REPRESENTATIONALLY FAITHFUL DISCLOSURES, ORGANIZATIONAL
DESIGN AND MANAGERS’ SEGMENT REPORTING DECISIONS
Christine Botosan
Professor of Accounting
University of Utah
Susan McMahon
Ph.D. Student
University of Utah
Mary Stanford
Professor of Accounting
Texas Christian University
March 2009
ABSTRACT: SFAS No. 131 governs disclosure about segments of an organization. Critics charge that
the latitude extended to managers under SFAS No. 14 led to segment disaggregation in which unrelated
or vaguely related business activities were combined under broad industry headings. By prescribing a
management approach to segment disaggregation, SFAS No. 131 calls for segmentation that more closely
reflects firms’ organizational design. The purpose of our study is to evaluate the success of SFAS No. 131
in inducing firms to disclose more about their organizational design via their segment definitions. Overall,
we find that firms that changed their segment disaggregation following the adoption of SFAS No. 131
increased the relatedness of operations combined in segments suggesting that their external reporting is
more aligned with their internal organizational structures. Our results also indicate that firms reporting a
single segment under SFAS No. 14, but multiple segments under SFAS No. 131, tended to take greater
advantage of the flexibility in SFAS No. 14 to “hide” certain operations. This is consistent with prior
research that suggests these firms faced greater proprietary, agency or other costs of disclosure. Finally,
we provide evidence that firms that did not change their segment definitions in response to SFAS No.
131, combine more dissimilar operations that likely deviate from their internal organizational structures to
a greater extent than firms that changed their segment definitions in response to SFAS No. 131.
Keywords: Segment, disclosure, SFAS No. 131, SFAS No. 14.
Data Availability: All data is publicly available.
I.
INTRODUCTION
The Sarbanes-Oxley Act of 2002 required the SEC to study standard setting in the U.S. The
congressional mandate charged the SEC with investigating the costs and benefits of concepts-based vs.
rules-based standard setting. The SEC concluded that both approaches pose challenges. Concepts-based
standards potentially present enforcement challenges, but properly applied should yield accounting
information consistent with the economic substance of transactions and events. In contrast, rules-based
standards can increase the perceived comparability and consistency of financial information, but
encourage transaction structuring to achieve desired financial reporting objectives regardless of the
economic substance of the underlying transaction. Ultimately, the SEC concluded that the benefits of
concepts-based standards outweigh the costs, and recommended a move to more concepts-based standard
setting (SEC, 2003).
In describing the desirable properties of a concepts-based standard, the SEC noted that such a
standard should be based on an improved and consistently applied conceptual framework. Consequently,
in October 2004, the FASB and IASB added to their agendas a joint project to develop an improved
conceptual framework. In the spring of 2008, this project produced an exposure draft, which elevates
faithful representation to the position of one of the two fundamental qualitative characteristics of
decision-useful financial reporting information; the other fundamental qualitative characteristic is
relevance. The proposed greater emphasis on faithful representation in the conceptual framework is
consistent with the prominence afforded to the communication of “economic substance” in the SEC’s
discussion of the primary benefit of a concepts-based standard.
Constituents, such as financial analysts, heavily criticized SFAS No. 14, Financial Reporting for
Segments of a Business Enterprise, for failing to provide decision-useful information. SFAS No. 14
dictated that managers use industry membership to delineate their segments, but critics charge that the
standard allowed managers undue discretion in defining industry boundaries, and consequently too much
discretion in determining their segment definitions.1 This lead to the concern that the resulting segment
disclosures failed to faithfully represent firms’ internal organization of business activities into operating
units, thereby allowing managers to obscure value-relevant information. Consistent with the argument
that motives other than faithful representation influenced segment definitions under SFAS No. 14, prior
research finds that managers’ segment definitions were associated with exposure to proprietary and/or
agency costs of disclosure (Harris, 1998; Botosan and Stanford, 2005; Berger and Hann, 2007).
