Are non-audit services associated with firm value? Evidence from financial information system-related services
We acknowledge helpful comments from two anonymous referees, Jere Francis, Simon Fung, Bill Messier, Dan Simunic, and seminar participants at City University of Hong Kong and The Hong Kong Polytechnic University. We also thank Robert Faff (the Editor). An earlier version of this paper titled, ‘The Association between Financial Information System Development and Implementation Costs and Firm Value: Further Evidence of Payoff from IT Investments’ was presented at the 2002 Annual Meeting of the American Accounting Association. We appreciate helpful comments from the participants and two anonymous annual meeting reviewers. We are grateful to the Accounting & Corporate Governance Centre, City University of Hong Kong, for financial assistance.
Abstract
The purchase of non-audit services from incumbent auditors has generated considerable attention. Surprisingly, limited empirical evidence exists on the association of non-audit services with firm value. Focusing on services related to financial information system (FIS), we find that the market value of equity is greater for firms that purchase FIS-related services from their incumbent auditors relative to firms that do not. The levels of FIS fees are also positively related to firm value after controlling for total other fees, or total other non-audit fees. Hence, despite the negative perception associated with non-audit services, investors regard FIS-related services as value-adding activities.
1. Introduction
The provision and the associated consequences of non-audit services by incumbent auditors to their clients have received considerable attention from regulators, auditors, investors, media and academic researchers. Much of the extant research investigating the consequences of incumbent auditors providing non-audit services has focused more on the negative consequences of such services, such as lower earnings quality, higher frequency of restatement of financial statements, lower propensity to issue going-concern opinions or lower reliability of reported earnings than the potential benefits (see DeFond et al., 2002; Firth, 2002; Frankel et al., 2002; Ashbaugh et al., 2003; Chung and Kallapur, 2003; Kinney et al., 2004; Reynolds et al., 2004; Krishnan et al., 2005; Gul et al., 2006). Surprisingly, limited empirical evidence exists on whether such services increase firm value for the investors. A focus on investors is important because investors, particularly equity investors, are major participants of the capital markets and primary users of financial statements. If they consider that the provision of non-audit services has overall net benefits to the firms, then the importance of auditor-provided non-audit services could not be overlooked.
The objective of this study is to provide empirical evidence on the linkage between market value of equity and a particular non-audit service – designing and developing a financial information system (FIS). We focus on this particular non-audit service for two reasons. First, it is a major type of non-audit service1 that was initially singled out to be disclosed, and subsequently been banned by the Securities and Exchange Commission (SEC, 2000).2 This move by the SEC is consistent with the notion that even provision of an immaterial amount of FIS-related services impairs auditor independence. Therefore, FIS services deserve attention on their own among the non-audit services. Second, despite the concern that an increasing amount of FIS fees creates the appearance of economic bond between the auditees and the auditors and the expectation that the auditor might tolerate greater earnings management contributing to a lower quality of earnings (DeAngelo, 1981), there are advantages to clients who employ their incumbent auditors for providing FIS-related services. FIS fees proxy for tangible and intangible benefits associated with a FIS – enhanced productivity and profitability in the form of reduced inventory, reduced headcount in accounting, improved business processes, and strategic planning capability – all of which can enhance firm value.3 Furthermore, there are potential benefits of knowledge spillover of auditor-provided non-audit services whereby knowledge gained in the provision of non-audit services could be beneficial to audit work or vice versa (Simunic, 1984). Therefore, FIS-related services provided by the incumbent auditors offer an interesting context to empirically examine two related issues concerning non-audit services. First, whether the provision of FIS-related services is valued by investors? Second, whether higher levels of FIS-related services increase firm value despite rising threat to auditor independence (economic bond)?
Our investigation has policy implications not only for the USA but also for other countries that are considering banning auditor-provided FIS services. The position of the SEC in banning the provision of FIS services is very strict since it represents an all-or-nothing approach. Such a position is reflected in the Sarbanes–Oxley Act of 2002 which prohibits the provision of more types of non-audit services. However, whether the provision of even an immaterial amount of FIS-related services impairs auditor independence is an empirical question. In addition, the business value of information technology (IT) is as important in other countries as in the USA. The benefits of FIS services are independent of jurisdiction. If auditor-provided FIS services increase firm value despite the concern for auditor independence in the USA, the same might also be true in other countries like Australia. This would mean that other countries like Australia should be cautious in adopting the all-or-nothing approach in regulating non-audit services.
