Is auditor switching associated with delayed accounting recognition of bad news?

Grayson, Michael M. ; Luehlfing, Michael S.

ABSTRACT

We study the association between auditor switching and the delayed accounting recognition of bad news about net income. Using a nonparametric sign test and a test of proportions, we analyze 305 auditor switches which occurred during the event period 1976 to 1994, a period which predated a significant increase in the number of financial-statement restatements (General Accounting Office 2002). The results (null hypothesis rejected at z > 5 for each test) suggest that some association exists between the fact of auditor switching (whether reported as resignation or dismissal) and the occurrence of decreases in net income from the year preceding the auditor switch (t-1) to the year following the auditor switch (t+1).

INTRODUCTION

Time-series research shows that accounting net income is normally expected to rise from year to year, not fall. We assert that when management expects accounting net income to fall, management may switch auditors in an attempt to delay accounting recognition of bad news. In this regard, we study the association between auditor switching and the delayed accounting recognition of bad news about net income.

The remainder of the paper is organized as follows. We begin by providing a brief background overview. Thereafter, we develop the research hypothesis. After explaining the methods used (including sample selection and statistical tests), we discuss the results, contributions and limitations. Finally, we provide several comments regarding future research.

BACKGROUND

As indicated in the Appendix, there are numerous possible reasons for auditor switching. However, as already stated, we focus on management's desire to delay the recognition of bad news about net income as a reason for auditor switching.

Knapp and Elikai's (1990) information suppression hypothesis assumes that management needs to suppress information permanently. In contrast, Kluger and Shields (1991, 255), suggest that auditor switches may be associated with attempts to "delay the release of unfavourable information." (emphasis added) Thus, management may be satisfied by suppressing information temporarily rather than permanently. Indeed, if management wants to delay accounting recognition of bad news, and succeeds in doing so, then the bad news would be recognized subsequent to the year of the auditor switch.

While managements have incentives to voluntarily disclose bad news (Skinner 1994 and 1997), they also have incentives not to disclose bad news; or, at least, to delay the disclosure. For example, by switching auditors, management may delay recognition of the bad news until such time that management is able to downwardly adjust market expectations so that investors will already be expecting bad news by the time the bad news becomes public. Management gains at least the delay involved in the switch, which is entirely contained within the annual reporting cycle (that is, year t, the year in which the auditor switch occurs). However, if recognition of the bad news can be delayed until the year following the auditor switch, management also has time to reset annual compensation and bonus targets so that managers still receive high compensation and bonuses even for meeting targets which were set lower than they otherwise would have been set. Healy (1985) demonstrated that managers change decisions as a result of compensation arrangements, including moving recognition in financial statements (both accelerating and delaying recognition) from one fiscal year to another. Knapp (1991, 41, Table 1) reported that managers do control the selection of auditors, either by selecting them outright or by giving the board of directors a list from which the board is permitted to select.

RESEARCH HYPOTHESIS

Ball and Watts (1972, 680) concluded that accounting net "income can be characterized on average as a submartingale or some similar process." That is, the net income for each year is expected to be greater than or equal to the preceding year's net income. Mathematically, [Y.sub.t] [greater than or equal to] [Y.sub.t-1]. Taking the mathematical expression literally, this is an extreme expectation because it never permits net income to be lower than in any preceding year.

Another extreme expectation is that accounting net income always moves randomly; that is, accounting net income moves down from one year to the next as often as it moves up. This--expectation of random movement of accounting net income is the opposite of the Ball and Watts (1972, 680) expectation. However, we know from the time-series literature in accounting that accounting net income does not move randomly (e.g., Brown and Rozeff 1979, Foster 1977, Griffin 1977, Watts 1975, and Bathke and Lorek 1984). Nevertheless, the existence of the time-series literature demonstrates that the true pattern of movement of accounting net incomes is somewhere along the continuum from completely random to the Ball and Watts (1972) expectation.

