Effect of Auditing on Variability of Returns and Trading Volume Charles J. P. Chen*
[email protected] Bin Srinidhi**
[email protected] Xijia Su*
[email protected]
April 2007
*
Department of Accountancy, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong ** School of Accounting and Finance, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Acknowledgment: This study was supported by a grant from the Research Grant Council of the Hong Kong Special Administrative Region, China (CityU 1101/02H). We express our gratitude to Jere Francis, Dan Dhaliwal and Dan Simunic for their excellent suggestions to improve the paper. The paper has benefited from the comments of workshop participants at American Accounting Association 2003 Annual Meeting, City University of Hong Kong and Hong Kong University of Science and Technology.
Effects of Auditing on Variability of Returns and Trading Volume
Abstract We develop a model to show that auditing can reduce information divergence among investors to the extent that it increases the relative weight they place on common financial statement information as opposed to diverse non-accounting information. Therefore, both variability of stock returns and trading volume are expected to be lower for observations with audited financial statements. Consistent with our predictions, we find that audited observations are associated with lower stock returns variability and lower volumes of trading volume than non-audited firms, subsequent to their announcement of semi-annual financial statements. These results are robust to variations in event window length and specification of empirical measures. Our findings show the benefits of auditing in that it reduces perceived information risk in audited financial statements as well as reducing information divergence among investors. Keywords: auditing, information asymmetry, variability of stock returns, trading volume. Data Availability: Data used in this study are available from identified sources.
2
I. INTRODUCTION Investment decisions are based partly on reported financial statement information. However, unreliable firm-specific information in financial statements could pose a nondiversifiable information risk to investors (see Francis et al., 2005 for the definition and description of information risk). Current reporting regimes in most countries seek to increase the reliability of financial statement information by a combination of many mechanisms. Two of the important mechanisms used in improving the reliability of financial statements are the auditing of annual financial statements by a third party, namely external auditors, and an interim reporting process requiring quarterly or semi-annual financial statement preparation and reporting. Recent studies such as Brown and Pinello (2007) have documented that audited annual financial statements more reliable than unaudited interim statements. A natural question that arises from this finding is whether external audits of interim statements can improve overall financial statement reliability. Potentially, external auditing can improve the reliability of financial statements by independently attesting compliance with prevailing accounting standards. The improved reliability of financial statement information reduces the information risk associated with it and increases the weight that investors place on such information relative to other non-accounting information. In effect, financial statement information which is common across investors is given greater weight in the assessment of equity value by investors. By the same token, the investors will place less weight on the diverse non-financial information. From this reasoning, to the extent that auditing increases the reliability of financial information, the inter-investor divergence
1
on the assessment of firm value should decrease. Using the same reasoning in
reverse, if measures of consensus between investors are higher in firms with audited interim financial statements auditing compared to firms with non-audited interim financial statements, ceteris paribus, we can conclude that auditing improves financial statement reliability.
A
comparison of inter-investor information divergence subsequent to the release of audited and unaudited financial statements in comparable firms can yield the necessary evidence on whether auditing indeed improves reliability and reduces inter-investor divergence. However, in regimes 1
We use the term “information divergence” instead of “information asymmetry” in this paper to denote interinvestor differences in information. The term “information asymmetry” has the connotation of information differences between managers and investors, which is not the focus of this paper.
3
where auditing of annual reports is mandatory and interim reports are not audited, 2 such a comparison is not possible. Our paper examines this issue using a unique window of time in which Chinese regulations allowed firms to get their interim semi-annual statements voluntarily audited by external auditors in addition to having their annual statements audited. There has been extensive prior research on whether and how auditing improves financial statement reliability. Becker et al. (1998) show that the Big 6 auditors constrain the extent of earnings management. Teoh and Wong (1993), Choi and Jeter (1992) and Loudder et al. (1992) show that earnings of firms that are audited by the large auditors exhibit higher stock return responses. These studies have focused on the effect of audit quality (proxied by auditor size) on investor confidence in financial statements and information risk for investors. However, there are two potential problems with this approach. The first is that audit quality is difficult to measure because of the relatively unobservable nature of the audit function (DeAngelo, 1981). As a result, auditor size is used to proxy for audit quality because larger auditors are perceived to need to maintain high reputation (reputation hypothesis) and because they have greater wealth at risk from litigation (deep pocket hypothesis – see Lennox, 1999 for an evaluation of the two hypotheses).
However, Bar-Yosef and Sarath (2005) show that
concentration of audit market share can occur even if well-capitalized auditors have no relative advantage in supplying higher quality audits. Little direct empirical evidence exists on whether the relationship between size and quality extends beyond the Big 4 auditors and whether it extends beyond the developed countries where reputation and litigation risks are significant. The second potential problem is whether the relationship between auditor size and information risk provides sufficient evidence on the role of auditing per se. Stated differently, the evidence from prior studies cannot distinguish between the reliability improvement brought about by the fact that the financial statements are independently audited vis-à-vis the reliability improvement that is brought about by the improved competence of a higher quality auditor. In this paper, we surmount these problems of indirect measurement by utilizing an opportunity in the Chinese auditing environment where about one third of semi-annual financial reports are audited and the rest are not. We show that auditing reduces the variability of stock 2
Auditor’s involvement in interim reports in the US is limited (Frankel et al. , 2002) to a review that is of more limited scope than audit, taking the form of inquiries and analytical review procedures rather than substantive testing (Brown and Pinello, 2007).