We examine the first issue by using a matched sample of firms. We select firms that buy FIS-related services from the incumbent auditors in year 2000 or 2001. Then, we match these firms with a control group of firms that did not purchase such services by year, two-digit SIC code, and total assets at the beginning of the year. We also control for self-selection and endogeneity issues. The results show that market value of equity is positively and significantly associated with the purchase of FIS-related services from the auditors. Therefore, the purchase of FIS services from incumbent auditors enhances firm value.
To examine the second issue, we use only those firms that report non-zero FIS fees paid to their auditors in year 2000 or 2001. These firms are not matched with firms that did not purchase FIS services from their auditors. Our results show that FIS fees are positively and significantly associated with market value of equity after controlling for other fees (total fees, or total non-audit fees, less FIS fees). Hence, despite rising threat to auditor independence at higher levels of FIS fees, our findings suggest that non-audit services pertaining to FIS are value-adding activities.
The contributions of this study are in two areas. First is the growing literature on the consequences of non-audit services provided by the incumbent auditors to which we contribute by providing empirical evidence on the linkage between firm value and non-audit services relating to FIS. This study is the first study that empirically documents that FIS-related services provided by the incumbent auditors are value-adding activities. While the results of Krishnan et al. (2005) and Gul et al. (2006) suggest that investors discount the reliability of reported earnings for firms with high levels of non-audit services over the short term, the results of the present paper suggest that firm value does increase for auditor-provided FIS services despite the concern for auditor independence. Note that concern for independence impairment is one of those factors that could affect firm value and the investors will take into account other factors in assessing auditor-provided FIS-related services (e.g. the value of FIS and the effects of knowledge spillover). Hence, our findings extend the results of Krishnan et al. (2005) and Gul et al. (2006).
Second is the literature on the business value of IT. In a recent review of research on returns on investments in IT, Dehning and Richardson (2002, p. 27) state, ‘never before has IT played such an important role in the existence of companies, yet the overall impact of IT on performance remains a largely unexplained puzzle. We see this as an opportunity for accounting researchers to contribute to both academic research and current business practice.’ Prior research (e.g. Hayes et al., 2001) on the business value of IT focuses on large IT systems, such as Enterprise Resource Planning (ERP) systems, the spending on which is capitalized in the financial statements. We examine whether investors value FIS fees paid to the auditors, a revenue expenditure as opposed to the large, capital expenditures, such as installation of ERP systems. Our findings suggest that even though FIS fees paid to the auditor are expensed under Generally Accepted Accounting Principles, such fees are favourably associated with market value of equity, a popular measure of firm performance.
The rest of the present paper is organized as follows. The next section describes the benefits and concerns of having auditor-provided FIS, and states the research questions. Section 3 develops the research design, the empirical model and describes the sample selection process. Results are in Section 4. Section 5 presents the conclusions.
2. Background
2.1. Benefits of an FIS
A well-developed FIS is a prerequisite for obtaining timely feedback on operating performance and for making corrective actions to preserve a firm's competitiveness. Jablonsky (2001) illustrates how Cisco Systems has developed a state-of-the-art FIS, dubbed ‘the virtual finance organization’, that enables Cisco to produce critical financial information on an hourly and daily basis. In the past, it took 14 working days to close its books. Today, the hard close is completed in less than 24 hours. Furthermore, by automating routine, repetitive tasks, financial professionals at the corporate level are being freed-up to work on more value-adding activities. The bottom line is productivity enhancement.
Another example of potential payoffs from implementing an FIS relates to the adoption of ERP systems. ERP is a suite of software applications designed to process an organization's transactions and facilitate integrated and real-time planning, production and customer response (O’Leary, 2000). Hayes et al. (2001) note that benefits from adopting ERP systems include enhanced productivity and profitability in the form of reduction in inventory, reduced headcount in accounting function, increased efficiency and effectiveness due to reengineered and automated processes, improved customer service, strategic planning capability and effective decision-making. Hayes et al. (2001) find that the stock market reacts favourably to ERP implementation announcements. This finding is consistent with the notion that the market views adoption of ERP systems as value-increasing investment in information technology.
2.2. Auditor-provided FIS services
When auditors provide FIS-related services to the clients, there are additional benefits to the clients. First, auditors know the needs and operation of the clients and, hence, could design a better FIS or design such a system more expeditiously. Clients would also benefit from the savings in terms of faster information processing and operational efficiency. Second, an auditor-provided FIS is more likely to meet the audit needs and objectives of audit and financial reporting. Auditors would save resources in tracing transactions and finding appropriate audit evidence from the systems they design.4 Overall, both the clients and their auditors gain if auditors provide FIS services to the clients. The literature calls these effects ‘knowledge spillover’ (see SimuNic, 1984).
