Aggregationand Accounting Informativeness

Daniel A. Bens

INSEAD – Accounting and Control Area

Boulevard de Constance

F-7705 Fontainebleau Cedex, 77305

France

Steven J. Monahan

INSEAD – Accounting and Control Area

Boulevard de Constance

F-7705 Fontainebleau Cedex, 77305

France

Logan B. Steele

University of Connecticut

School of Business, Accounting Department

2100 Hillside Road, Unit 1041

Storrs, CT 06269

July December 304, 2013

This paper has benefitted from comments and suggestions provided by an anonymous referee,from Peter Easton,Mozaffar Khan, S.P. Kothari (editor), Stephen Ryan, and workshop participants at the 2009 HEC-INSEAD Accounting Colloquium, London Business School,New York University, Tilburg University, and the University of Arizona. A previous version was titled “Intra-firm Income Offsetting and Asymmetric Timely Loss Recognition.” We are solely responsible for any errors.

1

Aggregationand Accounting Informativeness

Abstract

In a sample of U.S. multiple-segment firms we document a negative association between aggregation and the accounting system’s ability to communicate bad economic news. We argue thataggregation, as reflected in firms’ reported organizational structures (the definition and characteristics of their segments), results in some multiple-segment firms exhibitinglower cross-segment variation in profitability. We find that firms thatare engaging in such aggregation have accounting systems that provide less timely information about economic losses. Our results hold for a variety of approaches to measuring the extent to which accounting earnings reflectbad economic newsin a timely manner. Weprovide initial evidence supporting Beyer et al. (2010) and Berger’s (2011) contention that accounting informativeness is,in part, a function of managers’ aggregation choices.

Keywords: accounting choice; aggregation; conservatism;impairments; segment reporting; special charges

1

1.Introduction

We examine how a firm’s reported organizational structure affects the extent to which accountingearnings provide timely information about bad economic news. The issue of how aggregation influences accounting informativeness has received little attention in the accounting literature. Rather, the extant research generally focuses on how segment-level aggregation is determined (Harris (1998); Berger and Hann (2007)). We test whether theaggregation of heterogeneous operations into a single segment: (1) reduces the probability of reporting significant negative accruals when other indicators of underperformance are present and (2) results in firm-level earnings that exhibit a relatively weak association with contemporaneous bad economic news.

We argue thattwo forces work in concert to delay the accounting recognition of bad economic news. We refer to the first, and more fundamental,forceasex anteaggregation. As the name implies, ex ante aggregation occurs at the point in time whenthe managers of a firm define itssegments for external reporting purposes—i.e., before segment earnings are determined. Ex ante aggregation reflects the commingling ofheterogeneous operations into a single segment.

We refer to the second force as ex postoffsetting, which occurs at the point in time when segment earnings are determined. Ex post offsetting manifests itself in two ways. First, when gain reserves (i.e., assets that have fair values that exceed their book values) are liquidated and recognized in segment income in order to offset the underperformance of some of the segment’s activities. For example, suppose one of a segment’s product lines experiences an unexpected decrease in revenues. Ceteris paribus, this underperformance leads to lower segment income. However, if during the same accounting period the segment’s managers recognize a gain by selling an asset that has appreciated in value, the effect of the underperformance on segment-level income (and by extension firm-level income) is mitigated.

The second way in whichex post offsetting occursis when gain reservesarenot liquidated but rather are used to avoid recognizing a segment-level impairment loss. This occurs because unrecognized gain reserves can be offset against unrecognized losses when a group is tested for impairment. In particular, even if the fair values of the some assets in a group have fallen below their carrying values (i.e., there are unrecognized losses), an impairment loss is not recorded if the gain reserves exceed the unrecognized losses.

