The Strategic Reporting of Special Items: Does Management

The Strategic Reporting of Special Items: Does Management Presentation Reflect Underlying Firm Performance or Opportunism?

Edward J. Riedl *

Harvard Business School

Suraj Srinivasan University of Chicago

August 2006

ABSTRACT: This paper investigates whether manager discretion regarding presentation within the financial statements reflects the firm’s underlying economic performance or opportunistic motivations. In particular, we examine managers’ decision to present recognized special items either as a separate line item on the income statement or aggregated within another line item with identification only via footnote disclosure. Our study is motivated by standard-setting interest in financial statement presentation, as well as prior research investigating the causes and effects of management presentation behavior. We first confirm the findings of prior research that special items have lower persistence than other components of income (i.e., earnings before special items), and then provide three new insights. First, we demonstrate that special items presented separately on the income statement have lower persistence than those revealed only in the footnotes, consistent across the sign, magnitude, and category of reported special items. Second, we demonstrate that special items revealed only in the footnotes have similar persistence as earnings before special items. Finally, we provide similar inferences examining an alternative disclosure mechanism: annual press releases. Overall, our analyses are consistent with managers using this presentation decision primarily to reflect underlying firm performance, and thus provide an alternative perspective from prior research suggesting that management presentation (e.g., Schrand and Walther 2000) and classification (e.g., McVay 2006) of such items tends to be opportunistic. Keywords: special items, strategic reporting, presentation We thank the following individuals for their useful comments and discussions on previous versions of this manuscript: Mark Bradshaw, Michael Kimbrough, S.P. Kothari, Ben Lansford, Asis Martinez-Jerez, Greg Miller, Ray Pfeiffer, Doug Skinner, Mohan Venkatachalam, Jim Wahlen, Greg Waymire, two anonymous reviewers, and seminar participants at Harvard Business School, University of Massachusetts – Amherst, Michigan State, the UNC/Duke Fall Camp, and the AAA 2006 Annual meeting. We obtained analyst data from I/B/E/S. We gratefully acknowledge the contribution of I/B/E/S International Inc. for providing earnings-per-share forecast data, available through the International Brokers Estimate System. These data have been provided as part of a broad academic program to encourage earnings-expectation research. We also thank Claire Chiron, Maylene Han, Susanna Kim, and Grace Lin for excellent research assistance. * corresponding author: Morgan Hall 365 Boston, MA 02163 617.495.6368 617.496.7363 fax [email protected]

1

The Strategic Reporting of Special Items: Does Management Presentation Reflect Underlying Firm Performance or Opportunism?

ABSTRACT

This paper investigates whether manager discretion regarding presentation within the financial statements reflects the firm’s underlying economic performance or opportunistic motivations. In particular, we examine managers’ decision to present recognized special items either as a separate line item on the income statement or aggregated within another line item with identification only via footnote disclosure. Our study is motivated by standard-setting interest in financial statement presentation, as well as prior research investigating the causes and effects of management presentation behavior. We first confirm the findings of prior research that special items have lower persistence than other components of income (i.e., earnings before special items), and then provide three new insights. First, we demonstrate that special items presented separately on the income statement have lower persistence than those revealed only in the footnotes, consistent across the sign, magnitude, and category of reported special items. Second, we demonstrate that special items revealed only in the footnotes have similar persistence as earnings before special items. Finally, we provide similar inferences examining an alternative disclosure mechanism: annual press releases. Overall, our analyses are consistent with managers using this presentation decision primarily to reflect underlying firm performance, and thus provide an alternative perspective from prior research suggesting that management presentation (e.g., Schrand and Walther 2000) and classification (e.g., McVay 2006) of such items tends to be opportunistic.

2

The Strategic Reporting of Special Items: Does Management Presentation Reflect Underlying Firm Performance or Opportunism?

I. INTRODUCTION

This paper investigates whether manager discretion regarding presentation of recognized

special items within the financial statements reflects underlying firm performance or

opportunistic motivations. In particular, we focus on management’s decision to present

recognized special items either as a separate line item on the income statement (income

statement presentation) or aggregated into another line item with identification only via footnote

disclosure (footnote presentation).1

Prior research suggests that managers engage in “strategic” reporting behavior, with

evidence generally consistent with opportunistic behavior regarding the presentation and

classification of reporting items (e.g., Schrand and Walther 2000; Doyle et al. 2003; McVay

2006). We re-examine whether strategic reporting, in the context of presentation within the

financial statements, reflects economic or opportunistic motivations. By economic, we suggest

managers use income statement presentation as a mechanism to assist users in better identifying

and understanding the firm’s underlying performance: that is, managers highlight certain special

items on the income statement because they have differing properties from other items affecting

income (including special items aggregated into other reporting lines). By opportunistic, we

suggest managers use this presentation decision to influence perceptions of the firm’s

performance in a biased way. This dual perspective is consistent with prior research suggesting

that the reporting of special items and “pro forma” earnings is driven by both economic factors

as well as reporting incentives (e.g., Francis et al. 1996; Riedl 2004; Lougee and Marquardt

1 Note that under both presentation choices, the special item is recognized (that is, reflected in net income). Thus,

this presentation choice focuses on the extent to which management highlights the charge on the income statement for financial statement users. This is similar to the notion of classification shifting (e.g., McVay 2006).

3

2004; Bowen et al. 2005). In addition, a fuller understanding of managers’ presentation behavior

appears warranted, as prior research documents that presentation can affect users’ judgments

(e.g., Hirst and Hopkins 1998; Maines and McDaniel 2000; Libby et al. 2005).

We choose special items as our context as they have been shown to have differing

properties relative to other components of income (e.g., Lipe 1986), have been increasing in

frequency and magnitude (e.g., Elliott and Hanna 1996), and are heterogeneous across a number

of characteristics (e.g., Francis et al. 1996; Burgstahler et al. 2002). In addition, our examination

of manager presentation behavior regarding these items – which are often characterized as “non-

recurring” – may provide insights relevant to other current and future financial reporting

requirements likely to introduce similar non-recurring items (such as fair value accounting).

We conduct two sets of analyses examining the presentation of special items on the

income statement: we examine the determinants of income statement versus footnote

presentation; and then we examine whether the persistence of special items varies systematically

across this presentation decision. We first confirm the findings of prior research that special

items have lower persistence than other components of income (i.e., earnings before special

items). We then leverage our data and research design to provide two primary insights. First, we

demonstrate that special items receiving income statement presentation have lower persistence

than those receiving footnote presentation. Second, we demonstrate that special items presented

on the income statement have lower persistence than earnings before special items, while those

revealed only in the footnotes have similar persistence as earnings before special items. These

findings are consistent across a variety of characteristics relating to the special items, including

their magnitude, directional impact on net income (income-increasing or income-decreasing),

and category (restructuring, write-offs, other). In addition, managers appear to respond to

4

demands for transparent information, being more likely to highlight special items when

institutional ownership and/or analyst following is higher. Our only evidence supporting

opportunistic behavior in the presentation decision is that managers are more likely to separately

present special items that cause the firm to miss an earnings benchmark, consistent with

managers emphasizing favorable performance targets (e.g., Schrand and Walther 2000). This

limited evidence is in contrast to prior research, which offers general support for managers acting

opportunistically in their presentation decisions.

A third set of analyses examines the consistency of this presentation decision within an

alternative disclosure mechanism: press releases during annual earnings announcements. Results

provide corroborating evidence that the management decision to discuss/identify special items in

the press release is driven by similar economic factors as the income statement presentation

decision. In particular, special items highlighted in press releases are also less persistent than

those not discussed in press releases.

Overall, these results provide evidence that managers are able to identify ex ante those

special items charges that are more transitory in nature, and choose to highlight these on the

income statement. This is consistent with managers using discretion over presentation of a

substantial reporting item within the financial statements – special items – primarily to reflect

underlying firm performance. This provides an alternative perspective to prior research

documenting opportunistic behavior of managers in their presentation and classification of

reporting items (e.g., Schrand and Walther 2000; McVay 2006). These findings may be of

interest to standard-setters, as they continue to deliberate issues surrounding financial statement

presentation (e.g., FASB 2006; IASB 2006), as well as reporting topics likely to increase the

5

occurrence of so-called “non-recurring” charges (such as the increased use of fair values and

impairment testing).

The remainder of this paper is organized as follows. Section 2 discusses related prior

research. Section 3 presents our hypothesis development, research design, and empirical results

for analyses examining the causes underlying managers’ presentation decision for special items.

Section 4 presents the same for related analyses examining how the persistence of reported

special items varies across this presentation decision. Section 5 examines the consistency of this

presentation across alternative disclosure channels and through time. Section 6 concludes.

II. PRIOR RESEARCH

Our paper builds on two streams of literature: managers’ strategic reporting decisions

(including pro forma reporting), and the characteristics of special items. Several recent papers

examine management’s strategic reporting and presentation decisions, generally providing

evidence of managers engaging in opportunistic behavior. Much of this research centers on the

disclosure mechanism of press releases and related pro forma disclosures. Bradshaw and Sloan

(2002), using I/B/E/S analyst forecast data, documents an increase in the provision of “pro

forma” or street earnings, which often involves restating “core” earnings to exclude non-

recurring charges (particularly negative, or income-decreasing, charges). Schrand and Walther

(2000) documents that managers opportunistically select the prior-period earnings amount used

as a benchmark to evaluate current-period earnings. They find that managers are more likely to

separately announce a prior-period gain from the sale of assets than a loss, thereby reducing the

benchmark used to evaluate current earnings. Weiss (2001), examining the reporting effects of

the 1993 change in corporate income tax rates, similarly finds that managers are more likely to

6

separately disclose negative than positive non-recurring items in press releases, consistent with

managers attempting to highlight the negative items as transitory or non-core expenses. McVay

(2006) broadens this to a more general setting, providing evidence consistent with managers

opportunistically shifting reported expenses from core expenses (such as cost of goods sold) to

special items, thereby overstating “core” earnings. Doyle et al. (2003) provide evidence that so-

called “non-recurring” items excluded from GAAP earnings in the definition of “pro forma”

earnings have substantial predictive ability for future cash flows and thus implications for future

firm performance.2

In some contrast to the above studies, two papers provide evidence that “pro forma”

reporting reflects elements of both underlying economic performance and opportunistic

behavior. In particular, Lougee and Marquardt (2004) documents that firms with low GAAP

earnings informativeness are more likely to disclose pro forma earnings, consistent with

motivations to reflect the firm’s performance. However, the paper also reveals that the direction

of the GAAP earnings surprise is an important determinant in the decision to provide pro forma

earnings, consistent with opportunistic motivations. Similarly, Bowen et al. (2005) provide

evidence that managers emphasize metrics that portray more favorable firm performance;

however, these same metrics are also more value relevant.