However, auditor independence might be impaired if auditors also provide FIS-related services to the audit clients. Will the auditors be more (or less) likely to report detected material misstatements associated with clients for whom they design FIS? If investors are concerned about impairment of auditor independence, then auditor-provided FIS services might not be viewed as value-adding activities by investors. In other words, whether auditor-provided FIS services increase firm value is an empirical question. We examine this issue by testing whether the market value of equity is greater for firms that buy FIS-related services from their auditors relative to firms that do not buy such services. A follow-up question is whether the relation between investments in FIS-related services (proxied by FIS fees) and firm value hold in the presence of economic bond? In other words, do investors discount the value of FIS-related services due to concern over reduced auditor independence? Formally, the research questions in this paper are:
Research Question 1: Does the provision of FIS services by auditors increase firm value?
Research Question 2: Is firm value higher for firms that pay a higher amount of fees for FIS services to their auditors?
2.3. Prior studies
To our knowledge, there is no study that focuses on FIS fees paid to auditors, although there are studies that examine non-audit services in general. Nevertheless, as FIS-related service is one type of non-audit services, we draw on prior studies for discussion. Frankel et al. (2002), using accruals as a measure of earnings management, report that firms with high non-audit services fees are likely to have more earnings management. Hence, high non-audit services fees are associated with lower auditor independence. However, other studies (e.g. Ashbaugh et al., 2003; Chung and Kallapur, 2003) find no evidence of earnings management and, hence, loss of auditor independence using other measures of accruals. Using the issuance of going-concern opinion, DeFond et al. (2002) also do not find evidence of lower auditor independence for firms with high non-audit services fees. Hence, the majority of the evidence in the literature suggests that there is no loss of auditor independence in fact when clients pay large non-audit service fees to their auditors.
More closely related to this paper are Krishnan et al. (2005) and Gul et al. (2006) who found that firms with high non-audit service fees are associated with lower earnings response coefficient. Therefore, investors discount the reliability of reported earnings when non-audit service fees paid to the incumbent auditors are high, implying loss of perceived auditor independence.5 This paper extends Krishnan et al. (2005) and Gul et al. (2006) by examining whether FIS-related services are enhancing firm value despite investors’ concern for auditor independence. Krishnan et al. (2005) and Gul et al. (2006) do not distinguish between FIS-related services and other non-audit services. A focus on FIS-related services is important because the literature on the business value of information technology generally supports the notion that investments in information technology are associated with firm value (Dehning and Richardson, 2002). In addition, investors will take into account the value of FIS investments and the potential benefits of knowledge spillover as well as concern for auditor independence in assessing the value of auditor-provided FIS services. Our study provides evidence on the net benefits (over costs) of auditor-provided FIS-related services given the various potential benefits and costs of such systems.
3. Research design
3.1. Use of market measures
Our approach towards examining the association between firm value and auditor's FIS-related services is motivated by the literature on the business value of IT. A number of information system researchers have used stock market valuation as a measure of the business performance (Dos Santos et al., 1993; Brynjolfsson and Yang, 1996; Im et al., 2001). This approach overcomes the shortcoming associated with accounting measures of performance, such as return on assets or gross margin. For example, Generally Accepted Accounting Principles require immediate expensing of certain types of IT-related expenditures (such as fees paid to auditors for FIS-related consulting and employee training costs associated with new IT systems), when in fact their benefits could extend beyond the current year. In contrast, market measures such as stock prices represent a direct and arguably the best measure of stockholder value and reflect a rich, comprehensive information set, including information not captured by firms’ financial statements.
3.2. Empirical model
Studies that use a market measure have used a level model or a return model. Gu (2005) notes that there is no consensus regarding the choice between a level model and a return model and concludes that the models capture two different economic relations. As in Easton et al. (1993) and Kothari and Zimmerman (1995), we use both specifications. We first use a level model as a primary test and then examine the return model in sensitivity analyses (see Section 4.3). The level model is based on a commonly used valuation model developed by Ohlson (1995) that links market value of equity to net income, book value of equity, earnings growth and FIS purchases (see Barth et al., 1998; D'Souza and Jacob, 2001). Specifically, we estimate the following model (firm subscripts are suppressed) for our first research question:
(1)Where MVE is market value of common equity calculated at the end of 3 months after fiscal year-end divided by book value of equity at the beginning of the year. EARN is net income divided by book value of equity at the beginning of the year. After-tax amount of fees paid to the auditors for FIS-related services are added back to earnings.6 BE is book value of equity at the end of the year divided by book value of equity at the beginning of the year. EGROW is growth in earnings per share calculated over a 3-year period. FISD equals 1 for firms reporting FIS fees in either year 2000 or 2001 and 0 for firms that are not reporting such fees.7 The model also includes 34 industry-dummy variables DI representing two-digit SIC code numbers for the matched sample. The objective is to capture industry-specific commonalities in the dummy-variable coefficients and, hence, reduce correlations among regression residuals.
