Stewardship Value of Income Statement Classifications: An Empirical Examination

Abstract

This study investigates the use of earnings components in setting CEO compensation and explores how persistence and controllability affect it. The results indicate that compensation committees do accord differential treatment to earnings components based on their persistence and controllability. Among above the line items, income from continuing items, the most persistent item, also receives the most weight, followed by special items, which have smaller persistence. Furthermore, these weights vary across firms in different stages of life cycle in a manner that mirrors variations in their informativeness and persistence. Weights on special items also exhibit contextual sensitivity based on their type and frequency. Among below the line items, discontinued items, which are at least under partial control of the CEO, receive a positive weight. Except extraordinary items, which are largely uncontrollable, we find no evidence that CEOs are shielded from the income-decreasing effects of any of the earnings components.

Keywords

earnings persistence controllability principle special items discontinued operations extraordinary items income statement classification executive compensation Sarbanes–Oxley Act firm life cycle

Introduction and Literature Review

According to the Financial Accounting Standards Board (FASB, 1978), an important objective of the income statement is “. . . to help investors, creditors, and others assess the amounts, timing, and uncertainty of prospective net cash inflows to the related enterprise” (p. 2). To help achieve this objective, the income statement is presented in a classified form where recurring and nonrecurring components of net income are shown separately. This is done because these items differ in their implications for sustainable firm value. The first item, income from continuing operations, is recurring by nature. The remaining items—special items, discontinued operations, and extraordinary items—are considered nonrecurring as they represent transitory events or transactions that are not expected to recur. Many financial analysts ignore these nonrecurring items in their earnings forecasts. Within the nonrecurring group, items differ considerably in their controllability by the CEO. For example, while CEOs can choose among depreciation or inventory valuation methods, they have little discretion in classifying an item as extraordinary. This study investigates the use of earnings components in setting CEO compensation and explores how persistence and controllability affect it.

Prior research indicates that capital market participants assign a higher response coefficient to earnings components that exhibit greater persistence. Burgstahler, Jiambalvo, and Shevlin (2002) report that special items have smaller persistence than income from continuing operations. The response coefficient on special items is also smaller than income from continuing operations. This is expected if, by presenting them separately, firms convey information about the transient nature of special items to the market. As a result, stock prices reflect relatively more implications of special items for future earnings. Similarly, Elliott and Hanna (1996) report that the response coefficient related to write-offs is significantly smaller than the earnings before write-offs. Ballas (1999) reports that extraordinary items are capitalized at a much lower rate than income from ordinary activities. A separate presentation of nonrecurring items also enhances the predictive ability to forecast future earnings. Fairfield, Sweeney, and Yohn (1996) find that compared to net income alone, income statement classification produces more accurate “out of sample” predictions of future profitability. In their forecasting model, operating earnings are weighed most heavily followed by nonoperating earnings. Special items, though informative about future profitability, receive significantly less weight than either the operating or nonoperating income. Finally, extraordinary items and discontinued operations have no power to forecast future profitability.

Above studies show that a classified income statement is useful for stock valuation and income forecasting. Assuming this occurs because earnings components are informative about the differences in underlying economic events, it is natural to ask whether they contain information about managerial actions that can be used in setting compensation. Early research in this area looks at individual transactions, which typically lack persistence. For example, Healy, Kang, and Palepu (1987) report that when firms switch from the accelerated method to the straight-line method of depreciation, compensation committees do not adjust for the income-increasing effect of such a change. Defeo, Lambert, and Larcker (1989) find that accounting gains associated with equity-for-debt swaps are not excluded from earnings in the compensation function. Dechow, Huson, and Sloan (1994) examine whether restructuring charges are used differently in the compensation function. Their results indicate that restructuring charges are excluded when determining CEO’s cash compensation. Kren and Leauby (1997) report that compensation is shielded from the income-decreasing effects of SFAS No. 106 (Accounting For Post-Retirement Benefits Other Than Pension).

Baber, Kang, and Kumar (1998) find that compensation is positively related to earnings persistence. However, they do not examine individual components of earnings. We simultaneously assess the stewardship value of all four components of earnings—income from continuing operations, special items, discontinued operations, and extraordinary items. To the extent, earnings component differ in their persistence, they provide a practical way for compensation committees to incorporate the distinction between permanent and transitory earnings into compensation decisions. Because the individual components also vary in their potential for managerial manipulation, a separate examination of these items is also useful in assessing the extent to which the potential for manipulation reduces their usefulness as performance measures.

Gaver and Gaver (1998) find that while the gains arising from nonrecurring transactions such as extraordinary events and discontinued operations flow through to compensation, the losses arising from these same transactions do not affect compensation. However, they do not provide a theoretical rationale for this asymmetry. We generate and test hypotheses based on (a) persistence that has been shown to play a role in stock valuation, and (b) controllability that has been shown to play a role in performance evaluation. Reliance on this theoretical framework enables us to make predictions regarding relative magnitude of weights on earnings components in the compensation function.

Our results indicate that compensation committees do accord differential treatment to earnings components based on their persistence and controllability. Income from continuing operations, the most persistent item, also receives the most weight in compensation followed by special items, which have smaller persistence. Discontinued operations receive a positive weight because they arise from the decisions taken by the management. CEO compensation is insulated from the income-decreasing effects of largely exogenous extraordinary items. These findings indicate that by presenting income components separately, the structure of income statement provides a natural vehicle to convey information about their differential persistence and controllability that is utilized by compensation committees.

We extend the existing literature in several important ways. First, we explore the impact of firm life cycle stages on the cross-sectional variation in compensation weights. Life cycle stages succinctly capture shifts in the external environment, organizational strategy, and structure of compensation, all of which have a profound impact on the informativeness and persistence of earnings components and their stewardship value. When we group firms by life cycle stages, a distinct pattern emerges: weights on income from continuing operations and special items rise from introduction to growth and maturity stages and fall in the decline stage. This pattern closely mirrors variations in their informativeness and persistence. To our knowledge, we are the first to report this phenomena.

Second, we conduct a temporal analysis to uncover underlying trends in stewardship value of earnings components. The analysis shows a secular declining trend in the compensation weights on income from continuing operations and special items. A closer look at the period surrounding the passage of the Sarbanes–Oxley Act (SOX) reveals a small spike before the resumption of downward trend. It appears that either SOX did not have any lasting impact or the magnitude of the impact was too small to reverse the secular decline in compensation weights.

Third, we disaggregate special items and report stewardship weights of individual items. The results indicate that individual special items do not exhibit isomorphic relation to compensation and combining disparate individual special items understates their significance. We also find that the compensation weight of special items is conditioned by their frequency and historical pattern. If a firm reports a special item for the first time, it is filtered in the compensation function. However, for serial reporters of special items, subsequent special items affect compensation. This context-sensitive response to special items extends the findings of Adut, Cready, and Lopez (2003) regarding restructuring charges. To our knowledge, this is the first study to perform a detailed analysis of the stewardship value of special items.

Finally, while we find strong evidence supporting persistence and controllability hypotheses, we find little support for the entrenchment phenomena (CEO compensation increasing in profits and positive nonrecurring items but not declining in losses and negative nonrecurring items) documented by Gaver and Gaver (1998). Although differences in time period, sample size, and selection criteria make direct comparisons with their study difficult, many structural changes lead us to believe that the stewardship value of earnings components has changed from their sample period (1970-1993) to the more recent time period (1992-2012) examined in this study. The frequency of special items has risen dramatically over this period (Johnson, Lopez, & Sanchez, 2011). However, extraordinary items have slowly faded out of existence. With the rise of stock-based pay (from 25% in 1992 to 57% in 2008) and corresponding decline of cash compensation, composition of CEO pay has undergone drastic changes (Frydman & Jenter, 2010). Value relevance of earnings (Francis & Schipper, 1999) and earnings persistence have declined, whereas volatility has increased (Liu, 2012). The sample used in Gaver and Gaver’s (1998) study consists of 376 large cap firms that have at least 15 years of time-series data. In contrast, our study is based on a larger sample of 3,257 medium and large cap firms. The sample is also free of survivorship bias because we do not rely on time-series estimates.

The remainder of this article is organized as follows. Next section presents our hypotheses. This is followed by “Research Design” and “Sample” sections. After this, we present our empirical results including robustness checks in the “Results” section. We end the article by presenting summary and concluding thoughts.

