Accounting watchdogs have long warned that providing nonaudit services can compromise an auditor's independence. Now they have some numbers to support their case. In a new study, researchers at MIT, Michigan State University and Stanford conclude that auditors are more likely to condone earnings management when audit clients pay large nonaudit fees.
The relationship between earnings quality and nonaudit services has generated concern since the 1970s, when the U.S. Congress first considered limiting the services that auditors could provide. Since then, interest in the issue has intensified in response to two trends: the rapid growth of auditors providing nonaudit services and the emergence of shareholder value as the dominant yardstick by which companies are judged. Between 1993 and 1999, the use of nonaudit services, such as management and information systems consulting, increased at an annual rate of 26%, compared to 9% for audits. During that same period, the growing obsession with stock market performance — initially applauded for increasing CEO accountability —began to show its flaws. Critics worried that unscrupulous executives would do whatever it took to drive up their company's share price, just as auditors' incentives were becoming increasingly stacked in favor of turning a blind eye.
In response to these concerns, the U.S. Securities and Exchange Commission revised its guidelines for determining auditor independence in November 2000. Among other provisions, the new framework prohibited auditors from providing certain nonaudit services, such as bookkeeping, that inherently create conflicts of interest, and it required public companies to begin disclosing the audit and nonaudit fees they pay each year.
The study's authors took advantage of this ruling to construct a sample of 3,074 nonfinancial firms that filed proxy statements between February 5, 2001, when the new requirement went into effect, and June 15, 2001. Simply tabulating the fees disclosed in these statements revealed that nonaudit services were much more prevalent than previously believed. In formulating its new rules, the SEC relied on data suggesting that only a minority of audit clients — a little over one-fourth — purchased nonaudit services and that these transactions amounted to only 10% of revenues for the Big Five (Arthur Andersen, Deloitte & Touche, Ernst & Young, KPMG and PricewaterhouseCoopers). However, the first wave of disclosures painted a very different picture. More than 95% of the companies in this sample purchased nonaudit services from their auditors, and nonaudit fees constituted roughly half the total paid by the average firm.
Moreover, companies that purchased more nonaudit services were more likely to manage earnings, the researchers found. If nonaudit services had no impact on accounting practices, they reasoned, companies paying high and low nonaudit fees would be equally likely to report earnings that met, beat or missed analysts' expectations. But what they found was “dramatically different behavior within a very narrow range — one penny on either side [of the consensus forecast],” co-author Karen Nelson explains.
Companies that paid high nonaudit fees — either relative to their audit fees or in comparison to their auditor's other clients — were less likely to report earnings that fell just short of expectations and more likely to meet the consensus forecast or to beat it by $0.01 per share. (In this analysis, the authors controlled for several factors, such as litigation risk, that might have affected incentives to manage both earnings and expectations.) To illustrate the magnitude of this effect, the researchers divided their sample into fourths, on the basis of each company's ratio of nonaudit fees to total fees. Moving from the bottom quartile to the top quartile of this distribution would boost a representative company's chances of meeting or just beating the consensus forecast by 8%, Nelson explains.
Companies that purchased more nonaudit services also tended to have higher absolute levels of discretionary accruals — that is, accruals (net income minus cash from operations) that depart from what a model based on average behavior across the industry would predict. This evidence suggests that these firms were more likely to be engaging in practices such as accelerated revenue recognition or “cookie jar” accounting —taking a large charge that can be reversed later to boost earnings — that involve “playing around” within generally accepted accounting principles. While distinct from fraud, this behavior can still distort a firm's profit picture, Nelson points out.
A third test found no relationship between nonaudit fees and another benchmark: whether companies reported slight increases in earnings during the previous year. However, Nelson believes this result reveals more about the target of earnings management than about its incidence. “In this time period, the most important benchmark investors were looking at was whether you beat analysts' expectations,” she observes. Consequently, simply showing growth in earnings was not a top concern.
Nelson is quick to note that these findings do not disprove the argument that providing a broader range of services can improve audit quality, as advocates maintain. “With cross-selling of services, there may very well be knowledge transfer or increases in reputational capital,” she says. “We can't rule that out.” But on average, this study suggests that these benefits are outweighed by an increase in the auditor's incentives to acquiesce to a client's demands.
What's more, even in cases where audit quality remains high, there may be a downside to purchasing audit and nonaudit services from the same firm. After examining the market's first-day reaction to the fee disclosures, Nelson and her colleagues found that firms that reported larger than expected nonaudit fees tended to underperform their peers. This constitutes “modest evidence,” they believe, that investors are beginning to demand compensation for perceived increases in audit quality risk.
“The Relationship Between Auditors' Fees for Non-Audit Services and Earnings Quality,” by Richard M. Frankel, assistant professor of accounting at the MIT Sloan School of Management; Marilyn F. Johnson, associate professor of accounting at Michigan State's Eli Broad Graduate School of Management; and Karen K. Nelson, assistant professor of accounting at Stanford University's Graduate School of Business, will appear in a forthcoming special issue of The Accounting Review.