Initial evidence on the association between nonaudit fees and restated financial statements.

SYNOPSIS. An increasing number of firms have restated previously issued financial statements in recent years. Legislators, regulators, and others speculate that restatements are associated with fees received by auditors for nonaudit services (nonaudit fees). The current study provides empirical

We identify a sample of 110 firms that restated financial statements previously filed with the SEC for fiscal years 2000 or 2001, and provided relevant audit and nonaudit fee data. The SEC requires firms to disclose these fee data beginning in proxy statements filed on or after February 5, 2001. We compare the fees paid by the restatement sample with fee data for 3,481 firms that filed proxies with the SEC from February 5, 2001 to August 31, 2001 and develop benchmarks for expected nonaudit fees, fee ratio, and total fees. Using these benchmarks, we calculate the unexpected values for these measures and investigate whether restatement firms differ from the control firms. Our findings of no significant differences between the restatement and control samples for unexpected nonaudit fees, fee ratios, and total fees do not support concerns that either nonaudit fees or total fees inappropriately influence the audit and lead to restatements.

INTRODUCTION

In recent years, an increasing number of firms have restated their financial statements to bring them into conformity with GAAP (Moriarty and Livingston 2001). While serving as Chief Accountant of the SEC, Lynn Turner said that the SEC regards restatements to be audit failures (Turner 1999). Consistent with this position, Palmrose and Scholz (2000) find that auditor litigation is significantly associated with restatements involving core or recurring earnings. Legislators and others suggest that a relationship exists between the growing number of restatements and the increasing reliance of audit firms on fees from nonaudit services (U.S. House of Representatives 2002; U.S. Senate 2002). Turner also noted that "accounting firms have become too dependent on nonaudit fees and are unwilling to risk those revenues by challenging corporate managers who stretch accounting rules" (Berenson 200l). In response to concerns that the supply of nonaudit services to audit clients can inappropriately influence audit judgments, in 2000 the SEC attempted to restrict the supply of nonaudit services by auditors to their SEC audit clients. More recently, in July 2002 Congress enacted legislation prohibiting auditors from supplying specific types of nonaudit services (Sarbanes-Oxley Act 2002).

After the enactment of the Sarbanes-Oxley Act, the SEC has issued detailed rules related to the prohibition and disclosure of nonaudit fees (SEC 2003). Specifically, the SEC (2003) prohibited registrants from purchasing financial information systems design and implementation services or internal audit outsourcing from the incumbent auditor. However, registrants may still purchase many types of nonaudit services from the incumbent auditor including tax compliance and consulting, employee plan audits, consulting on accounting matters, merger and acquisition consulting, and consulting on new debt and equity issues.

Thus, recent actions of legislators and the SEC indicate concerns that nonaudit services provided by incumbent auditors can inappropriately influence audit judgments, make auditors less likely to enforce GAAP, and result in subsequent restatements. This study provides empirical evidence about a possible association between firms that restate financial statements and the fees paid to their incumbent auditors.

BACKGROUND AND RESEARCH QUESTION

The SEC contends that providing nonaudit services to audit clients " ... creates economic incentives that may inappropriately influence the audit" (SEC 2000b). This concern is based on the premise that nonaudit services increase the economic dependency of auditors on their clients. Others counter the SEC's concern by suggesting that market-based incentives help to ensure audit quality. For example, Benston (1975) and Watts and Zimmerman (1986) suggest that auditor concern for professional reputation is an incentive for maintaining audit quality. In addition, the risk of litigation arising from class action suits by investors alleging injury provides an incentive for auditors to act independently.

Some recent studies examine whether there is an association between nonaudit fees and financial-reporting quality. In general, these studies use estimates of discretionary accruals, which are believed to signal earnings management, as a proxy for financial-reporting quality. Frankel et al. (2002) find that firms with greater ratios of nonaudit fees to total fees are more likely to just meet or beat analysts' forecasts and to report larger discretionary accruals. Although their results provide some support to the SEC's position that nonaudit fees can influence auditor judgments, other studies provide contrary evidence. Ashbaugh et al. (2002) use an alternative estimate of discretionary accruals, find no association between income-increasing accruals and client fees, and also find no evidence that investors react to fee disclosures. Similarly, Chung and Kallapur (2002) report no statistically significant association between their estimate of discretionary accruals and two measures of client importance--the ratio of (1) audit fees and (2) nonaudit fees from a particular client to the audit firm's overall U.S. revenues.