Consistent with prior research a1, a2 and a3 are expected to be positive. We offer no prediction for the coefficient of interest, a4, for the following reasons. Benefits associated with a well-designed FIS represent an intangible asset. This intangible asset can contribute to future cash flows in the form of productivity gains, enhanced decision-making capability and lower operating costs. These benefits contribute to the ability of the firm to enhance its competitive advantage, earn economic rents and, therefore, increase firm value. Hence, the market participants are likely to assign a positive and statistically significant value to FISD. However, the provision of FIS-related services by the incumbent auditors might negatively influence the stock market participants’ concern about auditor independence and this could counterbalance the value of benefits generated by FIS-related services. As a result, a significantly positive (negative) coefficient of FISD would suggest that the provision of FIS-related consulting services by the auditor results in a net benefit (loss) to the firm. An insignificant result of FISD might suggest that the positive effect of the provision of FIS services might be cancelled out by the negative impact of investors’ concern over auditor independence. In summary, whether the perceived benefits arising from FIS-related services exceed the concerns about auditor independence is an empirical issue.
To investigate our second research question, we replace FISD in model 1 by LALLXFIS and LFISF. Specifically, we use the following regression model.
(2)FISF is the natural logarithm of one plus fees paid to the auditors for FIS-related services. LALLXFIS is the natural logarithm of one plus the sum of audit and non-audit fees less FIS fees. The purpose of using LFISF and LALLXFIS is to capture the effect of FIS-related fees on firm value while controlling for the effect of auditors’ other fees. We do not use fee ratio since Ashbaugh et al. (2003) suggest that fee ratio does not capture the economic significance of auditors’ economic bond accurately. Instead, they suggest the use of level of fees to capture economic bond. We do not have expectations on FISF since it could be positively or negatively related, or unrelated, to market value of equity, MVE, depending on investors’ perception of the value of auditor-provided FIS services. We also calculate an alternate measure of auditors’ other economic bond, LONAF, which is the natural logarithm of one plus total non-audit fees less FIS fees. LONAF will take the place of LALLXFIS in model 2.
3.3. Sample selection
Data for all variables except FIS and other fees are obtained from the PC version of Compustat. For FIS and other fees, we use a database of fees paid to auditors for audit and non-audit services, including IT-related services, compiled by Standard & Poor's from proxy statements.
For our first research question, we use a matched sample. We identify firms that report a non-zero value for FIS fees and match these firms with firms that did not report FIS fees by year, two-digit SIC code, and beginning total assets. The objective is to identify control firms that are similar to firms that purchase FIS services.8 As firms that do not purchase FIS services from their auditors may purchase the services from other providers, the control sample excludes those firms, which are identified from the Factiva database and IT professional journals. Altogether two such firms are excluded and the sample selection procedure yields 562 firm-year observations, consisting of 281 observations for each group. We refer to this sample as the matched sample.
For our second research question, we use a different sample. First, from the fee database compiled by Standard & Poor's we select firms that report a non-zero value for FIS fees. These firms are not matched with firms that do not purchase FIS services from their auditors. Next, we gather the relevant data from the Compustat database. This sample selection procedure yields 408 firm-year observations. We refer to this sample as the second sample.
4. Results
4.1. Does the stock market value auditor-provided FIS services?
Descriptive statistics for the matched sample for our first research question are reported in Table 1.9 Firms with FIS purchases (FISD = 1) have significantly higher firm value (both mean and median values of MVE) than firms without such purchases (FISD = 0). However, the control variables (EARN, BE and EGROW) are not statistically different between the two groups of firms.10 These results suggest that the differences in firm value are more likely to be due to the provision of FIS services than to some other factors that could drive firm value.