If lower net incomes do indeed occur systematically with respect to auditor switches, this suggests some association, causal or otherwise, exists between the two. If management is promptly reporting all bad news, then management has no incentive to switch auditors in order to delay any reporting of any bad news. On the other hand, if management is desirous of delaying the reporting of bad news, and management suspects that the auditors will both find the bad news and insist that its effects be recognized in the current year's financial statements, then management may decide that the delayed reporting of bad news is more important to them than the continued good relationship with their incumbent auditor. In such a case, management can then choose a course of action whereby it (1) attempts (whether successfully or unsuccessfully) to find new auditors who will not require that the bad news be recognized in the current year's financial statements, (2) issues guidance to lower investors' expectations for the following year, and (3) gives accounting recognition to the bad news in that following year (or later, or never). This conceptual sequence gives rise to the following null hypothesis:

H0: Among client companies that switched auditors, the net
income in the year after the auditor switch (t+1) was greater than
or equal to the net income in the year before the switch (t-1).

In substance, this hypothesis tests for an association between auditor switching and the delayed revelation of bad news. This hypothesis is based on the assumption that decreases in net income either are bad news themselves or else proxy for bad news.

METHODS

Sample Selection

The sample was selected from the Compustat data base (that is, a limited subscription to Compustat composed of 4,106 companies reported to have been randomly selected by Compustat from the full Compustat data base) for the years 1975 through 1994. Use of this period avoids the increased number of financial-statement restatements found by the General Accounting Office (2002) to have occurred in subsequent years, that is, 1997 to 2002, and therefore avoids a potential problem whereby some additional companies ought to have restated, but did not do so. See Tables 1 and 2 for companies included in the data base, but excluded from the sample. See Table 3 for information on the final sample.

To be selected, companies must have had different auditor codes reported by Compustat from one year to the next, plus have had two following years of data. If a company was taken over (Anderson, Stokes and Zimmer 1993) or merged (Hribar and Collins 2002), then it did not have two following years of data and was not included in this study. The purpose of using multiple years in which auditor switches occurred was to guard against any systematic effects in the economic environment causing auditor changes and/or declines in net incomes in particular calendar years. Auditor codes were not available from Compustat for banks, life insurance, or property and casualty companies (Compustat 1994, 5-26), so those companies were excluded from selection.

Unlike DeFond (1992, 23), this study included companies with multiple switches. Suppose a company were to switch auditors solely to obtain lower audit fees. Since the company would then be paying less to its outside vendor of audit services, its net income would be higher than otherwise. Because we hypothesize that auditor switching is associated with lower net incomes rather than higher net incomes, including multiple auditor switches tends to bias against finding lower net incomes. Thus, inclusion of companies with multiple switches biases against finding the hypothesized effect. Therefore, if significant results are found, they may understate what other researchers, who would have made different research design choices, would have found.

Kluger and Shields (1991, 263) tested two bankruptcy prediction models based on companies which either switched or did not switch auditors two to three years before bankruptcy, and found that the switched companies were more likely to go bankrupt. By selecting companies which survived two fiscal years beyond switching auditors, even though the tests were conducted only on one fiscal year beyond switching auditors, any possible effects attributable to attempts to stave off bankruptcy, or to suppress information concerning severe financial distress culminating in bankruptcy, were lessened.

Statistical Tests

We tested the null hypothesis using the Fisher distribution-free sign test (Hollander and Wolfe 1973, 39-40). This test analyzes the direction of movement, if any, and results in a pattern of ones (for movement in the hypothesized direction) and zeroes. Once the direction of movement is determined, the ones and zeroes are distributed according to a binomial distribution (Hollander and Wolfe 1973, 40). When the sample is large, Hollander and Wolfe (1973, 40) show that the test statistic [B.sup.*] is approximately normally distributed. The sample of 305 auditor switches is considered a large sample, and therefore the standard deviation and the z score were computed according to Devore and Peck (1986, 223).

To perform a test of significance on the direction of changes in net income, it is necessary to have an expectation for the number of times net income should move in the specified direction. The observed number of changes in net income is then compared as a proportion or number or percentage of the sample against an expected proportion or number or percentage. We performed two tests, using two expectations.

The expectation for net income according to Ball and Watts (1972, 680) is that accounting net "income can be characterized on average as a submartingale or some similar process." That is, the net income for each year is expected to be greater than or equal to the preceding year's net income. However, the Ball and Watts expectation can be characterized as an expectation which is extreme, because it allows zero lower net incomes.

Another extreme expectation is that accounting net income always moves randomly; that is, accounting net income moves down from one year to the next as often as it moves up. This expectation of random movement of accounting net income is the opposite of the Ball and Watts (1972, 680) expectation. However, we know from the time-series literature in accounting that accounting net income does not move randomly (e.g., Brown and Rozeff 1979, Foster 1977, Griffin 1977, Watts 1975, and Bathke and Lorek 1984). Nevertheless, the existence of the time-series literature demonstrates that the true pattern of movement of accounting net incomes is somewhere along the continuum from completely random to the Ball and Watts (1972) expectation.

