhardt et al, 2001). According to J. K. Reynolds & J.R. Francis (2001), larger companies in term of sales, being one of the operating characteristics, are generally correlated with smaller accruals. No matter whether the accruals are scaled by lagged company assets or are not. Similarly, clients’ financial condition is correlated with the level of earnings management, especially when there is financial distress. AZ can be the proxy for financial distress by using Altman Z-score (Al-Thuneibat et al, 2011; J. K. Reynolds & J.R. Francis, 2001).

Independent variable

As client size will be the independent variable in the estimation model, an appropriate proxy should be identified. According to P.M. Dechow & I.D. Dichev (2002), firms on stock exchange market with larger total assets generally show more stable accruals. This is the literature support that total assets can be used as proxy for client size. However in term of client size, it can be also explained as the firm value, which is the sum of total assets and total debts. Because firm value can reflect the economic impacts of clients and it will be

(14)

5

AZ

it

+

it

Where:

LV

it

:

The natural log of firm value of company i in year t.

Other variables are same as in prior equation definition above.

Results

Sample statistics

This study applies data from China because of two reasons basically. First, Chinese stock market is becoming an important part of global capital market. For example, according to China Briefing (2009), Shanghai Stock Exchange has become the world's 6th largest stock market in term of market capitalization by the end of Dec 2011. Secondly Chinese stock market carries its own unique characters. Unlike most stock exchanges over the world, the Shanghai Stock Exchange is still not entirely open to overseas investors because of the tight capital account controls run by the Chinese authorities (International Herald Tribune 2009; W.A. Thomas, 2001). CSMAR (China Stock Market & Accounting Research) is the comprehensive database for Chinese business research and it covers data on the Chinese stock market, financial statements and China corporate governance of Chinese listed firms. This study is planned to offer empirical evidence from Chinese capital market. Thus CSMAR, also a sub-database of WRDS database, would be the most appropriate data source for this study.

All the listed firms would be on either Shanghai stock exchange or Shenzhen stock exchange. This study chooses fiscal year from 2010 to 2012 and all the accounting figures should be disclosed on 31th December 2010, 31th December 2011 and 31th December 2012. This could lead to more accurate and valid accounting figures. The initials sample size is 7309, with 162 listed companies audited by big four auditor firms and 2360 listed companies audited by non-audit fours. Any companies without necessary data and unqualified for the model are

(16)

15

eliminated afterwards. Besides, all listed company should be audited by either big four or non-big fours, without auditor firm switches. After the sample selection process, all the rest samples carry sufficient data from year 2010 to 2012. The final sample size is 5097 (101 big 4 audited firms; 1598 non-big 4 firms). Such multi-principle sample selection will lead to more reliable, tenable and valuable data series, which eventually can contribute to the accuracy of hypotheses testing.

Table 1 illustrates sample frequencies. It is evidently that the majority of listed companies receive audit services from non-big 4 accounting firms (over 90 percent). This is consistent with the current auditing service market development in China.

Table 1. Frequencies statistics (both Big-4 audited & Non-Big-4 audited clients)

Big 4 Firms Non-Big 4 Firms Total

Number of Sample Companies 101 1598 1699 Corresponding Sample Size 303 4794 5097 Percentage 5.94% 94.06% 100.00%

Table 2.1 and Table 2.2 report the descriptive statistics of all variables in the model for both big-4 clients and non-big 4 clients respectively. No matter in term of client size, big 4 clients generally hold larger total assets size or total firm value size (natural logarithm index: 23.99 to 21.74 & 24.42 to 22.14 respectively). Although the average debt ratios are at same level (0.55 and 0.54 respectively), non-big 4 clients show greater fluctuate range with almost no debt (0.7%) or sky high debt (1890%) extreme figures. As for operating sales and operating Table 2.1. Descriptive statistics (Big-4 audited firms)

Big-4 Audited Firms

N Minimum Maximum Mean Std. Deviation

LA 303 20.62723 28.40521 23.99173 1.611098

LV 303 20.9488 28.78061 24.42333 1.664725

(17)