Hypothesis Development

From a theoretical standpoint, the use of earnings components in compensation depends on their stewardship value. According to agency theory, the stewardship value of a component depends on how informative it is about the CEO’s efforts. If it takes greater effort to produce persistent earnings, such earnings are likely to receive a higher weight in compensation function than transitory earnings. Baber et al. (1998) show that persistent earnings innovations lead to greater increases in compensation. Their analysis is based on identifying and estimating the time-series parameters of earnings. However, for many firms, the persistence parameter does not converge or there are not enough time-series observations available to estimate it.¹ Even if the persistence parameter can be correctly estimated, it is not clear whether compensation committees actually take the trouble of doing so. To the extent components of earnings displayed on the income statement differ in their degree of transience, they provide a natural and practical alternative vehicle for incorporating persistence into compensation decisions.

Burgstahler et al. (2002) report that income from continuing operations has greater persistence than special items. If compensation committees perceive persistent earnings components as having greater stewardship value, they will assign greater weight to income from continuing operations than to special items. Because other nonrecurring items are transitory, they are expected to be filtered in compensation. This leads to the following hypothesis:

Persistence Hypothesis: Income from continuing operations receives the most weight in compensation. Special items receive a smaller but positive weight. Other nonrecurring items are filtered in compensation.

Persistence hypothesis essentially implies that the weights on earnings components are ordered by persistence—with the most (least) persistent component receiving the highest (lowest) weight. However, to the extent these components are outside the control of the CEO, they also impose risk on the person. A risk-averse CEO needs to be paid higher expected compensation to assure reservation wages.

To reduce compensation cost while preserving incentives for effort, agency theory recommends that a manager should be shielded from the risk imposed by factors beyond his control, particularly, if they do not provide any information about his effort. Management accounting textbooks often recommend that the deviations attributable to uncontrollable factors should be separated during performance evaluation. The implication is that only controllable components of earnings should enter the compensation function. Components of earnings vary considerably in their degree of controllability by the CEO. Income from continuing operations is most susceptible to managerial actions. Extraordinary items are both unusual in nature and are not expected to recur in the foreseeable future. Because only rarely does an event or transaction satisfy both of these criteria, this narrow definition severely restricts the CEO’s ability to manage them and effectively puts them beyond his control.

Special items are either unusual or infrequent and result from a firm’s customary business activities. Companies disclose special items either as a separate line item above the line or in the footnotes of financial statements. These items are controllable in the sense that the CEO has discretion in designating a transaction as a separate line item or merging it with other expenses. This vertical shift does not change the bottom line earnings, but it conveys a distorted picture of economic reality. Prior studies have shown that managers opportunistically shift items within the income statement. Kinney and Trezevant (1997) find that income-decreasing items are more likely to be reported as separate line items whereas income-increasing items are either not reported separately or are reported in footnotes. In other words, managers tend to highlight the transitory nature of expenses, but not income. McVay (2006) reports that unexpected core earnings are increasing in special items. As core earnings exclude special items and depreciation, her finding is consistent with managers classifying core expenses as special items. This also helps managers meet consensus analyst forecast that typically excludes special items and other nonrecurring components of income.

Discontinued operations result when a company eliminates a product line or division from its ongoing operations and disposes of any related assets. If a division or product line continues to suffer from losses, the CEO has little choice but to discontinue it. On the contrary, one can argue that if the CEO takes credit for starting a new division or product line, he or she should also be held accountable for the discontinuation decision.

Although there is ambiguity about managerial motives behind the decision to discontinue operations, it is reasonable to say that they fall between special items and extraordinary items on the spectrum of controllability. If compensation committees weigh controllable items more heavily than uncontrollable items, income from continuing operations and special items will receive greater weight than discontinued operations whereas extraordinary items will receive a zero weight. This leads to the following hypothesis:

Controllability Hypothesis: Income from continuing operations receives the most weight in compensation. Special items receive a smaller weight followed by discontinued operations. Extraordinary items are filtered in compensation.

Both the persistence and controllability hypotheses have largely similar predictions. The only exception is discontinued operations. The controllability hypothesis predicts a positive weight on discontinued operations whereas the persistence hypothesis predicts that they will be filtered in the compensation function. Table 1 summarizes the predicted signs on components of earnings under these two hypotheses.

Table 1.

Predicted Signs on Earnings Components.

	Predicted sign
	Persistence hypothesis	Controllability hypothesis
IC	+	+
SPI	+^a	+^a
DO	0	+^b
XI	0	0

Weight smaller than IC.

Weight smaller than SPI.

Research Design

The focus of this study is to determine the differential impact of earnings components on compensation. CEOs have the greatest ability to influence earnings. If compensation committees make any distinction between earnings components, it is most likely to manifest at the CEO level. For this reason, we restrict the analysis to CEO compensation. To enable comparison with prior literature, compensation is defined as salary plus bonus. Two considerations motivate this choice of compensation measure. First, these elements, at least in part, are usually based on earnings or earnings-related measures (Baber et al., 1998; Sloan, 1993). Huang, Marquardt, and Zhang (2015), using a sample of S&P 500 firms during 1993-2007 period, show that EPS is the most commonly used performance metric (38%) and four of the top five performance measures are earnings- based. A recent survey conducted by a compensation consulting firm (Meridian Compensation Partners, 2016) affirms that “. . . profit measures (e.g., operating income and EPS) remain the most common metric used in annual incentive plans” (p.13). Second, salary and bonus still represent a significant part (45%) of total compensation. We exclude other elements of compensation such as stock options and restricted stock, which are primarily based on stock returns and are designed to impound the long-term consequences of managerial actions.

We use the implicit contract approach to assess the stewardship value of earnings components. In doing so, we follow the prior empirical executive compensation literature (Adut et al., 2003; Baber et al., 1998; Banker, Huang, & Natarajan, 2009; Darrough, Guler, & Wang, 2014; Gaver & Gaver, 1998). Under the implicit contract approach, compensation is regressed on performance measures to obtain their ex post weights. The limitation of this approach is that it ignores variations in earnings measure used by different companies such as EBITDA, ROE, or EVA. However, this only acts as a bias against finding support for our hypotheses. The advantage of this approach is that it factors in the discretion commonly exercised by compensation committees. It also makes it possible to use a large sample to increase the power of tests. Accordingly, we use the following basic model to test our hypotheses. The model is specified in “changes” form to control for factors that vary by firm, such as size and industry, and by CEO, such as qualifications, experience, and stock ownership that can potentially influence compensation.

SALBONU S_{it} = b_{0} + b_{1} I C_{it} + b_{2} NONRE C_{it} + ε_{it}

(1)

In the above equation, SALBONUS is the change in salary plus bonus of CEO. IC is the change in income from continuing operations (excluding special items) and NONREC is the sum of changes in special items, discontinued operations, and extraordinary items. Equation 1 posits that, other things being equal, changes in compensation are a function of changes in earnings components.

We use panel data, which exhibit cross-sectional and intertemporal dependence. To address this issue, we use two-way cluster robust errors (CL-2). Two-way clustering by firm and year allows for both within-firm (time-series) dependence and within-year (cross-sectional) dependence. In other words, the observation for firm j for year t can be correlated to the firm j's observation for year t+ 1 (serial dependence) or to firm k's observation for year t (cross-sectional dependence). Several econometric studies (Cameron, Gelbach, & Miller, 2011; Gow, Ormazabal, & Taylor, 2010; Thompson, 2011) show that CL-2 corrects for both time-series and cross-sectional correlations.

To the extent nonrecurring items in a classified income statement isolate the transitory component of income, the persistence hypothesis predicts 0<b₂<b₁. If compensation committees wish to insulate the CEO from uncontrollable items included in nonrecurring items such as extraordinary items, the controllability hypothesis also predicts b₂ to be smaller than b₁.