4
returns and the trading volume subsequent to the announcement of semi-annual reports.3 These findings are consistent with the notion that auditing reduces information divergence among investors by increasing their confidence in the financial statement information. In effect, our study differs from these earlier studies in that our setting allows us to directly compare the returns variability and trading volume between observations with audited semi-annual reports and those without them. We describe below the context in which semi-annual auditing is practiced in China. Under Chinese regulation, firms with poor performance records or financial positions, as well as firms that plan to issue rights or pay dividends in the second half of the year, are required to have their semi-annual reports audited. Other firms can have their semi-annual reports audited voluntarily.4 In recent years, about a quarter of semi-annual financial reports have been audited and over seventy percent of them were not required to be audited. Three points should be noted here. First, all of these firms with semi-annual audits also have their annual reports audited. One could argue that the effects of semi-annual audits in these firms could be negated by the annual audits via ex-post settling up of accounts. This reduces the likelihood of our detecting the information effects of semi-annual audits. Therefore, an empirical detection of reduced information divergence in this setting shows that it extends beyond the dilutive effects of ex-post settling by annual audit. Second, both the audit scope and reporting requirements of auditors of semi-annual financial statements in China are similar to that of annual audits. Auditors with whom we held follow-up interviews told us that the audit procedures they used in semi-annual audits were substantially the same as those used in annual audits. The similarities in the comprehensiveness and the consistency in audit procedures help generalize our results to annual audits. Third, we recognize the need to correct for the self-selection bias in our research design since semi-annual audits are not mandatory. In addition to employing the approach suggested by Heckman (1976), we have taken many steps in this study to address the issue of self-selection bias and these will be discussed in sensitivity test section. 3
The bid-ask-spread which is a common measure of inter-investor information divergence is not available in the Chinese market context. In a sensitivity test, we show that the high-low spread (which is reported) decreases for audited firms. 4
In general, voluntary auditing of semi-annual statements in China and quarterly statements in the US is not forbidden. However, our setting is different as some firms are required to have their semi-annual statements audited. This sensitizes investors and firms to the possibility and benefits of a semi-annual audit.
5
This paper contributes to the literature in two ways: by the context in which the study is conducted; and by the approach used to study the effects of auditing. In addition to directly comparing audited observations with non-audited ones, our context allows us to examine the difference in stock returns variability and trading volume between voluntarily audited and nonaudited observations. Our approach identifies the information risk and divergence reduction for audited firms relative to firms that were not audited. This is different from earlier approaches of examining earnings response coefficients or discretionary accruals. In our context of semiannual audits, the use of earnings response coefficients is not appropriate because a justifiable earnings expectation model is not available. The semi-annual income of the previous year is not a good expectation of the current year semi-annual income both because the 1990s represented a very high growth period for Chinese firms and because auditing affects earnings, book values, and other valuation-pertinent financial information not captured by the earnings response coefficient. The remainder of the paper is organized as below. Section II provides the background and literature review. Section III develops the theory for the proposition that auditing affects the variability of stock returns and trading volume, and presents research questions and propositions. Section IV gives the sample, research method and empirical results. Section V concludes the paper.
II. BACKGROUND AND LITERATURE REVIEW Auditing could improve the precision of financial statement information in two ways. First, by issuing qualified opinions to firms with unreliable financial statements, auditors enable investors to screen out such firms. We call this the attestation benefit. Attestation of financial statements by auditors makes the financial statements more reliable and consequently more beneficial to investors (Winters, 1975; Libby, 1979; Reckers and Panny, 1979; Pany and Smith, 1982; Johnson et al., 1983; Epstein and Pava, 1993; Hirst et al., 1995; Maines 1996; Hirst et al. 1999). The attestation benefits of auditing arise from an independent verification of the compliance of the financial statements with the prevailing GAAP. Further, by undertaking the audit, the prospective of receiving modified opinions motivates managers to be more disciplined in their financial reporting, which increases the reliability of financial statements.5 5
Auditors have an incentive to make managers more conservative in their reporting so that audit risk is reduced. In
6
In addition to the attestation benefits, in a voluntary audit regime, firms can signal higher quality financial statements by having them audited. In a mandatory audit regime, firms signal the quality of their financial statements by choosing an auditor who is perceived by investors to be of high quality. This results in a separating equilibrium because firms with lower financial statement quality find it more risky to be audited.6 The audit process increases the probability that the low quality of their financial statements would be detected, thus making them suffer the costly consequences of receiving a modified audit opinion. The firm would then trade-off the cost of auditing and the probability of qualified audit opinion, with the expected benefit from a clean opinion that results in its financial statements being perceived as being more reliable. A firm with reliable financial statements will find the trade-off worthwhile, but a firm whose financial statements are not reliable will not. Signaling and attestation effects, however, are interdependent. If the attestation effect is low, self-selection by firms into voluntary auditing becomes meaningless and signaling becomes ineffective. Alternatively, in a voluntary audit regime, if firms do not derive any signaling benefits from getting their financial statements audited, the demand for attestation will drop. The improvements in reliability brought about by both the attestation and signaling increase the weight that investors place on audited earnings relative to the weight they place on other information. Consistent with the valuation literature (Ohlson, 1995, Feltham and Ohlson, 1995), we assume that investors assess the value of the firm based on both the financial and nonfinancial information. Financial information is common across all investors and is disclosed in accordance with the generally accepted accounting principles. Disclosures that do not form part of financial statements are not necessarily disseminated to all investors. Therefore, non-financialstatement information could vary across investors. Placing a higher weight on the common financial statement information and relatively less weight on non-financial-statement information other words, gains and asset increases require a higher standard of verification than losses and liabilities (definition of conservatism in Watts, 2003) in the presence of an audit. Investors can therefore rely more on earnings that have undergone a higher standard of verification. 6
In reality, there are firms with a continuum of financial statement qualities. As a result, there is self-selection into auditing as well as into auditors of different perceived qualities. In the wake of Enron bankruptcy and subsequent revelations about Andersen, firms that had self-selected into Andersen based on audit quality started changing their auditors based on the changed perception. It is not realistic to believe that the actual quality of audit by Andersen changed immediately after the Enron bankruptcy – only the perception changed. See WSJ Online, March 31, 2002, “Saying Goodbye to Andersen” in url http://online.wsj.com/article/0,,SB1015604214815743120,00.html?mod=article-outset-box
7
will increase the “common information component” of valuation by different investors. As a result, the information divergence across investors should go down. This reduction should lead to a decrease in both stock return variance and investors’ trading volume. Conversely, if other variables are held constant, a decrease in the inter-investor information divergence (decrease in stock price variability and turnover) is consistent with increased weight being placed by investors on the common financial information and in turn, with increased investor confidence in audited financial statements. Our use of the stock volatility to measure inter-investor divergence is supported by extensive literature. Froot et al. (1992) argue that stock return volatility increases a firm’s perceived risk, thereby raising its cost of capital.