16 SALE 303 18.82535 28.65564 23.51489 1.805154 OCF 303 -0.28953 0.407699 0.063685 0.072543 Altman Z-score 303 0.164004 5.315265 1.692663 0.917197 DAs 303 -0.07847 0.292655 0.017597 0.036388 Valid N (list wise) 303

Table 2.2. Descriptive statistics (Non-big-4 audited firms)

Non-Big-4 Audited Firms

N Minimum Maximum Mean Std. Deviation

LA 4794 13.7633 26.32618 21.74349 1.262046 LV 4794 15.8405 26.90936 22.14557 1.301254 DEBT 4794 0.00708 18.93984 0.539775 0.69228 SALE 4794 9.044175 25.81119 21.12034 1.554257 OCF 4794 -1.56802 2.457275 0.037102 0.10211 Altman Z-score 4794 -465.743 65.28047 1.733701 7.407407 DAs 4794 -0.37249 9.927102 0.033504 0.257051 Valid N (list wise) 4794

cash flows per asset, big-4 clients still stand averagely higher level (1.81 for SALE, 0.07 for OCF and 1.55 for SALE, 0.10 for OCF respectively). In other words, big 4 clients normally run more sustainable, show greater operation and asset management abilities. Since it has been widely acknowledged that financial distress is a vital contributor to discretionary accruals, tis study calculate the sample’s Altman Z-scores from 5 dimensions: net working capital / assets, retained earnings / assets, earnings before interest and taxes / assets, book value of equity / liabilities and sales / assets. The average Altman Z-scores for both big 4 clients and non-big 4 clients stay at same level (1.69 and 1.73 respectively), while the variances are quite different. Non-big 4 clients again hold greater fluctuate range that the counterparts of big 4 clients.

(18)

17

Evidently these variables’ descriptive figures are averagely at same level, but the scenario is quite different when it comes to the variances of these variables. An obviously reason is that non-big 4 clients stand sixteen-fold sample size as big 4 clients (4794 to 303). This will lead to more extreme variables in term of both amounts and numbers. However it cannot be ignored that such variance difference between big 4 clients and non-big 4 clients can be affected by company themselves as well. Z. Guo (2011) cites the auditing data of top 100 accounting firm published by CICPA (Chinese Institute of Certified Public Accountants). Specifically, the year 2010 witnessed that big 4 accounting firms’ revenue account for 44% top 100 accounting firms’ total revenue in China. As for average auditor revenue annual revenue, auditors in big 4 accounting firms earn more than 6 times than auditors in non-big4 accounting firms. Such notable difference between big 4s and non-big 4s reflect that clients with higher financial capacities are more likely to accept audit services from big 4 accounting firms. In other words, big 4 clients generally stand larger in terms of company size, economic capability etc. because they have to afford relatively high audit fees. But for non-big 4 clients, it is different scenario that clients’ sizes are far more varied.

Hypotheses testing

Hypotheses H1 and H2 will be tested by testing all the sub-hypotheses at the first place. For all sub-hypotheses H1a, H1b, H2a and H2b, the absolute value is applied as dependent variables in the linear regression. Control variables in the linear regression equations are as follows: debt ratio, the natural logarithm of sales scaled by lagged total assets, the operating cash flow scaled by lagged total assets and the Altman Z-score.

For both hypotheses H1a and H2a, the natural logarithm of total assets is applied as independent variables. For both hypotheses H1b and H2b, the natural logarithm of firm value is applied as independent variables.

The linear regression results for both big 4 clients and non-big 4 clients are reported in 4 Table series: Table 3.X, Table 4.X, Table 5.X and Table 6.X (X= 1, 2 & 3). For any one particular table series, the first one would be the ANOVA test table, following by the Model Summary table and the last one would be the Coefficients table.

OLS regression estimations for testing H1a & H1b are in Table 3 series and Table 4 series, where client size is measured setting the level of total assets or firm as a proxy in sequence.

(19)

18

OLS regression estimations for testing H2a & H2b are in Table 5 series and Table 6 series, where client size is measured setting the level of total assets or firm value as a proxy in sequence.