A number of studies also examine the properties of special items. Work by Fairfield et

al. (1996) reveals that the disaggregation of earnings (e.g., into operating earnings, non-operating

earnings and taxes, and special items) improves forecasts of one-year ahead return-on-equity,

suggesting such classifications serve to improve the predictive content of reported earnings.

2 While the I/B/E/S data used in Bradshaw and Sloan (2002) and Doyle et al. (2003) is not a source of management

reporting choices per se, both papers indicate a high correlation exists between the I/B/E/S data and that contained in management press releases, suggesting the content is similar and thus likely reflects management’s own estimates. This is more directly confirmed in Bhattacharya et al. (2003).

7

Francis et al. (1996) document that short-window market reactions vary depending on the type of

special item charge; for example, the market reacts negatively to announcements of write-offs of

inventory, but positively to restructuring charges. Finally, Burgstahler et al. (2002) find that

special items are more transitory in nature compared to other earnings components. In addition,

they also find this varies across positive and negative special items.

Our paper adds to these literatures by examining whether the presentation of special

items within the financial statements reflects economic and/or opportunistic behavior by

managers. Such analysis provides further insights into motivations surrounding managers’

strategic reporting behavior, and increases our understanding of the properties of reported special

items.

III. THE PRESENTATION OF SPECIAL ITEMS

ON THE INCOME STATEMENT

Hypothesis Development and Research Design

As defined by Accounting Principles Board 30 – Reporting the Results of Operations,

special items are charges that are infrequent or unusual in nature.3 Proper identification and

labeling of this type of charge is likely relevant for financial statement users, as these items have

differing properties from other components of earnings (e.g., Lipe 1986). However, special

items are not a homogenous set of reporting events, ranging from write-offs to restructuring

charges to gains/losses on sales of assets (e.g., Francis et al. 1996). Further, prior research

indicates that management emphasis has the capacity to affect how users interpret and weight

presented financial information (e.g., Hirst and Hopkins 1998; Maines and McDaniel 2000).

3 Note that our definition of special items excludes discontinued operations, extraordinary items, and effects of

changes in accounting principle. All three qualify for specific treatment under US GAAP: each must be disclosed separately, net of applicable taxes, on the income statement below income from continuing operations.

8

Accordingly, we examine how management chooses to present special items within the financial

statements.4

Management presents special items in one of two ways: as a separate line item on the

income statement, with discussion in the footnotes (i.e., income statement presentation); or

aggregated within another line item on the income statement, with identification and discussion

of the special items only via the footnotes (footnote presentation).5 As argued above, this

presentation choice can potentially affect the costs to investors to identify, interpret, and weight

the implications of special items for the firm.6 Thus, we first examine the determinants of

management presentation of special items, classifying them into three groups: characteristics of

the special items, earnings management incentives, and demand for information.

First, we examine whether characteristics of the special items may influence this

presentation decision. We begin with the premise that income statement presentation serves to

alert financial statement users that the special item has characteristics that differ from other

reporting items. One such characteristic is the item’s usefulness in predicting the firm’s future

performance. Accordingly, we conjecture that managers – utilizing their available (public and

private) information and absent incentives to the contrary – are more likely to highlight those

special items having lower (expected) persistence. Accordingly, we predict the following:

4 In addition, even if these presentation decisions have no impact on users, managers may still make these

decisions in the belief that they can affect users. 5 We note that management has other disclosure methods to highlight reporting items (including special items),

such as through discussion within earnings announcements and/or during conference calls. Given the prominence of the income statement, we believe our heuristic relating to presentation choice on the income statement remains of interest. Nonetheless, we consider these alternative disclosure mechanisms in Section 5.

6 Note our paper differs somewhat from prior research examining the recognition versus disclosure of items (e.g., Aboody 1996, which examines the effects of macroeconomic events in the oil and gas industry under the full cost versus successful-efforts method; and Aboody et al. 2004, which examines the voluntary recognition of stock-option compensation expense). As explained in footnote 1, both presentation choices in our context include an item that is recognized and reflected in net income; the choice is the extent to which management highlights the charge on the income statement for financial statement users.

9

H1A: The likelihood of management presenting special items separately on the income statement is decreasing in the persistence of the special items.

Related, we argue that the effect of the special item on net income (i.e., whether it is a

positive special item (PSI) or negative special item (NSI)) affects this presentation decision.

This may reflect either economic or opportunistic motivations. Relating to economic

motivations, PSI may have differing properties than NSI (e.g., Burgstahler et al. 2002), which

management may wish to highlight. Relating to opportunistic motivations, management may

wish to downplay income-decreasing special items (i.e., “bad news” such as write-offs),

suggesting NSI are less likely to receive income statement presentation than PSI. This would be

consistent with management trying to soft-pedal negative events relating to the firm.

Alternatively, management may prefer to highlight NSI relative to PSI as a mechanism to label

the former as more transitory, suggesting NSI are more likely to receive income statement

presentation (e.g., Weiss 2001). This would also be consistent with prior research documenting

management’s use of negative special items in “big bath” type contexts (e.g., Riedl 2004). This

discussion results in the following two-tailed hypothesis:7

H1B: The likelihood of management presenting PSI as a separate line item on the income statement differs from that for NSI.

We then consider two proxies for reporting incentives, focusing on whether the special

item causes the firm to miss or beat prior year’s earnings, as prior research has documented that

the effect of special items on relevant benchmarks affects manager’s presentation in press

releases (Schrand and Walther 2000). We use prior year’s earnings (measured before special

items) as a benchmark, versus other constructs such as consensus forecasts, because there is

7 This analysis extends Kinney and Trezevant (1997), who document that NSI are more likely to be shown on the

face of the income statement than PSI. While these latter findings are suggestive of a one-tailed test (i.e., that NSI are more likely to receive income statement presentation), the latter paper performs only a univariate analysis on a limited sample of firms.

10

variation in how analysts include/exclude special items in their forecasts (e.g., Gu and Chen

2004). We include two proxies for reporting incentives. The first is an indicator variable

capturing whether the special item causes the firm to miss prior year’s earnings. The predicted

sign is positive, as management would wish to highlight that such an item is transitory, and thus

should not be considered a part of current year’s “core earnings” (e.g., McVay 2006). The

second is an indicator variable capturing whether the special item causes the firm to beat prior

year’s earnings. The predicted sign is negative, as management would wish to de-emphasize that

beating prior year’s earnings is attributable to a current year special item (similar to Schrand and

Walther 2000). Note that each variable captures specific distributions of the (net) NSI and PSI,

respectively. For example, while all NSI reduce earnings definitionally, only a subset of firms

reporting NSI will miss prior year’s earnings as a direct result of reported NSI. This discussion

leads to the following hypotheses:

H1C: The likelihood of presenting a special item separately on the income statement is higher if the special item causes the firm to miss an earnings benchmark.

H1D: The likelihood of presenting a special item separately on the income statement is lower if the special item causes the firm to beat an earnings benchmark.

Finally, we consider demand effects that may affect managers’ presentation of special

items. In particular, management presentation decisions may reflect their expectations regarding

users’ preferences for information and how it is presented. Institutional owners represent a

group of sophisticated users, which managers likely perceive as willing to penalize them if there

is any evidence of non-transparent reporting. Thus, we predict:

H1E: The likelihood of management presenting a special item separately on the income statement is increasing in the (perceived) demand for transparent information, proxied by institutional ownership.8

8 Alternatively, one could consider that institutional investors would prefer less transparent financial information,

as they may be better able to process and analyze information relative to non-institutional investors. We do not consider this in our hypothesis H1E (and thus do not propose a two-tailed test), as we think it more likely that

11

To test these hypotheses, we employ the following model:

SI_Sepjt = δ0 + δ1YEAR t + δ2SIZEjt + δ3SI_MAGjt + δ4SI_PERSISTjt

+ δ5NSIjt + δ6MISS_PYEjt + δ7BEAT_PYEjt + δ8INSTjt + φjt (1)

We run this model under two specifications. First, under an OLS specification, we define the

dependent variable SI_Sep as the percentage of special items reported within a separate line item

on the income statement for firm j in fiscal year t. Specifically, SI_Sep equals the absolute

amount of special items identified in a separate line item on the income statement, divided by the

absolute amount of total reported special items: thus the variable ranges from 0 to 1 in value,

inclusive. Second, owing to the properties of our dependent variable (i.e., which has significant

clustering of values at the end points of its distribution), we also examine a logistic specification.

In this regression, we define SI_Sep as an indicator variable equal to 1 when any reported special

items receive income statement presentation, and 0 when all reported special items receive

footnote presentation. This is akin to suggesting that management highlighting of any special

item serves as a “red flag” for users to look for other related items.9 Note that positive and

negative special items are not netted in calculating these dependent variables.

Our model includes three control variables. First, we include YEAR (the year) as the

likelihood of presenting special items separately on the income statement may be increasing over

time, coincident with the rise of pro forma reporting over our sample period (e.g., Bradshaw and

Sloan 2002). Second, we include SIZE (the log of firm j’s year t sales) to control for the impact

managers make presentation decisions reflecting concerns of potential downside risks (e.g., in the form of penalties for a lack of transparency) than expectations of potential upside from being rewarded by certain subsets of users for lower transparency.