Individual nonrecurring items differ in persistence and controllability. Income from continuing operations exhibits the most persistence while special items have a smaller degree of persistence. The remaining nonrecurring items lack persistence. CEOs have greater influence over above-the-line items than below-the-line items. They have some discretion over discontinued operations but little, if any, control over extraordinary items. To examine whether nonrecurring items are weighed disparately by compensation committees, we replace NONREC in Equation 1 with its constituents, namely, special items (SPI), discontinued operations (DO), and extraordinary items (XI) and estimate the following model:

SALBONU S_{it} = c_{0} + c_{1} I C_{it} + c_{2} SP I_{it} + c_{3} D O_{it} + c_{4} X I_{it} + ε_{it}

(2)

Under the null hypothesis that nonrecurring items do not differ in persistence or controllability, the prediction is c₂ = c₃ = c₄. The alternative hypothesis is that the weight depends on persistence and controllability of the item. Special items have smaller persistence than income from continuing operations. Accordingly, the persistence hypothesis predicts c₂<c₁. Because below-the-line nonrecurring items are considered transitory, the persistence hypothesis predicts that these will be filtered in compensation (c₃ = c₄ = 0). If income from continuing operations is considered the most controllable item and discontinued operations are perceived as partially but less controllable than special items, the prediction of the controllability hypothesis is c₃<c₂<c_1. If extraordinary items are perceived as uncontrollable, the controllability hypothesis also predicts c₄ = 0.

Sample

To select the sample, we begin with firms included in the EXECUCOMP database during the 1992-2012 period. To be included in the sample, a firm must also have relevant data on Compustat. Because we require one year of lagged data, our regressions are run for the1993-2012 period. The final sample consists of 3,257 firms (31,619 firm years). All continuous variables are winsorized at 1%.

Table 2 reports the frequency and dollar amount of nonrecurring items. Note that the income-decreasing nonrecurring items far exceed the income-increasing nonrecurring items both in frequency and magnitude. This pattern is observed (in terms of frequency and median) for every year for the sample period. Moreover, the dollar amounts are economically significant. As a percentage of income, median income-increasing (income-decreasing) amounts range from 3.5% to 13.8% (1.6%-10.5 %).

Table 2.

Distribution of Aggregate Nonrecurring Items.

Year	Income-increasing				Income-decreasing
Year	n	M	Median	% of income	n	M	Median	% of income
1993	66	88.9	28.9	11.1	185	300.7	55.0	5.9
1994	163	50.7	13.3	13.8	359	84.0	19.5	9.8
1995	217	83.4	8.5	10.3	575	127.2	17.4	8.3
1996	252	50.4	11.2	8.8	563	70.5	20.1	10.5
1997	243	56.8	12.8	10.9	627	109.9	21.4	8.4
1998	218	172.2	13.3	10.0	741	122.3	24.1	8.9
1999	339	167.8	16.7	9.9	681	86.7	17.3	8.4
2000	348	129.5	11.0	7.1	746	120.0	18.8	5.2
2001	211	99.8	9.7	6.1	1,008	203.4	25.5	4.1
2002	271	73.4	5.5	4.8	935	233.1	16.4	2.3
2003	346	130.3	8.4	4.8	885	101.8	14.6	4.2
2004	356	59.9	7.7	4.6	905	114.9	11.1	3.9
2005	400	65.0	7.7	4.6	887	104.0	12.8	3.8
2006	420	141.0	9.3	4.5	870	74.5	11.1	3.6
2007	440	91.4	9.1	4.9	940	114.8	10.5	3.1
2008	343	79.3	7.2	4.3	1,157	254.9	24.1	1.6
2009	359	126.6	7.9	5.1	1,115	101.5	13.0	2.9
2010	361	94.0	6.6	4.0	1,081	69.0	9.9	3.9
2011	347	61.1	6.7	3.8	1,124	95.6	10.8	3.8
2012	317	72.6	6.2	3.5	1,140	127.4	12.2	4.7
Total	6,017	96.7	8.8	5.9	16,524	128.5	15.4	4.3

Note. Aggregate nonrecurring items are defined as the sum of (a) special items, (b) discontinued operations, and (c) extraordinary items. Income-decreasing amounts are absolute values. All amounts are in millions of dollars and are adjusted for inflation based on 1993 Consumer Price Index.

Table 3 provides a breakdown of nonrecurring items. Special items result from customary business activities and are unusual or infrequent, but not both. Some examples of special items are—write-downs of assets, gain or loss on sale of assets, merger/acquisition costs, and restructuring charges. Panel A of Table 3 presents the descriptive statistics related to special items. The absolute mean (median) amount of income-decreasing special items is $126.2 ($14.9) million, which is much greater than $84.6 ($6.7) million, the mean (median) amounts of income-increasing items. Likewise, income-decreasing special items (16,110) far outnumber income-increasing special items (4,702). Year-by-year analysis shows a similar pattern. If both the income-increasing and income-decreasing items are equally likely to occur, the observed pattern implies that firms try to influence investors’ perception by showing income-decreasing transactions as separate line items. To put it differently, income statement classification provides another vehicle for earnings management.

A discontinued operation results when a company eliminates a component (e.g., a product line or a division) from their ongoing operations and disposes of any related assets. Panel B of Table 3 presents the descriptive statistics related to discontinued operations. In contrast to special items, income-increasing discontinued operations (3,132) outnumber income-decreasing items (2,786). Their magnitude, as measured by the mean (median), is also greater than the income-decreasing items in 16 (14) out of 20 years. This could be either because companies cherry pick components that can be sold profitably or the book value of assets sold is very low. There is an upward trend in the frequency of discontinued operations since 2001. One reason behind it is SFAS No. 144 (Accounting for the Impairment or Disposal of Long-Lived Assets), which requires that the gains or losses arising from disposal of assets be reported as discontinued operations in certain cases even though they may not meet the definition of a business segment.

Table 3.

Distribution of Nonrecurring Items by Type and Amount.

Year	Income-increasing				Income-decreasing
Year	n	M	Median	% of income	n	M	Median	% of income
Panel A: Special items
1993	58	71.0	26.6	10.1	166	294.3	58.0	7.6
1994	142	36.8	11.3	12.2	311	75.0	23.1	14.3
1995	175	45.9	7.7	8.2	513	100.9	19.0	9.3
1996	213	35.3	9.1	7.9	515	74.6	22.1	10.5
1997	206	46.4	9.9	8.9	567	114.3	26.3	9.9
1998	178	166.6	9.5	8.7	689	132.0	25.7	9.2
1999	269	147.9	15.6	8.4	638	90.3	17.7	8.7
2000	275	131.8	9.6	6.4	710	114.9	19.2	6.3
2001	172	72.5	7.7	4.7	967	214.5	26.5	4.9
2002	230	65.3	4.7	4.0	896	210.0	15.9	2.3
2003	269	133.1	5.3	3.5	872	98.0	13.6	3.8
2004	261	49.6	4.9	3.6	897	110.6	10.5	4.2
2005	289	41.7	5.2	3.4	893	102.5	12.6	3.8
2006	312	117.3	5.0	2.6	860	75.0	10.3	3.6
2007	316	67.8	4.3	2.7	955	114.9	10.5	3.2
2008	262	61.5	5.9	3.3	1,161	254.5	23.9	1.7
2009	308	121.4	6.8	5.3	1,098	98.1	13.5	3.0
2010	275	100.5	5.1	3.4	1,094	67.2	9.4	3.9
2011	267	48.7	5.1	2.3	1,137	96.1	10.8	4.0
2012	225	76.0	5.0	2.8	1,171	123.2	11.5	4.3
Total	4,702	84.6	6.7	4.4	16,110	126.2	14.9	4.3
Panel B: Discontinued operations
1993	22	120.9	52.7	8.3	21	146.6	81.5	2.9
1994	47	75.0	22.4	14.7	39	82.2	22.3	5.3
1995	74	158.1	13.7	16.4	66	71.2	11.1	7.8
1996	76	113.2	22.1	20.9	63	49.3	8.3	7.4
1997	81	92.7	18.5	15.5	51	57.4	12.2	11.1
1998	71	179.8	26.9	14.6	73	37.6	13.8	1.0
1999	110	181.5	15.9	10.9	67	60.2	16.5	0.5
2000	92	94.9	11.5	9.1	92	121.1	19.4	3.9
2001	87	223.7	14.1	9.6	82	120.8	18.0	0.3
2002	120	81.4	8.0	2.5	172	204.6	11.7	0.7
2003	171	85.5	11.3	4.1	190	57.5	9.1	1.7
2004	222	57.1	9.3	5.2	191	40.3	5.3	0.9
2005	267	73.1	7.7	3.9	184	33.7	5.5	1.6
2006	285	101.9	9.4	3.9	218	31.3	4.4	1.7
2007	316	86.3	9.1	2.8	215	30.1	5.1	0.6
2008	251	76.1	7.4	1.4	240	34.7	3.5	0.2
2009	190	54.5	3.6	1.0	234	23.1	3.7	0.4
2010	221	49.7	5.8	2.3	208	26.3	2.8	0.7
2011	213	61.9	8.2	3.6	191	19.5	3.1	0.6
2012	216	65.6	7.4	3.0	189	48.6	4.4	1.0
Total	3,132	88.0	9.4	4.1	2,786	53.9	6.4	0.7
Panel C: Extraordinary items
1993	3	29.2	22.3	2.2	48	97.8	15.3	2.8
1994	6	258.9	8.5	16.2	77	73.6	5.3	3.4
1995	14	12.5	2.0	0.2	95	194.5	4.6	5.1
1996	11	2.3	0.9	1.1	98	16.7	4.7	2.9
1997	14	36.7	15.5	2.5	133	37.4	6.5	2.7
1998	21	38.7	3.5	3.7	119	22.1	7.4	2.0
1999	42	63.2	3.3	0.7	106	27.4	6.1	2.1
2000	60	76.1	3.7	1.3	94	13.0	2.6	1.5
2001	67	54.2	5.8	−0.4	141	15.8	3.3	1.0
2002	32	37.1	3.9	−0.1	78	8.3	2.9	0.7
2003	8	125.8	4.4	3.4	5	12.2	5.5	0.1
2004	11	6.3	2.0	0.7	4	370.8	364.8	41.2
2005	11	13.0	6.7	3.9	3	68.1	37.0	2.1
2006	10	9.9	5.0	0.3	2	2.3	2.3	0.5
2007	6	11.7	2.2	6.0	4	66.7	73.2	3.0
2008	4	323.4	7.1	1.8	4	116.6	93.3	1.9
2009	4	40.6	33.3	2.6	3	829.2	3.6	0.2
2010	3	9.1	6.8	6.8	1	344.7	344.7	46.3
2011	4	159.4	130.3	6.2	0	0.0	0.0	0.0
2012	0	0.0	0.0	0.0	0	0.0	0.0	0.0
Total	331	56.5	4.3	0.9	1,015	49.7	5.2	2.0