Table 3.1 ANOVA (big-4 audited clients, client size proxy: total assets)

Model Sum of Squares df Mean Square F Sig. 1 Regression .059 5 .012 10.343 .000b

Residual .341 297 .001 Total .400 302

a. Dependent Variable: DA

b. Predictors: (Constant), AZ, SALE, OCF, DEBT, LA

Table 3.2 Model Summary (big-4 audited clients, client size proxy: total assets)

Model R R Square Adjusted R Square Std. Error of the Estimate 1 .385a .148 .134 .033862675 a. Predictors: (Constant), AZ, SALE, OCF, DEBT, LA

Table 3.3. Coefficients (big-4 audited clients, client size proxy: total assets)

Model Unstandardized Coefficients Standardized Coefficients

t Sig. B Std. Error Beta

(20)

19 LA -.012 .005 -.540 -2.473 .014 DEBT .018 .021 .093 .865 .388 SALE .010 .005 .494 2.162 .031 OCF -.140 .030 -.279 -4.702 .000 AZ .004 .005 .104 .829 .407 a. Dependent Variable: DA

Table 4.1. ANOVA (big-4 audited clients, client size proxy: firm value)

Model Sum of Squares df Mean Square F Sig. 1 Regression .059 5 .012 10.314 .000b

Residual .341 297 .001 Total .400 302

b. Predictors: (Constant), AZ, SALE, OCF, DEBT, LV

Table 4.2. Model Summary (big-4 audited clients, client size proxy: firm value)

Model R R Square Adjusted R Square Std. Error of the Estimate 1 .385a .148 .134 .033869795 a. Predictors: (Constant), AZ, SALE, OCF, DEBT, LV

Table 4.3. Coefficients (big-4 audited clients, client size proxy: firm value)

Model Unstandardized Coefficients Standardized Coefficients

t Sig. B Std. Error Beta

(21)

20 LV -.012 .005 -.555 -2.447 .015 DEBT .026 .020 .133 1.337 .182 SALE .010 .005 .492 2.139 .033 OCF -.140 .030 -.278 -4.696 .000 AZ .004 .005 .103 .818 .414 a. Dependent Variable: DA

For clients of big 4 accounting firms

The ANOVA test contains information of the linear regression model as a whole. The evidence in Table 3.1 (significance is 0.000) shows significant relationship between total assets and discretionary accruals. Simultaneously, in table 4.1 the significance is 0.000 as well when it comes to the relationship between firm value and discretionary accruals. Combining the evidence above, client size of big 4 accounting firms is evidently correlated with audit quality.

In Model Summary tables, the adjusted R square is applied to determine the explanatory power of the linear regression model in this study. Specifically, it verifies the level in percentage that independent variable and control variables collectively contribute to the variation of the dependent variable (discretionary accruals). In this study, both adjusted R square figures in Table 3.2 and Table 4.2 are 0.134. In other words, total assets or firm value associated with all the rest variables contribute to 13.4 percent of the level or the change level of the discretionary accruals in the regression model. Such relatively low percentage means that there are more factors that can contribute to the level of discretionary accruals. The linear regression model can be improved in term of accuracy, if other factors can be discovered and added to the model.

According to J.L. Devore (2011), in an integral part of statistical hypothesis testing, significance level can be used as a tool to determine whether a hypothesis can be accepted or rejected. P-values are often integrated to the level of significance, which is normally at 5% (0.05) (S. Sandra, 2007). Therefore, if a p-value was found to be less than 0.05, then the

(22)

21

standardized coefficient result can be regarded as significant statistically and the hypothesis can be accepted consequently (M. Steve, 2006). There are also other significance levels (0.1 & 0.01) that would be applied in this study towards the relation between client size and audit quality.

