The earnings management literature has considered many non-accrual approaches to estimate earnings quality. Ronen and Yaari (2008) identify three prominent ones: a single- statement item, examining the distribution of firms’ earnings, and examining the distribution of the second digit of the decimal in the earnings per share (EPS) calculation. Other alternative measures of earnings quality are also discussed.
The analysis of a single account focuses on the examination of a particular income statement line item. Generally, for a specific line item to be chosen its manipulation must have a material impact on earnings and managers must legally be able to alter the line item within the boundaries of accounting standards. Ronen and Yaari (2008) further note that the examination of a single account (rather than total accruals) can be advantageous as the analyst understands the accounting model and can easily estimate what the unmanaged expense ought to be. Daley and Vigeland (1983) and Aboody and Lev (1998) compare the treatment of research and development (capitalising or expensing) to companies’ dividend restrictions. Strong and Meyer (1987) and Cohen, Dey and Lys (2005) examine inventory write-offs whilst McNichols and Wilson (1988) investigate bad debt expense and Roychowdhury (2006) analyses the loss reserves of property and casualty insurance companies. Stubben (2010) investigates earnings management using a discretionary revenue measure (more specifically premature revenue recognition) and finds that the revenue model is more likely to detect revenue and expense manipulation than an accruals model. However, examining a single accounting item at a time gives a limited picture of a firm’s earnings management practices.
The distributional approach is supported by Burgstahler and Dichev (1997). The approach assumes that unmanaged earnings follow a Gaussian (more commonly known as normal) distribution, and the evidence for earnings management is the deviation of observed earnings
Page | 41 from the distribution. Studies that have used this approach, for example Jacob and Jorgensen (2007), show that there is a kink in the distribution function of the earnings management object around the benchmark.
Thomas (1989) considers the “rounding EPS” approach. In this approach, firms express their EPS as a decimal, with two digits after the point: xx.yz. Thomas (1989) considers that firms can alter EPS by changing z so that the EPS amount can be rounded up or down depending on whether they wish to increase or decrease EPS. Researchers test the distribution of z by comparing the actual object of earnings management with the hypothesised distribution free of earnings management. Das and Zhang (2003) and Kinnunen and Koskela (2003) consider similar methodologies using United States firms and on an international basis respectively. Das and Zhang (2003) find, after examining the first digit after the decimal point y, that firms adjust y by more or less four digits.
Beneish (1997) formulates a probit model designed to detect firms falling victim to earnings management by distinguishing manipulated from non-manipulated earnings using 1989 companies during the 1987-1993 period. The Beneish (1997) model is based on variables assessing the likelihood of detection and variables that are used as proxies for their ability and incentive to violate accounting standards. The variables assessing the likelihood of detection are: days’ sales in receivables, gross profit margin, asset quality, depreciation, sales, general and administration expense, and total accruals to total assets. The variables used as proxies for their ability and incentive to violate accounting standards are: capital structure, prior market performance, time listed, sales growth and a dummy variable for accrual and declining cash sales. Morgan Stanley (2011) uses a similar methodology to develop an earnings quality score.
Skinner (2008), on the other hand, uses dividend information to test the quality of earnings. Dividend theory would suggest that part of the basis for a firm’s dividend declaration is an evaluation of how sustainable its earnings will be. Because high quality earnings would exhibit high persistence, it would follow that dividend policy should have a good correlation with earnings quality. Skinner (2008) finds that firms that pay dividends have a stronger relationship between current and future earnings than firms that do not. The size of the dividends is also important; firms that tend to pay large dividends are inclined to have higher earnings quality. Furthermore, Skinner (2008) also provides evidence of a size effect; large
Page | 42 firms that pay high dividends have better earnings quality than smaller companies that pay high dividends.
DeFond (2010) provides a potentially attractive alternative to accruals as a proxy for earnings quality: earnings restatement and accounting, and auditing enforcement release. DeFond (2010, p.404) argue that an advantage of using companies subject to regulatory- imposed restatements is that they are direct proxies of earnings management since they are “actual events, rather than error terms from a statistical method that cannot be validated”. However, the disadvantage of this method is that it only captures poor earnings quality that has both been discovered and judged to merit restatement from governing bodies.
Lastly, Dichev, Graham, Harvey and Rajgopal (2013, p.2) use a different approach to evaluate earnings quality since “archival research cannot satisfactorily parse out the portion of managed earnings from the portion resulting from the fundamental earnings process”. The authors provide insights, through surveys and interviews from the direct producers of earnings quality, such as Chief Financial Officers (CFOs), investment managers and analysts. Dichev et al. (2013) show that CFOs consider that most earnings management occurs in an attempt to sway share prices in order to meet earnings benchmarks or to influence their compensation contracts. But they point out that about half of the factors responsible for earnings quality are beyond management control, for example business models, industry, and macro-economic conditions. Dichev et al. (2013, p.1) also find evidence that roughly 60 percent of earnings management is income-increasing and the remainder is income-decreasing.