9 We also run an alternative logistic specification, defining SI_Sep as an indicator variable equal to 1 if management presents all reported special items as a separate line item on the income statement, and 0 if management presents all reported special items only within the footnotes. Thus, this latter specification excludes observations wherein management adopts a mixed presentation (i.e., presents some special items separately on the income statement, and aggregates some within other line items with identification only via the footnotes). As discussed later, results are unchanged under this specification.

12

of differing investing and information environments across variously sized firms on the

presentation decision. Finally, we include SI_MAG (firm j’s total special items divided by total

assets at the beginning of year t), as the likelihood of reporting special items as a separate line

item should be increasing in the magnitude of the special item, consistent with materiality

affecting management reporting and presentation decisions.

Among our experimental variables, SI_PERSIST is the persistence parameter for special

items. As discussed above, this is intended to capture the notion of economic content in the

reported special item. To reflect this notion, we estimate a persistence parameter by industry-

year, which should capture the economics of similar firms in similar time periods.10 Specifically,

we use Compustat data to run annual regressions for each 3-digit SIC industry of our sample,

thus obtaining a vector of year-industry specific persistence parameters for special items. The

regressions are of earnings (before special items) for year t+1 regressed onto earnings (before

special items) for year t and special items for year t (all scaled by beginning market value of

equity).11 A negative coefficient would be consistent with managers being less likely to

separately present special items that have higher persistence.

10 Alternatively, we could attempt to estimate firm level parameters. However, the limited observations per firm in

our setting (a maximum of 10 observations per firm – representing the 10 years of our sample data) makes this latter estimation less desirable. Further, we conjecture that use of an industry level measure is consistent with managers considering the economic behavior of special items within peer firms to assess their persistence.

11 That is, we run the following regression by industry i and year t: E*jt+1 = λ0 + λ1E*jt + λ2SIjt + νjt where E*jt+1 (E*jt) is earnings before special items for firm j for year t+1 (year t), and SIjt is reported special

items per Compustat for firm j for year t. Thus, SI_PERSIST is the vector of λ2 obtained from the i X t industry-year regressions.

We use Compustat data to obtain this variable, versus our smaller sample of hand-collected special items data, principally to increase the sample size, enabling us to obtain industry level and year-specific parameters. This appears reasonable for constructing SI_PERSIST, as the correlation between total special items for Compustat and our hand-collected sample is very high (> 95%). Note, for industry-years having less than 20 observations, we aggregate these to the 2-digit SIC level. In addition, we winsorize SI_PERSIST at the top/bottom 1% levels. Results using alternative specifications to obtain SI_PERSIST (e.g., defining the dependent variable as earnings versus earnings before special items; obtaining coefficients at the 2-digit SIC level; not winsorizing the distribution of special items parameters) are similar to those reported.

13

We then include NSI, an indicator variable equal to 1 if the firm has negative special

items, and 0 otherwise. A positive (negative) coefficient is consistent with management being

more likely to present NSI (PSI) as a separate line item. MISS_PYE is an indicator variable

equal to 1 if the reported special items cause operating earnings to be lower than the prior year

operating earnings, 0 otherwise. A positive coefficient is consistent with management being

more likely to highlight a NSI that causes the current year’s earnings to fall below prior year’s

earnings. BEAT_PYE is an indicator variable equal to 1 if the reported special items cause

operating earnings to be higher than the prior year operating earnings, 0 otherwise. A negative

coefficient is consistent with management being more likely to downplay a PSI that causes the

current year’s earnings to be above prior year’s earnings. Finally, INST is the percentage of firm

j’s common shares outstanding owned by institutions, measured using data from Spectrum. A

positive coefficient is consistent with management being more likely to separately present

special items on the income statement in response to demand for more transparent financial

reporting.

Sample Selection

As discussed previously, we use special items as the context for examining strategic

presentation decisions by managers, in part because this represents a substantial subset of

reporting events (e.g., Elliott and Hanna 1996) likely having varying implications for the firm’s

performance (e.g., Francis et al. 1996; Burgstahler et al. 2002). Untabulated descriptive results

reveal that the frequency (and magnitude) of reported special items is increasing throughout our

sample period, rising from 20% in 1993 to over 60% of our sample firms in the final three years

of our sample period. This is consistent with special items being a substantial reporting

14

phenomena, and likely providing a strong context in which to examine managers’ presentation

decisions.

To provide richer analysis, we derive a sample of firms to obtain a full decomposition of

reported special items.12 We first identify all firm-years falling within the S&P 1500 during the

period 1993 – 2002. Our restriction to this subset enables us to capture a broad cross-section of

firms while focusing our analysis on a relatively large dollar value of the US economy. Due to

the cost of hand-collection of data, we randomly choose 500 firms from among all firms that fall

within the S&P 1500 at any point during our sample period. For these 500 firms, we then

include all available firm-years within the sample period, resulting in a sample of 4,734 firms-

years. Several criteria (e.g., eliminating bankruptcy years) reduces our final sample to 4,695

observations (see Table 1).13

For the 4,695 firm-years, we then hand-collect and categorize all special items using the

firm’s 10-Ks, annual reports, and/or 10-Qs. Our collection includes performing key word

searches within electronic source documents, as well as scanning management discussion and

analysis, the financial statements, and footnotes for indications of special items, regardless of

whether Compustat reports that the firm has a special item.14 We group reported special items

into three primary categories: restructuring charges, write-offs, and other.15 We also collect

managements’ presentation choice of special items (income statement or footnote presentation).

12 As discussed in the appendix, much of prior research has relied on the Compustat special items category, which

historically has aggregated all special items. Owing to the differing properties of special items (e.g., see Francis et al. 1996), as well as the need to obtain the presentation of special items on the income statement, we choose a smaller hand-collected sample.

13 Note that various analyses use a subset of these observations; these are noted where relevant. 14 However, as one verification of our hand-collected data, we find an over 95% correlation between our total

special items amounts and that indicated by Compustat. Reiterating our motivation for hand-collecting this data: Compustat does not identify the category of the special item (at least during our sample period), nets positive and negative special items together, and does not provide the special item’s presentation on the income statement.

15 Restructuring charges include employee severance, facility closing, other, and restructuring reversals. Write-offs include write-offs of goodwill, intangibles, PP&E, investments, oil and gas properties, software, leases,

15

Table 2 provides descriptive data for our sample. The average firm-year has total assets

of $ 8.4 billion, consistent with our selection criteria focusing on S&P 1500 firms. Over half of

the observations report special items (2,412 out of 4,695, or 51%), with the majority reporting

only negative special items (1,452 out of 2,412, or 60%). Special items are distributed widely

across the three primary categories of restructuring charges (46% of observations reporting

special items), write-offs (34%), and other (71%).16 Finally, there is substantial variance in

firms’ presentation of special items, with 55% (1,335 out of 2,412) presenting all special items as

separate line items on the income statement, 30% aggregating all special items in other line items

on the income statement with identification only via footnote disclosure, and 15% adopting

mixed presentation. Note that our paper focuses on this variance in the presentation decision, as

reflected in the bottom three rows and last two columns of Table 2.

Empirical Results

Table 3 presents descriptive statistics relating to the sample used to examine the decision

to present special items as a separate line item on the income statement versus via footnote

disclosure. This sample focuses on those observations reporting special items and having

available data for Equation (1) (N = 2,228). Panel A presents means and medians for the

regression variables. Special items are typically reported as a separate line item on the income

statement (mean of SI_Sep = 0.646), and on average represent approximately 5% of beginning

total assets (mean of SI_MAG = 0.048). Special items have an average persistence

(SI_PERSIST) of 0.21, and 86% of observations report a negative special item (NSI = 0.859).

Almost 21% of observations report negative special items that result in the firm missing prior

inventory, and other. Other includes gains on settlements (both legal and insurance charges), losses on settlements, in-process R&D, gains on sales of assets, losses on sales of assets, merger related costs, and other.

16 The percentages do not sum to 100%, as a firm may report multiple categories of special items in any given year.

16

year’s earnings (MISS_PYE = 0.208), while only 4% report positive special items that result in

the firm beating prior year’s earnings (BEAT_PYE = 0.039). This latter is consistent with the

generally conservative nature of how special items are reported – leading to a higher frequency

(and concurrent greater impact on benchmarks) of negative than positive special items. Panel B

presents Pearson correlations, with univariate associations generally consistent with our

previously discussed predictions.17

Table 4 presents results from our analyses examining management’s decision to present

special items as a separate line item on the income statement. Under the OLS regression (Model

1), of our control variables, SIZE is negatively associated with this decision (consistent with

larger firms being less likely to separately present special items) and SI_MAG is positively

associated with this decision (consistent with larger special items being more likely to receive

income statement presentation). YEAR is insignificant.

Regarding our experimental variables, SI_PERSIST is negative and significant as

predicted (coefficient = - 0.03, t-statistic = - 2.89), indicating managers are more likely to

provide income statement presentation for those special items having lower persistence. This is

consistent with H1A, and suggests managers use income statement presentation to assist users in

identifying more versus less persistent reporting items. Also as predicted, NSI is positive and

significant (coefficient = 0.12, t-statistic = 4.32), indicating managers are more likely to present

negative special items separately on the income statement. This is consistent with H1B, and

suggests managers are more likely to highlight negative than positive special items as being

transitory (controlling for the persistence and magnitude of the special items).18 Of our reporting

17 The correlations between independent variables in Panel B of Table 3, as well as various regression diagnostics

(e.g., variance inflation factors), are consistent with multi-collinearity not affecting our specifications. 18 In Section 4, we perform additional analyses to disentangle whether this reflects economic or opportunistic

motivations.

17

incentive variables, only MISS_PYE is significantly positive as predicted (coefficient = 0.14, t-

statistic = 5.78), indicating that managers are more likely to provide income statement

presentation for special items that cause current year’s earnings to fall below previous year’s

earnings. This supports H1C, and suggests managers use income statement presentation to

highlight negative special items that affect a relevant earnings benchmark. BEAT_PYE is

insignificant (t-statistic = 1.53) and of the wrong sign; thus, we fail to find support for H1D.