The correlation coefficients can be found in both Table 2.3 and Table 3.3. In Table 2.3, where total assets stands the proxy for clients size, the standardized coefficients for LA is – 0.540 (negative) and is highly significant at 0.014. As for LV in Table 3.3, the standardized coefficient is – 0.555 (negative) and is highly significant at 0.015. Therefore the high significance level of either LA or LV can support the correlation between client size and the level of discretionary accruals. This is statistical result is consistent with the counterparts in ANOVA tables, in term of relation between client size and audit quality. Thus both H1a and H2b are accepted as follows: There is a highly significant negative relation between total assets and discretionary accruals in big 4 audited clients. There is a highly significant negative relation between firm values and discretionary accruals in big 4 audited clients. In other words, it reveals the correlation between client size and audit quality. Given that the variable coefficients signs are both negative, it can be concluded that client size and audit quality are negatively correlated. The larger client size of big 4 accounting firms is generally refers to lower audit quality correspondingly. Thus H1 is accepted as follow: Under the circumstances clients are audited by big 4 audit firms, there is a highly significant negative relation between client size and audit quality.

Table 5.1. ANOVA (non-big 4 audited clients, client size proxy: total assets)

Model Sum of Squares df Mean Square F Sig. 1 Regression 1.122 5 .224 3.404 .005b

Residual 315.576 4788 .066 Total 316.697 4793

b. Predictors: (Constant), AZ, SALE, OCF, DEBT, LA

(23)

22

Model R R Square Adjusted R Square Std. Error of the Estimate 1 .060a .004 .003 .256728814 a. Predictors: (Constant), AZ, SALE, OCF, DEBT, LA

Table 5.3. Coefficients (non-big-4 audited clients, client size proxy: total assets)

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -.058 .065 -.887 .375 LA -.011 .006 -.054 -1.945 .052 DEBT -.003 .006 -.007 -.412 .681 SALE .016 .005 .095 3.425 .001 OCF -.073 .037 -.029 -1.969 .049 AZ .000 .001 -.005 -.289 .772 a. Dependent Variable: DA

Table 6.1. ANOVA (non-big 4 audited clients, client size proxy: firm value)

Model Sum of Squares df Mean Square F Sig. 1 Regression 1.146 5 .229 3.477 .004b

Residual 315.552 4788 .066 Total 316.697 4793

(24)

23 b. Predictors: (Constant), AZ, SALE, OCF, DEBT, LV

Table 6.2. Model Summary (non-big 4 audited clients, client size proxy: firm value)

Model R R Square Adjusted R Square Std. Error of the Estimate 1 .060a .004 .003 .256719099 a. Predictors: (Constant), AZ, SALE, OCF, DEBT, LV

Table 6.3. Coefficients (non-big 4 audited clients, client size proxy: firm value)

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -.059 .063 -.927 .354 LV -.011 .005 -.056 -2.036 .042 DEBT .000 .006 .000 .022 .982 SALE .016 .005 .097 3.501 .000 OCF -.076 .037 -.030 -2.035 .042 AZ .000 .001 -.005 -.293 .769 a. Dependent Variable: DA

For clients of non-big 4 accounting firms

As for ANOVA test, evidence in Table 3.1 (significance is 0.005) shows significant relationship between total assets and discretionary accruals. Simultaneously, in table 4.1 the significance is 0.004 when it comes to the relationship between firm value and discretionary accruals. Combining the evidence above, client size of non-big 4 accounting firms is evidently correlated with audit quality.

(25)

24

The correlation coefficients can be found in both Table 5.3 and Table 6.3. In Table 5.3, where total assets stands the proxy for clients size, the standardized coefficients for LA is – 0.540 (negative) and is significant at 0.052. As for LV, the standardized coefficient is – 0.560 (negative) and is significant at 0.042. Therefore the significance level of either LA or LV can support the correlation between client size and the level of discretionary accruals. This is statistical result is consistent with the counterparts in ANOVA tables, in term of relation between client size and audit quality. Thus it seems that both H2a and H2b are supposed to be accepted as follows: There is a significant negative relation between total assets and discretionary accruals in non-big 4 audited clients. There is a significant negative relation between firm values and discretionary accruals in non-big 4 audited clients.