Note. Income-decreasing amounts are absolute values. All amounts are in millions of dollars and are adjusted for inflation based on 1993 Consumer Price Index. A zero indicates insignificant amount.

Extraordinary items are both unusual in nature and are not expected to recur in the foreseeable future. Because it is rare that an item clearly satisfies both criteria, they are the smallest category among nonrecurring items. Panel C of Table 3 presents the descriptive statistics related to extraordinary items. One of the most common items in this category was gain or loss on debt extinguishment. The SFAS No. 145 (Rescission of FASB Statements No. 4, 44, and 64 and Technical Corrections), which became effective in 2003, eliminated the extraordinary treatment of this item. As a result, fewer items are reported in this category since 2003.

Figure 1 shows that special items constitute the bulk (74%) of all nonrecurring items. This is not surprising given the wide latitude firms enjoy in reporting this item. Among below the line items, discontinued operations (21%) are the most common. As might be expected, extraordinary items (5%) are the least common.

Figure 1.

Frequency of nonrecurring items (1993-2012).

Table 4 presents the descriptive statistics over the entire sample period. Compared to Compustat population, the sample firms are large as indicated by their mean (median) assets of $10.7 ($1.4) billion. This is because the Execucomp database covers only S&P firms, which are dominated by medium and large cap firms. With the exception of discontinued operations, the nonrecurring items are income-decreasing. The mean salary and bonus is close to one million dollars.

Table 4.

Descriptive Statistics.

Variable	Firm- years	M	SD	First quartile	Median	Third quartile
Assets	31,619	10,738.2	59,149.9	452.2	1,378.8	4,896.1
Earnings before nonrecurring items	31,619	285.8	1,197.7	16.7	59.3	202.0
Special items	31,619	−51.7	566.2	−15.8	−0.2	0.0
Discontinued operations	31,619	4.0	146.6	0.0	0.0	0.0
Extraordinary items	31,619	−1.0	55.8	0.0	0.0	0.0
Salary plus bonus	31,619	982.1	1,401.3	438.8	662.1	1,085.0

Note. All amounts are in millions of dollars and are adjusted for inflation based on 1993 Consumer Price Index. A zero indicates insignificant amount. Missing values have been set to zero.

Results

Tests of Hypotheses

The primary focus of this study is to investigate the stewardship value of income statement classification. To the extent nonrecurring items capture the transitory component of earnings, the persistence hypothesis predicts a smaller weight on them. Panel A of Table 5 presents the results. The coefficient on IC, income from continuing operations, is positive and significant. The coefficient on NONREC (the sum of special items, discontinued operations, and extraordinary items) is also positive and significant, though much smaller than IC, and the difference is significant at 0.01 level. Because the persistence of income from continuing operations is greater than nonrecurring items, the results are consistent with the argument that persistent earnings components receive a greater weight in compensation. The results also support the controllability hypothesis because nonrecurring items as a group are only partially controllable by the CEO.

Table 5.

Regression of CEO Compensation on Recurring and Nonrecurring Earnings.

Dependent variable: SALBONUS
	Panel A	Panel B		Panel C	Panel D
IC	2.108***	2.104***	IC	2.333***	2.301***
	(6.999)	(6.994)		(7.237)	(7.234)
NONREC	0.844***		ICNEG	−0.543	−0.493
	(6.524)			(−1.119)	(−1.044)
SPI		0.881***	NONREC	0.820***
		(5.236)		(4.315)
DO		1.741***	NONRECNEG	0.028
		(2.774)		(0.087)
XI		−7.328*	SPI		0.895***
		(−1.870)			(3.687)
Constant	0.036	0.036	SPINEG		−0.049
	(0.810)	(0.816)			(−0.133)
			DO		1.185**
					(2.210)
b ₁ –b ₂	1.265***		DONEG		1.099*
	(5.180)				(1.925)
c ₁ –c ₂		1.222***	XI		15.522*
		(5.510)			(1.700)
c₂–c₃		−0.860	XINEG		−48.615**
		(−1.240)			(−2.457)
c₁–c₃		−0.363	Constant	0.028	0.024
		(−0.49)		(0.532)	(0.460)
n	31,619	31,619		31,619	31,619
Adjusted R²	2.2%	2.2%		2.2%	2.3%
p value	.000	.000		.000	.000

Note. 1. Equations:

$Panel A : SALBONU S_{it} = b_{0} + b_{1} I C_{it} + b_{2} NONRE C_{it} + ε_{it},$

$Panel B : SALBONU S_{it} = c_{0} + c_{1} I C_{it} + c_{2} SP I_{it} + c_{3} D O_{it} + c_{4} X I_{it} + ε_{it},$

$Panel C : SALBONU S_{it} = b_{0} + b_{1} I C_{it} + b_{2} ICNE G_{it} + b_{3} NONRE C_{it} + b_{4} NONRECNE G_{it} + ε_{it},$

$Panel D : SALBONU S_{it} = c_{0} + c_{1} I C_{it} + c_{2} ICNE G_{it} + c_{3} SP I_{it} + c_{4} SPINE G_{it} + c_{5} D O_{it} + c_{6} DONE G_{it} + c_{7} X I_{it} + c_{8} XINE G_{it} + ε_{it} .$

Variable definitions appear in the appendix section.

Significance levels: *p≤ .10. **p≤ .05. ***p≤ .01. t statistics in parentheses.

In the above analysis, nonrecurring items with varying degrees of persistence and controllability have been grouped together. To see whether compensation committees distinguish among nonrecurring items, the variable NONREC is replaced by its constituents, namely, special items (SPI), discontinued operations (DO), and extraordinary items (XI) in Panel B, Table 5. The coefficients on income from continuing operations and special items are positive and significant. Consistent with the persistence hypothesis, the coefficient on SPI is smaller than IC and the difference is significant at 0.01 level.

The coefficient on DO is positive and significant. This is contrary to the persistence hypothesis, which predicts that it will be filtered in the compensation function. However, this is consistent with the controllability hypothesis, which predicts that discontinued operations will receive a positive weight. The coefficient on DO is smaller than IC but greater than SPI, though the differences are not statistically significant. This would imply that compensation committees view discontinued operations at least as controllable as other components of earnings. Contrary to the predictions of persistence and controllability hypotheses, the coefficient on XI is negative and weakly significant (0.10 level) indicating that CEO compensation is insulated to some extent from extraordinary items.² We discuss this further in the next section.

Above results show remarkable sophistication on the part of compensation committees. It appears that they go beyond the bottom line earnings to take into account both persistence and controllability when assigning weights to earnings components.