In discussing Bushee and Noe (2000),
Venkatachalam (2000) argues that when stock return volatility is high, investors demand to be compensated for both the systematic and unsystematic risks and consequently require higher returns (thereby increasing the cost of equity). In addition, he argues that increased stock price volatility might also increase the cost of debt. In effect, these studies suggest that the variability in stock returns is a good measure of systematic information risk faced by investors. Our second measure of information divergence across investors is the trading volume. This measure is also supported by a number of studies. Kim and Verrecchia (1991) use a twoperiod rational expectations model and show that the expected trading volume is positively associated with information divergence. Atiase and Bamber (1994) and Lobo and Tung (1997) provide empirical evidence that trading volume is associated with information divergence. This paper’s comparison of audited and non-audited observations in a voluntary audit regime is a direct test of the effects of auditing. The availability of a voluntary audit context in recent times is quite unique. The only other paper that uses a voluntary audit regime is Chow (1982). He takes advantage of a historical regulation in the US in 1926, prior to securities laws, when auditing was optional in public firms.
He studies the characteristics of firms that
voluntarily chose to have their financial statements audited, but he does not examine whether and how audited financial statements are different from non-audited ones. Other papers examine voluntary uses of auditor expertise in areas other than mandated auditing. For example, Givoly et al. (1978) focused on the review function (that is not mandated) and examined auditorreviewed and non-reviewed firms for systematic differences. They failed to document any difference, but found it difficult to make a definitive conclusion due to small sample and data
8
limitations. n a follow-up study Alford and Edmonds (1981) replicated the study by Givoly et al. (1978) and found consistent results.
As the scopes and the procedures of reviews are
substantially different from those of annual audits, the results from these two studies cannot be readily generalized to auditing. Another important feature of this paper is its focus on the Chinese context. The Chinese market has attracted increasing attention from accounting and auditing researchers. Chen et al. (1999) provide a descriptive analysis of the auditing requirements and environment in China. DeFond et al. (1999) present evidence that the frequency of modified audit opinions (MAOs) increased significantly after the adoption of the auditing standards in 1995, which was immediately followed by “flight from audit quality.” Chen et al. (2000) present empirical evidence on a negative market reaction to initial modified audit opinions in China. They also find that auditors are more likely to issue MAOs for regulation-induced earnings management. Haw et al. (2003) show that the timeliness of financial reporting is negatively associated with modified audit opinions. The reason as to why some firms undertake semi-annual audits while others do not is particularly intriguing in China because such semi-annual auditing, based on our investigations with local audit firms, typically costs 30 to 50 percent of annual audit fees. Moreover, the audit could lead to unfavorable opinions that are costly in terms of managerial reputation and regulatory scrutiny. For the firms that undertake voluntary semi-annual auditing, the expected benefits of auditing need to be higher than the expected cost that includes the possibility of negative consequences such as modified audit opinion. Even so, a large number of firms voluntarily opt for semi-annual audits. Understanding this self-selection is important to our study in two ways. It helps us in understanding the audit context in China and the motives for transparent and reliable disclosures. Second, we need to control for self-selection to isolate the effect of auditing on inter-investor divergence. In other words, we need to show that the factors which are inducing the firms to choose to be audited are not also leading to the results. In order to understand self-selection, we conducted a series of interviews with partners of audit firms and managers of listed firms for this project. We learnt that firms opted for semi-annual audits for several reasons. First, some firms wanted to improve their market image (signaling), which in turn could help in their future share issuance or business deals, such as strategic alliances and joint ventures.
9
Managers of a
Shanghai company told us, for example, that they were negotiating a joint venture with a multinational company and believed that a voluntary audit would make their company more transparent and attractive to the potential partner. Second, some firms selected semi-annual audits to make the annual audits less time consuming and more controllable. As each listed firm is assigned a date by the Stock Exchange to publish its annual report, it is important that they have the reports ready on time. A semi-annual audit can reduce the workload of the annual audit. The managers also suggested that this would be particularly useful if the auditor was small with limited resources and not one of the “Top 5” auditors in China. Third, external auditing is often employed to complement internal auditing, which is often not well developed. Fourth, we learnt that better performing firms that had significant increases in revenue and profits in the first half of the year, were likely to be audited voluntarily so that they could signal this information to the investment community. These interviews helped us understand the many different benefits of voluntary semi-annual audit and helped us identify a number of variables to be used in the probit model for estimating the inverse Mill’s ratio as will be discussed subsequently.
III. THEORETICAL DEVELOPMENT, RESEARCH QUESTION AND PROPOSITIONS The theoretical basis for the effect of auditing on returns’ variance (or stock prices’ variance) is obtained from the following sequence of reasoning that has been formally developed in the appendix: 1. In valuation decisions, each investor aggregates accounting information and non-accounting information.
Banker and Datar (1989) show that the relative weight placed on each
information component is proportional to its performance sensitivity and precision. 2. Independent auditors bring to bear their firm-specific knowledge, financial expertise and diligence to ensure compliance with the accounting standards. As a result, the audited accounting information will have less bias and greater precision. Better compliance with accounting standards also makes audited numbers more performance-sensitive compared to non-audited information. Investors expect these increases in performance-sensitivity and
10
accuracy and in turn, place a higher relative weight on audited information than on nonaudited information. This is shown in the first part of the appendix. 3. The accounting information, whether it is audited or not, is public and is therefore common to all investors. Non-accounting information can either be public (such as public disclosures of new product introductions, management changes, strategic initiatives etc.) or private (generated by the private insights of the analyst or the investor). Only the public part of nonaccounting information is common across investors and the private component is divergent across investors. In effect, when the investors place a greater relative weight on accounting information, they will be placing increased relative weight on common public information component vis-à-vis the private information component. Consequently, the inter-investor distribution of estimated stock price for the firm is more homogeneous (less variance). This is formally shown in the second part of the appendix. 4. When compared to non-audited firms, audited firms’ values are assessed more homogeneously across investors. This results in a smaller variability of stock returns and a lower trading volume for audited firms. Based on the above rationale, we examine whether the effect of auditing by comparing the variability of stock returns and trading volume between the audited and non-audited firms. Our hypotheses stated in alternate form are: H1: Audited firms exhibit a significantly lower variability in stock returns as compared with non-audited firms after the announcement of semi-annual reports. H2: Audited firms exhibit a significantly lower trading volume as compared with non-audited firms after the announcement of semi-annual reports.