In other words, it also reveals the correlation between client size and audit quality. Given that the variable coefficients signs are both negative, it can be concluded that client size and audit quality are negatively correlated. The larger client size of non-big 4 accounting firms is generally refers to lower audit quality correspondingly. Thus it seems that H2 is supposed to be accepted as follow: Under the circumstances clients are audited by non-big 4 audit firms, there is a significant negative relation between client size and audit quality.

However the statistics in Model Summary tables cannot support H2. In Model Summary tables for non-big 4 accounting firms, both adjusted R square figures in Table 5.2 and Table 6.2 are 0.003. In other words, total assets or firm value associated with all the rest variables barely contribute to 0.3 percent of the level or the change level of the discretionary accruals in the regression model. Such extremely low percentage means that there are far more factors that can contribute to the level or the change level of discretionary accruals. The linear regression model for non-big 4 accounting firms is inaccurate and need to be reestablished by adding more factors and more logical factor selection. The adjusted R square figures are quite different between big 4 clients and non-big 4 clients are applied in the regression model. Thus the regression model quality is different and obviously it is hardly useful for non-big 4 accounting firms researching the relation between client size and audit quality. Therefore, H2 should be rejected that there is no significant correlation between client size and audit quality for non-big 4 audited companies.

All in all, the data in non-big 4 clients sample cannot support testing H2 by using the regression model. Thus only H1 is accept and H2 is rejected, because there is insufficient evident to accept H2 in this study.

(26)

25

Additional analysis

Table 7. Correlations between control variables and DA Control Variable Big 4 Non-big 4 LA LV LA LV DEBT 0.093 0.133 -0.007 0.000 SALE 0.494 0.492 0.095 0.097 OFC -0.279 -0.278 -0.29 -0.030 AZ 0.104 0.103 -0.005 -0.005

As it is illustrated in Table 7, for clients of big 4 accounting firms, DEBT, SALE and AZ are positively correlated with DA while for OCF the relation is negative. This is consistent with the assumption in P.M. Dechow et al (2002), M. Defond & J. Jiambalvo (1994), W. Gebhardt et al, 2001), which means the larger for any one of DEBT, SALE or AZ, the more discretionary accruals will be. But SALE’s influence on DA is adverse with the arguments in J. K. Reynolds & J.R. Francis (2001), in which they claim more sales are always consistent with smaller discretionary accruals. Furthermore, the level of significance is extremely low (> 0.1) for DEBT and AZ: 0.388, 0.182, 0.407 and 0.414. However the significance levels of SALE and OCF can support the highly significant relations between SALE and DA, OCF and DA (0.031, 0.000, 0.033 and 0.000).

Overall, in this sample part (clients of big 4 accounting firms), SALE and OCF contribute to audit quality at highly significant level and both AZ and DEBT are not significantly related to audit quality.

As for clients of non-big 4 accounting firms, DEBT is positively correlated with DA while for OCF and AZ the relations are negative. Moreover, for DEBT the relations are totally inverse applying LA or LV as the proxy for client size. This is likely due to the inter-affects between LV and DEBT, while for LA such inter-affects should be less.

Furthermore, the level of significance is extremely low (> 0.1) for DEBT and AZ: 0.681, 0.982, 0.772, and 0.769. However the significance levels of SALE and OCF can support the highly significant relations between SALE and DA, OCF and DA (0.001, 0.000, 0.049 and 0.042). The insignificant relation between DEBT and DA may also explain part of the inverse relation for separately applying LA and LV, because some extreme errors or values may partially lead to the difference.

(27)

26

In sum, in this sample part (clients of non-big 4 accounting firms), SALE and OCF contribute to audit quality at highly significant level and both AZ and DEBT are not significantly related to audit quality.

Conclusions and Discussions

This study investigates the relationship between the client size (both big-4 audited companies and non-big 4 audited companies) and audit quality. The empirical evidence is based on China stock market (Shanghai stock market and Shenzhen stock market), applying the data from CSMAR database. The linear regression model for hypotheses testing is based on Jones Model with minor modifications. Two similar regression models are employed for LA and LV separately. Specifically the relationship between client size and audit quality is measured by the level of discretionary accruals and the level of total assets and firm value. The linear relation and significance level statistics of the model in this study are summarized in the follow Table 8.