If accounting net incomes were to move randomly, then the movement from year (t-1) to year (t) is either up or down, and, similarly, the movement from year (t) to year (t+1) is either up or down. Thus, the movement from year (t-1) to year (t+1) is either up up, up down, down up, or down down. Making a simplifying assumption of equal-sized movements in net income, then only in the "down down" pattern of net income will net income be lower in (t+1) than in (t-1). Thus, an expectation of random movement of accounting net incomes means that twenty-five percent of the time, net incomes should be lower in (t+1) than in (t-1), which allows for a different test of the same null hypothesis. This expectation that net income is lower twenty-five percent of the time in (t+1) than (t-1) actually overstates the percentage because the movement of accounting net incomes is upward rather than random; from the time-series literature findings of patterns rather than randomness, the true but unknown percentage has to be lower than twenty-five percent.

Use of a matched-pairs test is not appropriate because the relevant population characteristic (direction of changes in net income) is known from the Ball and Watts (1972) study and from the time-series literature (e.g., Brown and Rozeff 1979, Foster 1977, Griffin 1977, Watts 1975, and Bathke and Lorek 1984). Unlike Garsombke and Armitage (1993), we tested the direction of changes in net income, not the quantity of net income, for two reasons. First, we needed to be consistent with our hypothesis, which deals with direction--not magnitude--of change in net income. Second, even a small percentage change in net income by a very large company may, by itself, be larger than the combined changes in net incomes of many small companies.

RESULTS

H0: Among client companies that switched auditors, the net
income in the year after the auditor switch (t+1) was greater than
or equal to the net income in the year before the switch (t-1).

There were 121 observed lower net incomes (out of 305 switches). When the proportion of 121/305 was compared against the Ball and Watts (1972, 680) income expectation, the null hypothesis was rejected at a z value of 13.9. When the proportion of 121/305 was compared against an expectation of random movement of accounting net incomes, the null hypothesis was rejected at a z value of 5.1.

Given the results of both these tests, the null hypothesis was strongly rejected at a z value somewhere between 5.1 and 13.9. The z value of 5.1 is based on an extreme expectation that net income is lower in (t+1) than in (t-1) twenty-five percent of the time, which overstates the actual percentage. The z value of 13.9 is based on an expectation, which is extreme in the opposite direction, that net income is never lower in (t+1) than in (t-1), which understates the actual percentage. Therefore, any reasonable income expectation which falls between these two extremes would also result in the null hypothesis being rejected at some z value which falls between 5.1 and 13.9.

The =NORMSDIST(Z) function in an Excel spreadsheet was used to compute the cumulative area under the curve for these z values. At a z value of 5.1, the cumulative area is 0.9999998+, for a p value of <0.0000002. At a z value of 13.9, the cumulative area, correct to twenty-five decimal places, is 1, for a p value of <.0000000000000000000000001. This means that, much more frequently than expected, and with little probability of having reached an incorrect conclusion, net income is lower the year following an auditor switch (t+1) than the year preceding an auditor switch (t-1).

CONTRIBUTIONS AND LIMITATIONS

We discussed a possible sequence of events regarding possible attempts by management to delay accounting recognition of bad news, and from that formulated a testable hypothesis. We then documented a strong statistical association between the fact of auditor switching (whether reported as resignation or dismissal) and the occurrence of lower net incomes in the year following the auditor switch (t+1) than in the year preceding the auditor switch (t-1).

While we found strong evidence of an association, we cannot state with certainty what causes this association since we have not conducted a critical experiment--that is, an experiment which not only supports one causative factor, but also rules out all other possible causative factors. We also did not test whether managers' compensation targets are actually adjusted when companies' incomes decrease. Nevertheless, the large z score suggests that further research is warranted.

This study did not distinguish between auditor dismissals and auditor resignations because this study tested for an association between auditor switching (regardless of who initiated the switch) and bad news which may have become public after the switch. While some studies may have used sources such as 8-K filings to determine whether auditor switches were reported as dismissals or resignations (Nichols and Smith 1983, and Wells and Louder 1997), we ignore the characterization of the switch and concentrate instead on the fact of the switch. Krishnan and Stephens (1995, 180) referred to McConnell in acknowledging that it may be difficult to determine the true reason for an auditor switch, no matter what people or documents say. As seen below, McConnell (1984, 46) recognized that what is stated in 8-K filings or other sources may be false information.