In light of these concerns, SFAS No. 131, Disclosures about Segments of an Enterprise and
Related Information, superseded SFAS No. 14.2 A critical difference between SFAS No. 131 and SFAS
No. 14 is the adoption of the management approach to grouping operations into reportable segments. The
management approach identifies reportable segments based on the way management organizes business
activities within the firm for the purposes of operating decisions and assessing performance. A key
objective of SFAS No. 131 is to improve the decision-usefulness of firms’ segment disclosures by
faithfully representing the firm’s internal organization of business activities.
Typically it is quite difficult for research to provide evidence of the extent to which managers’
external disclosures faithfully represent the underlying economics of a business activity or event because
the underlying economics of a transaction or event are not determinable in a large sample setting.
Nonetheless, the adoption of SFAS No. 131 provides a unique opportunity to offer such evidence. We
investigate whether SFAS No. 131 is more successful than SFAS No. 14 in compelling managers to
delineate segments for external reporting purposes in a manner that faithfully represents their firms’
internal organizational structure.
The most direct test of our research question would involve a comparison of firms’ internal
organizational structures to their groupings of business activities for segment reporting purposes, but
these internal organizational structures are not directly observable. Nonetheless, existing research in
accounting, finance, and strategy identifies several aspects of organizational design that maximize firm
1
Under SFAS No. 14 firms were required to provide line-of-business and geographic segment disclosures. For
purposes of our study we focus on line-of-business disclosures only.
2
SFAS No. 131 is effective for fiscal year ends beginning after December 15, 1997.
2
value. Specifically, strategic management theory suggests that firms’ real diversification decisions reflect
a tradeoff between “the threat of losing focus and the opportunity to grow and exploit synergies” (Adner
and Zemsky, 2006). Consistent with this, studies in finance show that related diversification mitigates the
diversification discount and more recently results in a diversification premium (Berger and Ofek, 1995;
Villalonga 2004).
We assume that firms structure their activities to maximize firm value, and rely on the strategic
management literature to identify the aspects of organizational design that achieve this objective.
According to this literature, segmentation – or the deliberate grouping of tasks and activities into business
units that best achieve a firm’s strategic objectives and operating efficiencies – is an important aspect of
organizational design (Brickley, Smith and Zimmerman, 2004 (BSZ)). Specifically, the strategy literature
concludes that relatedness among activities is a critical dimension of a firm’s “optimal” segmentation
strategy (Rumelt, 1974; Palepu 1985), and economic or environmental pressures such as competition and
market opportunities also shape a firm’s segmentation strategy (BSZ).
Using a sample of 1,081 firms that altered their segment definitions in response to SFAS No. 131,
we examine whether the segment definitions employed after the adoption of the new standard reflect
greater relatedness among operations than the segment definitions employed under SFAS No. 14. We also
examine whether the similarity of business units’ competitive environments and growth opportunities
increased within each segment following the implementation of the new standard. Finally, we test whether
the diversity of these attributes across segments increased post-SFAS No. 131.
Focusing on firms that changed their segment definitions in response to SFAS No. 131 excludes
firms that continue to employ the same definitions after adoption of the new standard as before. This latter
group of firms is interesting since their lack of response to the new standard suggests that their segment
definitions might reflect their internal organizational structures even under SFAS No. 14. In other words,
these firms may have “voluntarily” provided representationally faithful definitions despite the inherent
flexibility in SFAS No. 14 that would have allowed them to do otherwise. Alternatively, SFAS No. 131
might be sufficiently costly for these firms that they choose to mitigate their segment disclosure costs by
3
altering their internal organizational structures or choosing not to comply with SFAS No. 131. In a small
sample study, Street, Nichols and Gray (2000) find little evidence of internal restructuring in response to
SFAS No. 131. However, they also find that a “significant minority” continue to report segment
information on a basis that is inconsistent with other parts of the annual report. Thus, we investigate these
issues by examining the segment definitions employed by a sample of 2,690 firms that did not alter their
segment definitions in response to SFAS No. 131, hereafter referred to as the no-change firms.
We find that firms reporting a single segment under SFAS No. 14, but multiple segments under
SFAS No. 131 alter their segment definitions toward a more faithful representation of their firms’ internal
organization of business activities. Similarly, we find that firms reporting multiple segments under both
standards, but employing different groupings under SFAS No. 131, also alter their segment definitions
toward more faithful representation of their firms’ internal organizational structures. Moreover, the single
segment change firms group together considerably more diverse operations under SFAS No. 14 than the
multiple segment change firms, suggesting that the single segment firms took greater advantage of the
flexibility inherent in SFAS No. 14 and provided segment disclosures that deviated from the economic
substance of their internal organizational structure to a greater extent than the multiple segment change
firms.