| Variable | Mean | t-statistic | Median | z-statistic | ||
|---|---|---|---|---|---|---|
| FISD = 1 | FISD = 0 | FISD = 1 | FISD = 0 | |||
| MVE | 3.180 | 2.755 | 1.99** | 2.291 | 2.109 | 2.12** |
| EARN | 0.098 | 0.091 | 0.38 | 0.137 | 0.124 | 1.38 |
| BE | 1.118 | 1.122 | −0.11 | 1.058 | 1.065 | 0.17 |
| EGROW | 0.154 | 0.017 | 1.03 | 0.164 | 0.135 | 0.99 |
- Firms reporting non-zero financial information system (FIS) fees are matched with a control group of firms not reporting such fees by year, two-digit SIC code, and total assets at the beginning of the year. Total number of observations for each group equals 281. Data are for years 2000 and 2001. ** Indicates two-tailed statistical significance at the 0.05 level. The variables are defined as follows: FISD is an indicator variable set equal to 1 for firms reporting fees paid to the auditor for FIS-related services, and 0 otherwise; MVE is market value of common equity at the end of 3 months after fiscal year-end divided by the book value of equity at the beginning of the year; EARN is net income divided by the book value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for FIS-related services are added back to earnings; BE is book value of equity at the end of the year divided by beginning book value of equity; and EGROW is growth in earnings per share calculated over a 3-year period. Tests are two-tailed. t-statistics are from t-tests of the differences in the means and z-statistics are from Wilcoxon two-sample tests.
Table 2 provides the correlation matrix of the variables. The results suggest that MVE is positively and significantly correlated with the control variables: EARN, BE and EGROW. These results are consistent with expectation. More importantly, MVE is also positively and significantly correlated with FISD, suggesting that the provision of FIS services increases firm value, and this correlation is consistent with the result in Table 1.
| Variable | (1) | (2) | (3) | (4) | (5) |
|---|---|---|---|---|---|
| (1) MVE | 1.000 | ||||
| (2) EARN | 0.404*** | 1.000 | |||
| (3) BE | 0.430*** | 0.322*** | 1.000 | ||
| (4) EGROW | 0.316*** | 0.557*** | 0.280*** | 1.000 | |
| (5) FISD | 0.084** | 0.016 | −0.005 | 0.043 | 1.000 |
- Total number of observations equals 562. *** and **indicate, respectively, two-tailed statistical significance at the 0.01 and 0.05 levels. The variables are defined as follows: MVE is market value of common equity at the end of 3 months after fiscal year-end divided by the book value of equity at the beginning of the year; EARN is net income divided by the book value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for financial information system-related services are added back to earnings; BE is book value of equity at the end of the year divided by beginning book value of equity; EGROW is growth in earnings per share calculated over a 3-year period; and FISD is an indicator variable set equal to 1 for firms reporting fees paid to the auditor for financial information system-related services, and 0 otherwise.
Table 3 reports the results of regression for model 1. Reported t-statistics are adjusted for heteroscedacity (White, 1980). The control variables –EARN and BE– are positive as expected and significant at the 0.01 level or better. Although EGROW is not statistically significant at conventional level for a two-tailed test, its sign is positive as expected and it is significant for a one-tailed test. The variable of interest, FISD, is positive and significant at the 0.05 level, indicating that the provision of FIS-related services by the incumbent auditors is favourably valued by the stock market.11
| Independent variable (expected sign) | Coefficient | t-statistic |
|---|---|---|
| Intercept (?) | 0.601 | 0.92 |
| EARN (+) | 3.193 | 4.22*** |
| BE (+) | 2.031 | 8.23*** |
| EGROW (+) | 0.112 | 1.53 |
| FISD (?) | 0.395 | 2.43** |
| Adjusted R2 | 0.385 |
- Total number of observations equals 562. *** and ** indicate, respectively, two-tailed statistical significance at the 0.01 and 0.05 levels. The variables are defined as follows: MVE is market value of common equity at the end of 3 months after fiscal year-end divided by the book value of equity at the beginning of the year; EARN is net income divided by the book value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for financial information system-related services are added back to earnings; BE is book value of equity at the end of the year divided by beginning book value of equity; EGROW is growth in earnings per share calculated over a 3-year period; and FISD is an indicator variable set equal to 1 for firms reporting fees paid to the auditor for financial information system-related services, and 0 otherwise. The model also includes 34 industry dummy variables representing two-digit SIC code numbers comprising the sample (coefficients not reported). Reported t-statistics are White (1980) heteroscedacity-consistent t-statistics.
The positive association between MVE and FISD is also consistent with the literature on business value of information technology. For example, Brown et al. (1995) find that the stock market reacts favourably to the announcements of investment in strategic information systems and in years subsequent to the investment in IT, and that the sample firms tended to be more profitable than otherwise comparable firms in their industries.