Income-Increasing Versus Income-Decreasing Earnings Components

Nonrecurring items can be either income-increasing or income-decreasing. Although not supported by any theory, it is possible that income-increasing items receive a differential treatment in compensation than income-decreasing items. As Gaver and Gaver (1998) argue, this could happen if, due to their superior bargaining power or influence on the board of directors, CEOs manage to immunize their compensation against income-decreasing transactions while benefiting from income-increasing transactions. This entrenchment argument implies that while positive nonrecurring items will increase compensation, negative items are either filtered or receive a smaller weight.

To see whether the direction of income effect of nonrecurring items has any impact on compensation weights, two variables are added to Equation 1: income from continuing operations (ICNEG) and nonrecurring items (NONRECNEG) when they are negative and zero otherwise. As shown in Panel C of Table 5, the coefficients on both ICNEG and NONRECNEG are insignificant. There is no indication that negative items receive any different weight than positive items.

As nonrecurring items differ in their degree of controllability, it is possible that compensation is shielded from some negative items but not others. To examine this possibility, four variables are added to Equation 2: income from continuing operations (ICNEG), special items (SPINEG), discontinued operations (DONEG), and extraordinary items (XINEG) when they are negative and zero otherwise. The results, displayed in Panel D of Table 5, indicate that the coefficients on ICNEG and SPING are indistinguishable from zero. Thus, two major components of earnings, income from continuing operations and special items, flow through to compensation unaffected by sign.

Contrary to the entrenchment argument, the coefficient on DONEG is positive and significant at 0.10 level indicating that the weight on discontinued operations is greater when they are negative. The coefficient on XINEG is negative and significant at 0.05 level indicating that the weight on extraordinary items is reduced when they are negative. As this category includes uncontrollable losses such as natural disasters and expropriation of assets, the interpretation is that compensation committees shield CEO compensation from exogenous losses to reduce the risk imposed on the CEO.

To summarize, we do not find any evidence that the CEO is insulated from the income-decreasing impact of special items and discontinued operations. This implies that either CEOs are not entrenched, or they do not use their influence over the board of directors to engage in opportunistic behavior.

Impact of Firm Life Cycle on Stewardship Value of Earnings Components

Life cycle theory posits that a firm’s evolution follows distinct stages, each with a unique set of attributes (e.g., technology, customer demand, and competition). According to the tenets of optimal contracting theory, the board will adjust the stewardship weights on performance measures to motivate the CEO to adapt strategy, structure, and decision processes to meet the unique challenges of the life cycle stage. As the firm transitions from one life cycle stage to another, shifts in the external environment, organizational strategy, and structure of product market alter the informativeness and persistence of earnings. If compensation committees value these properties, then we expect concomitant shifts in the role of earnings in compensation. Below we describe different life cycle stages and how they affect the stewardship value of earnings.

Wang and Singh (2014) describe four distinct stages³ in the life of a firm (introductory, growth, maturity, and decline) and their impact on compensation design. Introductory firms have small or nonexistent sales volume. They spend significant resources on commercializing the innovation and prospecting customers. At this stage, accounting earnings are often negative or lack persistence and, given their historical focus, are yet to reflect the managerial efforts. Therefore, earnings receive a lower weight in compensation. Growth firms invest heavily in productive capacity and distributional channels as they broaden product offerings and enter regional and national markets. At this stage, earnings exhibit strong upward trajectory and the firm may even generate small profits. The weight on accounting earnings increases as they become persistent and provide a better signal of managerial efforts. However, they do not yet fully capture the investments in incremental innovation, brand development, distribution channels, and other growth-focused activities. Mature firms achieve increased profitability as investments made in the growth stage bear fruit and economies of scale are realized. They sustain profitability by seeking operational efficiencies and cost reductions. Earnings now fully reflect the impact of these measures. Earnings persistence is also at peak driven by stable demand and competitive advantages obtained in earlier stages. This is corroborated by Dickinson (2011) who finds that mature firms exhibit the highest level of persistence of return on net operating assets (RNOA). Thus, we expect that earnings will also receive the most weight at this stage. Firms in the decline stage suffer from loss of market share and decreasing profits. The CEO is now tasked with reviving the firm by undertaking measures such as controlling costs, overhauling existing product lines, and rebuilding competitive advantage. As turnaround efforts are not reflected in earnings of the current period, and decline in sales and profits weakens earnings persistence, we expect a diminished reliance on earnings in the compensation function at this stage.

To summarize, informativeness and persistence of earnings rise during introduction, growth, and maturity stages, and fall in the decline stage. This phenomena has also been confirmed by prior studies. Jenkins, Kane, and Velury (2004) find that the information content of earnings, as measured by the earnings response coefficient (ERC), increases from growth to maturity stage and declines in the stagnant stage. Drake (2015) reports that earnings persistence rises during introductory, growth, and maturity phases and then falls as firms enter the decline phase.⁴ If compensation committees value these properties, then we expect stewardship weights also to vary with the life cycle stage of the firm in a manner that corresponds to the pattern of variations in informativeness and persistence. In other words, we expect weight on earnings to increase during introductory, growth, and maturity phases and to fall in the decline stage.

To test it empirically, we categorize sample firms in life cycle stages using a cash flow proxy⁵ based on Dickinson (2011). Her proxy has the advantage of parsimoniously capturing the life cycle stage based on publicly available information about operating, investing, and financing cash flows and allows nonsequential movement of firms between life cycle stages. Next, we estimate the regression model of Equation 2 separately for each life cycle group. Our main interest is in the persistent components of earnings, namely, income from continuing operations and special items. The results are presented in Table 6. The first column reports results for the entire sample, which are similar to those reported in Panel B, Table 5. However, when we group firms by their life cycle stage, a distinct pattern emerges. The weight on income from continuing operations is 1.148 at the introductory stage and rises to 3.034 for growth firms. The increase is significant at 0.01 level. The coefficient dips to 2.592 for maturity firms. Although the growth firms seem to reward income from continuing operations more than the mature firms, the difference is statistically insignificant. From the maturity to decline stage, the coefficient falls rather sharply from 2.592 to 1.301. The decrease is significant at 0.05 level. The weight on special items, which have smaller persistence, follows the predicted pattern. However, only the difference between maturity and decline stage is statistically significant at 0.01 level. To sum up, the evidence is consistent with the notion that the weights assigned to income from continuing operations and special items by compensation committees are conditioned on the informativeness and persistence of earnings at any given life cycle stage.

Table 6.

Regression of CEO Compensation on Recurring and Nonrecurring Earnings by Firm Life Cycle Stages.

Dependent variable: SALBONUS
	All	Introduction	Growth	Maturity	Decline
IC	2.155***	1.148***	3.034***	2.592***	1.301***
	(7.169)	(4.109)	(5.974)	(5.091)	(5.683)
SPI	0.883***	0.532*	0.897***	1.555***	0.400**
	(5.232)	(1.645)	(3.455)	(5.065)	(2.176)
DO	1.886***	4.315***	1.486	2.209***	1.214
	(3.192)	(2.994)	(1.129)	(2.760)	(1.481)
XI	−6.995*	23.843**	−19.097**	−4.489	−2.385
	(−1.794)	(2.437)	(−2.182)	(−0.574)	(−0.257)
Constant	0.033	0.037	0.030	0.020	0.053
	(0.729)	(0.886)	(0.721)	(0.408)	(1.055)
n	30,608	1,450	10,155	14,941	4,062
Adjusted R²	.024	.013	.038	.027	.012
p value	.000	.000	.000	.000	.000

Note. 1. Life cycle stages are based on Dickinson (2011).

Equation:

$SALBONU S_{it} = c_{0} + c_{1} I C_{it} + c_{2} SP I_{it} + c_{3} D O_{it} + c_{4} X I_{it} + ε_{it} .$

Variable definitions appear in the appendix.

Significance levels: *p≤ .10. **p≤ .05. ***p≤ .01. t statistics in parentheses.