DeFond et al. (2002) note that the auditor's influence on the client's accruals and earnings characteristics is indirect and that prior research indicates that there are empirical problems in measuring discretionary accruals (Guay et al. 1996; McNichols 2000). They argue for a more direct test that focuses on audit opinions, which clearly and directly reflect auditors' judgments. Their study finds no significant association between auditors' propensity to issue going-concern-modified audit opinions on financially distressed firms and (1) nonaudit service fees, (2) audit fees, (3) total fees, or (4) the fee ratio.

Whisenant et al. (2003) use data filed in 2,666 proxy statements with the SEC in 2001 and find no association between audit and nonaudit fees using a simultaneous specification of the fee system. Their results suggest that prior studies that document an association between audit and nonaudit fees by using single-equation estimations to model the audit and nonaudit fees suffer from simultaneous-equations bias. Since audit and nonaudit fees influence each other, considering them as independent ignores the influence of the two types of fees on each other and leads to incorrect inferences. This finding casts doubt on the SEC's "economic bonding" argument.

Research Question

We believe that examining nonaudit fees paid by firms in the context of subsequent restatements or client failures constitutes a direct test of the association between nonaudit fees and audit quality. The recent SEC requirement for firms to disclose audit and nonaudit fees, beginning in proxy statements filed on or after February 5, 2001, makes these fee data available for analysis. Seidler (2002) notes that "specific instances of substandard [audit] performance are disclosed only when revealed by some other event such as a restatement or bankruptcy." (1) Thus our study is another direct test bearing on nonaudit services and audit quality as it examines whether an association exists between firms that restate financial statements and the nonaudit fees paid to incumbent auditors during periods that required restatement. In short, the research question examined here is:

RQ: Are firms that pay higher fees for nonaudit services more likely to restate their financial statements?

METHOD AND DATA

Restatement and Control Samples

We searched the EDGAR Online database for firms that restated previously issued financial statements using keyword searches, such as restate, adjust, or error. The search covered restatements of financial statements filed initially with the SEC for fiscal years 2000 or 2001, and included the period up to June 30, 2002. This sample enables us to obtain, for identified restatement firms, the incumbent auditor fee data required to be disclosed by registrants under recent SEC regulations.

Table 1 indicates that our initial search identifies 177 financial restatements by firms for fiscal year 2000, and 133 for fiscal year 2001. We eliminate "technical" restatements resulting from changes in the application of GAAP, such as a change from LIFO and converting to equity method accounting for investments, or from adoption of EITF rulings that affect classification but do not impact earnings. These include: EITF No. 00-10, Accounting for Shipping and Handling Fees and Costs; EITF No. 00-14, Accounting for Certain Sales Incentives; and EITF No. 00-25, Vendor Income Statement Characterization of Consideration from a Vendor to a Retailer (FASB 2000a, 2000b, 2000c). As noted later, we require financial and stock-price data for our regression analyses; because of missing data in the Compustat and CRSP databases, we delete 43 observations. As done in prior research, we use log-transformed values of nonaudit fees and total fees in our analysis. Because the log of zero is not defined, we delete six restatement firms that reported zero nonaudit fees. Our final sample includes 110 observations.

To examine our research question, we require a control sample of firms to compare with the nonaudit fees of the restatement sample firms. Using the list of firms included in the Compustat database, we searched proxy statements filed by August 31, 2001 with the SEC. This search identified audit and nonaudit fee data for 3,621 firms. To be consistent with the restatement sample, we deleted 140 observations with zero nonaudit fees, leaving 3,481 observations in the control sample.

Expected and Unexpected Nonaudit Fees

Prior research seeks to explain nonaudit fees with variables that proxy for various firm characteristics, such as size, financial and stock price performance, complexity, and special circumstances such as merger or issuance of equity, and auditor type (Firth 1997; Frankel et al. 2002; Parkash and Venable 1993). Drawing upon prior research, and considering additional variables expected to be important fee determinants, DeFond et al. (2002) and Whisenant et al. (2003) estimate the determinants of nonaudit fees. This study uses variables that were employed in the latter two studies to estimate the following regression model for nonaudit fees:

   log(NONAUDIT FEE) = [[Beta].sub.0] + [[Beta].sub.1](log(ASSETS)) +
                       [[Beta].sub.2](BIG 5) + [[Beta].sub.3](ROA) +
                       [[Beta].sub.4](RETURN)
                       + [[Beta].sub.5](LEV) + [[Beta].sub.6]
                       (INSTITUTIONAL) + [[Beta].sub.7](SPECIAL ITEMS)
                       + [[Beta].sub.8](BOOK
                       TO MKT) + [[Beta].sub.9](SEGS) + [[Beta].sub.10]
                       (FOROPS) + [[Beta].sub.11](EMPLAN) +
                       [[Beta].sub.12](INITIAL
                       YEARS) + [[Beta].sub. 13](MERGER) +
                       [[Beta].sub.14](FINANCE) + [[Beta].sub.15]
                       (SALES GROWTH) + e

where:

   log(NONAUDIT FEE) = the natural logarithm of nonaudit fees paid to
                       the auditor;
          log(ASSET) = natural logarithm of total assets at the end of
                       the year in millions of
                       dollars;
               BIG 5 = an indicator variable set equal to 1 when the
                       auditor is a Big 5 firm, and
                       0 otherwise;
                 ROA = return on assets defined as operating income
                       divided by total assets;
              RETURN = the firm's stock return over the fiscal year;
                 LEV = total liabilities divided by total assets at the
                       end of the fiscal year;
       INSTITUTIONAL = the percentage of shares held by institutions at
                       the end of the fiscal
                       year;
       SPECIAL ITEMS = the absolute value of negative special items
                       divided by total assets, and
                       0 if no negative special items;
         BOOK TO MKT = book value of common equity divided by market
                       capitalization on the
                       last day of the fiscal year;
                SEGS = the number of operating segments disclosed in
                       the segment footnote;
              FOROPS = an indicator variable set equal to 1 if the firm
                       has foreign operations,
                       and 0 otherwise;
              EMPLAN = an indicator variable set equal to 1 if the firm
                       has a pension or post-retirement
                       plan, and 0 otherwise;
       INITIAL YEARS = an indicator variable set equal to 1 if it is
                       the initial two years of the
                       audit engagement, and 0 otherwise;
              MERGER = an indicator variable set equal to 1 if the
                       client acquired a firm during
                       the fiscal year, and 0 otherwise;
             FINANCE = an indicator variable set equal to 1 if the firm
                       issued equity or debt
                       greater than $10 million during the fiscal year,
                       and 0 otherwise; and
        SALES GROWTH = percentage change in sales over the prior year.

Large firms can be expected to have greater needs for nonaudit services, so we expect the coefficient of log(ASSETS) to be positive in the above regression. Similarly, the demand for nonaudit services may increase with the complexity and growth of firms leading us to predict positive coefficients for SEGS, FOROPS, EMPLAN, and SALES GROWTH and a negative coefficient for BOOK TO MKT. Based on evidence from prior research, we posit that firms with good operating and stock price performance have lower needs for nonaudit services and predict negative coefficients for ROA and RETURN. The Big 5 audit firms offer a broader array of services than other auditors, so we expect a positive coefficient for BIG 5. Firms may be less likely to purchase nonaudit services from the incumbent auditor during the initial years of an audit engagement (or, auditors may be more likely to market more nonaudit services after becoming more familiar with the client), so we predict a negative coefficient for INITIAL YEARS. Special situations, such as mergers, issuance of equity or debt, and unusual or nonrecurring transactions like restructuring charges and write-downs, likely lead to greater demand for nonaudit services, so we predict a positive coefficient for MERGER, FINANCE, and SPECIAL ITEMS. Firms with greater agency costs are more likely to curtail nonaudit purchases, suggesting a negative coefficient for LEV. Because agency costs may be mitigated in the presence of large institutional shareholders, we predict a positive coefficient for INSTITUTIONAL.

We use the above regression model to predict the expected nonaudit fees for each firm. By subtracting these predicted values from the actual nonaudit fees, we calculate the unexplained portion of the nonaudit fees for each observation in the restatement and control samples. We then investigate whether the unexplained portion of nonaudit fees differs between the restatement and control samples.

Examining the "Fee Ratio"

The SEC (2000a, 2000b) and others focus on the relative magnitude of nonaudit fees, expressed as a proportion of either the audit fee or the total fees. Consequently, we also examine the distribution of the fee ratio for the restatement sample. Consistent with DeFond et al. (2002) and Whisenant et al. (2003), we measure the fee ratio as follows:

FEE RATIO = Total Nonaudit Fees/Total Fees (Nonaudit + Audit)

Nonaudit fees are the sum of "Financial Information Systems Design and Implementation Fees" and "Other Fees." As before, we use the variables in those two studies to estimate the determinants of the fee ratio. We predict the fee ratio for the restatement sample and calculate the unexplained portion of the fee ratio. We then investigate whether the unexplained portion of the fee ratio differs for the restatement and control samples.

