Auditor Changes and Restatements

By Wanda A. Wallace

E-mail Story Print Story

Policy Questions and Data

Debate continues about the benefits and costs of auditor rotation. In 2004, the GAO issued a study showing that roughly half of the companies surveyed had been with their current auditing firm at least 10 years and 88% were satisfied with the firm.

The auditor of record in any given year may or may not have been associated with the financial statements requiring restatement. (The auditor is usually not the discovery source for a restatement.) The GAO report indicated that in about half of the cases, the restating company itself was reported as responsible for recognizing the previous misstatements. Other sources included the SEC (4.5% of the cases analyzed), the external auditor (2.5%), or some other external party (9%); in about one-third of the restatements, the source was unknown.

Audit committees are charged with overseeing the selection and ratification of external auditors. In doing so, a question may arise as to whether a change in auditors precedes, occurs with, or follows restatement phenomena experienced by other companies. To explore how the profile of auditors among restating companies compares to that of other companies, the author analyzed the companies tracked in the Standard & Poor’s Research Insight CompustatPC database for 1996 through mid-2003, comparing the relative incidence and the nature of auditor changes observed for the list of 731 entities recording the 919 restatements identified in the GAO report covering 1997 to June 2002. In addition, the author profiled the audit reports issued to restating companies relative to those of the other CompustatPC entities.

Aggregate Perspective

Exhibit 1 presents an aggregate picture of the auditor distribution per year for all restating entities relative to other CompustatPC companies per year, including the count for entities coded within the database as unaudited. Certain patterns emerge. For example, Arthur Andersen (AA) held from 15% to 20% of the entities both among restating and nonrestating entities for 1996 through 2001; this dramatically declines as a result of the firm’s dissolution. Likewise, Coopers & Lybrand (CL) held between 10% and 19% of the entities in 1996–1997, which disappeared when it merged with Price Waterhouse (PW). Interpreting Exhibit 1 requires looking at the lines connecting the restatement and no-restatement columns for each year; if they are parallel, as in 1996 for Ernst & Young (EY), Deloitte & Touche (DT), and PricewaterhouseCoopers (PWC), then the firm’s relative proportion of restatements is similar to that observed for other CompustatPC firms in that year. On the other hand, a growing width between columns, as for KPMG Peat Marwick (KPMG) and “other” firms, suggests that the proportion of restating companies is less than that observed for other CompustatPC companies.

A way to recast the data is to take the total number of companies for each CPA firm or group of firms from CompustatPC and determine the proportion of the entities audited by that firm in a given year with restatements within the time frame of the GAO study. This entails dividing the number of entities restating anytime from 1996 through 2003 for a given CPA firm, based on the GAO information set, by the total number of observations analyzed in that same year for that firm.

This analysis would indicate that, in relative terms, Coopers & Lybrand had the largest proportion of the restating entities in a given year (11%); until its merger with PWC, which then experienced a greater relative proportion (8%), from 1999 to 2002. Of further interest is the observation that the “other” group experienced far lower relative incidence of restatement (2 to 4%) than did any of the large firms (total ranging from 6 to 7%). Annual statistics must be interpreted in light of the variation across time in the number of entities for which information is provided by CompustatPC, as well as for potential effects of a survivorship bias within the restatement sample due to the real costs, such as litigation, that often accompany restatements. That is, those restating entities that go out of business or are acquired fall out of the data set if they are no longer retrievable from CompustatPC. Exhibit 1 helps illustrate this phenomenon by providing counts.

The auditor distribution in Exhibit 1 implies that changes in auditors have occurred in both subsets of companies, although effects of initial public offerings and mergers could also affect the distribution, as well as data availability through CompustatPC. Moreover, two types of changes might reasonably be set aside: the shifts caused by the PricewaterhouseCoopers merger in 1998 (59 restatements and 489 no restatements) and the shifts attributable to Andersen’s dissolution (five restatements and 83 no restatements).