Becauseex ante aggregation increases the heterogeneity of a segment’s operations, it increases the likelihood ofex post offsetting. As a segment’s operations become more heterogeneous, there is a greater likelihood that the segment will simultaneously have unrecognized gain reserves and: (1) some operations that are underperforming or(2) some assets with carrying values below their fair values—i.e., unrecognized losses. Hence, managers who define segments broadlyex ante have greater ability to carry out ex post offsetting.

An important consequence of ex ante aggregation and ex post offsetting is that there is less cross-segment variation in profitability. We refer to this as the “variance effect.” Suppose the managers of a multiple-segment firm define the firm’s segments broadly by either commingling unrelated activities within segments or by forming fewer segments.This represents ex anteaggregation and it causes the firm’s segment-level profits to exhibit relatively low dispersion—i.e., ceteris paribus, an increase in within-segment heterogeneity implies a decrease in cross-segment heterogeneity. Multiple-segment firms will also have relatively low dispersion in segmentprofits if managers conduct ex post offsetting within the segments they manage.

We take advantage of the variance effect to develop a new empirical measure of aggregation for a sample of multiple-segment firms.[1]In particular, for each multiple-segment firm in our sample, wecreate a proxy that is a decreasing function ofthe firm’s cross-segment variationin profitability. Hence, higher values of our proxy, which we refer to as R_SR, imply greater aggregation.[2]Next, we evaluate whether R_SR has a negative associationwith the timely accounting recognition of bad economic news—an issue that is vital for both the valuation and stewardship roles of financial accounting. We also form an alternative measure of managers’ aggregation choices by employing the implementation of SFAS 131 as a setting where previously “hidden” segments were revealed. This alternative measure has the benefit of comparing each firm with itself across time.

As we further detail in section two, both the recent review paper by Beyer, Cohen, Lys and Walther (2010), and the discussion of that paper by Berger (2011), highlight the lack of research examining the consequences of manager’s data aggregation choices. Since one of the primary functions of financial accounting is to summarize—i.e., aggregate—the population of a firm’s transactions, we are answering Beyer et al. (2010) and Berger’s (2011) call for more research in this area.

Our evidence strongly supportsour hypothesis. First, we evaluate the relation between R_SR and various “special” charges including asset impairments, restructurings, and special items. We study special charges because they are a set of observable accounting accruals that directly communicate bad economic news to financial statement users. Using a Probit model, and controlling for various fundamentals that determine special charges, we show that there is a negative association between R_SR and the likelihood that a firm reportsa special charge in a particular fiscal year.

We then broaden our view of accounting beyond special charges and examine how R_SR influences the association between negative stock returns and accounting earnings. Basu (1997) first documented that earnings exhibit a stronger association with bad economic news (captured by negative stock returns) vis-à-vis good economic news (positive stock returns). We demonstrate that this association between earnings and negative economic news declines as R_SR—i.e., the degree of aggregation—increases.[3]This evidence is corroborated by our alternative measure of aggregation, as we find that the association between earnings and bad economic news increases for firms that, upon adopting SFAS 131, revealed segments that were “hidden” under the previous segment reporting rules (i.e., SFAS 14).

Next, we evaluate the effect of aggregation on the interaction between accruals and cash flow changes. As discussed in Ball and Shivakumar (2006), accruals play two roles: (1) a smoothing role in which they smooth out transitory changes in cash flows and (2) a stewardship role in which they complement bad news in cash flows. We show that as R_SR increases accruals smooth out both good and bad news in cash flows, which implies that the stewardship role of accruals is diminished as aggregation increases. Finally, we show that the income-effects associated with bad economic events are smoothed over longer time periods in the face of aggregation: increases in R_SR are accompanied by increases in the time-series persistence of negative earnings changes.

We contribute to the extant literature in two ways. First, our results shed light on how a firm’s reported organizational structure—i.e., the definition and characteristics of its segments—affects the extent to which accounting earningsreflect contemporaneous bad economic news. Hence, we add to the body of literature that considers how conservative accounting varies across accounting and regulatory regimes (e.g., Ball, Kothari and Robin (2000), and Bushman and Piotroski (2006)), as well as operating and financing environments (e.g., Khan and Watts (2009)). We also respond to the call from Beyer et al. (2010) and Berger (2011) for more research on the effects of managers’ aggregation policies.