Analysis Based on Total Fees

The argument that "nonaudit fees impair audit quality" can logically be extended to any fees received by the auditor. That is, if the economic bonding created by nonaudit fees is argued to adversely affect audit quality, the same argument can be applied to all fees received by the auditor including the audit fee. Thus, an investigation of the difference between the magnitude of total fees received by the auditor and firms that restate financial statements is also pertinent.

For the analysis based on total fees, we again rely on the models in DeFond et al. (2002) and Whisenant et al. (2003). The total fee model has LOG (TOTAL FEE) as the dependent variable. In addition to the variables considered previously for nonaudit fees, we use these explanatory variables:

   INVREC     = inventory plus accounts receivable divided by total
                assets at fiscal year end;
   REPORT LAG = number of days between fiscal year end and earnings
                announcement date;
   VOLATILITY = the variance of the residual from the market model over
                the fiscal year.

Because prior research suggests that the audit fee increases with the proportion of assets in the form of receivables and inventory, we include INVREC in the regression for total fees and expect a positive coefficient. REPORT LAG is a measure of audit delay and a proxy for the presence of issues involving extended testing or negotiations; we believe it will have a positive coefficient in the regression for total fees. Finally, because audit fees should increase with client risk, we predict a positive coefficient for VOLATILITY.

With respect to the other variables included in the regression, we predict that measures of size, complexity, growth, performance, auditor size, auditor tenure, and special events will have an association with audit fees similar to that for nonaudit fees. Consequently, we expect the signs of the coefficients for variables that proxy for the above factors in the regression for total fees to be similar to that in the nonaudit fee model. However, agency costs are generally expected to increase audit fees whereas nonaudit fees are expected to decrease with increases in agency costs. Consequently, we do not make ex ante predictions about the sign of the coefficient for LEV and INSTITUTIONAL in the total fees model.

We use the predicted values from this expanded model to compute the unexpected value of the total fees paid by each firm. We then investigate whether the unexplained portion of total fees differs between the restatement and control samples.

RESULTS

Descriptive Statistics

Panel A of Table 2 presents details about the industry composition of the restatement and control groups. The industry distribution of the control group is similar to that of the restatement group, except that the control group includes a much higher proportion of observations from the financial services industry and a much lower proportion from the retail industry. As noted later, we delete the financial services observations and obtain results substantively similar to those reported.

Panel B of Table 2 provides descriptive statistics about the restatement and control sample of firms examined. The data indicate that the restatement sample is generally larger than the control sample, in terms of total assets, sales, audit fees, nonaudit fees, and total fees. Interestingly, the fee ratio is almost identical for the restatement and control sample of firms.

Regression Results Based on Nonaudit Fees, Fee Ratio, and Total Fees

Table 3 reports the results from the three different regressions for nonaudit fees, fee ratio, and total fees for the full sample. The explanatory power of the models, as measured by the adjusted [R.sup.2], is comparable to that obtained by DeFond et al. (2002) and Whisenant et al. (2003). Consistent with expectations, the results indicate that nonaudit fees, nonaudit fee ratio, and total fees are positively associated with (1) firm size, (2) growth, (3) complexity, (4) using a Big 5 auditor, and (5) special situations including mergers, issuance of debt or equity, and other unusual or nonrecurring items. Further, as expected, each of these three fee measures is inversely associated with (1) profitability, (2) stock price performance, and (3) auditor tenure. We now use the estimated coefficients from these models to calculate for each firm the expected values of nonaudit fees, fee ratio, and total fees paid to the auditor.

Panel A of Table 4 indicates that the mean (median) values of LOG(NONAUDIT FEE) for the restatement and control samples are 13.05 (12.83) and 12.29 (12.26), respectively. The table also provides information about the unexpected nonaudit fees for the restatement and control samples. A comparison of the unexpected nonaudit fees for the two groups indicates that there is no statistically significant difference between the two groups (p-value = .99). (2)

Panel B of Table 4 indicates that the mean (median) fee ratio for the restatement sample is 0.51 (0.51). The corresponding values for the control group are 0.50 (0.51). Panel B of Table 4 provides information about the actual and unexpected fee ratios for the restatement and control samples. Once again, the two sets of distributions appear to be similar and there is no statistically significant difference between the two groups (p-value = .25).