Finally, consistent with H1E, the coefficient on INST is significantly positive (coefficient = 0.16,

t-statistic = 3.56). This indicates that greater institutional ownership leads to an increased

likelihood of presenting special items separately on the income statement, consistent with

managers responding to demands for more transparent information.

Turning to the logistic specification, in Model 2 we find similar results where

observations adopting a mixed presentation for special items are coded as if all special items

receive income statement presentation (i.e., akin to suggesting that presenting any special items

on the income statement serves as a “red flag” for users to look for more special items). In

Model 3, we again find similar results using the sample that excludes observations in which the

firm adopts a mixed presentation for special items.

Sensitivity Analyses

We conduct a number of sensitivity analyses to validate the above regressions, as well as

provide additional insights into our findings. First, we examine if the magnitude of special items

affects our results beyond the inclusion of the control variable SI_MAG. Accordingly, we re-run

the regressions using only observations reporting large special items (N = 1,228), defined

consistent with prior research as those exceeding 1% of lagged total assets (e.g., Elliott and Shaw

18

1988; Elliott and Hanna 1996). Inferences are unchanged from those reported, except

SI_PERSIST (t-statistic = -0.74) and NSI (t-statistic = 0.94) remain of the correct sign but are

now insignificant. Inferences using the subset of observations where special items are less than

1% of lagged total assets (N = 1,000) are unchanged from those reported. Second, to provide an

alternative control for temporal effects, we drop the variable YEAR and instead include year

fixed-effects; inferences are unchanged (and slightly stronger than those reported). Related, we

also drop the variable YEAR and run ten annual (versus one pooled) regressions. Inferences are

unchanged, though slightly weaker (primarily due to lower power).19 Third, to control for

possible interactive effects that could occur across the sign of special items, we include an

additional variable interacting SI_MAG with NSI. No differences arise, except SI_MAG is now

insignificant (t-statistic = 1.32) and SI_MAG * NSI is positive but insignificant (t-statistic =

0.79). Fourth, to examine the robustness of our proxy for demand for transparent information,

we also include analyst following (which, like institutional ownership, has a predicted positive

sign – as managers of firms with greater analyst following are expected to make presentation

decisions more likely reflective of transparent reporting). Results are unchanged from those

reported; the coefficient on analyst following is positive as expected (t-statistic = 5.48).

Next, we conduct additional analyses to validate the finding that NSI are more likely to

receive income statement presentation than PSI. These analyses focus on the subset of

19 In particular, the ten annual regressions resulted in the following:

Variable Predicted Sign

# years Positive

# years Negative

# years Significant

SIZE + / - 1 9 3 SI_MAG + 10 0 8 SI_PERSIST - 2 8 4 NSI + / - 8 2 4 MISS_PYE + 10 0 7 BEAT_PYE - 7 3 2 INST + 9 1 5

19

observations reporting both PSI and NSI, which enables the firm to act as its own control. We

obtain the percentage of observations reporting PSI as a separate line item on the income

statement, and perform univariate comparisons to this same percentage for NSI using tests of

means. Across a range of cutoffs, we find that NSI are consistently more likely than PSI to

receive income statement presentation (conditional on reporting both), providing additional

support for H1B.20

We also leverage our decomposition to examine why NSI are more likely to receive

income statement presentation. Specifically, we examine whether the category of special item

affects management’s presentation decision. This analysis is motivated by prior research

documenting that special items are heterogeneous with differing implications for the firm (e.g.,

Francis et al. 1996). Thus, we replicate equation (1) including three additional variables to

capture the three primary categories we classify special items into: RESTR (total restructuring

charges), WO (total write-offs), and OTHER (total other special items).21 Untabulated results

reveal that the coefficient on RESTR (2.36, t-statistic = 4.22) is significantly higher at the less

than 1% level than either WO (coefficient = 0.10, t-statistic = 1.74) or OTHER (0.26, t-statistic =

3.81). Inferences on the remaining variables are unchanged from those reported. This builds on

our above finding that managers are more likely to separately present negative than positive

special items; in particular, the income statement presentation of negative special items appears

20 In particular, we examine the following sub-samples of observations reporting both PSI and NSI: all observations

(N = 611); observations with PSI > 1% of earnings before special items (N = 559) ; observations with PSI > 1% of lagged total assets (N = 193); observations with both PSI and NSI (individually) > 1% of earnings before special items (N = 557); and observations with both PSI and NSI (individually) > 1% of lagged total assets (N = 140). NSI are consistently more likely to receive income statement presentation than PSI, with differences always significant at the less than 5% level.

21 Note we drop SI_MAG and NSI from the regression when we include the categorical decomposition, as the categories are perfectly collinear with SI_MAG and NSI (e.g., write-offs are all definitionally negative special items). RESTR, WO, and OTHER are all scaled by beginning total assets (similar to SI_MAG).

20

primarily driven by restructuring charges. In the next section, we leverage our analyses further

to discern if this is indicative of opportunistic or economic behavior.

Finally, we examine the subset of firms reporting only write-offs to investigate if an

additional determinant affects the likelihood of presenting the special item separately on the

income statement: the uncertainty regarding the asset. In particular, uncertainty is generally

increasing in the life of the asset (owing to longer horizons, and generally lower liquidity as the

asset’s economic life increases). Management may wish to highlight resolution to this

uncertainty. To test this hypothesis, we obtain the sub-sample of firms reporting only write-offs

as NSI in the following sub-categories: inventory, investments, software/hardware, PPE,

intangibles, and goodwill. We then include an additional variable in Equation (1), LIFE. This is

a discrete variable that is increasing in the estimated (pre-write-off) economic life of the asset

written off: it takes the value 1, 2, 3, 4, 5, or 6 if the primary asset written off by the firm is

inventory, investment, software or hardware, PP&E, intangibles, or goodwill, respectively. This

ordering is chosen to reflect the presentation found on the balance sheet, which corresponds to

decreasing liquidity. A positive coefficient would be consistent with income statement

presentation of write-offs increasing with the uncertainty regarding the asset written off (where

the life of the asset proxies for uncertainty). Untabulated results confirm this intuition. In

addition to results that are unchanged from those reported earlier (except that institutional

ownership is no longer significant), write-offs are more likely to receive incomes statement

presentation as the life of the asset written off increases (coefficient on LIFE = 0.04; t-statistic =

1.73). This is consistent with managers using the presentation decision to resolve uncertainty.22

22 Descriptive statistics also confirm this association: the percent receiving income statement presentation is

generally monotonically increasing in the life of the asset as follows: inventory (9%); investments (49%); software/hardware (80%); PPE (48%); intangibles (55%); and goodwill (68%). We note that the regression results are strongest when inventory is included, though the directional association remains regardless.

21

Overall, the results of these analyses suggest managers’ income statement presentation

decision regarding special items reflects both economic and opportunistic motivations.

Managers are more likely to present on the income statement special items that are larger and

have lower persistence, consistent with managers using presentation to assist users in identifying

and understanding properties of reporting items that may differ from other operating items.

Managers also appear to consider demands for transparent information, being more likely to

provide income statement presentation of special items when institutional ownership is high.

However, managers are also more likely to emphasize “bad” news (i.e., negative special items)

as transitory relative to “good” news (i.e., positive special items), are more likely to emphasize

negative special items that cause the firm to miss an earnings target, and are more likely to

emphasize restructuring charges over other special items. It is unclear whether these latter

results suggest motivations to better reflect the firm’s underlying performance or opportunistic

behavior. In the next section, we provide further analyses to disentangle these competing

explanations.

IV. THE PERSISTENCE OF SPECIAL ITEMS

CONDITIONAL ON THEIR PRESENTATION

Hypothesis Development and Research Design

To better understand the factors underlying the decision regarding the presentation of

special items on the income statement, we now examine whether the persistence of special items

varies across this presentation. This will allow further insights into our above finding that

managers are more likely to provide income statement presentation for less persistent special

items. This analysis seems warranted, as much of the discussion regarding user interpretation of

income statement reporting elements surrounds the correct weighting investors should apply with

22

respect to predicting future performance (e.g., FASB 1980). Further, note that special items may

not be fully transitory, as indicated by prior research (e.g., Burgstahler et al. 2002; Doyle et al.

2003). In particular, the special items themselves may recur.23 In addition, the special items

reported in year t may also affect other aspects of the firm’s future operating performance. If

managers are able to correctly identify (ex ante) which special items are more likely to be

transitory, and they use income statement presentation of special items as a signaling mechanism

to highlight for users components of earnings that are more transitory, then special items

receiving income statement presentation should have lower persistence than those receiving

footnote presentation. This results in the following hypothesis:

H2: Special items receiving income statement presentation are more transitory than those receiving footnote presentation.

To test this hypothesis, we conduct an analysis similar to Burgstahler et al. (2002), using

the following regression:

Ejt+1 = α0 + α1E*jt + α2SI_ISjt + α3SI_FNjt + γjt (2)

where Ejt+1 is earnings for firm j for year t+1; E*jt is earnings before special items for year t;

SI_IS jt is special items presented separately on the income statement for firm j for year t; and

SI_FN jt is special items aggregated within another line item and identified only via the footnotes

for firm j for year t. Note for this analysis, we use signed (versus absolute) special item amounts.

All variables are scaled by market value of equity at the beginning of year t. This analysis

examines the extent to which current period earnings and special items are predictive of (one-

period ahead) future earnings. If managers correctly ex ante identify those special items that are

more transitory, and emphasize them via income statement presentation, H2 predicts that α2 < α3.

23 As a specific example, see Arner (2004) regarding Kodak’s reporting of restructuring charges for multiple years;

more generally, see Elliott and Hanna (1996) for cross-sectional evidence of serial reporting of special items.