11
V. THE SAMPLE, RESEARCH METHOD AND RESULTS Sample Selection We selected years 1997 to 2000 as our sample period because many observations had missing values before 1997 and quarterly financial reporting became mandatory after 2000. Table 1 summarizes the auditing status of listed firms during this period. Firms in China could either be restricted to domestic ownership (A shares) or could have both domestic and foreign ownership (A and B shares). Firms that are cross-listed in Hong Kong also issue H-shares to trade in Hong Kong. Firms issuing B or H-shares could be motivated to seek semi-annual voluntary audits for different reasons than those of A-share-only firms. For example, they could get their interim financial reports audited to attract foreign investors and to minimize their cost of capital. Therefore, we limited our sample to firms that issue only A-shares. Our sample was retrieved from the A-share file of the Taiwan Economic Journal database. A total of 3,679 firmyear observations were available for sampling. We excluded 396 non-A-share-only observations and 414 with missing values, and were left with 2,869 firm-year observations, of which 951 were audited and 1,918 were not audited. To examine the effect of voluntary auditing, we removed 101 observations of Special Treatment (ST) and Particular Treatment (PT) firms and 374 observations for firms with rights issues during the year where semi-annual auditing is mandatory. This filtering left us with a final sample of 2,394 firm-year observations, of which 684 observations were voluntarily audited.
[Insert Table 1 here]
Control for Self-selection Bias
12
We examine the effect of voluntary semi-annual auditing on the variance of stock returns and trading volume. However, firms with certain firm-level and contextual characteristics {Wi } may self-select to be audited. In other words, if the relevant characteristics can be represented as composite variable z, and z* is a threshold level of z, firms with z > z* are more likely to voluntarily choose audit than firms with z ≤ z.* Thus, the sample of audited firms selected for comparisons of post-announcement return variability and trading volume, is likely to be truncated in the region z ≤ z.* Likewise, the sample of non-audited firms is likely to be truncated in the region z>z.* The truncations in the two samples cause self-selection bias. To correct for this self-selection bias, we build a probit model to identify variables that are likely to elect semi-annual audits. We compute the Inverse Mills Ratio (IMR) from the probit model based on Heckman (1976) to correct for the bias caused by certain firms’ self-selection into the audited sample (Johnston and DiNardo, 1997; Heckman, 1976). The probit model is represented as Pr( z i = 1) = φ (γ 'Wi ) , where zi = 1(0) represents the range of the distribution in which z>z* (z ≤ z*) and (γ 'Wi ) represent the predicted value from the prediction model. From this probit
φ (γˆ ' wi ) model, the IMR for audited firms is estimated as λˆ = where the denominator is the Φ (γ ' wi ) cumulative probability distribution and wi is the vector Wi scaled by the standard deviation of the error term in the probit model. Similarly, the IMR for the non-audited firms is estimated as
λˆ =
φ (γˆ ' wi ) . Similarly to Heckman (1976), we use the IMR as an additional control 1 − Φ (γ ' wi )
variable in the models that compare the effects of audited and non-audited firms, to control the bias that results from the truncation. We discuss below the development of the probit model.
13
Variables used in empirical choice models could be based on the results of earlier empirical tests or on analytical models (Palepu, 1986). While there is no published study that models the choice between getting audited and not getting audited, there are two prior studies are helpful in the choice of the variables. The first study is Francis et al. (1999) that examines the choice of a Big 6 auditor in the US by firms that need to signal better financial statements. Signaling by voluntary choice of auditing is somewhat similar to signaling by voluntary choice of a high quality auditor and to that extent, we could rely on their findings. However, because their study is conducted in the US context, we also seek to identify choice variables that are more reflective of the Chinese context. For this, we rely on Chen et al. (2001) who find that earnings management incentives in China might motivate voluntary audit decisions. In addition to these two studies, Chow (1982) and Ettredge et al, (1994) provide additional guidance in the choice of variables. In order to gain additional insights in our particular context, we also interviewed the managers of listed firms who had the choice to be audited and chose either to be audited or not; and auditors who audited some of those firms. 7
Based on both the prior studies and our
interviews, we developed the following probit model to control for self-selection: Pr( zit = 1)it = γ 0 + γ 1OPCYCLEit −1 + γ 2CAPINTit −1 + γ 3 Sizeit −1 + γ 4 Leverageit −1 + γ 5 PEit −1 + γ 6 ROAit −1 + γ 7 Lossit + γ 8Top5it −1 + γ 9TACCRit −1 + γ 10 SalesGrwthit −1 + γ 11`Betait −1 + γ 12 Nontradeit −1 + γ 13 y98 + γ 14 y99 21
+ γ 15 y 00∑ γ 16 k INDik + uit k =1
(1) 7
Our interviews of partners of audit firms and managers of listed firms yielded significant insights: (i) improvement in the firms’ market image (signaling) that could help in their future share issuance, strategic alliances and joint ventures – For example, managers of a Shanghai company told us, for example, that they were negotiating a joint venture with a multinational company and believed that a voluntary audit would make their company appear more transparent and attractive to the potential partner; (ii) reducing the annual audit time so as to meet the deadline for publishing the annual report before the deadline assigned by the Stock Exchange; (iii) reduction of the workload of the annual audit (particularly small auditors with limited resources not classified among the “Top 5” auditors in China are stretched to the limit during annual audits and welcome an opportunity to smooth the workload over the year); (iv) complementing internal audits that are often not well developed; and (v) signaling of better performance by firms with significant increase in revenue and profits during the first half of the year.