We find that firms not changing their segment definitions in response to SFAS No. 131 differ in
the pre- and post-SFAS 131 period from firms changing their segment definitions in response to the new
standard. Under SFAS No. 14 the single and multiple segment no-change firms group more similar
operations than the single or multiple segment change firms. This suggests that under SFAS No. 14, the
no-change firms provided relatively more representationally faithful segment disclosures than the change
firms. Nonetheless, after the adoption of SFAS No. 131, the single and multiple segment change firms
group significantly more similar operations than either subset of no-change firms. This suggests that the
no-change firms tend to provide relatively less representationally faithful segment definitions when
compared to the benchmark established by SFAS No. 131. Accordingly, our results suggest that SFAS
4
No. 131 is partially successful in achieving a key objective of enhancing the relevance of firms’ segment
disclosures by faithfully representing the firm’s internal organization of business activities.
We believe that the adoption of SFAS No. 131 provides an interesting context for examining the
question of the extent to which firms’ disclosures are representationally faithful for several reasons. First,
a primary objective of SFAS No. 131 was to enhance the faithful representation of firms’ internal
organization of business activities via their segment disclosures. Second, the management approach is
expected to be more reliable and less costly to follow since it employs information generated for internal
use. Thus, we expect managers and auditors to share a cost effective signal of the underlying economics
the disclosures should reflect. This is important because we would expect non-representational faithful
reporting to be even more of an issue when the underlying economics are difficult for auditors to observe.
Accordingly, to the extent we provide evidence of non-representational faithful reporting, we contend that
our evidence should be interpreted as a lower bound of such activities. Third, the strategy literature
suggests various attributes we expect to characterize firms’ otherwise unobservable internal
organizational structures. Thus, the context we examine is unique in that we have a theoretically
supported approach for capturing the underlying economics of the business activities the disclosure
purports to convey.
Our paper contributes to the disclosure literature by undertaking an initial attempt to assess the
representational faithfulness of firms’ disclosures in a large sample setting. We believe standard setters
will be interested in our evidence regarding the extent to which SFAS No. 131 achieves its objective of
compelling firms to disclose segment information that faithfully represents their internal organizational
structures. Analysts and other users might also be interested in the results of our research since
representationally faithful disaggregation provides information to allow users to better assess firms’ future
cash flows. In addition, our paper contributes to the disclosure literature by linking managers’ disclosure
choices to the strategic management and organization design literature. Finally, we introduce multiple
measures of the relatedness of operations that have not been examined in prior accounting research.
5
The remainder of our paper proceeds as follows. In Section II, we summarize the related literature
and develop our hypotheses. Section III describes our research design and empirical proxies. In Section
IV, we outline our sample selection procedures and provide descriptive statistics regarding our sample
firms. Section V presents the results of our analyses. Finally, Section VI offers a summary of our findings
and conclusions.
II.
RELATED LITERATURE AND HYPOTHESES DEVELOPMENT
Factors Motivating Managers’ Segment Definitions under SFAS No. 14
Prior literature finds that the economic consequences of disclosure influence managers’ segment
definitions under SFAS No. 14 (Harris, 1998; Botosan and Stanford, 2005; Berger and Hann, 2007).
Studies show that firms facing greater proprietary costs of disclosure are less likely to provide finer
segmentation. That is, managers seek to protect competitive advantages (i.e. business units generating
abnormal profits) by including a wider variety of business activities within reported segment categories
(Harris, 1998; Botosan and Stanford, 2005).
Hayes and Lundholm (1996) provide a theoretical model that supports the proprietary cost
argument. The model predicts that firms are more likely to report activities separately when the results of
operations are more alike, and group activities together when their results are more dissimilar. Hayes and
Lundholm suggests that managers prefer to disaggregate similar activities because the finer disclosure has
less potential to inform rivals. By combining dissimilar activities mangers can protect sensitive
information. Interestingly, this prediction is opposite to that found in the strategy literature, which
suggests that firms capitalize on operational efficiencies and economies of scale by combining similar
activities for operating purposes.