23

We then further decompose special items according to their sign and category. This

follows similar decomposition performed in our previous section examining the presentation

decision, as well as prior research documenting differences in special items according to their

sign (Burgstahler et al. 2002) and category (Francis et al. 1996). In particular, we estimate the

following regressions:

Ejt+1 = β0 + β1E*jt + β2PSI_ISjt + β3PSI_FNjt + β4NSI_ISjt + β5NSI_FNjt + ηjt (2a)

Ejt+1 = χ0 + χ1E*jt + χ2RESTR_ISjt + χ3RESTR_FNjt +

χ4WO_ISjt + χ5WO_FNjt + χ6OTHER_ISjt + χ7OTHER_FNjt + ιjt (2b)

Ejt+1 and E*jt are as defined in Equation (2). Equation (2a) examines if differential persistence

exists across the presentation choice for special items, conditional on the signed effect of the

special item on reported income (i.e., positive special items (PSI) and negative special items

(NSI)). Thus, holding the signed effect constant, we examine whether the persistence of special

items receiving income statement presentation (i.e., the variables with the “_IS” suffix) is lower

than those receiving footnote presentation (i.e., the variables with the “_FN” suffix) by

comparing β2 < β3 and β4 < β5. Similarly in Equation (2b), we examine if differential persistence

exists across the presentation of special items, conditional on the category (i.e., restructuring

(RESTR), write-off (WO), and other (OTHER)). Thus, holding the category constant, we

compare the coefficients χ2 < χ3, χ4 < χ5, and χ6 < χ7. As above, all variables are scaled by

beginning market value of equity. Finally, because managers may base their presentation

decisions primarily on the effect of special items for continuing earnings (versus earnings

inclusive of special items), we alternatively replace the dependent variable in equations (2), (2a),

and (2b) with E*jt+1, earnings before special items for firm j for period t+1 (also scaled by

beginning market value of equity).

24

Empirical Results

Results are presented in Table 5 (N = 3,698). Models (1) – (4) under Panel A present

results using lead earnings (Et+1) as the dependent variable; Models (5) – (6) under Panel B

present results using lead earnings before special items (E*t+1) as the dependent variable.

Model 1 provides a benchmark, revealing that special items (SIt) are less predictive of

one-period ahead earnings than earnings before special items (E*t), consistent with prior research

(Burgstahler et al. 2002). Model 2 then reveals that special items presented in the footnotes

(SI_FNt) are more persistent (coefficient = 0.52) than those receiving income statement

presentation (SI_ISt) (coefficient = 0.29) (F-value on difference = 6.38, with a p-value < 0.01).

This is consistent with our previously discussed finding that special items with lower persistence

are more likely to receive income statement presentation. Model 3 presents a further

decomposition on the sign of special items, revealing higher persistence for footnote presentation

for negative special items (F-value on difference = 6.06, with a p-value < 0.01); persistence is

also higher for positive special items receiving footnote presentation – but this difference is

insignificant (F-value = 0.90, with a p-value = 0.17). Of further note, the coefficient on PSI_FN

(0.88) is not significantly different from that on E* (1.01), suggesting these have similar

persistence. These latter findings provide additional insights into our previous finding that

managers are more likely to present negative special items. In particular, the results in Model 3

indicate this is consistent with economic (versus opportunistic) motivations, as negative special

items are less persistent than positive special items.

Finally, Model 4 presents special items according to their category and presentation.

Special items presented in the footnotes are more persistent for restructuring (F-value on

25

difference = 5.63, with a p-value = 0.01) and other special items (F-value = 4.04, with a p-value

= 0.02), but not for write-offs (F-value = 0.02, with a p-value = 0.90). Interestingly, Model 4

also reveals that restructuring and other special items that are presented in the footnotes have

similar persistence (RESTR_FN = 0.92; OTHER_FN = 1.02) as earnings before special items (E*

= 1.01), consistent with managers providing footnote presentation for special items within these

two categories when these special items have similar persistence to earnings before special items.

However, restructuring charges and other special items presented in the income statement

(RESTR_IS = 0.13; OTHER_IS = 0.54) have significantly lower persistence than E*, as do write-

offs whether presented in the income statement (WO_IS = 0.35) or in the footnotes (WO_FN =

0.37). Again, this provides additional insights to our previous findings that restructuring charges

are the most likely category to receive income statement presentation. This also appears

consistent with managers using income statement presentation for economic (versus

opportunistic) reasons, as restructuring charges receiving income statement presentation

(coefficient = 0.13) have lower persistence compared to write-offs (0.35) and other special items

(0.54) also receiving income statement presentation.24

We then conduct a battery of analyses to validate the above results. Table 5, Panel B

presents results for Models (5) and (6), using lead earnings before special items (E*t+1) as the

dependent variable, with inferences unchanged from those discussed above.25 In addition,

untabulated results using variables scaled by lagged total assets or unscaled variables are similar

to those reported. Results are also unchanged when only large special items observations (again, 24 We also estimate regressions using total restructuring, write-offs, and other special items (i.e., not decomposing

these into their respective components receiving income statement and footnote presentation). Results confirm that restructuring charges have the lowest persistence compared to either write-offs or other special items charges.

25 While the coefficient on WO_IS is now higher than WO_FN (inconsistent with our findings on restructuring and other charges), we note that the primary driver of this is likely the very insignificant coefficient on WO_FN (coefficient is of the wrong sign at -0.06, t-statistic = -0.39). Inferences on regressions comparable to Models (2) and (3) in Table 5 are unchanged from those reported in Panel A.

26

defined as those greater than 1% of lagged total assets) are included in the regressions. Finally,

we examine firms reporting both PSI and NSI with necessary data (N = 447). We find that PSI

have higher persistence (coefficient = 0.50, t-statistic = 7.41) than NSI (coefficient = 0.16, t-

statistic = 3.34) with a difference that is significant at the < 1% level. This is again consistent

with our previous finding that firms are more likely to provide income statement presentation for

NSI (for firms reporting both NSI and PSI) for economic versus opportunistic reasons.

In summary, the results suggest that managers are more likely to provide income

statement presentation for special items having lower persistence. Additional analysis reveals

this behavior holds across both positive and negative special items, as well as for both

restructuring and other special items categories. However, we fail to find a difference in how

managers present write-offs. Overall, the results support H2, and are consistent with managers

having both the ability to identify more versus less persistent special items ex ante, as well as

using income statement presentation as a mechanism to alert financial statement users about

those special items having lower persistence. This is consistent with manager presentation

decisions regarding special items on the income statement reflecting economic versus

opportunistic motivations.

V. HOW CONSISTENT IS THE INCOME STATEMENT PRESENTATION?

Consistency Across Alternative Disclosure Media: Press Releases

Because managers have multiple avenues for providing disclosure and information, the

presentation decision on the income statement (or any other financial statement) may be

correlated with, or subsumed by, a presentation/disclosure decision affecting these alternative

27

mechanisms.26 The most common forms of these alternative disclosures include press releases

associated with earnings announcements and conference calls. Due to data availability, we focus

on the former.27

Accordingly, we hand-collect annual earnings press releases for our observations

reporting special items to provide a comparative analysis of managers’ decision to discuss

special items in their press releases relative to their decision to present special items on the

income statement. Table 6 Panel A presents our sample selection process. We begin with all

observations having special items (N = 2,412), and retain only those observations in which the

source of the available press release from Factiva is either Business Wire (BW) or Press Release

Newswire (PRN). We impose this restriction in order to focus our analysis on direct (i.e.,

unedited) management disclosures as press releases from other sources (such as Dow Jones or

Reuters) may experience editing of the original disclosure, with unclear implications for our

analysis. This leads to a sample of 1,514 available observations.28 Due to the cost of hand-

collection, we randomly select approximately one-half of the available observations (N = 762),

reading the annual press releases to identify if management provides any discussion of the

special items in the press release (e.g., in the headline, in the text discussion, or in any included

tabulated financial statements). This measure of alternative disclosure regarding special items

will capture (in a binary sense) managers’ highlighting of special items in the press release. We

then perform three analyses on these observations: we examine univariate relationships; we re-

26 We thank both referees for suggesting this complementary line of inquiry. 27 In particular, an analysis of conference call data relating to the presentation/discussion of special items would

require availability of transcripts to identify the nature of management discussion regarding these items. However, conference call transcripts are not widely available for most of our sample period.

28 Test of means comparing the sample of observations with BW or PRN press releases to those without BW or PRN press releases reveal no significant differences across characteristics of either the firm (e.g., size, profitability) or special items (e.g., magnitude, income statement presentation).

28

examine our presentation decision analysis of Section 3; and we re-examine our persistence

analysis of Section 4.

Table 6 Panel B presents our univariate analysis, comparing the decision to discuss the

special item in the press release (PR – No versus PR – Yes) with the decision to present the

special item as a separate line item on the income statement. We group special items into three

size categories (Small SI, Medium SI, Large SI) to hold the magnitude of the special items

constant. Three insights in this panel warrant discussion. First, the likelihood of discussing

special items in the press release generally increases across the presentation in the income

statement (e.g., focusing on Small SI, the percentage of observations having PR – Yes is 73% for

footnote presentation, 84% for mixed presentation, and 91% for income statement presentation).

This is true across all three magnitude categories of the special items. Second, similar to the

presentation on the income statement, the likelihood of discussion in the press release is also

generally increasing in the magnitude of the special item (e.g., in the last column of Panel B,

82% of Small SI, 79% of Medium SI, and 93% of Large SI are discussed in the press release).

Finally, the absolute likelihood of discussing special items in the press release is higher than

providing separate presentation on the income statement (e.g., the overall likelihood of

discussing special items in the press release is 87% of the 762 observations).29 Overall, these

descriptive statistics indicate a positive association between discussion in the press release and

presentation on the income statement – consistent with managers undertaking similar actions

across these disclosure mechanisms.

Panel C of Table 6 presents a revised logit analysis, now examining the decision to

discuss the special items in the press release. Similar to our previous equation (1), we estimate:

29 This is not surprising, as there are more opportunities to identify special items in the press release (e.g., via

discussion in the text or highlighting on the income statement that is included in the press release).