14
We give below the definitions and then discuss the rationale for selection of the variables in the above model. Definitions: OPCYCLE = Operating Cycle: [365*(average inventory/cost of goods sold) + 365*(average accounts receivable/sales)]/30; CAPINT = Capital Intensity: Gross PP&E /sales Size: natural logarithm of total asset Leverage: total long-term debt to total asset ratio PE = P/E Ratio: Stock price over EPS ROA = semi-annual net income over beginning total asset Loss: 1 for net income less than 0 and 0 otherwise TACCR = Total Accrual: annual total accrual Top5 = Top 5 Auditor: 1 if the auditor is among the top 5 in China (by market share) and 0 otherwise SalesGrwth = Sales growth: (sales in year t – sales in year t-1)/sales in year t-1 Beta: Beta estimated by the market model Nontrade: Percentage of non-tradable shares outstanding y98, y99 y00, indicator variables for years 1998, 1999 and 2000, respectively IND: Twenty-one Industry dummies based on Chinese industry classification i: firm indicator t: interim period indicator, t- 1 for beginning of the year The inclusion of OPCYCLE, CAPINT, Size, Leverage, PE and Loss in the probit model is based on Francis et al. (1999). Firms with longer operating cycles need to develop estimates over a longer time horizon and have more complex accounting procedures to determine
15
appropriate accruals. This increases skepticism among the investors, which, in turn increases the need felt by the firm for sending a positive audit signal to the market. Accordingly, we expect firms with longer operating cycles to opt more frequently for voluntary auditing. Firms with high capital intensity (defined as the gross property, plant and equipment divided by sales) have relatively high depreciation, and their managers can choose the depreciation method as well as the estimated useful asset lives. Here again, auditing is likely to improve the credibility of reported earnings and asset values.8 We include firm size in our model to control for firms’ difference in innate credibility and their information environment. We expect large firms to show less need for semi-annual auditing, ceteris paribus, since their financial reports are more likely than the financial reports of smaller firms to be scrutinized both by investors and regulators. As a result, their financial reports are generally perceived to be more reliable. As a firm’s debt level increases, its debt holders may need to monitor its management team. Therefore, firms with high leverage ratios are more likely to employ semi-annual audits. Firms with low Price Earnings (PE) ratios are often undervalued. Managers of these firms are more likely to resort to external auditing in their attempts to communicate to investors that their firms are good investment opportunities. Thus, we expect firms with lower PE ratios to opt more frequently for auditing. The selection of four of the variables, namely ROA, Top5, SalesGrwth, and Loss were based on knowledge obtained through our interviews with partners of audit firms and managers of listed companies. ROA is the ratio of the semi-annual period income over the previous yearend’s total asset. Some partners suggested that firms that do well in the first part of the year get
8
As our test context is different that of Francis et al. (1999), who employed operating cycle and capital intensity to examine Big 6 auditors’ role in the credible reporting of accruals, we do not expect all variables adopted from their model to affect the choice of semi-annual auditing in the same that they affect the choice of Big 6 auditors.
16
audited to signal the good news to the market. Based on this rationale, we expect the audited firms to have a significantly higher ROA than non-audited firms. Large (Top5) auditors are more independent, have high reputation and are more likely to issue modified audit opinions (DeFond et al., 1999; DeFond et al., 2002; Ashton and Kennedy, 2002). In anticipation of being held to higher standards by the Top 5 auditors, firms might be less willing to be voluntarily audited by them. Another factor of interest is the workload. Non-Top5 auditors usually have limited resources that are stretched to the limit during annual audits and might encourage their clients to opt for semi-annual audits so as to smooth out their workload. At the same time, the voluntary audit choice signal will be even more powerful and the benefits might be seen to be higher if a Top5 auditor is chosen. Therefore, we do not predict a sign on this variable but recognize that it is an important control variable. Firms without sales growth or with losses reported in the most recent fiscal period will be less willing to have their semi-annual reports audited. Furthermore, we expect high-risk firms (those with high accruals and high beta values) to assess the overall costs to be higher than the incremental benefits of auditing; but we expect low risk firms to be more willing to get audited, in general. A variable that is particular to the Chinese context is the percentage of outstanding non-tradable shares which proxies for government control of the firm. Usually, managers in government-controlled firms have less need to communicate with investors, as these firms depend less on the market for finance and receive government protection from regulators and investors. Therefore, we expect managers of firms with more non-tradable shares to show a lower propensity to have their semi-annual reports audited. In our model, we employ an indicator variable for each year and each industry to control for industry and year effects.
17
In order to construct a parsimonious model, we exclude variables that are trivial in our sample or not reported to be significant in prior studies. For example, the proportion of common stock owned by officers and directors is not included because both the mean and median values of this variable in our sample is so low that it is not expected to affect the audit choice decision. The ratio of inventory over total assets and the number of reportable business segments are not included, because Ettredge et al (1994) did not find them to be significant. Moreover, the ratio of receivables over total assets is dropped because total accruals in our model can capture its effect. We report the probit model results (other than the industry dummy variable results) in Table 2. The results are generally consistent with our expectations. They show that the decision for semi-annual auditing is negatively associated with PE ratio, loss reported in the previous year, Top5, risk (Beta) and percentage of non-tradable shares outstanding. Leverage, profitability (ROA) and sales growth are positively associated with choice for semi-annual auditing. Firm size has a negative coefficient but is not statistically significant, and the likelihood ratio is very significant, which indicates that the probit model effectively differentiates between audited and non-audited observations.
[Insert Table 2 here]
The Heckman (1976) correction for self-selection bias is an appropriate method to use in this particular context for the following reasons: First, the method is robust in cases where the two sets of variables are the same (one used for the probit model and the other to determine the effect of audit on information divergence). Johnston and DiNardo (1997) argue that this
18
correction is less sensitive to normality assumptions when these two sets of variables differ. In this study, the variables that affect the outcome include the variance of the returns prior to the announcement and other variables that differ from variables used in the probit model. This makes the IMR method less sensitive to normality assumptions. Second, in most situations, it is difficult to find variables that affect probability but do not factor in the equation that tests the differences (Johnston and DiNardo, 1997). In our study, we use a number of variables that affect stock market variability and trading volume but do not necessarily predict the choice of voluntary auditing. For example, random arrival of value relevant information may affect both return variability and trading volume. However, it is not expected to affect the choice for semiannual audit9.