The results from the analyses based on total fees appear in Panel C of Table 4. The unexpected values of LOG (TOTAL FEE) fees for the restatement and control samples appear to be similar, and there is no significant difference between the two groups (p-value = .44).

Thus, the results indicate that firms with restatements are not likely to have had unexpectedly high nonaudit fees, fee ratios, or total fees. This empirical evidence does not support assertions that restatements are more likely to occur in firms that paid higher than normal nonaudit fees or total fees to their auditors.

Additional Analyses Based on Matched-Pair Designs

We also use two different matched-pair designs to examine the association between restatements and fees. In the first approach, we match each restatement firm with a control firm on the basis of auditor, industry (SIC code), size (sales within 20 percent), and time period. We are able to obtain matched pairs for 78 observations. In 42 of these 78 matched pairs (54 percent) the restatement firms have lower fee ratios, but the observed proportion is not statistically significant from 50 percent--the expected proportion under a random distribution--using a binomial proportions test. (3)

In the second approach, we match on the basis of auditor type, industry code, and size (based on quartiles), obtaining 84 matched pairs. For these 168 observations we use the coefficient estimates in Table 3 to calculate the unexplained portions of nonaudit fees, fee ratio, and total fees. We then investigate whether the unexplained portions of the three fee measures differ between the restatement and control samples, using median and Wilcoxon signed-rank tests. We obtain results substantively similar to those reported in Table 4 (p-value > .20 in each instance). Thus the results from the matched-paired tests are consistent with previously reported regression-based results, and the results of both types of analyses do not indicate an association between restatements and nonaudit fees, fee ratios, or total fees.

Sensitivity Analyses

We perform the following additional analyses as part of our sensitivity tests. First, because firms in the financial sectors (SIC codes 60-69) have different financial statement characteristics than nonfinancial-services firms, we delete all observations with primary SIC codes of 60-69 from the restatement and control samples. Second, we restrict the analyses to only those firms that report a restatement of fiscal year 2000 financial statements. Third, we restrict the analyses to firms with a Big 5 auditor. In all of these additional analyses, the results remain similar to those reported for the full sample and there are no statistically significant differences in the unexpected values of nonaudit fees, the fee ratio, or total fees between the restatement and control samples (p-value > .20 in each instance).

SUMMARY AND CONCLUSIONS

The SEC, legislators, and others speculate that restatements are associated with fees received by incumbent auditors for audit and nonaudit services. This study examines the association between firms that issued restated financial statements and the magnitude of nonaudit fees, fee ratio, and total fees paid to the auditor.

Our analyses compare the fees of 110 firms with restatements and 3,481 firms without restatements. Using fee models similar to those in DeFond et al. (2002) and Whisenant et al. (2003), we obtain expected values for nonaudit fees, fee ratio, and total fees, and use these values to calculate the unexpected values for each of these three measures for each firm. The results, based on a comparison of the distribution of unexpected values, indicate no statistically significant differences across the two groups for any of the three fee measures. Thus, these findings are not consistent with concerns that nonaudit fees or total fees paid to the auditor inappropriately influence the audit such that financial statements are subsequently restated.

Any empirical study is subject to limitations. As with any study that fails to reject a null hypothesis, we acknowledge that a potential limitation of our research design is that our tests lack power. To address this concern, we use all available observations at the time of the data collection to obtain the highest possible consistency in the parameter estimates, and employ a matched-paired research design in our tests of differences. In addition, our analyses rely on data reported by firms in their proxy statements and such disclosures include some subjective judgments related to the classification of the fees paid as audit or nonaudit fees.

Finally, our measure of nonaudit fees includes fees for tax services as well as other nonaudit services. Such a classification is consistent with the regulations promulgated by the SEC when the Commission issued the new rules on auditor independence in November 2000. Our measure of nonaudit fees is consistent with the regulations in effect during our sample period. However, on January 28, 2003 the SEC (2003) acted again on the issue of fee disclosures. The new rules expand the category of "other fees" to three distinct categories: (1) audit-related fees, (2) tax fees, and (3) all other fees. This distinction also reflects the views of the Congress; the Sarbanes-Oxley Act of 2002 suggests that tax services are viewed as less of a concern than other nonaudit services. Also, our study only includes restatements relating to fiscal years 2000 and 2001, the first two years for which fee data are available. An interesting issue for future research is to examine whether there is any association between restatements and fees paid to auditors in other periods, or whether the first two years are special in any respect. Also, future research can test for associations between the separate subcategories of nonaudit fees as well as other measures of financial-reporting quality.