29

SI_Sep_PRjt = δ0 + δ1YEAR t + δ2SIZEjt + δ3SI_MAGjt + δ4SI_PERSISTjt

+ δ5NSIjt + δ6MISS_PYEjt + δ7BEAT_PYEjt + δ8INSTjt + φjt (3)

Thus, the dependent variable is now SI_Sep_PR, defined as an indicator variable equal to 1 if the

special items are explicitly discussed/identified anywhere in the annual press release (e.g., either

in the text or in any included financial statements), and 0 otherwise. The remaining variables are

as previously defined in Equation (1). If managers approach the decision to discuss a special

item in the press release similarly to the decision to present it separately on the income

statement, our predicted signs on all of the variables are unchanged from equation (1). Using the

sample of observations with available data (N = 725), we find similar results to those previously

presented. In particular, NSI and MISS_PYE are significant at the less than 10% level. SI_MAG

and SI_PERSIST have the correct sign, but are insignificant. INST now has the incorrect sign.

Finally, Panel D of Table 6 presents analyses of the persistence of special items

according to their discussion in the press release. To provide further evidence of whether

managers use the press release discussion for economic or opportunistic reasons, we examine

whether special items discussed in the press release have lower persistence than those not

discussed in the press release. This parallels our development of H2: if managers are able to

identify ex ante those special items that have lower persistence, and they use discussion in the

press release as a mechanism to highlight less persistent items for financial statement users, then

special items highlighted in the press release should have lower persistence. Accordingly, we

estimate the following regression:

Ejt+1 = π0 + π1E*jt + π2SI_in_PRjt + π3SI_Not_in_PRjt + ξjt (4)

where Ejt+1 is earnings for firm j for year t+1; E*jt is earnings before special items for firm j for

year t; SI_in_PRjt is special items for firm j for year t that are discussed/identified in the firm’s

annual earnings press release; and SI_Not_in_PRjt is special items for firm j for year t that are not

30

discussed/identified in the firm’s annual earnings press release. As with equation (2), all

variables are scaled by beginning market value of equity. Similar to our Section 4 analysis, if

managers use the press release to highlight special items that are less persistent then π2 < π3.

Consistent with this expectation, special items identified in the press release are significantly less

persistent than those not discussed in the press release (significant at the < 1% level). The same

results occur using earnings before special items for year t + 1 as the dependent variable.

Consistency in Presentation Across Time

We also examine whether the presentation decision varies over time for a given firm, as

there may be stickiness in such decisions. We examine this is two ways. First, we rerun our

analysis examining the presentation decision (i.e., Equation (1)) including an additional control

variable: the presentation of special items in the most recent year having a special item

(LAG_SI_Sep). Untabulated results are stronger than those reported on all variables, except

INST is of the same sign but no longer significant (t-statistic = 1.34). In addition (and not

surprisingly), LAG_SI_Sep is positive and highly significant (t-statistic = 20.60).30

Second, we also reexamine the presentation analyses using only the first time a firm

reports a special item (N = 448). Results remain consistent with those previously reported,

except BEAT_PYE is now of the correct sign, though it remains insignificant (coefficient =

- 0.09, t-statistic = - 0.77).

In summary, these results suggest that managers are more likely to discuss special items

in press releases for similar reasons as present them separately on the income statement (though

30 We do not include LAG_SI_Sep in our primary model as we are attempting to model the underlying determinants

of the presentation decision, versus document serial behavior.

31

the absolute occurrence of discussion items within the press release is quite higher). In addition,

the decision to discuss these items in the press release appears similarly motivated by economic

versus opportunistic reasons, as special items discussed in the press release have lower

persistence than those not discussed in the press release. Finally, the income statement

presentation decision appears consistent over time.

VI. CONCLUSION

This paper examines whether manager discretion in the presentation of recognized

special items within the income statement reflects underlying firm performance or opportunistic

motivations. This analysis is motivated by prior research, which provides evidence that

managers engage in opportunistic reporting in their presentation and classification decisions

(e.g., Schrand and Walther 2000; McVay 2006), and that presentation decisions can affect users’

judgments (Hirst and Hopkins 1998; Maines and McDaniel 2000). It is also motivated by two

recent areas of interest in financial standard setting: performance reporting and financial

statement presentation (FASB 2006; IASB 2006), and the movement towards reporting standards

likely to increase the occurrence of “non-recurring” type charges, such as fair value changes and

impairment testing.

Overall, our results suggest that manager discretion in the presentation of special items

primarily reflects economic motivations. In particular, managers provide separate income

statement presentation for those special items that are less persistent, which is consistent with

using separate presentation as a mechanism to highlight non-recurring items for financial

statement users. These results hold across various characteristics of special items, including their

magnitude, signed effect on net income, and category. Additional analyses provide

32

corroborating evidence in the context of an alternative disclosure mechanism: press releases. We

do find limited evidence of opportunistic behavior, as managers are more likely to separately

present special items that cause the firm to miss an earnings benchmark, analogous to results

found in Schrand and Walther (2000). Nonetheless, our overall findings that manager

presentation decisions within the financial statements are primarily consistent with economic

(versus opportunistic) motivations expands the concept of “strategic reporting” to encompass

reporting behavior that may benefit users. In addition, our results provide a useful alternative

perspective to prior research, which suggests that managers engage in opportunistic behavior in

their presentation and classification decisions.

33

APPENDIX - Definition of Special Items

The term “special items” has been used in the academic and practitioner literatures to

refer to a broad range of economic phenomena. Special items have often been used

interchangeably with other terms such as “non-recurring,” “non-operating,” and “one-time,” with

negative special items also referred to as “write-offs,” “write-downs,” or “charge-offs.” These

terms allude to these economic phenomena generally being thought of as having lower

persistence (i.e., being more transitory) than other components of firms’ earnings.

This paper focuses on special items as defined by Accounting Principles Board 30,

Reporting the Results of Operations (1973, paragraph 26): items that are unusual or infrequent

(but not both). These items must be included in income from continuing operations, and (when

material) reported either separately on the income statement or via footnote disclosure. Our

starting point for what constitutes a special item is the Compustat definition (data #17 in the

annual file, #32 in the quarterly file), which encompasses a broad range of phenomena. Among

the twenty-three items listed under the description are: “any significant non-recurring items,”

“natural disaster losses,” “impairment of goodwill,” “restructuring charges,” “nonrecurring profit

or loss on the sale of assets,” “purchased research and development,” “reserve for litigation,” and

“write-downs or write-offs of receivables and intangibles.” These descriptions allude to the

variety of items classified within this data field. Prior to 2001, Compustat aggregated these

items into a single field designated “special items,” with an undisclosed note field to indicate the

types of charges making up the amount. However, a specific breakdown of special items is not

available on a broad cross-sectional basis. From 2001, Compustat provides some breakdown of

these items through its expanded data series.31

31 Using Compustat’s expanded data fields, the percentage of firms reporting any special item in a given year is 3%

for 1999, 6% for 2000, 49% for 2001 and 54% for 2002. The low percentages for 1999 and 2000 (compared to the percent of firms reporting special items per the aggregated data #17), coupled with the dramatic increase in percentage from 2000 to 2001, suggest these expanded data fields are not complete until 2001.

34

REFERENCES

Aboody, D. 1996. “Recognition versus Disclosure in the Oil and Gas Industry.” Journal of Accounting Research 34 (Supplement): 21-32.

Aboody, D., M. Barth, R. Kasznik. 2004. “Firms’ Voluntary Recognition of Stock-Based

Compensation Expense.” Journal of Accounting Research 42 (May): 123-150. Accounting Principles Board (APB), 1973. Reporting the Results of Operations – Reporting the

Effects of Disposal of a Segment of A Business, and Extraordinary, Unusual and Infrequently Occurring Events and Transactions. Opinion No. 30. Norwalk, CT: FASB.

Arner, F. 2004. “Kodak’s Fuzzy Numbers: The company has taken ‘one-time’ charges every

year for the past twelve.” Business Week, February 9. Bhattacharya, N., E. Black, T. Christensen, and C. Larson. 2003. “Assessing the Relative

Informativeness and Permanence of Pro Forma Earnings and GAAP Operating Earnings.” Journal of Accounting and Economics 36 (December): 285-319.

Bowen, R., A. Davis, and D. Matsumoto. 2005. “Emphasis on Pro Forma versus GAAP

Earnings in Quarterly Press Releases: Determinants, SEC Intervention, and Market Reactions.” The Accounting Review 80: 1011-1038.

Bradshaw, M., and R. Sloan. 2002. “GAAP Versus The Street: An Empirical Assessment of

Two Alternative Definitions of Earnings.” Journal of Accounting Research 40 (March): 41-66.

Burgstahler, D., J. Jiambalvo, and T. Shevlin. 2002. “Do Stock Prices Fully Reflect the

Implications of Special Items for Future Earnings?” Journal of Accounting Research 40 (June): 585-612.

Doyle, J., R. Lundholm, and M. Soliman. 2003. “The Predictive Value of Expenses Excluded

from Pro Forma Earnings.” Review of Accounting Studies 8: 145-174. Elliott, J., and J. Hanna. 1996. “Repeated Accounting Write-offs and the Information Content

of Earnings.” Journal of Accounting Research 34 (Supplement): 135-155. Elliott, J. and W. Shaw. 1988. “Write-offs as Accounting Procedures to Manage Perceptions.”

Journal of Accounting Research 26 (Supplement): 91-119. Fairfield, P., R. Sweeney, and T. Yohn. 1996. “Accounting Classification and the Predictive

Content of Earnings.” The Accounting Review 71 (July): 337-355. Financial Accounting Standards Board. 1980. Statement of Financial Accounting Concepts No.

2: Qualitative Characteristics of Accounting Information. FASB, Stamford, Connecticut.

35

Financial Accounting Standards Board. 2006. Financial Statement Presentation – Joint Project of the IASB and FASB (formerly known as Financial Performance Reporting by Business Enterprises). FASB, Stamford, Connecticut.