Second, we provide additional evidence on the reporting choices made by managers of multiple-segment firms. As discussed in section two, these firms are economically relevant andare the subject of a sizeable body of research. For example, they serve as the unit of analysis in studies examining line-of-business diversification (e.g., Lang and Stulz (1994); Berger and Ofek (1995); Denis, Denis and Sarin (1997); etc.), and numerous studies evaluate these firms’ disclosure choices (e.g., Harris (1998); Berger and Hann (2003); Bens, Berger and Monahan (2011); etc.) and the valuation implications of these choices (e.g., Bens and Monahan (2004) and Berger and Hann (2007)). We extend this literature by evaluating the effect that aggregation has on an attribute of firm level profitabilitythat is of first-order importance: the ability to communicate bad economic news.

2.Hypothesis Development

If bad economic events impose a lower cost on “insiders” (e.g., managers, large block holders, etc.) than minority/non-controlling investors (i.e., “outsiders”), insiders will exploit their information advantage and withhold, or delay the release of,information about these events. Outsiders react to this information problem by price protecting. For example, they reduce the expected value of manager’s compensation, they discount securities issued by the firm, etc. This is suboptimal, however, as it leads to deadweight costs. One way of mitigating these deadweight costs is for insiders to commit to provide outsiders with more timely information about bad economic news.[4]

Although such a commitment is valuable, perfect timeliness will not be adopted for at least two reasons. First, outsiders do not want it. Perfect timeliness implies that all bad economic news is published in a timely manner in the firm’s financial reports. Although this sort of transparency increases outsiders’ ability to monitor insiders’ actions, it also provides the firm’s rivals with valuable information they can use for competitive purposes. In light of this fact, outsiders and generally accepted accounting principles, GAAP, give managers discretion over when economic losses are recognized in accounting earnings.[5] Second, the communication of bad economic news is also imperfect if insiders and outsiders cannot write perfect, complete contracts. Ex post it is infeasible for outsiders to verify whether insiders provided full information. Moreover, even if outsiders are able to prove that the manager dissembled, full ex post settling up is not always possible given horizon problems and the fact that managers have limited liability. This implies that managers have discretion about what and when they disclose.

One way for managers to exercise their discretion is by aggregatingdissimilar activities into a single segment, which we refer to as ex ante aggregation. This allows them to hide the sources of their firm’s earnings from the managers ofrival firms (e.g., Hayes and Lundholm (1996);Harris (1998); etc.). Managers might also aggregate in order to hide information from outside monitors (e.g., Berger and Hann (2007);Bens, Berger and Monahan (2011); etc.).

The literature cited aboveevaluates whether unresolved agency problems or proprietary costs affect the manner in which the firm’s internal activities are aggregated and reported to outsiders. We extend this literature by examining how aggregationchoices affect firm-level reported earnings. In particular, we hypothesize that aggregating heterogeneous operations into a single segment reduces the extent to which reported firm-level earnings communicate bad economic news in timely manner. This delayed recognition is important because, as discussed in Ball (2001) and Watts (2003a, b), it can cause the firm’s shareholders to make suboptimal personnel decisions (e.g., retain underperforming executives) and adopt suboptimal investment strategies.

We focus on the implications of ex ante aggregation for management’s ability to conduct ex post offsetting. Higher ex ante aggregation implies a diversification effect—i.e., as a business unit’s activities become more heterogeneous, it becomes more likely that the unit will contain someunrecognized gain reserves. These gain reserves can be used in two ways. First, they can be recognized in order to offset the underperformance of other activities. This sort of offsetting can take various forms such as the strategic timing of asset sales (Bartov (1993)), discretionary cuts in R&D spending (Bushee (1998)), strategic management of working capital (Roychowdhury (2006)), etc.