TABLE 1

Sample Selection Details

Number of Restatements
                  FY 2000                    177
                  FY 2001                    133
                                            _____
Total Restatements                           310
Less:
Technical Restatements
                  GAAP Restatements   69
                  EITF Restatements   42    (111)
                                     ____   _____
                                             199
Repeat Observations (2001 deleted)           (40)
Financial/Market Data Not Available          (43)
Zero Nonaudit Fees                            (6)
                                            _____
Observations in Restatement Sample           110

We obtained the list of restatement firms from keyword searches of the EDGAROnline database. We restricted the search to restatements of financial statements filed initially with the SEC for fiscal years 2000 or 2001, and covered the period up to June 30, 2002.

TABLE 2

Descriptive Statistics

Panel A: Industry Distribution--Number of Firms in Restatement and
Control Samples Industry

                               Restatement       Control
                                  Sample          Sample

Pharmaceuticals                 4     (4%)     193     (5%)
Extractive                      5     (4%)     100     (3%)
Durable Manufacturers          24    (22%)     655    (19%)
Computers                      24    (22%)     587    (17%)
Transportation and Utilities    8     (7%)     287     (8%)
Retail                         20    (18%)     312     (9%)
Financial                       8     (7%)     760    (22%)
Services                       14    (13%)     285     (8%)
Other                           3     (3%)     302     (9%)
    Total                     110   (100%)   3,481   (100%)

Industry membership is determined by SIC code as follows: pharmaceuticals (2830-2836), extractive (2900-2999, 1300-1399), durable manufacturers (3000-3999, excluding 3570-3579 and 3670-3679), computers (7370-7379, 3570-3579, 3670-3679), transportation (4000-4899), utilities (4900-4999), retail (5000-5999), financial (6000-6799), services (7000-8999, excluding 7370-7379).

Panel B: Size and Fee Statistics

                                           25th                  75th
Item                 Group       Mean   Percentile   Median   Percentile

Total Assets      Restatement   6,243      133         676      2,826
($ in millions)     Control     4,891       86         346      1,431

Net Sales         Restatement   4,432       44         580      1,977
($ in millions)     Control     2,021       42         171        820

Audit Fees        Restatement   1,355      185         440      1,099
($ in 000)          Control       567      102         191        434

Nonaudit Fees     Restatement   3,050       77         343      1,788
($ in 000)          Control     1,387       64         214        720

TABLE 3
Coefficient Estimates for Nonaudit Fees, Fee Ratio, and Total Fees
Regressions
Based on Full Sample

                    Predicted        Log                         Log
Variable              Sign      (NONAUDIT FEE)   FEE RATIO   (TOTAL FEE)

log(ASSETS)                         0.63 **        0.05 **       0.55 **
BIG 5                  +            0.43 **        0.05 **       0.30 **
ROA                    -           -0.43 **       -0.03 **      -0.30 **
RETURN                 -           -0.11 **       -0.01 *       -0.07 **
LEV                   -/?          -0.22 **       -0.08 **      -0.03
INSTITUTIONAL         +/?           0.01 **        0.03 *        0.01 **
SPECIAL ITEMS          +            0.32 **        0.02          0.21 **
BOOK TO MKT            -           -0.01          -0.00 *       -0.01 **
SEGS                   +            0.19 **       -0.01 **       0.21 **
FOROPS                 +           -0.01          -0.00          0.03
EMPLAN                 +            0.14 **       -0.02 *        0.12 **
INITIAL YEARS          -           -0.10 *        -0.02 *       -0.09 *
MERGER                 +            0.28 **        0.03 **       0.16 **
FINANCE                +            0.45 **        0.06 **       0.25 **
SALES GROWTH           +            0.01 **        0.01 **       0.01 **
REPORT LAG             +                                         0.01 **
INVREC                 +                                         0.01
VOLATILITY             +                                        32.12 **
Adjusted [R.sup.2]                   .66            .27           .76

*, ** p-value <.05, <.01, respectively.

The predicted signs foe LEV and INSTITUTIONAL are only for nonaudit fees and the fee ratio. We do not make prediction for these variables in the regression for total fees.