In contrast, Berger and Hann (2007) interpret the evidence of firms’ aggregation of diverse
activities under SFAS No. 14 as indicating agency issues. For example, managers might be reluctant to
reveal underperforming segments. Evidence provided in Botosan and Stanford (BS, 2005) is not
consistent with this interpretation, however. BS examine the characteristics of business activities
6
previously hidden by firms that reported as single segment firms in the pre-SFAS No. 131 era. BS find
that managers withhold segment information under SFAS No. 14 to protect profits in less competitive
industries, not to conceal poor performance. Regardless, the overall conclusion drawn from this stream of
research is that disclosure costs tend to motivate managers’ disaggregation choices under SFAS No. 14.
The Advent of SFAS No. 131
In 1994, the Special Committee on Financial Reporting issued a report identifying five areas of
concern with SFAS No. 14.3 First, users believed that companies were reporting an inadequate number of
segments. Second, they were concerned about the limited amount of information disclosed about each
segment. Users also complained about a lack of consistency between firms’ segment grouping for
purposes of their segment disclosures versus the grouping emphasized elsewhere in the annual report such
as that used in the MD&A. Fourth, users expressed a preference for segmentation that corresponds to
internal management reports. These latter two complaints are symptomatic of a disconnect between firms’
segment groupings for external reporting purposes and their internal groupings of activities for operating
purposes. Finally, users called for the disclosure of segment information in firms’ quarterly financial
reports.
The FASB issued SFAS No. 131, which requires a “management approach” to define segments,
to address these perceived weaknesses. Under the management approach, reportable segments are defined
based on the way management organizes business activities within the firm for making operating
decisions and assessing performance. Since SFAS No. 131 was issued to address users’ concerns, key
objectives of SFAS No. 131 include increasing the number of reportable segments, increasing the amount
of information disclosed about each segment, increasing the timeliness of the disclosures, and increasing
the representational faithfulness of the segment definitions to the underlying organization of firms’
business activities.
3
This board was formed in 1991 by the AICPA to make recommendations on how to improve the relevance and
usefulness of financial reporting (FASB 1997). Their conclusions, subsequently incorporated into FASB’s exposure
draft, closely resemble those of a position paper addressing concerns regarding segment disclosure prepared by the
Association for Investment Management and Research (AIMR).
7
Existing research examines the extent to which SFAS No. 131 achieves the goal of improving
segment disclosures along some of the dimensions summarized in the preceding paragraph. For example,
extant research finds that disaggregation (specifically, the number of segments being reported) and the
amount of information provided about each segment increased after SFAS No. 131 (Herrmann and
Thomas, 2000; Berger and Hann, 2003). In addition, Street et al. (2000) find that SFAS No. 131 segment
information is more consistent with other information provided in the annual report. Ettredge, Kwon,
Smith and Zarowin (2005) show that the forward earnings response coefficient increased in the SFAS No.
131 period. In addition, Ettredge, Kwon, Smith and Stone (2006) find an increase in the across segment
variability of reported segment profits post SFAS No. 131.
Taken together this stream of research suggests that the quality of segment disclosures improved
post-SFAS No. 131. Nonetheless, the actual degree of improvement depends on the extent to which
managers’ segment definitions faithfully represent the underlying economics of their business operations.
In SFAS No. 131, the FASB explicitly addresses the importance to users of defining segments based on
internal organizational structure.
Almost all of the users and many other constituents who responded to the
Exposure Draft or who met with Board and staff members agreed that defining
segments based on the structure of an enterprise’s internal organization would
result in improved information. They said that not only would enterprises be
more likely to report more detailed information but knowledge of the structure of
an enterprise’s internal organization is valuable in itself because it highlights the
risks and opportunities that management believes are important.4
Thus, notwithstanding the evidence that managers provide greater and more consistent
information about a larger number of segments post-SFAS No. 131, users might still be hampered in their
ability to understand firms’ various risk exposures and future prospects if segment definitions fail to
faithfully represent the internal organizational structure. The importance of this question extends beyond
that of segment reporting to include the broader question of whether managers generally provide
representationally faithful disclosures.