Impact of SOX on Stewardship Value of Earnings Components

In 2002, Congress passed the landmark SOX, which requires CEOs and CFOs of publicly held companies to certify the accuracy and reliability of financial statements issued by their company. Those intentionally certifying false statements face severe civil and criminal penalties. To the extent these severe consequences reduce management’s incentives to manipulate earnings, the reported earnings quality is expected to increase in the post-SOX period. Ceteris paribus, an improvement in earnings quality in the post-SOX period should lead to an increase in stewardship weights. Consistent with this, Carter, Lynch, and Zechman (2009) find that bonus weights are higher for earnings increases in the post-SOX period. Chen, Jeter, and Yang (2015) report that pay-performance sensitivity, using either market-based or accounting-based measures of performance, increased significantly following the passage of SOX. However, Chang, Choy, and Wan (2012) find that pay-performance sensitivity decreased substantially in the post-SOX period indicating a weaker incentive alignment between shareholders and the CEO. One explanation for these conflicting findings is that managers use accruals-based earnings management (AEM) and real earnings management (REM) as substitutes (Zang, 2012). The heightened scrutiny of earnings in the post-SOX period may have encouraged managers to substitute AEM with REM to achieve earnings benchmarks (Cohen, Dey, & Lys, 2008). As REM causes deviations from optimal production, investment, and financing decisions and has an adverse impact on future profits and cash flows (Vorst, 2016), compensation committees are likely to decrease reliance on earnings to discourage the use of REM. Thus, stewardship weights can decline even in the presence of reduced earnings management.

In view of the conflicting findings of prior studies and the ambiguous impact of SOX on stewardship value of earnings, we do not offer any directional predictions. Our main interest is to assess what impact, if any, this landmark legislation had on the stewardship value of earnings components. To accomplish this, we divide our sample into two subperiods: 1992-2005 and 2006-2012, as SOX became fully effective only after 2005.⁶ We create a dummy for SOX that equals 1 if the year is 2006 or greater and zero otherwise. We interact this dummy with the variables of interest in Equations 1 and 2.

The results, presented in Table 7, indicate that in the post-SOX period, the weights on income from continuing operations (IC) and nonrecurring items (NONREC) have decreased by 2.337 and 0.646, respectively, and the decrease is significant at 0.01 level (Panel A, Table 7). An analysis of individual nonrecurring items shows that there is no change in the weights on discontinued operations and extraordinary items (Panel B, Table 7). Therefore, the decline in the weight on NONREC observed in Panel A of Table 7 is entirely driven by special items.

Table 7.

Regression of CEO Compensation on Recurring and Nonrecurring Earnings—Pre- and Post-SOX Period.

Dependent variable: SALBONUS
	Panel A	Panel B
IC	2.995***	2.991***
	(13.078)	(13.232)
NONREC	1.100***
	(6.146)
SOX	−0.212**	−0.213**
	(−2.340)	(−2.343)
SOX × IC	−2.337***	−2.341***
	(−7.963)	(−8.028)
SOX × NONREC	−0.646***
	(−2.776)
SPI		1.217***
		(5.097)
DO		2.008**
		(2.242)
XI		−7.912*
		(−1.658)
SOX × SPI		−0.817***
		(−2.677)
SOX × DO		−0.569
		(−0.521)
SOX × XI		19.639
		(1.070)
Constant	0.120***	0.120***
	(6.509)	(6.589)
n	31,619	31,619
Adjusted R²	4.3%	4.4%
p value	.000	.000

Note. 1. Equations:

$Panel A : SALBONU S_{it} = f_{0} + f_{1} I C_{it} + f_{2} NONRE C_{it} + f_{3} SO X_{it} + f_{4} SO X_{it} * I C_{it} + f_{5} SO X_{it} * NONRE C_{it} + ε_{it},$

$\begin{matrix} Panel B : SALBONU S_{it} = g_{0} + g_{1} I C_{it} + g_{2} SO X_{it} + g_{3} SO X_{it} * I C_{it} + g_{4} SP I_{it} + g_{5} D O_{it} + g_{6} X I_{it} + g_{7} SO X_{it} * SP I_{it} + \\ g_{8} SO X_{it} * D O_{it} + g_{9} SO X_{it} * X I_{it} + ε_{it} . \end{matrix}$

Variable definitions appear in the appendix.

Significance levels: *p≤ .10. **p≤ .05. ***p≤ .01. t statistics in parentheses.

A few studies have documented a decline in stewardship weights on earnings during the period overlapping with our study. Banker et al. (2009) report that both the value relevance and stewardship value of earnings decreased from the 1993-1997 period to the 1998-2003 period. Liu (2012), covering a longer time period of 1977-2006, finds that the stewardship weights on earnings have declined over time. This raises the possibility that the observed post-2005 shift in the slope coefficients on IC and SPI is a reflection of a secular declining trend rather than the passage of SOX. To investigate this, we run annual regressions of Equation 2. We then compute three-year average of coefficients to smooth out year-to-year fluctuations. Only IC and SPI coefficient averages are computed as no clear trend is discernible in DO and XI coefficients. The plot, presented in Figure 2, shows a steady decline in stewardship weights that begins long before SOX and continues after the passage of SOX. A closer examination of annual coefficients during the period surrounding the passage of SOX reveals a small spike before the resumption of downward trend.

Figure 2.

Three-year average coefficients—IC and SPI.

Overall, the analysis presented here indicates that, at best, SOX only had a transient impact on stewardship weights, and the small effect size was unable to overcome their long-term downward trend. We do not investigate the reasons for this decline as this is beyond the scope of our study.

Stewardship Value of Individual Special Items

Prior to 2001, Compustat reported special items as one category, which included many different components. Beginning in 2001, Compustat provides a detailed breakdown and separately reports eight types of special items: (a) acquisition/merger costs (AQP), (b) gain/loss on sale of assets (GLP), (c) impairment of goodwill (GDWLIP), (d) litigation/insurance settlement costs (SETP), (e) restructuring costs (RCP), (f) write-downs (WDP), (g) extinguishment of debt (DTEP), and (h) in-process research and development (RDIP). Remaining special items are combined together as other special items (SPIOP).

Prior research has focused primarily on individual special items, such as restructuring charges and goodwill impairment, partly due to data constraints. In this section, we provide preliminary evidence on the stewardship value of different types of special items described above. As the primary focus of this study is on persistence and controllability of earnings components, a detailed examination of special items, which happen to be numerous and require firm-specific context for any prediction, is beyond the scope of this study. To see which special items have stewardship value, we replace SPI with its constituents and repeat our tests for 2001-2012 period.

The results, presented in Table 8, indicate that individual special items exhibit considerable variation in their stewardship weights ranging from positive (RCP, WDP, and SPIOP) to negative (AQP and DTEP). Therefore, combining disparate special items into one category, as has been done in prior research, significantly understates their true economic impact on compensation.

Table 8.

Regression of CEO Compensation on Recurring and Nonrecurring Earnings With Special Item Categories (2002-2012).

Dependent variable: SALBONUS
IC	1.374***
	(3.827)
AQP	−4.815***
	(−2.917)
GLP	2.477
	(1.637)
GDWLIP	0.317
	(1.043)
SETP	−0.452
	(−0.386)
RCP	3.239***
	(4.282)
WDP	1.434***
	(2.582)
DTEP	−5.383**
	(−2.152)
RDIP	4.600
	(0.859)
SPIOP	4.068***
	(3.424)
DO	2.094***
	(3.759)
XI	−9.204
	(−0.910)
Constant	−0.009
	(−0.124)
n	18,043
Adjusted R²	1.3%
p value	.000

Note. 1. Equation:

$\begin{matrix} SALBONU S_{it} = h_{0} + h_{1} I C_{it} + h_{2} AQ P_{it} + h_{3} GL P_{it} + h_{4} GDWLI P_{it} + h_{5} SET P_{it} + h_{6} RC P_{it} + h_{7} WD P_{it} + \\ h_{8} DTE P_{it} + h_{9} RDI P_{it} + h_{10} SPIO P_{it} + h_{11} D O_{it} + h_{12} X I_{it} + ε_{it} . \end{matrix}$

Variable definitions appear in the appendix.

Significance levels: *p≤ .10. **p≤ .05. ***p≤ .01. t statistics in parentheses.