 Not acknowledging known unreported disagreements in his exhibit
 letter reviewing client disagreement assertions allows an auditor
 to avoid potential litigation, as well as the stigma of being "a
 poor loser."

Wells and Loudder (1997, 140) stated, "empirical evidence suggests that resignations are associated with unfavorable events within a firm." There was no reason to infer that the bad news being examined herein would differ across auditor dismissals and auditor resignations, particularly when both dismissals and resignations could each be disguised as the other. Auditors who do not wish to continue as auditors may raise their fees to such an extent that clients decide to switch. Clients who do not wish to continue their relationship with their incumbent audit firm can decide to have disagreements with the auditors or otherwise to make life difficult for the auditors.

The use of the Compustat data base caused a bias against finding the hypothesized effect. This bias occurred because the Compustat data base companies tend to be larger than the average of all publicly-owned companies. Those larger companies are more closely followed by market participants who may have, or seek, private information which may bear on price movements of the companies' securities. Consequently, the extra effort devoted to obtaining information about those companies made it less likely that management could have known material bad news and kept it from the market for an extended period.

FUTURE RESEARCH

There are several avenues for future research. First, cash flows could be examined in lieu of net incomes or other income-statement items. Second, differences among industries could be examined. Third, differences among audit firms could be examined.

In addition, methodology could be varied. For example, managements' incentives and resulting actions can be addressed not only by capital-markets studies, but also by behaviorally-based studies, analytical modeling, and studies which use experimental markets. This would provide triangulation as suggested by Abdel-khalik and Ajinkya (1979).

APPENDIX
Some Reasons for Auditor Switching

Burton and Roberts (1967) as accounting standards disputes
cited by Chow and Rice (1982, changes in management
327-328) demand for additional services
 needs arising out of new financing

Fried and Schiff (1981, 327) changes in auditor fee structure
 client's need for additional services
 regular auditor rotation policy

Chow and Rice (1982, 328) qualified opinion rendered by audit
 firm

Schwartz and Menon (1985) and attempts to influence the auditor
Teoh (1992)

Kluger and Shields (1987, also information suppression
cited in Haskins and Williams
1990)

Simunic and Stein (1987) and agency-theory-based Big 8
Francis and Wilson (1988), versus non-Big 8 product
both cited in Haskins and differentiation approach
Williams (1990)

Williams (1988) a management change in which new
 management wishes to deal with the
 firm it previously dealt with (247)
 fast growth causing a need for a
 larger audit firm (250) industry or
 other specialization available in the
 audit firm (250)

Francis and Wilson (1988, 668) diffusion of ownership
 leverage
 new securities issues

Knapp (1988, 42, who cited corporate managers' desire to
Bedingfield and Loeb 1974, 67; manipulate their firm's reported
Klott 1984; and Laventhol & financial condition
Horwath 1985)

Haskins and Williams (1990), opinion shopping
citing Chow and Rice (1982)

DeFond (1992, 17) agency conflicts

Anderson, Stokes, and Zimmer corporate takeovers
(1993)

Garsombke and Armitage timeliness of service
(1993, 95) lack of responsiveness

Dhaliwal, Schatzberg, and economic factors
Trombley (1993)

Krishnan (1994, 210-211) dissatisfaction with service
 dissatisfaction over fees
 disagreements over accounting issues
 (opinion shopping)
 management change
 change of engagement partner
 resignation of the account
 initial public offerings
 rapid growth
 search for "credible auditors"

REFERENCES

Abdel-khalik, A. R., and B. B. Ajinkya. 1979. Empirical research in accounting: a methodological viewpoint. Sarasota, FL: American Accounting Association.

Anderson, D., D. Stokes, and I. Zimmer. 1993. Corporate takeovers and auditor switching. Auditing: A Journal of Practice & Theory 12 (Spring): 65-73.

Ball, R., and R. Watts. 1972. Some time series properties of accounting income. Journal of Finance (June): 663-81.

Bathke, A. W., Jr., and K. S. Lorek. 1984. The relationship between time-series models and the security market's expectation of quarterly earnings. The Accounting Review Volume 59, Issue 2 (April): 163-176.