To provide further insight, we perform an additional analysis to examine which auditors are more likely to provide FIS services that add value to the clients. We repeat the sample selection procedures by matching on auditors also (non-Big Five (now non-Big Four) auditors are treated as a group due to small number of observations). This sample consists of 290 firm-year observations. Model 1 is re-run by splitting FISD into dummies for auditors in the sample. The results (available from the authors) show that Deloitte Touche Tohmatsu and Pricewaterhouse-Coopers are more likely to provide value-enhancing FIS services. Therefore, auditors are different in their ability to provide valuable FIS services.
4.2. Impact of FIS services on firm value
While findings in Table 3 suggest that market participants perceive FIS-related services as value-enhancing, does this association hold in the presence of economic bond? To put it differently, does fee dependence (i.e. when auditors receive a large amount of fees for FIS-related services) cause market participants to discount the benefits of FIS-related services?
We examine this issue by using the second sample (discussed earlier). In this sample, more than 96 per cent of the observations are audited by Big Five auditors. The sample firms come from 51 industry categories and, hence, model 2 also includes 50 industry dummy variables DI representing two-digit SIC code numbers comprising the sample (coefficients not reported). Purchases of FIS-related services appear to be more common in utilities, financial services, business services, and machinery and computer equipment industries than in the other industries.
The results of model 2 are in Table 4. In column (3), we control for all other fees. The variables EARN and BE are positive and significant as in Table 3. The variable of interest LFISF is also positive and significant. This result suggests that investors do value auditor-provided FIS services at high level of FIS fees despite concern for economic bond that might lead to lower auditor independence. Column (4) controls for all other non-audit fees. Again LFISF is positive and significant.
| Independent variable (expected sign) | Controlling for all other fees Coefficient (t-statistic) | Controlling for all other non-audit fees Coefficient (t-statistic) |
|---|---|---|
| Intercept (?) | 1.505 (1.35) | 1.732 (1.58) |
| EARN (+) | 3.289 (4.04***) | 3.311 (4.05***) |
| BE (+) | 1.574 (5.01***) | 1.567 (4.95***) |
| EGROW (+) | 0.117 (1.23) | 0.112 (1.18) |
| LALLXFIS (?) | 0.438 (3.52***) | |
| LONAF (?) | 0.425 (3.28***) | |
| LFISF (+) | 0.112 (1.69*) | 0.120 (1.82*) |
| Adjusted R2 | 0.402 | 0.400 |
- Total number of observations equals 408 in both regressions.*** and *indicate, respectively, two-tailed statistical significance at the 0.01 and 0.10 levels. The variables are defined as follows: MVE is market value of common equity at the end of 3 months after fiscal year-end divided by the book value of equity at the beginning of the year; EARN is net income divided by the book value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for financial information system-related services are added back to earnings; BE is book value of equity at the end of the year divided by beginning book value of equity; EGROW is growth in earnings per share calculated over a 3-year period; LALLXFIS is natural logarithm of one plus total fees paid to the auditor less fees for financial information system-related services; LONAF is natural logarithm of one plus other non-audit fees (non-audit fees paid to the auditor less fees paid to the auditor for financial information system-related services); and LFISF is natural logarithm of one plus fees paid to the auditor for financial information system-related services. The model also includes 50 industry-dummy variables representing two-digit SIC code numbers comprising the sample (coefficients not reported). Reported t-statistics are White (1980) heteroscedacity-consistent t-statistics.
Overall, the results in Table 4 are consistent with results in Table 3 and suggest that FIS-related consulting services provided by auditors are positively associated with firm value. These findings suggest that investors regard FIS-related services as value-enhancing activities despite possible concern for impairment of auditor independence due to fee dependence.
4.3. Sensitivity tests – return model
Next, we use a return specification to assess the robustness of our findings. Specifically, for our first research question, we estimate a regression of stock returns on unexpected earnings, market-to-book ratio, earnings persistence, and beta (see Collins and Kothari, 1989), and FISD:
(3)Where RET is annual buy-and-hold return calculated over a 12 month period ending 3 months after fiscal year-end. Following Easton and Harris (1991) and Cheng et al. (1996), we use a change-and-levels specification for unexpected earnings. The sum of c1 + c2 is the earnings response coefficient and is expected to be positive. MB is market-to-book ratio, a proxy for growth, calculated as market value of equity over book value of equity. No prediction is offered for c3 because the evidence is mixed on the association between earnings response coefficient and growth.12 PERSIST is an indicator variable set equal to 0 for firms with low earnings persistence and 1 for firms with high earnings persistence. Following Ali (1994), persistence is measured as follows: observations are partitioned into 10 groups based on the absolute value of change in income before extraordinary items. Observations in the four extreme deciles (top two deciles and bottom two deciles) are classified as low-persistence firms and observations in the middle six deciles are classified as high-persistence firms; c4 is expected to be positive.13 BETA is common stock beta and c5 is expected to be negative. The variable of interest, FISD, is the same as defined before. The model also includes industry dummy variables DI representing two-digit SIC code numbers comprising the sample (coefficients not reported).