4
SFAS No. 131, ¶59.
8
To our knowledge, existing research does not examine the extent to which managers’ SFAS No.
131 segment definitions convey the underlying economics of their internal organizational structures.
Moreover, we are aware of no prior research that attempts to ascertain the extent to which managers’
SFAS No. 14 segment definitions did or did not reflect their internal organization structures. We address
these issues by examining whether managers’ segment definitions parallel their internal organizational
structures to a greater extent following the adoption of SFAS No. 131, than under SFAS No. 14.
Factors Motivating Managers’ Internal Organizational Design Choices
Strategic management theory suggests that firms’ diversification decisions reflect a tradeoff
between “the threat of losing focus and the opportunity to grow and exploit synergies” (Adner and
Zemsky 2006). By structuring a diversified firm to enhance relatedness among operations, similarity of
competitive environments and market opportunities, managers minimize the potential costs of
diversification, and maximize the benefits.
We assume that managers structure their organizations to achieve operational efficiencies and
maximize firm value. Literature in accounting, finance, and strategy identifies aspects of organizational
design that maximize firm value. We rely on this literature to infer the types of operations firms group
together to achieve operational efficiencies and maximize firm value. Thus, although we do not observe
our sample firms’ internal organizational structures directly, we infer it based on this literature. We then
examine whether firms’ SFAS No. 131 segment definitions are more consistent with these inferred
groupings than firms’ SFAS No. 14 segment definitions. In the following paragraphs, we provide our
hypotheses and elaborate on the characteristics of organization design identified by the extant literature.
Industry Relatedness
The expertise required to manage related operations is presumably less than the expertise required
to manage operations that are more diverse. Baiman et al. (1995) find that management task expertise
(derived from product line relatedness), and the relative importance of a business unit to the firm, are
important determinants of organizational design. In a 2006 working paper, Adner and Zemsky model the
relationship between diversification and firm performance, and conclude that greater relatedness increases
9
the likelihood that diversification is profitable. Related literature in finance finds that greater focus, or a
move to divest unrelated operations, is associated with higher stock returns (Comment and Jarrell, 1995).
Berger and Ofek (1995) find that operating in related industries mitigates the “diversification discount”
(i.e. the negative market implications of corporate diversification). More recently, Villalonga (2004)
documents an overall diversification premium and presents evidence of a discount to unrelated
diversification offset by a predominate premium to related diversification. Research examining
acquisitions and subsequent divestitures find that firms are far more likely to divest unrelated acquisitions
than related acquisitions (Kaplan and Weisbach, 1992) and that post-merger performance is higher when
target and acquiring firms share a high level of business overlap (Healy, Palepu and Ruback, 1992).
This literature demonstrates that expansion into related activities has a more positive effect on
firm performance than diversification into unrelated activities. In related markets, firms are able to exploit
resources and cross-utilize technical and managerial skills, and knowledge (Rumelt, 1974). Additionally,
in related industries, firms more easily establish and maintain their reputation, as well as economies of
scale and scope (Nayyar, 1993). Accordingly, we expect managers to structure operations to exploit the
operational benefits accruing from greater industry relatedness. To the extent that SFAS No. 131
increases the representational faithfulness of firms’ segment definitions vis-à-vis their internal
organizational structures, we expect the degree of within segment industry relatedness to increase postSFAS No. 131. This yields our first hypothesis, stated below.
H1:
Following the implementation of SFAS No. 131, there is a significant increase in industry
relatedness among the operations grouped as segments.
Competitive Environment and Growth Opportunities
As noted previously, the proponents of the management approach to defining segments believe
that knowledge of the structure of an enterprise’s internal organization is valuable in itself because it
highlights the risks and opportunities that management believes are important. A major source of
operating risk derives from the firm’s competitive environment. In addition, growth opportunities are
likely among the opportunities management considers important to firm value. This suggests that
10
managers are more likely to structure their organizations to group operations that face similar risks and
opportunities and is consistent with a notion of economic “relatedness.”