Second, gain reserves can be used to either avoid or reduce the magnitude of impairment charges. The reason is that they can be offset against unrecognizedlosses on other activities in the unit. Absent ex ante aggregation these losses would be recognized; yet, they arenot in this case because they have been aggregated with other assets providing recoverable cash flows to such an extent that the combined asset group is not considered to be impaired per U.S. GAAP.

That is, impairment tests are inherently tied to the aggregation of assets used in the test; and, U.S. GAAP provides managers with a great deal of discretion over how this aggregation is done. For example, tangible fixed assets and identified intangibles subject to amortization are to be tested at the “asset group” level, with charges recorded only when the aggregate carrying amount of the group exceeds its aggregate undiscounted future cash flows (ASC 360-10-35-17). Moreover, managers have discretion over what to include inanasset group. Although the term asset group is defined in the ASC glossary as “the lowest level for which identifiable cash flows are largely independent of the cash flows from other groups of assets and liabilities,”practitioners have noted that this definition is challenging to implement. For example, Nurnberg and Dittmar (1996) writing in theJournal of Accountancy state (italics added):

Determining the lowest level of grouping requires considerable judgment. As groupsbecome broader, more judgment is involved and the opportunity is greater to combineassets with fair values in excess of carrying amounts with those that are impaired. Auditors should consider this when evaluating the reasonableness of groupings.

The standards regarding goodwill asset impairments also afford managers a great deal of discretion in aggregation. Goodwill is assigned to a “reporting unit” and assessed for impairment when the fair value of the unit falls below its carrying value. Although measuring fair value of non-traded assets is challenging enough, as described in ASC 350-20-35-37, there is also considerable ambiguity regarding how to define the unit containing the goodwill (italics added):

Reporting units will vary depending on the level at which performance of the segment is reviewed, how many businesses the operating segment includes, and the similarity of those businesses. In other words, a reporting unit could be the same as an operating segment, which could be the same as a reportable segment, which could be the same as the entity as a whole.

Practitioner-oriented journals give advice about strategies for allocating and reporting goodwill, such as advocating “establish[ing] a strong and well-managed reporting unit to be used as a basis for establishing goodwill to minimize the ‘impairment problem’” (Martinson 2002).

In light of the above, whether and to what extent aggregationaffectsthe information content of accounting earningsare relevant empirical questions. Nonetheless, these issues have received relatively little attention in the accounting literature. Rather, the extant research mainly focuses on how segments are determined (e.g., Harris (1998) and Berger and Hann (2007)).

Two recent academic review papers highlight the importance of aggregation and the relatively scant amount of research about it. Beyer et al. (2010) provide a comprehensive review of several facets of the financial reporting environment. One topic they address is the role of conservatism in financial reporting. Moving beyond the general debate over whether conservatism increases or decreases economic welfare, Beyer et al. point out the difficulty of accurately measuring it in the presence of aggregation (italics added):

In our view, obtaining a better understanding of the informational properties of conservative accounting numbers requires joint consideration of conservatism and aggregation, another salient feature of accounting. The reason is that ignoring good news when such news lack verifiability can reduce the informativeness of accounting reports. This is always true when gains and losses are reported separately. However, when gainsand losses are aggregated, less informative gains offset more informative losses such that the aggregate report may overall be less informative when gains are recognized than when gains are not recognized (p. 317).

Berger (2011) reviews Beyer et al., and also highlights the need for a better understanding of aggregation:

The archival data we observe are almost always the output of a considerable amount of aggregation. To really understand the decisions that lead to this aggregated output, we need to learn more about how managers use their discretion in going from the disaggregated data they observe internally to arrive at the highly aggregated items that get reported externally (p. 216).

Practitioners are also interested in the subject of aggregation. For example, Scott Taub, former Acting Chief Accountant of the SEC, writes about the “unit-of-account” issue, which is identical to what we refer to as aggregation: