Ebook Earnings Management over the Business Cycle
In this paper, we compare the tendency for firms to upwardly manage earnings during good versus bad economic times. In our analysis, earnings management may be achieved by manipulating either accruals or real transactions (e.g., by decreasing discretionary expenditures such as R&D). While there are other forms of earnings managementthan earnings enhancing, such as smoothing or big bath, we focus on upward management because it is the greatest concern to both academics and practitioners. Our investigation is important, because while earnings management has been extensively studied, virtually all research has focused on firm-specific factors, and there has been a dearth of research on how earnings management varies with the state of the economy. Understanding how the tendency to boost earnings changes with macroeconomic conditions would enhance our understanding of this important phenomenon.
Based on the analysis in Conrad et al. (2002) and on relative performance evaluation of managers, we predict that firms have a greater tendency to manage earnings upward during good times. Conrad et al find that investor reaction to earnings disappointments is more adverse during good times. Thus, firms face greater incentives to avoid poor earnings when the economy is up, so they are more prone to boost earnings at such times. As Conrad et al. explain, bad news in good times has such a negative stock price impact, both because it causes investors to revise downward their assessment of the economy, and because it leads to greater uncertainty about economic conditions, resulting in higher discount rates.
Relative performance evaluation of managers also provides a motivation for earnings management when the economy is strong. If managers’ compensation is indeed tied to the performance of specific peer groups, one can argue that when the economy as a whole is performing well, managers will be under pressure to report positive results. Therefore, managers at firms that perform worse than their peers in ‘good times’ will be penalized by the capital markets if their reported earnings fail to meet expectations.
Three central issues in our research are the identification of earnings management, the measure of business conditions, and the presence of time-varying firm characteristics. For earnings management, most studies use an abnormal accruals metric, based on the cross-sectional Jones (1991) model or one of its derivations. We do not use this method for a number of reasons. First, it is infeasible for us, because it produces a mean level of earnings management that is zero in each period, by construction. While using a firm specific time-series version of the Jones model is technically feasible, it has been shown to be less accurate than the cross-sectional version, and eliminates many firms without adequate time-series observations (Dechow et al., 1995; Subramanyam, 1996; Skinner, 1996). More important, any metric based on abnormal accruals ignores earnings management by real activities manipulation, which firms prefer to accrual adjustments (Graham et al., 2005), and which has been shown to be used by firms to achieve earnings targets (Roychowdhury, 2006). Due to these deficiencies in abnormal accruals based proxies, rather than attempt to measure earnings management directly, we infer earnings management based on the tendency of firms to meet or beat targets such as analysts’ consensus forecasts, zero earnings, and the previous quarter’s earnings. As Graham et al. discuss, managers consider meeting or beating benchmarks to be very important, especially the analysts’ consensus and the previous quarter’s earnings. Thus, we follow DeGeorge et al. (1999), Burgstahler and Dichev (1997), Matsumoto (2002), Brown and Caylor (2005), and Roychowdhury (2006), and identify earnings management from firms’ meeting or beating benchmarks.
For business conditions, many studies use measures based on business cycle data
such as growth rates of GDP or industrial production (e.g., Fama, 1981; Lev and
Thiagarajan, 1993; Veronesi, 1999, Chordia and Shivakumar, 2002; Johnson, 1999;
Kothari et al. 2001, and Klein and Marquardt, 2006). A potential problem with such measures for our study is look ahead bias (i.e., when do these figures become publicly available), because macroeconomic data for a given quarter might not be known until many weeks after the end of the quarter, and managers must know the state of the economy for it to affect their earnings management behavior. We address this issue in two ways. First, in addition to using contemporaneous measures of economic growth, we also use lagged measures. Second, we use Conrad et al’s (2002) measure based on the aggregate level of the stock market, relative to next year’s forecasted earnings. Since both components of this ratio are contemporaneously observable, there is no look ahead bias. The P/E ratio may also be a superior proxy for our purposes for two additional reasons. First, if incentives to manage earnings are based primarily on stock price effects, as Graham et al find, a price based measure is a better proxy for earnings management motivations than economic growth rates are. Second, economic growth relates only to P/E’s denominator, but the ratio reflects the market’s valuation of future growth. Conrad et al. (footnote 4) suggest that this may be why P/E performs better than earnings growth in their tests.
Time-varying firm characteristics can be a problem for our tests, because it is well known that certain firm characteristics are associated with the propensity to manage earnings upward, and the presence of these characteristics may be correlated with the state of the economy. For example, Teoh et al. (1998a, b) show that IPOs and SEOs are associated with upward accrual adjustments. These stock issuances are positively correlated with the level of the stock market (Ritter, 2003). Thus, any relation we find between earnings management and economic conditions might be due not to the state of the economy per se, but to time-varying firm characteristics; i.e., our tests may suffer from a correlated omitted variable bias. To address this issue, we control for the level of SEO activity.
Using quarterly data from 1984 thru 2006, we construct measures of earnings management, the market P/E ratio, and production growth rates. Consistent with our predictions, we find that the percentage of firms that meet or beat earnings benchmarks in a given quarter is positively correlated with the aggregate level of the market that quarter, and that this relation is not due to SEO activity. If benchmark beating is due to earnings management, as is widely believed, our results indicate that earnings management activity is positively associated with aggregate market conditions. This is consistent with Conrad et al’s evidence that the market penalty for poor earnings news is greatest when the market is up, and that managers try to avoid this penalty.
The rest of the paper is organized as follows. Section 2 discusses our hypothesis and related research. Section 3 discusses our data and tests. Section 4 reports our results. Section 5 concludes.
Download
PDF Ebook Earnings Management over the Business Cycle
Posted in :
