The aim of this paper is to investigate the extent to which benchmark beating by Australian firms can be reliably interpreted as evidence of earnings management. Researchers’ interest in evidence of benchmark beating likely reflects increasing interest in capital market incentives for earnings management (Dechow and Skinner, 2000). Managers frequently claim that they manage earnings in order to avoid so called “torpedo effects” associated with failing to achieve simple benchmarks such as negative rather than positive earnings, or avoiding a negative earnings change or surprise.
Several United States-based studies indicate that an unusually large number of firms manage to report a small (i.e., positive) income or a small increase in earnings. Despite differences in firm size, and institutional differences such as accounting standards, enforcement procedures, minimum financial reporting period and the extent of analyst following, Holland and Ramsay (2003) report similar results for Australian firms, and interpret this as evidence of earnings management by Australian firms.
However, recent United States–based evidence has called into question the extent to which benchmark beating can, in fact, be attributed to earnings management. For example, Dechow, Richardson and Tuna (2003) show that benchmark beating firms do not have unusually positive accruals, which is difficult to reconcile with earnings management-based explanations for apparent benchmark beating. Beaver, McNichols and Nelson (2003b) and Durtschi and Easton (2004) provide evidence consistent with several alternative explanations for discontinuities of the type highlighted by Holland and Ramsay (2003). These explanations include the use of price as a scaling mechanism, sample selection biases, the asymmetric treatment of gains and losses for tax purposes and the effect of special items. Given the various differences between Australian and United States financial reporting, we believe it is important to subject Australian evidence of benchmark beating to careful analysis. Following Dechow et al., our primary focus is on measures of unexpected accruals, although we also consider the possible effect of scaling on the extent of possible discontinuities in earnings levels and changes.
For the period 1993-2002, we initially demonstrate evidence comparable to Holland and Ramsay (2003) of an unusually large number of Australian firms reporting a small profit or small increase in earnings. More importantly though, we provide evidence that suggests caution is warranted before such evidence can be interpreted as being attributable to earnings management. Although we find that benchmark beaters have unusually high unexpected accruals relative to all other firms, we also find that firms that just miss the relevant benchmarks have similarly large unexpected accruals when compared with other firms. When we compare the unexpected accruals of the benchmark beaters and just miss groups, there are no statistically significant differences. Our results are robust to several methods for estimating unexpected accruals, as well as alternative interval widths. In most respects, our results confirm the United States-based evidence of Dechow et al. (2003), who also show that benchmark beating firms have similar unexpected accruals to those that just miss the target.
We effectively test a joint hypothesis, namely that any kink in the earnings distribution around a benchmark is evidence of earnings management, and that our measures of unexpected accruals are able to detect such earnings management. Hence, the findings we report could be because our models of expected accruals lack sufficient power to detect earnings management that is driving the kink. Alternatively, our models of expected accruals correctly identify earnings management, and earnings management is not significantly associated with the kink. To help distinguish between these explanations, we utilise several alternative models of expected accruals, as well as examining the persistence of unexpected accruals and the future performance of firms that have unusually high unexpected accruals. We also consider the distribution of earnings per share and unscaled (i.e., raw) earnings. These additional tests provide some support for our view that the kink in earnings distributions is a relatively poor proxy for earnings management, as well as suggesting that that the kink is at least partially attributable to the scaling of earnings by measures such as lagged assets or price.
Our paper also provides important evidence on the applicability of several models of expected accruals to Australian data, especially when they are estimated in cross-section and confined to same-industry data. There have been relatively few studies of earnings management by Australian firms using models of expected accruals of the type developed by Jones (1991), and they have typically used small samples or focussed on very specific events, such as earnings management around CEO changes (Wells, 2002). Our evidence suggests that Australian-based models of expected accruals have higher explanatory power to those used in extant United States-based studies, although as with these studies we also note that a minority of estimations yield counter intuitive coefficient signs. We also provide some preliminary evidence on the time series behaviour of benchmark beating, and show that the frequency of benchmark beating has declined for both earnings levels and changes.
The remainder of the paper proceeds as follows. Section two provides some background discussion of approaches to identifying earnings management, and especially the potential difficulties associated with the benchmark beating approach. Section three explains our data sources and measures of unexpected accruals, as well as presenting evidence of benchmark beating similar to Holland and Ramsay (2003). In section four we provide results of tests that examine unexpected accruals for various groups of firms, particularly benchmark beaters and just miss firms. We also provide some evidence on the properties of unexpected accruals, and the time-series behaviour of benchmark beating. Section five concludes and considers implications for both existing and future research.
Download
PDF Ebook Is “Benchmark Beating” by Australian Firms Evidence of Earnings Management?