We repeat the matching procedure discussed earlier for model 3. The number of observations available is 818. Regression results of model 3 are in Table 5. The sum of earnings level and changes, representing unexpected earnings, is 0.021 (–0.420 + 0.441), which is not significantly different from zero (p = 0.839). Among other control variables, MB and PERSIST are significant at the 0.01 level. The variable of interest, FISD, is positive and significant at the 0.10 level. Overall, the findings are consistent with the notion that after controlling for unexpected earnings, market-to-book ratio, persistence, beta and industry-related factors, FIS purchases are associated with higher stock returns. When market-adjusted returns are used as the dependent variable, the coefficient for FISD is 0.059 and significant at the 0.10 level (t-statistic = 1.78). In summary, results based on the return specification are consistent with the results based on Ohlson model presented earlier.
| Independent variable (expected sign) | Coefficient | t-statistic |
|---|---|---|
| Intercept (?) | 0.030 | 0.67 |
| EARN (+) | −0.420 | −5.45*** |
| ΔEARN (+) | 0.441 | 2.43** |
| MB (?) | 0.001 | 2.60*** |
| PERSIST (+) | 0.110 | 3.18*** |
| BETA (−) | −0.013 | −1.49 |
| FISD (+) | 0.058 | 1.76* |
| Adjusted R2 | 0.340 | |
- Total number of observations equals 818. ***, ** and *indicate, respectively, two-tailed statistical significance at the 0.01, 0.05 and 0.10 levels. The variables are defined as follows: RET is annual buy-and-hold return calculated over a 12 month period ending 3 months after fiscal year-end; EARN is net income divided by the market value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for financial information system-related services are added back to earnings; ΔEARN is change in net income from year t − 1 to t divided by the market value of equity at the beginning of the year; MB is market-to-book ratio calculated as market value of equity over book value of equity; PERSIST is an indicator variable set equal to 0 for firms with low earnings persistence and 1 for firms with high earnings persistence. Persistence is measured as follows: observations in each year are partitioned into 10 groups based on the absolute value of change in income before extraordinary items. Observations in the four extreme deciles (top two deciles and bottom two deciles) are classified as low-persistence firms and observations in the middle six deciles are classified as high-persistence firms; BETA is beta; and FISD is an indicator variable set equal to 1 for firms reporting fees paid to the auditor for financial information system-related services, and 0 otherwise. The model also includes 36 industry-dummy variables representing two-digit SIC code numbers comprising the sample (coefficients not reported). Reported t-statistics are White (1980) heteroscedacity-consistent t-statistics.
We also use a return specification for our second research question. This specification is similar to model 3 except that instead of FISD, we include LFISF and LALLXFIS or LONAF. The variables LFISF, LALLXFIS and LONAF are as previously defined for model 2. The sample consists only of firms that report non-zero FIS fees for year 2000 or 2001. The number of observations is 486. The results are in Table 6. The changes in earnings (ΔEARN) and persistence (PERSIST) are significant at the 0.01 level. More importantly, LFISF is positive and significant after controlling for all other fees (LALLXFIS) and all other non-audit fees (LONAF). When market-adjusted returns are used as the dependent variable, the coefficient for LFISF is 0.053 and significant at the 0.05 level (t-statistic = 2.01) for regression controlling for LALLXFIS. Corresponding figures for regression controlling for LONAF are 0.058 and 2.15 for value of LFISF and its t-statistic respectively. Therefore, the results are consistent with the results reported in Table 4 that higher levels of FIS fees are valued by the market participants despite the concern over economic bond.