Bedingfield, J. P., and S. E. Loeb. 1974. Auditor changes--an examination. Journal of Accountancy (March): 66-69. Cited in Knapp, M. C. 1988. Toward a more coherent regulatory policy for auditor changes: an empirical study of their impact on financial statement credibility. Research in Accounting Regulation 2: 41-59.

Brown, L., and M. Rozeff. 1979. Univariate time-series models of quarterly accounting earnings per share: a proposed model. Journal of Accounting Research (Spring): 179-189. Cited in Bathke, A. W., Jr. and K. S. Lorek. 1984. The relationship between time-series models and the security market's expectation of quarterly earnings. The Accounting Review Volume 59, Issue 2 (April): 163-176.

Burton, J. C., and W. Roberts. 1967. A study of auditor changes. Journal of Accountancy (April): 31-36. Cited in Chow, C. W., and S. J. Rice. 1982. Qualified audit opinions and auditor switching. The Accounting Review 57: 327-35.

Chaney, P. K., D. C. Jeter, and P. E. Shaw. 1997. Client-auditor realignment and restrictions on auditor solicitation. The Accounting Review (July): 433-453.

Chow, C. W., and S. J. Rice. 1982. Qualified audit opinions and auditor switching. The Accounting Review 57: 327-35.

DeFond, M. L. 1992. The association between changes in client firm agency costs and auditor switching. Auditing: A Journal of Practice & Theory 11 (Spring): 16-31.

Devore, J., and R. Peck. 1986. Statistics: The Exploration and Analysis of Data. St. Paul: West Publishing Company.

Dhaliwal, D. S., J. W. Schatzberg, and M. A. Trombley. 1993. An analysis of the economic factors related to auditorclient disagreements preceding auditor changes. Auditing: A Journal of Practice & Theory 12 (Fall): 22-38.

Foster, G. 1977. Quarterly accounting data: time series properties and predictive ability results. The Accounting Review (January): 1-21. Cited in Bathke, A. W., Jr. and K. S. Lorek. 1984. The relationship between time-series models and the security market's expectation of quarterly earnings. The Accounting Review Volume 59, Issue 2 (April): 163-176.

Francis, J. R., and E. R. Wilson. 1988. Auditor changes: a joint test of theories relating to agency costs and auditor differentiation. The Accounting Review 63 (October): 663-682. Cited in Haskins, M. E., and D. D. Williams. 1990. A contingent model of intra-Big Eight auditor changes. Auditing: A Journal of Practice & Theory 9 (Fall): 55-74.

Fried, D., and A. Schiff. 1981. CPA switches and associated market reactions. The Accounting Review 56 (April): 32641.

Garsombke, H. P., and J. L. Armitage. 1993. Competitive bidding, fees and auditor changes: additional evidence. Journal of Applied Business Research 9 (Summer): 93-99.

Griffin, P. 1977. The time series behavior of quarterly earnings: preliminary evidence. Journal of Accounting Research (Spring): 71-83. Cited in Bathke, A. W., Jr. and K. S. Lorek. 1984. The relationship between time-series models and the security market's expectation of quarterly earnings. The Accounting Review Volume 59, Issue 2 (April): 163-176.

Haskins, M. E., and D. D. Williams. 1990. A contingent model of intra-Big Eight auditor changes. Auditing: A Journal of Practice & Theory 9 (Fall): 55-74.

Healy, Paul M. 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting & Economics 7(1/2/3): 85-107.

Hollander, M., and D. A. Wolfe. 1973. Nonparametric Statistical Methods. New York: John Wiley & Sons.

Hribar, P., and D. W. Collins. 2002. Errors in estimating accruals: implications for empirical research. Journal of Accounting Research 40:1 (March), 105-134.

Klott, G. 1984. Auditors feel the heat of new scrutiny. New York Times, May 13, pp. 1 and 26 (Section 3). Cited in Knapp, M. C. 1988. Toward a more coherent regulatory policy for auditor changes: an empirical study of their impact on financial statement credibility. Research in Accounting Regulation 2: 41-59.

Kluger, B. D., and D. Shields. 1991. Managerial moral hazard and auditor changes. Critical Perspectives on Accounting 2 (September): 255-272.