Francis, J., D. Hanna, and L. Vincent. 1996. “Causes and Effects of Discretionary Asset Write-

Offs.” Journal of Accounting Research 34 (Supplement): 117-134 Gu, Z. and T. Chen. 2004. “Analysts’ Treatment of Nonrecurring Items in Street Earnings.”

Journal of Accounting and Economics 38: 129 – 170. Hirst, E., and P. Hopkins. 1998. “Comprehensive Income Reporting and Analysts’ Valuation

Judgments.” Journal of Accounting Research 36 (Supplement): 47-75. International Accounting Standards Board. 2006. Board Decisions on International Financial

Reporting Standards: Update March 2006. IASB, London, UK. Kinney, M., and R. Trezevant. 1997. “The Use of Special Items to Manage Earnings and

Perceptions.” Journal of Financial Statement Analysis 3 (Fall): 45-53. Libby, R., M. Nelson, and J. Hunton. 2005. “Recognition v. Disclosure and Auditor

Misstatement Correction: The Cases of Stock Compensation and Leases.” Working paper, Cornell University.

Lipe, R. 1986. “The Information Contained in the Components of Earnings.” Journal of

Accounting Research 24 (Supplement): 37-64. Lougee, B., and C. Marquardt. 2004. “Earnings Informativeness and Strategic Disclosure: An

Empirical Examination of “Pro Forma” Earnings.” The Accounting Review 79 (October): 769-795.

Maines, L., and L. McDaniel. 2000. “Effects of Comprehensive-Income Characteristics on

Nonprofessional Investors’ Judgments: The Role of Financial-Statement Presentation Format.” The Accounting Review 75 (April): 179-207.

McVay, S. 2006. “Earnings Management Using Classification Shifting: An Examination of

Core Earnings and Special Items.” The Accounting Review 81 (October): 501-531. Riedl, E. 2004. “An Examination of Long-Lived Asset Impairments.” The Accounting Review

79 (October): 823-852. Schrand, C., and B. Walther. 2000. “Strategic Benchmarks in Earnings Announcements: The

Selective Disclosure of Prior-Period Earnings Components.” The Accounting Review 75 (April): 151-177.

36

Weiss, I. 2001. “Managerial Responses to a Transitory Earnings Shock: Strategic Manipulation Between the Core and Non-Core Components of Earnings.” University of Chicago, unpublished dissertation.

37

TABLE 1

Sample Selection a Firm- Years Firms Available observations from Compustat Annual, 1993 – 2002 72,473 11,557 Observations designated as S&P 1500 by Compustat 1993 – 2002 14,400 b 2,466 All available observations for 1993 – 2002 for any firm designated as within the S&P 1500 during the sample period 20,198 2,466 Random selection of 500 firms, including all available years for 1993 – 2002 4,734 500 Less: start-up firm years, bankruptcy, other (missing data) 39 0 Final Sample 4,695 500 a This table shows the sample selection process. We begin with all firm-years available on Compustat

for the period 1993 – 2002. Using observations classified as within the S&P 1500, we identify all firms that are within the S&P 1500 at any point during our sample period. From these firms, we randomly choose 500. For the 500 firms selected, we then include all available firm-years for the sample period, leading to 4,734 observations. We exclude 39 firm-year observations for the reasons indicated above, resulting in a final sample of 4,695 firm-years representing 500 firms.

b Per discussion with Compustat, there are only 900 firms classified within the S&P 1500 for 1993; this results in 600 fewer available firm-years than expected (i.e., 15,000 – 14,400 = 600).

38

TABLE 2 Descriptive Data a

N

Unique Firms

Total

Assetst

NIt

NIt

(pre SI)

SIt

|SIt| / Total

Assetst-1

% SI listed separately

on IS

% SI identified only in FN

All Observations 4,695 500 8,449 207 271 (64) 2.5% 64% 36% By Existence of SI: No SI 2,283 477 5,461 171 171 - - - - SI 2,412 475 11,277 241 365 (124) 4.8% 64% 36% By Magnitude/Sign of SI: SI > 1% Total Assetst-1 1,279 400 4,951 94 296 (203) 8.3% 76% 24% Only NSI 1,452 451 10,185 202 329 (127) 5.5% 67% 33% Only PSI 349 215 8,419 342 295 46 3.3% 51% 49% Both NSI and PSI 611 267 15,496 277 492 (215) 4.0% 66% 34% By Category of SI: Restructuring 1,108 354 16,764 241 469 (229) 4.8% 74% 26% Write-off 836 359 7,465 76 315 (239) 6.5% 61% 39% Other 1,723 450 11,618 269 399 (130) 5.3% 63% 37% By Presentation of SI: All SI listed separately on income statement 1,335 399 12,265 171 317 (146) 6.1% 100% 0% All SI identified only in footnotes 733 309 9,274 340 383 (42) 1.6% 0% 100% Mixed presentation 344 196 11,687 304 516 (213) 6.5% 64% 36% a This table provides descriptive data for the samples used in our analyses. N is the number of observations. Unique firms is the number of

unique firms within each grouping. We then report means of the following measures. Total Assetst is end-of-year total assets. NIt is net income before extraordinary items. NIt (pre SI) is net income before extraordinary items and special items. SIt is net reported special items, measured

39

using hand-collected annual data from firms’ 10-Ks. The previous four variables are denoted in $ millions. |SI| / Total Assetst-1 is the absolute net reported special items divided by beginning-of-year total assets. % SI listed separately on IS is the average percent of the absolute total special items (not netted) that are presented on the firm’s income statement in a separate line item. % SI identified only in FN is the average percent of the absolute total special items (not netted) that are aggregated within another line item on the income statement and identified only via footnote disclosure.

We present the above measures for five groupings of observations. All Observations provides data for the pooled observations. No SI and SI divide the pooled observations into those reporting no special items, and those reporting non-zero special items. We then present the measures for observations where special items exceed 1% of lagged total assets (SI > 1% Total Assets t-1), observations reporting only income-decreasing (i.e., negative) special items (Only NSI), observations reporting only income-increasing (i.e., positive) special items (Only PSI), and observations reporting both income-decreasing and income-increasing special items in the same fiscal year (BOTH NSI and PSI). We then report the measures for observations reporting any restructuring charges (Restructuring), write-offs (Write-off), or other categories of special items (Other). Finally, we report the measures for observations wherein all special items are reported in separate line items on the income statement (All SI listed separately on income statement), all special items are aggregated in other line items on the income statement and identified only via footnote disclosure (All SI identified only in footnotes), and observations wherein some special items are listed separately on the income statement and others are aggregated into other line items (Mixed presentation).

40

TABLE 3 Management Presentation of Special Items as a Separate Line Item on the Income Statement:

Descriptive Statistics and Correlations a

Panel A: Descriptive Statistics (N = 2,228)

Variable Mean Median Standard Deviation

SI_Sep 0.646 1.000 0.455 YEAR 1,998 1,999 2.722 SIZE 7.109 7.091 1.676 SI_MAG 0.048 0.013 0.223 SI_PERSIST 0.218 0.140 0.823 NSI 0.859 1.000 0.348 MISS_PYE 0.208 0.000 0.406 BEAT_PYE 0.039 0.000 0.195 INST 0.681 0.693 0.223

Panel B: Correlations (N = 2,228)

SI_Sep YEAR SIZE SI_MAG SI_PERSIST NSI MISS_PYE BEAT_PYE YEAR -0.017 SIZE -0.079 *** 0.104 *** SI_MAG 0.101 *** 0.013 -0.099 *** SI_PERSIST -0.070 *** -0.055 *** 0.021 0.032 NSI 0.120 *** 0.092 *** 0.004 0.025 -0.058 *** MISS_PYE 0.147 *** -0.067 *** 0.053 ** 0.082 *** -0.035 * 0.208 *** BEAT_PYE -0.001 -0.025 0.011 0.077 *** 0.025 -0.248 *** -0.104 *** INST 0.073 *** 0.278 *** 0.063 *** 0.009 -0.049 0.085 0.009 -0.066 ***

a This table presents descriptive statistics (Panel A) and Pearson correlations (Panel B) for the variables used in the analysis examining

management’s decision to present special items as a separate line item on the income statement versus aggregate them within another line item with identification only via footnote disclosure. The sample is comprised of observations reporting special items and having available data (N = 2,228). SI_Sep is the absolute amount of (non-netted) special items presented in a separate line item on the income statement, divided by the

41

absolute amount of total (non-netted) reported special items: thus it ranges in value from 0 to 1, inclusive. YEAR is the year. SIZE is the log of the firm’s year t sales. SI_MAG is the absolute value of the total (non-netted) special items, divided by the firm’s beginning-of-period total assets. SI_PERSIST is the persistence parameter for year t special items on one-period ahead earnings (before special items); the parameter is calculated by year and industry at the 3-digit SIC level. NSI is an indicator variable equal to 1 if the firm reports negative (income-decreasing) special items, 0 otherwise. MISS_PYE is an indicator variable equal to 1 if the net special items cause the firm’s earnings to be below prior year earnings, 0 otherwise. BEAT_PYE is an indicator variable equal to 1 if the net special items cause the firm’s earnings to be above prior year’s earnings, 0 otherwise. INST is the percentage of the firm’s outstanding common shares owned by institutions at year end.