Table 8. Correlations between independent variables and DA Independent

Variable

Standardized Coefficients Level of Significance

Big 4 Non-big 4 Big 4 Non-big 4

LA -0.540 -0.540 0.014** 0.052* LV -0.555 -0.560 0.015** 0.042** a. Level of significance : *** P value < 0.01 ** P value < 0.05 * P value < 0.1

Barely seen from Table 8, for big 4 clients, both LA and LV are negatively correlated with DA. The high level of significances (0.014** & 0.015**) can testify the accuracy and reliance of the sample data. This will lead to the negative correlation between client size and audit quality. Thus hypotheses H1 is accepted consequently. For non-big 4 clients, although both LA and LV are negatively correlated with DA as well, the significant levels (0.052* &

(28)

27

0.042**) is notably lower than the counterparts of big 4 clients. Besides, based on the prior analysis with Model Summary tables, it is evidently that the regression model is not fit for client sample of non-big 4 accounting firms. The correlation between audit size and audit quality is not significant under such circumstances. Thus hypotheses H2 is rejected.

In sum, evidence from Chinese stock market indicates that for big 4 accounting firms, client size is significantly correlated with audit quality. Furthermore, the larger client size the lower audit quality will be. This result is consistent with some prior literature (L.E. DeAegelo, 1981; M.C. Knapp, 1985) that client size might be one of the contributors to audit quality with empirical evidence support. As for non-big 4 accounting firms, client size shows no significant correlation with audit quality. This result is consistent with E. Carson et al (2004) and F. Liu et al, (2009) that client size cannot significantly influence the auditor judgement in term of audit quality with empirical evidence.

This empirical study in fact contributes to prior studies and audit related researches. Firstly, in term of audit quality, there is not many researches focus on client size as a contributor. Thus this study might be supplement for the audit quality literature. In addition, the result confirms the “economic incentive” do exist in real life even for big 4 accounting firms, which is introduced by J. K. Reynolds & J.R. Francis (2001). Because client size’s influence toward audit quality of big 4 accounting firms is significant in this research. Secondly the proxies for auditing client size vary among literatures. Total sales, total assets and auditing fee level etc. have been employed in prior literatures and in this study firm value (the sum of total assets and total debt) is employed. Although the result is not satisfying for firm value, at least it brings an innovative attempt to measure client size by another proxy. Thirdly, this study provides empirical reference for individuals focusing on Chinese stock market. Thus empirical evidence can contribute to the research concentrating on Chinese stock market. Especially the latest 3-year-data (from 2010 to 2012) is collected for the empirical research, there the results is at least time efficient.

However, the limitation of this study cannot be ignored. No doubt that empirical evidence from China cannot be generalized worldwide, even if the result is valuable to Chinese stock market stakeholders. This is mainly due to the development level and unique characteristics (governmental policies, regulations and legislations) of Chinese stock market itself. The linear regression model for big 4 accounting firms need to be improved by adding more contributors. As for the regression model for non-big 4 accounting firms, it needs to be totally

(29)

28

restructured since the model can hardly explain the level of discretionary accruals. Not only the malfunctions caused by interplays between control variables LV and DEBT, but also the deficiencies that more valuable factors are missing. One of the possible contributors to audit quality would be the years that a listed company has been on stock market. As for prior literature support, J.H. Anthony & K. Ramesh (1992) come up with another variable, the number of years for clients in stock market. This can control the variation of companies’ accruals in different years. Additionally, the longer years for a client stand in stock market, the less risky will be generally (J. Boone et al, 2008). According to this, A. A. Al-Thuneibat et al. (2011) successfully employed this factor into the estimation of discretionary accruals. While in CSMAR database, I could not export appropriate data for this type of data. Thus the missing potential favourable data will lead to the limitation of regression model establishment. Moreover, in study the client size are measured separately for big 4 accounting firms and non-big 4s, because I potentially consider the interaction between client size and audit quality may be varied among different auditing service providers. Maybe take all clients as a whole sample for this study may generate more convincing and reliable results for the relationship between client size and audit quality.