Legend: The variables are defined as follows:

  log (NONAUDIT FEE) = the natural logarithm of nonaudit fees paid to
                       the auditor;
            FEERATIO = ratio of nonaudit fees relative to total fees
                       received from the client;
      log(TOTAL FEE) = the natural logarithm of total fees paid to the
                       auditor;
         log(ASSETS) = natural logarithm of total assets at the end of
                       the year in millions of dollars;
               BIG 5 = an indicator variable set equal to 1 when the
                       auditor is a Big 5 firm, and 0 otherwise;
                 ROA = return on assets defined as operating income
                       divided by total assets;
              RETURN = the firm's stock return over the fiscal year;
                 LEV = total liabilities divided by total assets at the
                       end of the fiscal year;
       INSTITUTIONAL = the percentage of shares held by institutions at
                       the end of the fiscal year;
       SPECIAL ITEMS = the absolute value of negative special items
                       divided by total assets, and 0 if no negative
                       special items;
         BOOK TO MKT = book value of common equity divided by market
                       capitalization on the last day of the fiscal
                       year;
                SEGS = the number of operating segments disclosed in
                       the segment footnote;
              FOROPS = an indicator variable set equal to 1 if the firm
                       has foreign operations, and 0 otherwise;
              EMPLAN = an indicator variable set equal to 1 if the firm
                       has a pension or post-retirement plan, and 0
                       otherwise;
       INITIAL YEARS = an indicator variable set equal to 1 if it is
                       the initial two years of the audit engagement,
                       and 0 otherwise;
              MERGER = an indicator variable set equal to 1 if the
                       client acquired a firm during the fiscal year,
                       and
                       0 otherwise;
             FINANCE = an indicator variable set equal to 1 if the firm
                       issued equity or debt greater than $10
                       million during the fiscal year, and 0 otherwise;
        SALES GROWTH = percentage change in sales over the prior year;
          REPORT LAG = number of days between fiscal year-end and
                       earnings announcement date; and
              INVREC = inventory plus accounts receivable divided by
                       total assets at fiscal year end;
          VOLATILITY = the variance of the residual from the market
                       model over the fiscal year.

TABLE 4

Comparison of Restatement and Control Samples

Based on Nonaudit Fees, Fee Ratio, and Total Fees

Panel A: Comparison of Nonaudit Fees (Log [Nonaudit Fees])

Statistic          Actual Nonaudit Fees      Unexpected Nonaudit Fees

             Restatement                    Restatement
               Sample      Control Sample     Sample      Control Sample
              (n = 110)      (n = 3,481)     (n = 110)     (n = 3,481)

Mean             13.05          12.29            0.10           0.00
25th
percentile       11.48          11.05           -0.62          -0.68
Median           12.83          12.26            0.06           0.02
75th
percentile       14.60          13.46            0.77           0.71

                     Median Test Statistic = 0.00; p-value = .99.

Panel B: Comparison of Fee Ratios

Statistic          Actual Fee Ratio             Unexpected Fee Ratio

             Restatement                    Restatement
               Sample      Control Sample     Sample      Control Sample
              (n = 110)      (n = 3,481)     (n = 110)     (n = 3,481)

Mean             0.51           0.50           -0.03           0.00
25th
percentile       0.30           0.31           -0.15          -0.14
Median           0.51           0.51           -0.03           0.01
75th
percentile       0.69           0.68            0.11           0.14

                    Median Test Statistic = 1.15; p-value = .25.

Panel C: Comparison of Total Fees (Log [Total Fees])

Statistic          Actual Total Fees           Unexpected Total Fees

             Restatement                    Restatement
               Sample      Control Sample     Sample      Control Sample
              (n = 110)      (n = 3,481)     (n = 110)     (n = 3,481)

Mean             13.85          13.14            0.13           0.00
25th
percentile       12.70          12.15           -0.42          -0.45
Median           13.67          12.97            0.02          -0.02
75th
percentile       15.06          13.94            0.77           0.42

                     Median Test Statistic = 0.77; p-value = .44.

(1.) Fuerman and Sawyer (1997) and Jones and Weingram (2001) report an association between class action lawsuits and restated financial statements. Palmrose and Scholz (2000) document that auditors were more likely to be sued over (1) economic restatements--misstatement of core or recurring income statement items such as revenue or cost of sales--than (2) technical restatements--departures from accounting guidelines in recording noncore or nonrecurring items such as restructuring charges.

(2.) The median test compares the number of observations that are above and below the median for the two samples. An alternative nonparametric test, the Wilcoxon signed-rank test, also indicates no significant difference between the unexpected fee measures for restatement and control samples.

(3.) A binomial proportions test compares the likelihood of an observed pattern occurring against the expected pattern based on an assumed probability of occurrence of an event. In this instance, the number of observations with higher fee ratios (36 out of 78) is not significantly different from that expected under the assumption of 50 percent for each type of occurrence.