***, **, * represent significance at the less than 1%, 5%, and 10% levels for two-tailed tests, respectively.

42

TABLE 4 Management Presentation of Special Items as a Separate Line Item

on the Income Statement: Regression Analysis a

Variable Predicted Sign Hypoth Model 1

OLS Model 2

Logit – “Red Flag” Model 3

Logit

Intercept 11.91 ( 1.64) 1.60 ( 0.01) 41.42 ( 1.14) YEAR + / - - 0.01 (-1.56) - 0.01 ( -0.01) - 0.02 ( -1.14) SIZE + / - - 0.02 (-3.73) *** - 0.01 ( -0.14) - 0.08 ( -5.81) ** SI_MAG + 0.16 ( 3.81) *** 18.25 (67.92) *** 15.60 (51.57) *** SI_PERSIST - H1A - 0.03 (-2.89) *** - 0.14 ( -5.87) ** - 0.14 ( -5.00) ** NSI + / - H1B 0.12 ( 4.32) *** 0.57 (17.61) *** 0.30 ( 4.66) ** MISS_PYE + H1C 0.14 ( 5.78) *** 0.32 ( 4.52) ** 0.35 ( 5.01) ** BEAT_PYE - H1D 0.08 ( 1.53) 0.01 ( 0.00) 0.01 ( 0.01) INST + H1E 0.16 ( 3.56) *** 0.74 (10.67) *** 0.74 ( 9.92) *** # of Obs 2,228 2,228 1,833 Adj-R2 0.051 Wald Statistic 153.09 *** 117.89 ***

a This table presents results from regressions examining the management decision to present special

items as a separate line item on the income statement (“income statement presentation”) versus aggregate them in another line item with identification only via footnote disclosure (“footnote presentation”). N = 2,228 for the OLS regression (Model 1) and for the logistic regression labeled “Red Flag” (Model 2), where observations adopting a mixed presentation for special items (i.e., some special items receive income statement presentation, and some receive footnote presentation) are coded as if all special items receive income statement presentation. N = 1,833 for the logistic regression (Model 3), where observations adopting a mixed presentation for special items are excluded. All special items are measured using hand-collected data from firm’s 10-Ks. The dependent variable is SI_Sep, the absolute amount of (non-netted) special items receiving income statement presentation, divided by the absolute amount of total (non-netted) reported special items: thus the dependent variable has a range of 0 to 1 for the OLS regression, and a value of either 0 or 1 for the logistic regressions. YEAR is the year. SIZE is the log of the firm’s year t sales. SI_MAG is the absolute value of the total (non-netted) special items, divided by the firm’s beginning-of-period total assets. SI_PERSIST is the persistence parameter for year t special items on one-year ahead earnings before special items; the parameter is calculated by year and industry at the 3-digit SIC level. NSI is an indicator variable equal to 1 if the firm reports negative (income-decreasing) special items, 0 otherwise. MISS_PYE is an indicator variable equal to 1 if the net special items cause the firm’s current year earnings to be below prior year’s earnings, 0 otherwise. BEAT_PYE is an indicator variable equal to 1 if the net special items cause the firm’s current year earnings to be above prior year’s earnings, 0 otherwise. INST is the percentage of the firm’s outstanding common shares owned by institutions at year end. t-statistics (Wald Chi-Square statistics) are in parentheses for the OLS (Logistic) regressions.

***, **, * represent significance at the less than 1%, 5%, and 10% levels for the indicated one- or two-tailed tests, respectively. Standard errors are robust to heteroskedasticity and clustered by company.

43

TABLE 5 The Persistence of Special Items Conditional on Their Presentation a

Dependent Variable:

Panel A. Lead Earnings (Et+1) Panel B. Lead Earnings Before Special Items (E*t+1)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Intercept 0.01 ( 4.34) *** 0.01 ( 4.37) *** 0.01 ( 3.82) *** 0.01 ( 3.97) *** 0.02 ( 6.45) *** 0.02 ( 6.22) *** E*t 1.00 (54.63) *** 1.00 (54.64) *** 1.01 (54.84) *** 1.01 (54.55) *** 1.03 (66.66) *** 1.02 (66.18) *** SIt 0.30 (11.75) *** ^^^ 0.17 ( 7.83) *** ^^^

SI_ISt 0.29 (10.11) *** ^^^ SI_FNt 0.52 ( 6.19) *** ^^^

PSI_ISt 0.61 ( 7.23) *** ^^^ PSI_FNt 0.88 ( 3.16) *** NSI_ISt 0.28 ( 9.35) *** ^^^ NSI_FNt 0.50 ( 5.83) *** ^^^

RESTR_ISt 0.13 ( 1.17) ^^^ 0.07 ( 0.72) ^^^ RESTR_FNt 0.92 ( 2.89) *** 0.60 ( 2.30) ** WO_ISt 0.35 ( 5.37) *** ^^^ 0.28 ( 5.27) *** ^^^ WO_FNt 0.37 ( 2.16) *** ^^^ -0.06 (- 0.39) ^^^ OTHER_ISt 0.54 ( 9.19) *** ^^^ 0.20 ( 4.17) *** ^^^ OTHER_FNt 1.02 ( 4.39) *** 0.99 ( 5.12) *** Adj-R2 0.470 0.470 0.473 0.473 0.561 0.562 Tests of H2 Comparing _IS (Income Statement) Coefficients to _FN (Footnote) Coefficients (p-values for one-tailed F-test): SI 6.38 ∀∀∀ PSI 0.90 NSI 6.06 ∀∀∀ RESTR 5.63 ∀∀∀ 3.83 ∀∀ WO 0.02 4.92 OTHER 4.04 ∀∀ 16.04 ∀∀∀

44

a This table presents regressions examining the persistence of reported special items in predicting one-period ahead earnings. N = 3,698. In Panel A, Models (1) – (4) use as the dependent variable Et+1, or earnings for year t+1. In Panel B, Models (5) – (6) use as the dependent variable E*t+1, or earnings before special items for year t+1. E*t are earnings before special items for year t. SI are reported special items, coded from our hand-collected data from firms’ 10-Ks. PSI are reported positive (i.e., income-increasing) special items. NSI are reported negative (i.e., income-decreasing) special items. RESTR are restructuring charges. WO are write-offs. OTHER are other special item charges. The “_IS” (“_FN”) suffix indicates the respective special item charges are reported as a separate line item (aggregated within another line item) on the income statement. All variables are scaled by beginning market value of equity. Observations with studentized residuals greater than 4 (a total of 47 observations, or 1.25% of the initial sample having available data for this analysis) have been deleted.

***, **, * represent significance at the 1%, 5%, and 10% levels for two-tailed tests, respectively. ^^^, ^^, ^ represent significance at the 1%, 5%, and 10% levels for two-tailed tests comparing the coefficient against that for E*t, respectively. ∀∀∀, ∀∀, ∀ represent significance at the 1%, 5% , and 10% levels for one-tailed tests, respectively. All standard errors are robust to heteroskedasticity and clustered by company.

45

TABLE 6 Consistency in Disclosure: Press Releases a

Panel A: Sample Selection

Initial Obs

Obs with no PR from BW or PRN

Available Obs

Random Sample Obs

Small SI 472 (166) 306 84 Medium SI 674 (253) 421 286 Large SI 1,266 (479) 787 392 Total 2,412 (898) 1,514 762

Panel B: Presentation of Special Items on the Income Statement versus in Press Release

Income Statement Presentation FN Mixed IS Total

Small SI: PR – No 9 (27%) 3 (16%) 3 (9%) 15 (18%) PR – Yes 24 (73%) 16 (84%) 29 (91%) 69 (82%)

Medium SI: PR – No 45 (34%) 1 (4%) 13 (10%) 59 (21%) PR – Yes 88 (66%) 25 (96%) 114 (90%) 227 (79%)

Large SI: PR – No 18 (14%) 4 (3%) 5 (4%) 27 (7%) PR – Yes 111 (86%) 122 (97%) 132 (96%) 365 (93%)

All Obs: PR – No 72 (24%) 8 (5%) 21 (7%) 101 (13%) PR – Yes 223 (76%) 163 (95%) 275 (93%) 661 (87%) Total Obs 295 171 296 762

Panel C: Logit Analysis: The Decision to Discuss Special Items in the Press Release

Variable Predicted Sign Hypoth Model 1

OLS

Intercept - 253.00 ( 9.39) *** YEAR + / - 0.13 ( 9.36) *** SIZE + / - 0.22 ( 8.27) *** SI_MAG + 1.45 ( 0.99) SI_PERSIST - H1A - 0.13 ( 1.15) NSI + / - H1B 0.53 ( 3.18) ** MISS_PYE + H1C 0.46 ( 1.70) * BEAT_PYE - H1D 0.27 ( 0.17) INST + H1E - 0.47 ( 0.88) # of Obs 725 Wald Statistic 31.47 ***

46

Panel D: The Persistence of Special Items within Press Releases

Predicted Dependent Variable Variable Sign Et+1 E*t+1

E*t + 0.99 *** 0.97 *** SI_in_PRt + 0.25 *** 0.22 *** SI_Not_in_PRt + 0.69 *** 0.55 *** Test of SI_in_PR – SI_Not_in_PR - (0.44) *** (0.33) *** # of Obs 559 559 Adj.-R2 0.42 0.52

a This table presents results from analyses examining management discussion of special items in annual

press releases (PR). All analyses use special items (SI) measured via hand-collected data from firm’s 10-Ks and the corresponding press release reported on Business Wire (BW) or Press Release Newswire (PRN). Panel A presents the sample selection, using three size strata of special items: Small SI (0 < absolute total special items < 0.25% of lagged total assets); Medium SI (0.25% < absolute total special items < 1% of lagged total assets); and Large SI (absolute total special items > 1% of lagged total assets). Panel B provides univariate comparisons of the presentation of special items on the income statement to whether the special items are discussed in the press release (PR – Yes) or not discussed in the press release (PR – No). FN is for observations wherein all reported special items are aggregated in another line item on the income statement and revealed only in the footnotes; Mixed is for observations where in some reported special items are presented separately on the income statement; and IS is for observations wherein all reported special items are presented separately on the income statement. Panel C presents logit results examining the decision to discuss special items in the press release. The dependent variable is SI_Sep_PR, an indicator variable equal to 1 if the special items are discussed/identified anywhere in the annual press release, 0 otherwise. The independent variables are as defined in Table 4. Panel D presents analyses examining the persistence of special items discussed/identified in the press release. SI_in_PR (SI_Not_in_PR) are special items discussed/identified (not discussed/identified) in the annual press release, scaled by beginning market value of equity. All other variables are as defined in Table 5.

***, **, * represent significance at the 1%, 5%, and 10% levels for one or two-tailed tests where indicated, respectively.