Empirical Models We focus on the variability of returns and average daily trading volume to examine the effect of auditing on the improvement of information convergence. We use the following model to compare the standard deviations of the risk-adjusted abnormal returns between audited and non-audited firms, around the earnings announcement dates:
v post = α 0 + α 1 Audit + α 2 v pre + α 3 Size + α 4 v annual + α 5 IMR + α 6 ABS _ CAR + α 7 y98 + α 8 y99 + β 9 00 + ε (2)
where v post = standard deviation of firm’s returns after semi-annual audit
Audit = 1 if audited and 0 if not audited 9
Larcker and Rusticus (2005) show limitations of using instrumental variables in accounting research. As an exercise of caution in interpreting our results, extensive robustness checks are performed and discussed in a subsequent section.
19
v pre = standard deviation of firm’s returns before semi-annual audit
vannual = standard deviation of firm’s returns after announcement of annual earnings made prior to each semi-annual audit Size = natural logarithm of equity’s beginning market value IMR = Inverse Mills Ratio from the probit model ABS_CAR = absolute value of risk-adjusted cumulative abnormal return over the post-announcement period
y98, y99, y00 = indicator variables for the years 1998, 1999 and 2000, respectively A significant negative coefficient on the indicator variable, Audit, indicates that audited semi-annual financial statements are associated with less return variability than non-audited firms. In addition, we employ the following model to examine the effects of auditing on average daily trading volume around the three windows defined earlier: TV
post
= β 0 + β 1 Audit + β 2 TV
pre
+ β 3 Size + β 4 TV
annual
+ β 5 IMR
+ β 6 MTV + β 7 ABS _ CAR + β 8 y 98 + β 9 y 99 + β 10 y 00 + ε
(3)
Where TVpost = Average daily trading volume after semi-annual announcement Audit = 1 if audited and 0 if not audited TVpre = Average daily trading volume before semi-annual announcement Size = natural logarithm of equity’s beginning market value TVannual = Average daily trading volume after annual announcement IMR = Inverse Mills ratio from the probit model MTV = Average daily market trading volume ABS_CAR = absolute value of risk-adjusted cumulative abnormal return over the post-announcement period
20
y98, y99, y00 = indicator variables for the years 1998, 1999 and 2000, respectively
Measures of Information Convergence Our first measure of information convergence is post-semi-annual-announcement returns variability and the second is post-semi-annual-announcement trading volume.
Post-
announcement return variability is measured by the standard deviation of risk-adjusted10 daily abnormal returns over three different time windows after the semi-annual earnings announcement dates (+1 to +7, +1 to + 15 and +1 to +30). Likewise, pre-announcement return is measured by the standard deviation of risk-adjusted abnormal returns over three different time windows before the semi-annual earnings announcement dates (-7 to -1, -15 to -1 and -1 to -30). The announcement date is excluded from both pre- and post-announcement periods.
We
measure trading volume as the average daily percentage of outstanding shares traded for a given firm. The market-wide trading volume is the average daily total number of all trades divided by the total number of all outstanding shares for the stock exchange.
Control Variables for the Return Variability Analysis We control for market size (natural logarithm of the market value of equity at t = -30) because larger firms resemble diversified portfolios and consequently have lower return variance. The pre-announcement standard deviation of returns (respectively over windows -7 to -1, -15 to 1 and -30 to -1) is a control for other firm-specific factors that affect the variability of returns. It also captures the level of pre-announcement information asymmetry among investors as Atiase
10
We compute the beta of the stocks using the price data over prior five years and use the market model to compute the risk-adjusted returns. There is some concern on whether the reasonableness of market model in China and other emerging markets. We have repeated all the tests with market-adjusted and raw return data and find similar results in all the analyses. This analysis is also provided in the additional analysis section.
21
and Bamber (1994) find it to be positively related with trading volume reaction to announcement of accounting information.
Additionally, Atiase and Bamber (1994) also find that trading
volume reaction is positively associated with the absolute value of cumulative abnormal returns during the announcement period. Therefore, ABS_CAR is included to control for this effect. As a further control for firm-specific factors, we control for the post-annual-announcement return variability of the previous year. We control for year-specific effects by year dummies. Finally, we control for the self-selection bias by including the IMR from the probit model as an additional control variable.
Control Variables for the Trading Volume Analysis We control for the firm-specific effects by including the natural logarithm of the market value of equity, the pre-announcement trading volume and the post-annual announcement trading volume in the regression. We use market-wide average daily trading volume to control for the market trading activity’s intensity effect on the trading volume of the firm. We employ TVpre and ABS_CAR to control for the effect of the positive association between these two variables and trading volume reaction to disclosure of accounting information as reported in Atisae and Bamber (1994). Finally, we use the years’ indicator variables, and include the IMR to control for the fixed effects and self-selection bias, respectively.
Univariate Analysis Comparison of Variables prior to and after Semi-Annual Announcement Panel A of Table 3 gives the descriptive statistics of the variables v pre , v post , vannual (standard deviations of firm returns before and after semi-annual earnings announcement, and
22
after the previous annual announcement, respectively) and market size. The returns variability is typically higher for non-audited firms than for audited firms after (but not before) the announcement of semi-annual reports. The absolute value of cumulative abnormal returns is significantly (p γ 1u and γ 2 a < γ 2u . In effect, audited financial statement numbers are weighted more than non-audited ones relative to the weighting of non-accounting information. Step 2: Auditing reduces the variance in the stock valuation by investors Heretofore, we have focused on one investor. We will now examine the divergence among investors. The valuation of the stock by the ith investor is given by (2):
Vi = γ 1 z R + γ 2 u i The first term is common to all investors. The second term consists of non-accounting information, which could be different for different investors. When we take the variance of Vi across investors, we have
Variance (Vi ) = γ 2 Variance (u i ) (8) From step 1, we know that γ 2 a < γ 2u . Therefore, from (8) we see that the expected variance of 2
the stock values perceived by investors is less for audited firms than for non-audited firms, ceteris paribus. Further, for a given market price of the previous period, (Pt-1), the expected return on the stock is given by (Vi-Pt-1)/Pt-1. The expected return will be equal to the market return. The expected variance of 2
the market return will be equal to [Variance (Vi )] / Pt −1 . Therefore, we expect audited firms to have a lower variance of market returns relative to non-audited firms. The differences in valuation by different investors also lead to a greater trading volume. Therefore, after we control for other determinants of trade volume, we expect the trade volume for audited firms to be less than the expected trade volume for nonaudited firms.