| Independent variable (expected sign) | Controlling for all other fees Coefficient (t-statistic) | Controlling for all other non-audit fees Coefficient (t-statistic) | ||
|---|---|---|---|---|
| Intercept (?) | −0.068 | (−0.51) | −0.052 | (−0.40) |
| EARN (+) | 0.000 | (0.00) | 0.005 | (0.01) |
| ΔEARN (+) | 0.549 | (2.25***) | 0.548 | (2.24***) |
| MB (?) | 0.007 | (1.02) | 0.007 | (1.04) |
| PERSIST (+) | 0.138 | (2.41***) | 0.131 | (2.36***) |
| BETA (−) | −0.003 | (−0.66) | −0.003 | (−0.67) |
| LALLXFIS (?) | −0.002 | (−0.09) | ||
| LONAF (?) | −0.012 | (−0.59) | ||
| LFISF (+) | 0.047 | (1.81*) | 0.053 | (1.96*) |
| Adjusted R2 | 0.194 | 0.194 | ||
- Total number of observations equals 486 for each regression. *** and *indicate, respectively, two-tailed statistical significance at the 0.01 and 0.10 levels. The variables are defined as follows: RET is annual buy-and-hold return calculated over a 12 month period ending 3 months after fiscal year-end; EARN is net income divided by the market value of equity at the beginning of the year. After-tax amount of fees paid to the auditor for financial information system related services are added back to earnings; ΔEARN is change in net income from year t − 1 to t divided by the market value of equity at the beginning of the year; MB is market-to-book ratio calculated as market value of equity over book value of equity; PERSIST is an indicator variable set equal to 0 for firms with low earnings persistence and 1 for firms with high earnings persistence. Persistence is measured as follows: observations in each year are partitioned into 10 groups based on the absolute value of change in income before extraordinary items. Observations in the four extreme deciles (top two deciles and bottom two deciles) are classified as low-persistence firms and observations in the middle six deciles are classified as high-persistence firms; BETA is beta; LALLXFIS is natural logarithm of one plus total fees paid to the auditor less fees for financial information system-related services; LONAF is natural logarithm of one plus other non-audit fees (non-audit fees paid to the auditor less fees paid to the auditor for financial information system-related services); and LISF is natural logarithm of one plus fees paid to the auditor for financial information system-related services. The model also includes 50 industry-dummy variables representing two-digit SIC code numbers comprising the sample (coefficients not reported). Reported t-statistics are White (1980) heteroscedacity-consistent t-statistics.
In short, the results support the notion that firms with increases in FIS fees are associated with positive stock returns. Overall, results from the above additional sensitivity analyses are consistent with findings in Tables 3 and 4 that FIS purchases and higher spending on such purchases have a favourable impact on firm value.
4.4. Limitations
Our study is subject to some limitations. First, despite our search for firms with non-auditor provided FIS services, not all those firms could be identified since firms are not statutorily required to disclose purchases of FIS services from non-auditors. To the extent that the control sample includes such firms, the firm value of the control sample is higher and this will bias against (but not in favour of) finding results that support the conjecture that firms with auditor-provided FIS services have higher firm value than the control firms. However, at the same time, the existence of non-auditor-provided FIS services in the sample might make the inferences of the results less precise. Second, the results in Table 3 should be interpreted cautiously because of self-selection and endogeneity issues. Self-selection refers to the possibility that firms that purchase FIS services might be fundamentally different from firms that do not, and it is such differences that affect the decision and, hence, firm value. Endogeneity refers to the chance that firms with higher market value of equity are more likely to invest in FIS. In an unpublished appendix (available from the authors), we control for both issues by following a two-stage estimation approach used by Lee (1978) and explained in Maddala (1983).14 The results show that the average predicted value of market value of equity is greater for firms that purchase FIS services after controlling for self-selection and endogeneity issues. However, it is still possible that these issues might affect our results despite our best efforts to control for them.
5. Conclusion
The provision of non-audit services by incumbent auditors remains a controversial issue. While auditees benefit from such services, they appear to increase the strength of the economic bond between the auditees and the auditors. This study contributes to the growing literature on the consequences of auditor-provided non-audit services by examining the association between fees paid to the incumbent auditors for FIS design and development and firm value.
A well-developed FIS is a prerequisite for obtaining timely feedback on operating performance and for making appropriate corrective actions to preserve a firm's competitive advantage. Other benefits of an FIS include enhanced productivity and profitability in the form of reduction in inventory, reduced headcount in accounting function, and strategic planning capability. Furthermore, auditor-provided FIS services have the effects of knowledge spillover that benefit both the clients and the auditors. However, the provision of FIS services by incumbent auditors might also create fee dependence on auditors and lower auditor independence. Hence, whether the provision of FIS services by auditors creates firm value is an interesting empirical issue.
We find that market value of equity is greater for firms that purchase FIS-related services from their incumbent auditors relative to a control group of firms without such purchases, matched by year, two-digit SIC code and total assets at the beginning of the year. We also find that after controlling for the economic bond between the audit clients and the auditors, fees paid to the auditors for FIS-related purchases are positively associated with the market value of equity. Our results are consistent with the notion that the value of FIS-related services outweighs the negative perceptions associated with the joint provision of audit and non-audit services by the incumbent auditors and results in a net benefit to the firms.