Further analyzing the pooled data set (including all years of restatement), Exhibit 2 shows the relative annual change among restating and nonrestating firms. Anecdotal accounts of interactions of auditors and auditees suggest that multiple auditor changes can arise in cases of restatement. To explore this possibility, auditor changes by the year of restatement were profiled for entities experiencing multiple changes in auditors between 1996 and 2002. As many as four changes are reported for five separate companies, with restatements; a greater number experienced two or three changes. The proportion of restating firms with multiple changes is reported in Exhibit 2 to compare unique changes with multiple changes. For restating entities, the implication is that much of the apparent difference in auditor change rates shown in Exhibit 2 stems from the multiple-change entities. The incidence of change before, during, and after restatements can be estimated by considering the timing of the GAO restatements in percentage terms (e.g., 7.64% of the restatements analyzed by the GAO occurred in 1997, when changes in auditors were observed among restating firms at the rate of 3.91%, of which 2.42% experienced more than a single change of auditors). The first column in Exhibit 2, representing each year’s relative share of total restatements, generally indicates a rising number of restatements, which is not mirrored by the proportion of auditor changes each year. In other words, the growth in restatements did not spawn a similar growth in auditor turnover.

Beyond multiple auditor changes by companies, the author considered multiple restatements. More than half of those entities with two or three restatements within the period changed auditors, whereas only 40% of those with a single restatement changed auditors. The implication is that the propensity for auditor change increases with the number of restatements by a company.

Disaggregated View Across Time, Reflecting the Year of Announced Restatement

A shortcoming of the aggregate analyses in Exhibits 1 and 2 is that all restating companies are pooled, which does not permit a focus on restatements in a given year and their auditor selections across the period under study. As a result, the distribution of auditors before restatement, during restatement, and after restatement cannot be discerned. To address this, the author aligned the year in which a restatement is announced according to the GAO report with the auditor identified in that year, to profile the mix of auditors within each restatement year, as depicted in Exhibit 3 for the Big Five and “other” firms. For example, of all restatements announced in 1997 for which the auditor was specified (including “other”), 35.48% were audited by PWC in the year of announcement. A caveat in interpreting Exhibit 3: The announcement of the restatement need not correspond to the period being restated. In other words, the auditor during the year of restatement may not correspond to the period being restated or the auditor within such period. Exhibit 3 includes a benchmark of the average proportion of no restatements for each auditor during the 1997–2002 period.

Consideration of the Audit Opinion

The ComputstatPC database permits a comparison of the profile of audit reports, as shown in Exhibit 4. Because CompustatPC has more detailed information on audit opinions than on the audit firm, the column totals do not agree with Exhibit 1. A comparison of the restatement and no-restatement data (the parallel lines connecting adjacent columns) indicates a consistent mix of audit opinion types from 1996 through 1998. In 1999 the incidence of unqualified opinions with additional language began to increase among restatement entities, a trend that continued through 2002. The incidence of unqualified opinions without additional language is greater for restating companies than for others, although the difference becomes negligible by 2002. No restated opinion was qualified or adverse, however, nor does the data set include any “no-opinion” cases, which may be due in part to the survival bias cited earlier and the effect of delayed reporting on delistings that potentially would remove information from CompustatPC.

Implications

Anecdotal accounts in the press speculate about the relationship between the selection of auditors and the incidence of auditor changes and restatements. The author analyzed the GAO and CompustatPC data to profile auditor selection, the incidence and timing of changes in auditors, and audit opinions. Auditor changes among restating companies, once multiple changers are set aside, is quite comparable to that observed in nonrestating companies. Audit opinions are reasonably similar in their distribution, although there is some evidence that additional language within unqualified reports became more common after 1999.

Analysis does not support speculation about the high incidence of Andersen’s clients restating results. Exhibit 1 suggests that the proportion of restating companies audited by Andersen is comparable to the proportion of restating companies overall from 1996 to 1998, with a slight increase after 1999. A similar balance is observed for Ernst & Young and Deloitte & Touche for 1996 through 2001. PricewaterhouseCoopers (and Coopers & Lybrand before it) exhibited a greater proportion of restating companies throughout the period. The auditors may or may not have been involved in the original restated financial statements, the discovery of the restatement, or the aftermath. Moreover, the size of client portfolios, industry mix, the effect of certain restatements on particular types of companies, and the volatility of the stock market may all offer explanations for the variations above.

Is there a relationship between restatements and auditor changes? The data indicate that auditor changes are more likely to occur surrounding restatements. This effect, however, is attributable to companies that experience multiple auditor changes, particularly entities experiencing multiple restatements. If variations in restatement rates attributable to companies that changed auditor multiple times during the 1996–2002 period under study are set aside, the overall incidence of auditor change among the remaining restating companies is actually lower, except in 1998 and 2002.