31
REFERENCES Alford, M. R. and Edmonds, T. P. (1981) “A Replication: Does Auditor Involvement Affect the Quality of Semi-Annual Report Numbers?” Journal of Accounting, Auditing and Finance, 4: 255-264. Ashton, R. and Kennedy, J. (2002) “Eliminating Recency with Self-Review: The Case of Auditors’ ‘Going Concern’ Judgments”, Journal of Behavioral Decision Making, 15(3): 221- 231. Atiase, R. K., and Bamber, L. S. (1997) “Trading Volume Reactions to Annual Accounting Earnings Announcements”, Journal of Accounting and Economics, 17(3): 309-329. Bamber, E. M., and Stratton, R. A. (1997) “The Information Content of the UncertaintyModified Audit Report: Evidence from Bank Loan Officers”, Accounting Horizons, 11(2): 1-11. Banker, R. D., and Datar, S. M. (1989) “Sensitivity, Precision, and Linear Aggregation of Signals for Performance Evaluation”, Journal of Accounting Research, 27(1): 21-39. Bar-Yosef, S., and Sarath, B. (2005) “Auditor Size, Market Segmentation and Litigation Patterns: A Theoretical Analysis”, Review of Accounting Studies, 10(1) : 59-92. Becker, C. L., DeFond, M. and Subramanyam, K. R. (1998) “The Effect of Audit Quality on Earnings Management”, Contemporary Accounting Research, 15(1): 1-24. Blackwell, D. W., Noland, T. R. and Winters, D. B. (1998) “The Value of Auditor Assurance: Evidence from Loan Pricing”, Journal of Accounting Research, 36: 57-70. Brennan, M. and Subrahmanyam, A. (1996) “Market Microstructure and Asset Pricing: On the Compensation for Illiquidity in Stock Returns”, Journal of Financial Economics, 41(3): 441-464. Brockman, P. and Chung, D. Y. (2003) “The Inter-Temporal Behavior of Informed and Uninformed Traders”, Review of Quantitative Finance and Accounting, 21(3): 251-265. Brown, L. D., and A. S. Pinello. 2007. To What Extent Does the Financial Reporting Process Curb Earnings Surprise Games? Journal of Accounting Research December. Chen, J. P., Chen, S. and Su, X. (2001) “Profitability Regulation, Earnings Management and Modified Audit Opinions: Evidence from China”, Auditing: Journal of Practice & Theory, 20(2): 9-30. Chen, K. C. W., and Yuan, H. Q. (2004) “Earnings Management and Resource Misallocation: Evidence from China’s Accounting-based Regulation of Rights Issue”, The Accounting Review, 79(3): 645-665. 32
Chow, C. W. (1982) “The Demand for External Auditing: Size, Debt and Ownership Influences”, Accounting Review, 57: 272-291. China Securities Regulatory Commission Shenzhen Office, CSRC Shenzhen (2000) The Latest Collection of Securities Laws & Regulations, Guangzhou: Guangdong Economics Publications. Choi, S. K., and Jeter, D. C. (1992) “The Effect of Qualified Audit Opinion on Earnings Response Coefficients”, Journal of Accounting and Economics, 14: 229-247. DeFond, M. L., Raghunandan, K. and Subramanyam, K. R. (2002) “Do Non-Audit Service Fees Impair Auditor Independence? Evidence from Going Concern Audit Opinions”, Journal of Accounting Research, 40(4): 1247-1274. DeFond, M. L., Wong, T. J. and Li, S. H. (1999) “The Impact of Improved Auditor Independence on Audit Market Concentration in China”, Journal of Accounting and Economics, 28: 269-305. Easley, D. and O’Hara, M. (2004) “Information and the Cost of Capital”, The Journal of Finance, 59(4): 1553-1583. Easton, P. D. (1999) “Security Returns and the Value Relevance of Accounting Data”, Accounting Horizons, 13(4): 399-412. Epstein, M., and Pava, M. (1993) The Shareholders’ Use of Annual Reports, Connecticut: JAI Press. Ettredge, M. L., D. T. Simon, D. B. Smith and M. S. Stone (2000) “The Effect of the External Accountant’s Review on the Timing of Adjustments to Quarterly Earnings”, Journal of Accounting Research, 38(1): 195-207. Francis, J. R., Maydew, E. L. and Sparks, H. C. (1999) “The Role of Big 6 Auditors in the Credible Reporting of Accruals”, Auditing: A Journal of Practice and Theory, 18: 17-34. Givoly, D., Ronen, J. and Schiff, A. (1978) “Does Auditor Involvement Affect the Quality of Semi-Annual Report Numbers?” Journal of Accounting, Auditing and Finance, 1: 361-372. Goel, A. M. and Thakor, A. V. (2003) “Why Do Firms Smooth Earnings?” The Journal of Business, 76(1): 151-192. Heckman, J. (1976) “The Common Structure of Statistical Models of Truncation, Sample Selection, and Limited Dependent Variables and a Simple Estimator for such Models”, Annals of Economic and Social Measurement, 5: 475-492.