The empirical results of this study need to be discussed further. As we have acknowledged from Z. Guo (2011) that big 4 accounting firms earn far more times revenue than any one of non-big 4 accounting firms. Big 4 accounting firms have better reputation on audit quality and are supposed to deliver auditing services to many clients. Thus big 4 firms are more likely to have less “economic incentives” and more “bargain power” as result. However the result in this study indicates that the client size still negatively correlated with audit quality. This may explained by the empirical evidence in this study itself. As we can see from Table 1, less than 8% listed companies are audited by big 4 auditor firms (101 for big 4; 1598 for non-big 4s). Thus one possible reason might be that big 4 accounting firms are trying to attract more clients at the cost of lower audit quality, compromising to clients’ wills to audit reports for instance. From another perspective, it can be referred to the growing Chinese auditing market. Thus more researches towards China auditing market would be valuable contribution. Moreover, this is consistent with (M.J. Brennan et al., 1996; W. Gebhardt et al., 2001; J. Boone et al., 2008), less monitoring from investors other than from Chinese authorities may contribute such negative relation between client size and audit quality. This study can also suggest future research investigating audit quality by accruals can improve the

(30)

29

regression model by adding more potential contributors. Moreover, based on the results for this study, public monitor tension may be the promising contributor for this.

References:

1. A. A. Al-Thuneibat et al. (2011), “Do audit tenure and firm size contribute to audit quality? – Empirical evidence from Jordan”, Managerial Auditing Journal, Vol. 26, No. 4, pp. 317-334.

2. A. Craswell et al. (2002), “Auditor independence and fee dependence”, Journal of Accounting & Economics 33, pp. 253-275.

3. C. Becker et al. (1998), “The effect of audit quality on earnings management”, Contemporary Accounting Research, Vol. 15, pp. 1-24

4. C. W. Chow & S. J. Rice. (1982), “Qualified audit opinions and auditor switching”, The Accounting Review 57, April, pp. 326-335.

5. China Briefing. (2009), “Foreigners Now Allowed to List on Shanghai Stock Exchange”- retrieved on January 21.

6. D.F. Prawitt et al. (2009), “Internal Audit Quality and Earnings Management”, The Accounting Review, Vol. 84, No. 4, pp. 1255-1280.

7. E. Carson et al. (2004), “Audit Fees and Market Segmentation-Further Evidence on How Client Size Matters within the context of Audit Fee Models”, International Journal of Auditing, 8, pp. 79-91.

8. F. Liu et al. (2009), “Auditors Size, Client Size and Audit Quality-Empirical evidence on cross-listed mainland companies in HKSE”, Journal of Audit Research, No. 3, pp. 45 – 54.

9. H. Arnett & P. Danos. (1979), “CPA firm viability”, University of Michigan, Ann Arbor, MI, pp. 8-56.

(31)

30

10. International Herald Tribune. (2009), “China further loosens its capital controls” retrieved on January 21.

11. J. F. Mutchler. (1984), “Auditors’ perception of the going-concern opinion decision”, Auditing: A Journal of Practice & Theory 3, Spring, pp. 17-30.

12. J. J. Jones. (1991), “Earnings management during import relief investigations”, Journal of Accounting Research 29, pp. 193-228.

13. J. K. Reynolds & J.R. Francis. (2001). “Does size matter? The influence of large clients on office-level auditor reporting decisions”, Journal of Accounting and Economics 30, pp. 375-400.

14. J. V. Carcello. & T. L.Neal. (2000), “Audit Committee Composition and Auditor reporting”, The Accounting Review, Vol. 75, No. 4, October, pp. 453-467.

15. J.H. Anthony & K. Ramesh. (1992), “Association between accounting performance measure and stock prices”, Journal of Accounting and Economics, Vol. 15, pp. 203-227.

16. J.L. Devore. (2011). “Probability and Statistics for Engineering and the Sciences (8th edition.)”, Boston, MA: Cengage Learning. pp. 300–344.