Looking at individual items, a positive and significant coefficient on restructuring costs (RCP) indicates that these costs flow through to compensation. This differs from the results reported by Dechow et al. (1994) but is consistent with Adut et al. (2003) who find that the CEO is only partially shielded from these charges. The significant negative coefficient on acquisition/merger costs (AQP) is somewhat puzzling as it would mean that compensation is adjusted upward for incurring such costs. One possible explanation is that CEOs often receive handsome bonuses explicitly tied to successful completion of merger/acquisition deals (Grinstein & Hribar, 2004). These rewards are given to mitigate the underinvestment by risk-averse managers so that they undertake potentially value-enhancing but risky merger/acquisitions. A closer look at the accrual-based special items reveals an interesting pattern. Write-downs (WDP), for which independent sources of measurement (e.g., market values or appraiser estimates) are more readily available, flows through to compensation. However, impairment of goodwill (GDWLIP), which is based on management’s estimates of fair value and lacks independent or market-based measures, is filtered in compensation. Thus, we are unable to confirm the findings of Darrough et al. (2014) who report that goodwill impairment leads to a significant reduction in CEO cash compensation. To summarize, the evidence is consistent with the view that compensation committees calibrate their response to special items based on their attributes rather than a “one size fits all” approach.

Initial and Subsequent Special Items

Among nonrecurring items, special items occur most frequently (Panel A, Table 3), representing almost three-fourths of all nonrecurring items (Figure 1). The evidence in Adut et al. (2003) indicates that the weight placed by compensation committees on restructuring charges, one of the special items, is based on its frequency and historical pattern. The initial restructuring charge is completely filtered. If another restructuring charge occurs within two years of a prior restructuring charge (subsequent proximate restructuring charge), it flows through to compensation. Any restructuring charge, after the initial charge, taken after a gap of two years (subsequent nonproximate restructuring charge) is only partially filtered in compensation. To see whether special items exhibit a similar contextual sensitivity, we replace SPI in Equation 2 with three variables—INI_SPI, PROXIMATE_SPI, and NONPROXIMATE_SPI. INI_SPI is the first special item reported by a firm during the sample period.⁷ PROXIMATE_SPI is the special item reported within two years of reporting another special item by the same firm. All other special items are classified as NONPROXIMATE_SPI. This results in the following model:

SALBONU S_{it} = d_{0} + d_{1} I C_{it} + d_{2} INI_SP I_{it} + d_{3} PROXIMATE_SP I_{it} + d_{4} NONPROXIMATE_SP I_{it} + d_{5} D O_{it} + d_{6} X I_{it} + ε_{it}

(3)

The results are reported in Table 9. The coefficients on proximate and nonproximate special items are significant and positive whereas the coefficient on initial special items is insignificant. This indicates that compensation committees modify the income number used in compensation for special items based on their frequency and historical pattern. When a firm reports special items for the first time, it is filtered in the compensation function. However, when it habitually reports special items, compensation committees do not shield the CEO compensation against them. A plausible explanation that is also consistent with the persistence hypothesis is that occasional special items are transitory and, therefore, are filtered in the compensation. Frequent special items, on the contrary, have some degree of persistence and, therefore, receive a positive weight in compensation. These findings extend the results reported by Adut et al. (2003) as they show that the contextual response of compensation committees extends to the entire category of special items.

Table 9.

Regression of CEO Compensation on Recurring and Nonrecurring Earnings with Initial, Proximate, and Nonproximate Special Items.

Dependent variable: SALBONUS
IC	2.096***
	(6.934)
INI_SPI	0.408
	(0.966)
PROXIMATE_SPI	0.835***
	(5.002)
NONPROXIMATE_SPI	1.279*
	(1.705)
DO	1.728***
	(2.748)
XI	−6.966*
	(−1.816)
Constant	0.037
	(0.831)
n	31,619
Adjusted R²	2.1%
p value	.000

Note. 1. Equation:

$\begin{matrix} SALBONU S_{it} = d_{0} + d_{1} I C_{it} + d_{2} INI_SP I_{it} + d_{3} PROXIMATE_SP I_{it} + d_{4} NONPROXIMATE_SP I_{it} + d_{5} D O_{it} + \\ d_{6} X I_{it} + ε_{it} . \end{matrix}$

Variable definitions appear in the appendix.

Significance levels: *p≤ .10. **p≤ .05. ***p≤ .01. t statistics in parentheses.

Robustness Checks

Annual regressions

In the analysis presented above, there is potential for bias in coefficient estimates because of possible cross-sectional dependence in error terms of the regression. Bernard (1987) suggests that if each annual regression is independent, the coefficients of annual regressions can be used for tests of significance. In other words, the bias arising from cross-sectional dependence will not affect the statistical significance of the coefficients. Following his suggestion, Models 1 and 2 are estimated annually. Means of the annual regression coefficients are used as parameter estimates. Tests of significance are conducted using standard error of the time series of estimates. The results are qualitatively similar to those reported for the pooled data.

Recession years

As our sample period includes periods of recession, we test our hypotheses excluding recession years 2001, 2008, and 2009.⁸ Our primary inferences in Table 5 remain unaffected. In fact, the magnitude of all coefficients (except XI) increases.

Exclusion of compensation from income

Payment of executive compensation reduces profitability. Thus, regressing compensation on earnings suffers from a mechanical negative dependence. As an empirical matter, this negative mechanical dependence should lead to negative coefficients. Most of the reported coefficients, however, are either significantly positive or nonnegative. To mitigate any residual concerns, we run regressions after adding back compensation to IC. The (untabulated) results do not lead to any change in our inferences. The only difference is that now ICNEG becomes much smaller and significant.

Total compensation.⁹

In this section, we report regression results based on total compensation, the broadest measure of compensation. One rationale for combining all elements of compensation is that current theories of compensation are not rich enough to specify which components of compensation will be influenced by a specific performance measure. The drawback of this approach is that by design, components of compensation are structured to respond differentially to earnings and other performance measures. Aggregating them into a single measure of total compensation weakens the relation between compensation and performance measures.¹⁰ The (untabulated) results indicate that the coefficient on income from continuing operations is positive and significant and its magnitude is now much greater. This is expected as the dependent variable is of much greater magnitude. The coefficients on SPI, DO, and XI are now statistically insignificant indicating their impact on total compensation is undetectable.

Summary and Conclusions

This study examines how components of earnings reported on a classified income statement enter the compensation function and the influence of persistence and controllability on their stewardship value. The results correspond to our intuition regarding persistence and controllability of earnings components. Income from continuing operations is the most persistent component and it also receives the highest weight in compensation. Special items have smaller persistence than income from continuing operations and consequently receive a smaller weight. This is also consistent with the controllability perspective because the CEO has great latitude in designating special items. Discontinued items receive a positive weight because they are controllable to some extent by the CEO. Except extraordinary items, which are largely uncontrollable, we find no evidence that CEOs are shielded from the income-decreasing effects of any of the earnings components.

We also find that the compensation weights on income from continuing operations and special items have been declining over time and the passage of SOX did not have any lasting impact on this secular downward trend. When we group firms by life cycle stages, we find that the compensation weights on income from continuing operations and special items rise from introduction to growth and maturity stages and fall as firms enter the decline stage, a pattern that mirrors variations in their informativeness and persistence. This evidence is consistent with the notion that stewardship value of these two components is conditioned by the informativeness and persistence of earnings at a given life cycle stage.

As special items are the most prominent among nonrecurring items, we select them for further examination. Similar to Adut et al. (2003), we find that compensation committees modify weight on special items based on their frequency and historical pattern. When we disaggregate special items by type, we find considerable variation in their stewardship weights.

In conclusion, the structure of income statement conveys information about the persistence and controllability of earning components that is utilized by compensation committees. This highlights a much more nuanced role for income statement classification than reported in prior literature. Our findings imply that adding special items as a separate category may enhance the stewardship value of income statement classification and eliminating extraordinary items, as done by FASB,¹¹ may reduce it.

Footnotes

Appendix

Variable Definitions.

AQP	Change in acquisition/merger costs
DO	Change in discontinued operations
DONEG	Change in discontinued operations if negative, zero otherwise
DTEP	Change in extinguishment of debt
GDWLIP	Change in impairments of goodwill
GLP	Change in gain/loss on sale of assets
IC	Change in income from continuing operations (excluding special items)
ICNEG	Change in income from continuing operations (excluding special items) if negative, zero otherwise
INI_SPI	Change in special items if it is the first special item reported by the firm during the sample period, zero otherwise
NONPROXIMATE_SPI	Change in special items if the firm did not report a special item in previous two years and it is not the first special item reported by the firm during the sample period, zero otherwise
NONREC	Change in nonrecurring earnings (defined as the sum of special items, discontinued operations, and extraordinary items)
NONRECNEG	Change in nonrecurring earnings (defined as the sum of special items, discontinued operations, and extraordinary items) if negative, zero otherwise
PROXIMATE_SPI	Change in special items if the firm reported a special item in previous two years, zero otherwise
RCP	Change in restructuring costs
RDIP	Change in in-process research and development expense
SALBONUS	Change in salary plus bonus of CEO
SETP	Change in litigation/insurance settlement costs
SOX	1 if the financial year>2005, zero otherwise
SPI	Change in special items
SPINEG	Change in special items if negative, zero otherwise
SPIOP	Change in other special items
WDP	Change in write-downs
XI	Change in extraordinary items
XINEG	Change in extraordinary items if negative, zero otherwise

Note. All variables are deflated by prior year total assets except SALBONUS, which is deflated by prior year total salary.