33
Hirst, D., Koonce, L. and Miller, J. (1999) “The Joint Effect of Management’s Prior Forecast Accuracy and the Form of its Financial Forecasts on Investor Judgment”, Journal of Accounting Research, 37: 101-124. Hirst, D., Koonce, L. and Simko, P. (1995) “Investor Reaction to Financial Analysts’ Research Reports”, Journal of Accounting Research, 33(2): 335-351. Johnson, D., Panny, K. and White, R. (1983) “Audit Reports and Loan Decision: Actions and Perceptions”, Auditing: A Journal of Practice and Theory, Spring: 38-51. Johnston, J. and DiNardo, J. (1997) Econometric Methods, New York: McGraw Hill Companies. Keshk, O. M. G. (2003) “CDSIMEQ: A Program to Implement Two-Stage Probit Least Squares”, The Stata Journal, 3: 1-11. Kim, O., and Verrecchia, R. E. (1991) “Trading Volume and Price Reactions to Public Announcements”, Journal of Accounting Research, 29(2): 302-321. Larcker, D and T. O. Rusticus (2005) “On the Use of Instrumental Variables in Accounting Research” working paper. University of Pennsylvania. Lennox, C. S. (1999) “Audit Quality and Auditor Size: An Evaluation of Reputation and Deep Pockets Hypotheses”, Journal of Business Finance & Accounting, 26: 779-805. Libby, R. (1979) “Bankers’ and Auditors’ Perceptions of the Message Communicated by the Audit Report”, Journal of Accounting Research, 17: 99-122. Lobo, G. J., and Tung, S. (1997) “Relation Between Predisclosure Information Asymmetry and Trading Volume Reaction Around Quarterly Earnings Announcements”. Journal of Business Finance and Accounting, 24(6): 851-867. Maines, L. (1996) “An Experimental Examination of Subjective Forecast Combination”, International Journal of Forecasting, 12: 223-233. Manry, D., Tiras, S. L. and Wheatley, C. M. (2003) “The Influence of Semi-Annual Auditor Reviews on the Association of Returns and Earnings”, The Accounting Review, 78(1): 251274. Ohlson, J. A. (1995) “Earnings, Book Values, and Dividends in Equity Valuation”, Contemporary Accounting Research, 11(2): 661-687. Palepu, K. G. (1986) “Predicting Takeover Targets”, Journal of Accounting and Economics, 8(1): 3-35. Pany, K. and Smith, C. (1982) “Auditor Association with Quarterly Financial Information: An Empirical Test”, Journal of Accounting Research, 20(2): 472-481.
34
Reckers, P. and Panny, K. (1979) “Quarterly Statement Reliability and Auditor Association”, Journal of Accountancy, 47: 97-100. Shillinglaw, G. 1961. “Concepts underlying interim financial statements”, The Accounting Review, 36(2): 222-231. Teoh, S. H. and Wong, T. J. (1993) “Perceived Auditor Quality and the Earnings Response Coefficient”, Accounting Review, 68(2): 346-366. Watts, R. L. (2003) “Conservatism in Accounting Part I: Explanations and implications”, Accounting Horizons, 17(3): 207-221. Winters, A. (1975) “Banker Perceptions of Unaudited Financial Statements”, The CPA Journal, 45: 29-34.
35
Table 1 Auditing Status of Listed A-Share Firms and Sample Selection Results *: A firm is publicly labeled as a Special Treatment (ST) firm if it has reported losses for two consecutive years, or when its net asset per share falls below par value. If an ST company continues to report losses in the third year, its label will change to Particular Treatment (PT) and its shares will be traded only once a week, on Fridays. All ST and PT firms are required to have their semi-annual reports audited. **: Firms must have their semi-annual reports audited if they plan to issue rights in the second half of the year. Firms that issue rights in the first half of the year do not have to be audited.
No. of listed firms (Non-A-Share-only firms) No. of A-Share firms (with missing values) Firms available for sampling Non-audited Audited
(PT or ST)* Non-audited Audited
(Rights offerings)** Non-audited Audited Mandatory audit Non-audited sample Voluntary audit sample Total Sample
1997 746 (92)
1998 880 (96)
1999 973 (100)
2000 1,080 (108)
Total 3,679 (396)
654 (164)
784 (96)
873 (72)
972 (82)
3,283 (414)
490 322 168
688 450 238
801 570 231
890 576 314
2,869 1,918 951
(0) 0 0
(21) 0 21
(380) 2 36
(420) 5 37
(101) 7 94
(83) 51 32
(85) 49 36
(99) 51 48
(107) 50 57
(374) 201 173
32
57
84
62
235
271 136 407
401 181 582
517 147 664
521 220 741
1710 684 2,394
36
Table 2 Probit Regression Results Dependent variable is an audit choice indicator: 1 for audited interim financial statements and 0 otherwise. All independent variables are measured at the beginning of the year except ROA which is semi-annul net income over total assets at the last year end. OPCYCLE: 365*(average inventory/cost of goods sold) + 365*(average accounts receivable/sales)/30; CPINT: Gross PP&E /sales; Size: Natural logarithm of total assets; Leverage: Total debt to total asset ratio; PE: Stock price over EPS; Loss: 1 for net income less than 0 and 0 otherwise; TACCR: Annual total accrual; Top5: 1 if the auditor is among the top 5 in China and 0 otherwise; SalesGrwth: (sales in year t – sales in year t-1)/sales in year t-1; Beta: Beta estimated by the Market Model; Nontrade: Percentage of non-tradable shares outstanding; y98, y99, y00: Indicator variables for years 1998, 1999 and 2000, respectively. *, **, ***: Significant at 10%, 5% and 1%, respectively, two-tailed
Intercept OPCYCLE CAPINT Size Leverage PE ROA Loss Top5 TACCR SalesGrwth Beta Nontrade y98 y99 y00
Wald Estimate Chi-Square 0.908 5.41** 0.001 0.04 -0.074 2.76* -0.045 1.78 0.741 12.07*** -0.112 17.93*** 6.379 39.84*** -0.700 17.48*** -0.413 5.12** -0.057 0.13 0.088 3.53* -0.583 32.63*** -1.144 26.37*** -0.121 1.85 -0.422 22.08*** -0.170 3.76*
(21 Industry indicator variables not tabulated) Likelihood ratio test: 1,288 N = 1,710 for non-audited group and 684 for audited group
37
Table 3 Descriptive Statistics of Variables Used in the Variance and Trading volume Models Vpost: Standard deviation of risk-adjusted abnormal turns after semi-annual announcement (0 < t < 8, 0 < t