17. J.R. Francis, (2011). “A Framework for Understanding and Researching Audit Quality”, Auditing: A Journal of Practice & Theory. Vol. 30, No 2, pp. 125-152. 18. J.R. Francis. (2004). “What do we know about audit quality”, The British Accounting

Review 36, pp. 345-368.

19. K. Raman et al. (2008), “Audit tenure and the equity risk premium”, Journal of Accounting, Auditing, and Finance, Winter, pp. 115-140.

20. L. E. DeAngelo. (1981), “Auditor size and audit quality”, Journal of Accounting and Economics 3, pp. 183-199.

21. L. Zhang. (2005) “Related analysis for earnings management and audit quality of listed companies”, Finance & Trade Economics, No.4, pp. 33-39

22. M. C. Knapp. (1985), “Audit Conflict: An Empirical Study of the Perceived Ability of Auditors to Resist Management Pressure”, The Accounting Review, Vol. 60(2), pp. 202-211.

23. M. DeFond & C. Park. (1997), “Smoothing income in anticipation of future earnings”, Journal of Accounting and Economics 23, pp. 115-139.

(32)

31

24. M. Defond & J. Jiambalvo. (1994), “Debt covenant violation and manipulation of accruals”, Journal of Accounting and Economics 17, pp. 145-176.

25. M. Geiger et al. (1998), “Costs associated with going-concern modified audit opinions: An analysis of auditor changes, subsequent opinions and client failures”, Advances in Accounting 16, pp. 117-139.

26. M. Steve. (2006). "Probability helps you make a decision about your results". Statistics Explained: An Introductory Guide for Life Scientists (1st edition.). Cambridge, United Kingdom: Cambridge University Press, pp. 44–56.

27. M.J. Brennan. (1996), “Market microstructure and asset pricing: on the compensation for illiquidity in stock returns”, Journal of Financial Economics, Vol. 41, pp. 441-464. 28. P. Dechow et al. (1995), “Detecting earnings management”, The Accounting Review

70, pp. 193-225.

29. P. Dechow et al. (2002), “The quality of accruals and earnings: the role of accrual estimation errors”, Accounting Review, Vol. 77, pp. 35-59 (supplement).

30. R.M. Frankel et al. (2002), “The relation between auditors’ fees for non-audit services and earnings management”, Accounting Review, Vol. 77, pp. 71-105.

31. S. Sandra (2007). “Elementary Statistics Using JMP (SAS Press) (PAP/CDR edition.).” Cary, NC: SAS Institute. pp. 166–169.

32. V. E. Johnson et al. (2002), “Audit-firm tenure and the quality of financial reporting”, Contemporary Accounting Research, Vol. 19, Winter, pp. 637-660.

33. W. A. Thomas. (2001), “Western Capitalism in China: A History of the Shanghai Stock Exchange”. Aldershot: Ashgate Pub Ltd.

34. W. Chi & H. Huang. (2004), “Discretionary accruals, audit firm tenure and audit partner tenure: empirical evidence from Taiwan”, working paper, National Chengchi University, Taipei.

35. W. Gebhardt et al. (2001), “Toward an implied cost-of-capital”, Journal of Accounting Research, June, pp. 135-176.

36. Z. Guo. (2011), “International Big Four Accounting Firms and High Auditing Quality – Evidence from China’s Securities Market”, Journal of Audit Research, 2011, No. 1, pp. 98-107.

(33)

32

37. Z. U. Palmrose. (1986), “Audit fees and auditor size: further evidence”, Journal of Accounting Research, Spring, 1986, Vol. 24 (1), pp. 97-110.

38. Z. U. Palmrose. (1986), “The Effect of Non-Audit Services on the Pricing of Audit Services: Further Evidence”, Journal of Accounting Research, Vol. 24 (2), pp. 405-411.

39. Z. U. Palmrose. (1997), “Audit litigation research: Do the merits matter? An assessment and directions for future research”, Journal of Accounting and Public Policy, Winter, pp. 355-378.

Does client size contribute to audit quality? : empirical evidence from China