Acknowledgements

We appreciate the helpful comments of workshop participants at the University at Buffalo, Morgan State University, University of Michigan, and the American Accounting Association Annual Meeting.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Adut

Cready

W. H.

Lopez

T. J.

(2003). Restructuring charges and CEO cash compensation: A reexamination. The Accounting Review, 78, 169-192.

Baber

W. R.

Janakiraman

Kumar

K. R.

(1996). Investment opportunities and the structure of executive compensation. Journal of Accounting and Economics, 21, 297-318.

Baber

W. R.

Kang

S. H.

Kumar

K. R.

(1998). Accounting earnings and executive compensation: The role of earnings persistence. Journal of Accounting and Economics, 25, 169-193.

Ballas

A. A.

(1999). Valuation implications of exceptional and extraordinary items. The British Accounting Review, 31, 281-295.

Banker

R. D.

Huang

Natarajan

(2009). Incentive contracting and value relevance of earnings and cash flows. Journal of Accounting Research, 47, 647-678.

Bernard

V. L.

(1987). Cross-sectional dependence and problems in inference in market-based accounting research. Journal of Accounting Research, 25, 1-48.

Burgstahler

Jiambalvo

Shevlin

(2002). Do stock prices fully reflect the implications of special items for future earnings?Journal of Accounting Research, 40, 585-612.

Cameron

A. C.

Gelbach

J. B.

Miller

D. L.

(2011). Robust inference with multiway clustering. Journal of Business & Economic Statistics, 29, 238-249.

Carter

M. E.

Lynch

L. J.

Zechman

S. L. C.

(2009). Changes in bonus contracts in the post-Sarbanes–Oxley era. Review of Accounting Studies, 14, 480-506.

10.

Chang

Choy

H. L.

Wan

(2012). Effect of the Sarbanes–Oxley Act on CEOs’ stock ownership and pay-performance sensitivity. Review of Quantitative Finance and Accounting, 38, 177-207.

11.

Chen

Jeter

Yang

(2015). Pay-performance sensitivity before and after SOX. Journal of Accounting and Public Policy, 34, 52-73.

12.

Cohen

D. A.

Dey

Lys

T. Z.

(2008). Real and accrual-based earnings management in the pre- and post-Sarbanes–Oxley periods. The Accounting Review, 83, 757-787.

13.

Darrough

M. N.

Guler

Wang

(2014). Goodwill impairment losses and CEO compensation. Journal of Accounting, Auditing & Finance, 29, 435-463.

14.

Dechow

Huson

M. R.

Sloan

R. G.

(1994). The effect of restructuring charges on executives’ cash compensation. The Accounting Review, 69, 138-156.

15.

Defeo

V. J.

Lambert

R. A.

Larcker

D. F.

(1989). Executive compensation effects of equity-for-debt swaps. The Accounting Review, 64, 201-227.

16.

Dickinson

(2011). Cash flow patterns as a proxy for firm life cycle. The Accounting Review, 86, 1969-1994.

17.

Drake

K. D.

(2015). Does firm life cycle inform the relation between book-tax differences and earnings persistence? (Working paper). Retrieved from https://ssrn.com/abstract=2547778

18.

Elliott

J. A.

Hanna

J. D.

(1996). Repeated accounting write-offs and the information content of earnings. Journal of Accounting Research, 34(Supplement), 135-155.

19.

Fairfield

P. M.

Sweeney

R. J.

Yohn

T. L.

(1996). Accounting classification and the predictive content of earnings. The Accounting Review, 71, 337-355.

20.

Financial Accounting Standards Board. (1978). Objectives of financial reporting by business enterprises (Statement of Financial Accounting Concepts No. 1). Norwalk, CT.

21.

Francis

Schipper

(1999). Have financial statements lost their relevance?Journal of Accounting Research, 37, 319-352.

22.

Frydman

Jenter

(2010). CEO compensation (Working paper). Cambridge, MA: MIT Sloan School of Management.

23.

Gaver

J. J.

Gaver

K. M.

(1998). The relation between nonrecurring accounting transactions and CEO cash compensation. The Accounting Review, 73, 235-253.

24.

Gow

I. D.

Ormazabal

Taylor

D. J.

(2010). Correcting for cross-sectional and time-series dependence in accounting research. The Accounting Review, 85, 483-512.

25.

Grinstein

Hribar

(2004). CEO compensation and incentives: Evidence from M&A bonuses. Journal of Financial Economics, 73, 119-143.

26.

Gupta

Y. P.

Chin

D. C. W.

(1994). Organizational life cycle: A review and proposed directions. The Mid-Atlantic Journal of Business, 30, 269-294.

27.

Healy

P. M.

Kang

S. H.

Palepu

K. G.

(1987). The effect of accounting procedure changes on CEOs’ cash salary and bonus compensation. Journal of Accounting and Economics, 9, 7-34.

28.

Huang

Marquardt

C. A.

Zhang

(2015). Using sales revenue as a performance measure (Working paper). Retrieved from https://ssrn.com/abstract=2636950

29.

Jenkins

D. S.

Kane

G. D.

Velury

(2004). The impact of the corporate life-cycle on the value-relevance of disaggregated earnings components. Review of Accounting and Finance, 3(4), 5-20.

30.

Johnson

P. M.

Lopez

T. J.

Sanchez

J. M.

(2011). Special items: A descriptive analysis. Accounting Horizons, 25, 511-536.

31.

Kinney

Trezevant

(1997). The use of special items to manage earnings and perceptions. The Journal of Financial Statement Analysis, 3, 45-53.

32.

Kren

Leauby

(1997). The effect of SFAS No. 106 on chief executive compensation (Working paper). Milwaukee, WI: University of Wisconsin–Milwaukee.

33.

Levie

Lichtenstein

B. B.

(2010). A terminal assessment of stages theory: Introducing a dynamic states approach to entrepreneurship. Entrepreneurship Theory and Practice, 34, 317-350.

34.

Liu

(2012). Changes in the role of earnings in compensation over the past three decades. The University of Texas at San Antonio. San Antonio, TX: Available from ProQuest Dissertations.

35.

McVay

S. E.

(2006). Earnings management using classification shifting: An examination of core earnings and special items. The Accounting Review, 81, 501-531.

36.

Meridian Compensation Partners. (2016). 2016 Trends and Developments in Executive Compensation Survey. Retrieved from http://www.meridiancp.com/insights/thought-leadership/2016-trends-developments-executive-compensation/

37.

Sloan

R. G.

(1993). Accounting earnings and top executive compensation. Journal of Accounting and Economics, 16, 55-100.

38.

Thompson

S. B.

(2011). Simple formulas for standard errors that cluster by both firm and time. Journal of Financial Economics, 99, 1-10.

39.

Vorst

(2016). Real earnings management and long-term operating performance: The role of reversals in discretionary investment cuts. The Accounting Review, 91, 1219-1256.

40.

Wang

Singh

(2014). The evolution of CEO compensation over the organizational life cycle: A contingency explanation. Human Resource Management Review, 24, 144-159.

41.

Zang

A. Y.

(2012). Evidence on the trade-off between real activities manipulation and accrual-based earnings management. The Accounting Review, 87, 675-703.

Stewardship Value of Income Statement Classifications: An Empirical Examination

Abstract

Keywords

Introduction and Literature Review

Hypothesis Development

Research Design

Sample

Results

Tests of Hypotheses

Income-Increasing Versus Income-Decreasing Earnings Components

Impact of Firm Life Cycle on Stewardship Value of Earnings Components

Impact of SOX on Stewardship Value of Earnings Components

Stewardship Value of Individual Special Items

Initial and Subsequent Special Items

Robustness Checks

Annual regressions

Recession years

Exclusion of compensation from income

Total compensation. 9

Summary and Conclusions

Footnotes

Appendix

Acknowledgements

Declaration of Conflicting Interests

Funding

Notes

References

Total compensation.⁹