Kluger, B. D., and D. Shields. 1987. Auditor changes, information suppression and bankruptcy prediction. Working paper, Rice University, February 25, 1987. Cited in Haskins, M. E., and D. D. Williams. 1990. A contingent model of intra-Big Eight auditor changes. Auditing: A Journal of Practice & Theory 9 (Fall): 55-74.

Knapp, M. C. 1988. Toward a more coherent regulatory policy for auditor changes: an empirical study of their impact on financial statement credibility. Research in Accounting Regulation 2: 41-59.

Knapp, M. C. 1991. Factors that audit committee members use as surrogates for audit quality. Auditing: A Journal of Practice & Theory 10 (Spring): 35-52.

Knapp, M. C., and F. M. Elikai. 1990. Auditor changes and information suppression. Research in Accounting Regulation 4 (Fall): 3-20.

Krishnan, J. 1994. Auditor switching and conservatism. The Accounting Review 69 (January): 200-215.

Krishnan, J., and R. G. Stephens. 1995. Evidence on opinion shopping from audit opinion conservatism. Journal of Accounting and Public Policy 14: 179-201.

Laventhol & Horwath. 1985. Investor protection is reason for SEC's 'opinion shopping' proposals. SEC Newsletter (November-December 1985): 1-2. Cited in Knapp, M. C. 1988. Toward a more coherent regulatory policy for auditor changes: an empirical study of their impact on financial statement credibility. Research in Accounting Regulation 2: 41-59.

McConnell, Jr., D. K. 1984. Auditor changes and related disagreements. Auditing: A Journal of Practice & Theory 3 (Spring): 44-56. Cited in Krishnan, J., and R. G. Stephens. 1995. Evidence on opinion shopping from audit opinion conservatism. Journal of Accounting and Public Policy 14: 179-201.

Nichols, D. R., and D. B. Smith. 1983. Auditor credibility and auditor changes. Journal of Accounting Research 21 (Autumn): 534-544.

Schwartz, K. B., and K. Menon. 1985. Auditor switches by failing firms. The Accounting Review 60 (April): 248-61. Simunic, D. A., and M. T. Stein. 1987. Product Differentiation in Auditing: Auditor Choice in the Market for Unseasoned New Issues. Vancouver, B.C.: The Canadian Certified General Accountants' Research Foundation. Cited in Krishnan, J. 1994. Auditor switching and conservatism. The Accounting Review 69 (January): 200-215.

Skinner, Douglas J. 1997. Earnings disclosures and stockholder lawsuits. Journal of Accounting & Economics 23 (November): 249-282.

Skinner, Douglas J. 1994. Why firms voluntarily disclose bad news. Journal of Accounting Research 32:1 (Spring): 3860.

Standard & Poor's Compustat. 1994. Compustat User's Guide. Englewood, CO: McGraw-Hill, Inc.

Teoh, S. H. 1992. Auditor independence, dismissal threats, and the market reaction to auditor switches. Journal of Accounting Research 30 (Spring): 1-23.

United States General Accounting Office. 2002. Financial Statement Restatements: Trends, Market Impacts, Regulatory Responses, and Remaining Challenges, GAO report GAO-03-138 dated October 4, 2002. Available at the GAO's Web site at www.gao.gov.

Watts, R. 1975. The time series behavior of quarterly earnings. Unpublished paper, University of New Castle. Cited in Bathke, A. W., Jr. and K. S. Lorek. 1984. The relationship between time-series models and the security market's expectation of quarterly earnings. The Accounting Review Volume 59, Issue 2 (April): 163-176.

Wells, D. W. and M. L. Louder. 1997. The market effects of auditor resignations. Auditing: A Journal of Practice & Theory 16:1 (Spring): 138-144.

Williams, D. D. 1988. The potential determinants of auditor change. Journal of Business Finance & Accounting 15 (Summer): 243-261.

Michael M. Grayson, Jackson State University

Michael S. Luehlfing, Louisiana Tech University

Table 1: Companies with only one switch which were excluded from
sample of auditor switches

Company Reason(s) for exclusion

American Stores Co. Auditor was switched simultaneous with change
 in end of fiscal year.

Chiquita Brands Compustat indicates four changes in fiscal
 year end during 1975 to 1994.

Craig Corp. Auditor was switched simultaneous with change
 in end of fiscal year.

Decorator Industries Fiscal year end was changed within two years
 following auditor switch.

