CHAPTER FOUR RESEARCH DESIGN
FIGURE 1 Event Periods Analysed
4.3 Measures and models for hypothesis testing
4.3.1 Measuring classification shifting
I follow McVay (2006) and measure opportunistic classification shifting by estimating unexpected core earnings, and unexpected change in core earnings, and then investigating the relationship between these measures and firms’ reporting decisions.
70 Recall from Chapter Two that the prohibition on the separate reporting of profit sub-totals before and after abnormal items became operative for financial periods ending on or after 30 June 2001. However, there was concern, particularly by ASIC that firms continued to separately report abnormal items as such or under a different name on the income statement. Subsequently, AASB1018 was further amended by restricting firms from presenting additional ‘material’ specific items and/or subheadings before and more prominently than profit/loss from ordinary activities; net profit/loss; and net profit/loss attributable to members. A sub-total of profit/loss should not be presented immediately before ‘material’ specific items. This second amendment became operative for financial periods ending on or after 30 June 2002.
88
The focus on measures of ‘core earnings’, reflects their importance to market participants. Professional investors and analysts focus on core earnings when estimating the value of the firm because this measure is believed to better represent the firms' ability to generate distributable cash flows in the future (Kinney and Trezevant 1997; Bradshaw and Sloan 2002; Gu and Chen 2004). Also, there is considerable evidence that analysts typically exclude items reported as ‘abnormal’ from their measure of forecast earnings (Fan et al. 2010) and so managers have taken an active role in defining ‘core’ or ‘street’ earnings when communicating their results to investors (Bradshaw and Sloan 2002; Christensen et al. 2011). Hence, managers may have incentive to take advantage of the market’s focus on core earnings instead of bottom line GAAP earnings to misclassify some core expenses opportunistically. As explained in Chapter One, my measure of core earnings is different from McVay’s (2006) core earnings in that the non-recurring items (i.e. abnormal items and discontinued operations) I exclude from core earnings are determined by analysts, whereas McVay’s (2006) measure reflects exclusions of a standard list of items classified as non-recurring (i.e. special items) by COMPUSTAT. Thus, my measure of core earnings is determined by analysts whereas McVay’s (2006) is a function of amounts reported on the face of the income statement and therefore reflects firms’ reporting behavior. Consistent with my measure of core earnings, I examine ‘abnormal items’ as identified by the Morningstar analyst assigned to the firm.
Following McVay (2006), I focus on income-decreasing abnormal items (i.e. expenses), because firms who wish to improve core earnings will more often classify core expenses as abnormal items rather than shifting abnormal revenues upward to operating revenue to be netted against core expenses,71 and because firms are significantly more likely to present income-decreasing abnormal items (AI) on the face of the income statement while income-increasing AI are more often disclosed in the notes to the financial statements (Kinney and Trezevant 1997; Weiss 2001). I expect core earnings for firms that opportunistically misclassify core expenses as AI to be overstated in the period that the AI are reported. Thus, I estimate a model of core earnings and anticipate unexpected
71 McVay (2006: 502) reports evidence consistent with “managers classifying core expenses as special items, increasing both core earnings and income-decreasing special items”. She leaves classification shifting using income-increasing special items to future studies. Houghton (1994) and Cameron and Gallery (2008) present anecdotal evidence that Australian firms reported more income-decreasing abnormal items than income-increasing abnormal items following the change in the definition of extraordinary items. Moreover, Chapter Six of this thesis presents statistical evidence that Australian firms report significantly more income-decreasing abnormal items than income-increasing abnormal items.
89
core earnings (reported core earnings less predicted core earnings) in year t to increase with the magnitude of income-decreasing AI. However, while a positive association between unexpected core earnings and AI is consistent with opportunistic classification shifting, it is also consistent with immediate benefits of real economic events such as efficient restructuring (McVay 2006). To extricate these real economic effects from managers’ opportunistic behaviour, I examine whether the higher than expected core earnings in the current period (year t) reverse in the subsequent period (year t+1). A reversal of unexpected core earnings in year t+1 is consistent with firms engaging in classification shifting because core expenses that are misclassified as abnormal in year t,
should be recorded as core expenses in year t+1 when there are no AI reported. To identify this effect, a model of change in core earnings in year t+1 is estimated where the unexpected change in core earnings in year t+1 (reported change in core earnings less predicted change in core earnings) is expected to decrease with AI.
In the next subsection I present the models of the unexpected core earnings and changes therein, which are the dependent variables for my subsequent hypotheses tests. I then present the regressions to test Hypotheses 1a (H1a) in Sub-section 4.3.3, 1b (H1b) in Sub-section 4.3.4, and 1c (H1c) in Sub-section 4.3.5. Next, I describe models used to identify alternative samples for tests of H1(a-c) and main tests of H2(a-c) in Sub-section 4.3.6. I then discuss in Sub-section 4.3.7 models required to estimate the proxies for alternative EM behaviour (i.e. accrual-based and real earnings management) which are the main explanatory variables in tests of H2(a-c). Models for tests of H2a(i) through to H2b(iii) are then provided in subsection 4.3.8 followed by the model to test H2c(i-iii) in Sub-section 4.3.9. Finally, I discuss sensitivity tests in Sub-section 4.3.10.