REFERENCES

Ashbaugh, H., R. LaFond, and B. Mayhew. 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review (July): 611-639.

Benston, G. 1975. Accountants' integrity and financial reporting. Financial Executive (August): 10-14.

Berenson, A. 2001. Watching the firms that watch the books. New York Times (December 5): C1.

Chung, H., and S. Kallapur. 2003. Client importance, nonaudit fees, and abnormal accruals. The Accounting Review (October) (forthcoming).

DeFond, M. L., K. Raghunandan, and K. R. Subramanyam. 2002. Do nonaudit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of Accounting Research (September): 1247-1274.

Financial Accounting Standards Board (FASB). 2000a. Accounting for Shipping and Handling Fees and Costs. EITF No. 00-10. Norwalk, CT: FASB.

--. 2000b. Accounting for Certain Sales Incentives. EITF No. 00-14. Norwalk, CT: FASB.

--. 2000c. Vendor Income Statement Characterization of Consideration from a Vendor to a Retailer. EITF No. 00-25. Norwalk, CT: FASB.

Firth, M. 1997. The provision of nonaudit services by accounting firms to their audit clients. Contemporary Accounting Research (Summer): 1-21.

Frankel, R., M. Johnson, and K. Nelson. 2002. Auditor independence and earnings quality. The Accounting Review (Supplement): 71-105.

Fuerman, R. D., and F. Sawyer. 1997. Naming auditor defendants in securities class actions. Journal of Legal Economics (Spring/Summer): 72-91.

Guay, W., S. Kothari, and R. Watts. 1996. A market-based evaluation of discretionary accrual models. Journal of Accounting Research (Supplement): 83-105.

Jones, C. L., and S. E. Weingram. 2001. The effects of insider trading, seasoned equity offerings, corporate announcements, and SEC enforcements actions on 10b-5 litigation risk. Working paper no. 139, Stanford Law School, John M. Olin Program in Law and Economics, Stanford, CA.

McNichols, M. 2000. Research design issue in earnings management studies. Journal of Accounting and Public Policy (Winter): 313-345.

Moriarty, G., and P. Livingston. 2001. Quantitative measures of the quality of financial reporting. Financial Executive (July): 17.

Palmrose, Z-V., and S. Scholz. 2000. Restated financial statements and auditor litigation. Working paper, University of Southern California, Los Angeles.

Parkash, M., and C. F. Venable. 1993. Auditee incentives for auditor independence: The case of nonaudit services. The Accounting Review (January): 113-133.

Sarbanes-Oxley Act. 2002. Public Law No: 107-204. Washington, D.C.: Government Printing Office.

Securities and Exchange Commission (SEC). 2000a. Proposed Rule: Revision of the Commission's Auditor Independence Requirements. Washington, D.C.: SEC.

--. 2000b. Final Rule: Revision of the Commission's Auditor Independence Requirements. Washington, D.C.: SEC.

--. 2003. Strengthening the Commission's Requirements Regarding Auditor Independence. Washington, D.C.: SEC.

Seidler, L. 2002. Testimony before the U.S. Senate Committee on Banking, Housing and Urban Affairs (March 6). Washington, D.C.: U.S. Senate.

Turner, L. 1999. 20th century myths. Speech delivered at New York University, November 15. Available at http://www.sec.gov/news/speech/speecharchive/1999/spch323.htm.

U.S. House of Representatives. 2002. Hearings before the Committee on Financial Services (March 13). Washington, D.C.: U.S. House of Representatives.

U.S. Senate. 2002. Oversight Hearing on Accounting and Investor Protection Issues Raised by Enron and Other Public Companies. U.S. Senate Committee on Banking, Housing and Urban Affairs (February 12, 26, and 27; March 5, 6, and 14 2002). Washington, D.C.: U.S. Senate.

Watts, R., and J. Zimmerman. 1986. Positive Accounting Theory. Englewood Cliffs, NJ: Prentice Hall.

Whisenant, S., S. Sankaraguruswamy, and K. Raghunandan. 2003. Joint determination of audit and nonaudit fees. Journal of Accounting Research (forthcoming).

K. Raghunandan is a Professor at Florida International University, William J. Read is a Professor at Bentley College, and J. Scott Whisenant is an Assistant Professor at the University of Houston.

We gratefully acknowledge many useful comments received from two anonymous reviewers.

Submitted: May 2002

Accepted: April 2003

Corresponding author: William J. Read

Email: wread@bentley.edu