Dole Food Co Inc. Auditor was switched simultaneous with change
 in end of fiscal year.

FoxMeyer Health Corp. The company changed its fiscal year the year
 after switching auditors.

General Motors Class E The company changed its fiscal year the year
 after switching auditors.

Hondo Oil & Gas The auditor switch occurred when Pauley
 Petroleum bought 81% of Hondo.

Placer Dome Income and auditor data were incorrect.

Ranger Oil Ltd. The auditor codes per Compustat were 9, 4, 4,
 and 6 for years ended 12/86 through 12/89.
 Per Moody's, the auditors were Thorne Riddell
 for 1985, Thorne Ernst & Whinney for 1986 to
 1988, and Peat Marwick Thorne for 1989. This
 appears to be the same Canadian audit firm,
 but the audit firm changed affiliations
 during the period.

Table 2: Companies excluded (E) or data changed (C)

Part A. Differences in Auditor between Compustat and Moody's

Auditor per Company FYE Moody's Auditor per Compustat

Avon Products (E) 12/88 6 Coopers & Lybrand
Crystal Oil Co. 12/86 8 Touche Ross
Crystal Oil Co. (C) 12/87 6 Touche Ross
Crystal Oil Co. (C) 12/88 8 Peat Marwick
Crystal Oil Co. 12/89 6 Peat Marwick
CSS Industries Inc. (E) 1/80 8 Arthur Andersen
Gerber Scientific (E) 4/75 N/A Peat Marwick
Gerber Scientific (E) 4/76 7 Peat Marwick
Gerber Scientific (E) 4/77 6 Peat Marwick
Gerber Scientific (E) 4/78 6 Peat Marwick
Lee Pharmaceuticals (E) 9/80 5 Deloitte Haskins & Sells
Lee Pharmaceuticals (E) 9/81 7 Deloitte Haskins & Sells
Lehigh Group Inc. (E) 12/91 1 KPMG Peat Marwick in
 summary listing, but
 mentions Arthur
 Andersen's audit report
Lehigh Group Inc. (E) 12/92 11 Arthur Andersen
Noble Affiliates 12/88 7 Price Waterhouse
Noble Affiliates (C) 12/89 1 Price Waterhouse
Noble Affiliates 12/90 1 Arthur Andersen
Placer Dome (E) 12/87 7 Clarkson Gordon; Price
 Waterhouse
Placer Dome (E) 12/88 14 Clarkson Gordon; Price
 Waterhouse
Placer Dome (E) 12/89 4 Ernst & Young; Price
 Waterhouse
Placer Dome (E) 12/90 7 Price Waterhouse
Ply-Gem Industries (E) 12/82 9 (not determined)
Ply-Gem Industries (E) 12/83 7 Weinick, Sanders & Co.
Ply-Gem Industries (E) 12/84 9 (not determined)
Republic Gypsum (E) 6/88 2 Arthur Young
Republic Gypsum (E) 6/89 4 Ernst & Young
Republic Gypsum (E) 6/90 1 Ernst & Young
Republic Gypsum (E) 6/91 1 Ernst & Young
Scherer, R. P. (E) 3/89 1 (not determined)
Scherer, R. P. (E) 3/90 3 (unable to locate in
 Moody's)
Scherer, R. P. (E) 3/91 1 (not determined)

Part B: Differences in Net Income between Compustat and Moody's

Company FYE Compustat Moody's

Placer Dome (E) 12/87 121.672 C$158.2
Placer Dome (E) 12/88 219.996 C$262.4
Placer Dome (E) 12/89 108.036 C$125.1
Placer Dome (E) 12/90 164.583 C$191.0
Placer Dome (E) 12/91 -236.200 C$236.2 loss
Placer Dome (E) 12/92 111.000 C$111.0

Note: C$ indicates amounts in Canadian dollars. Data sources give
net income in millions of dollars.

Table 3: Analyses of audit switches by Compustat variable DNUM
(SIC code) and by year of switch

Part A. Audit switches by Compustat variable DNUM
(four-digit SIC code)

less than 1000 0
1000 series 24
2000 series 52
3000 series 110
4000 series 16
5000 series 52
6000 series * 21
7000 series 20
8000 series 9
9000 series 1
Total 305

* Auditor codes are not available for banks, life insurance, or
property and casualty companies (Compustat 1994, 5-26), so those
companies are excluded.