We expect firms to group together operations characterized by similar economic risks and
opportunities. Accordingly, we expect similarity in competitive environment and growth opportunities to
play a role in firms’ internal organizational structures. To the extent that SFAS No. 131 increases the
representational faithfulness of firms’ segment definitions vis-à-vis their internal organizational
structures, we expect to observe a significant increase in the similarity of competitive environment and
growth opportunities among the operations grouped together to form a segment, following the adoption of
SFAS No. 131. This forms the basis of our second and third hypotheses, stated below.
H2:
Following the implementation of SFAS No. 131, there is a significant increase in the
similarity of competitive environment among the operations grouped as segments.
H3:
Following the implementation of SFAS No. 131, there is a significant increase in the
similarity of growth opportunities among the operations grouped as segments.
Hypotheses Pertaining to Cross-Segment Diversity
If firms structure their organizations by grouping related operations to achieve operational
efficiencies, grouped operations should be similar to one another. This motivates our hypotheses stated
above, which focus on the similarity of operations within a given segment. At the same time, if firms
structure their organizations by grouping related operations, operations grouped together should be
distinct from other parts of the organization. This motivates a second set of hypotheses stated below,
which focus on an increase in the diversity of industry relatedness, competitive environment, and growth
opportunities across segments post-SFAS No. 131.
H4:
Following the implementation of SFAS No. 131, there is a significant increase in the
diversity of industry relatedness across segments.
H5:
Following the implementation of SFAS No. 131, there is a significant increase in the
diversity of competitive environments across segments.
H6:
Following the implementation of SFAS No. 131, there is a significant increase in the
diversity of growth opportunities across segments.
11
III.
RESEARCH DESIGN AND EMPIRICAL PROXIES
This section describes our empirical proxies for industry relatedness, competitive environment,
and growth opportunities. Our measures of industry relatedness reflect horizontal and vertical integration
as well as related and unrelated diversification. Below we describe our variables and measurement
procedures in detail.
Proxies for Industry Relatedness
Within-Segment Tests (H1)
Compustat assigns a primary and, if applicable, secondary SIC code, to each segment
based on information in the firm’s 10-k or annual report. For single-activity segments,
Compustat assigns a primary SIC code but no secondary SIC code. A Compustat representative
stated that the main reason Compustat assigns a primary SIC code only is that a single SIC code
fully describes the activities of the segment as reported in the company footnote.
These single-activity segments are important to our analyses since, by definition, they
capture “perfectly” related activities. Since many of our relatedness measures require both a
primary and secondary SIC code we set the secondary SIC code equal to the primary SIC code
for all single-activity segments. We apply this procedure to the data in the pre- and post-SFAS
No. 131 periods. We believe our procedure captures the economics of the single-activity
segments, and the Compustat representative we spoke with agreed that equating the secondary
SIC code to the primary SIC code is appropriate in these cases.5
We base our first four industry relatedness measures on these data. We compute our first industry
relatedness measure by taking the absolute value of the difference between a segment’s primary and
5
If managers strategically describe more single-activity segments that are not actually single-activity operations
post-SFAS No. 131 this procedure could bias in favor of finding support for our first four hypotheses. The tests of
our second set of hypotheses, which look across the primary SIC codes of the firm’s segments, are not vulnerable to
this issue, however.
12
secondary 3-digit SIC codes. We refer to the resulting variable as Threediff.6 The smaller the difference
between the primary and secondary 3-digit SIC codes the more similar the grouped industries.
Consequently, Threediff is decreasing in industry relatedness. Since H1 predicts a significant increase in
industry relatedness following the adoption of SFAS No 131, a significant decline in Threediff provides
support for our first hypothesis.
This approach to capturing industry relatedness follows prior research (e.g. Palepu, 1985; Berger
and Hann, 2003; Chen and Zhang, 2003; Bens and Monahan, 2004). Nonetheless, it relies on a somewhat
questionable cardinal interpretation of the distance between SIC codes. For example, a Threediff value of
one suggests greater industry relatedness than a Threediff value of 100, but it is inappropriate to conclude
that the operations grouped together in the first instance are 100 times more closely related than the
operations grouped together in the second instance.
To mitigate this issue, we also employ a variation of a measure of industry dissimilarity
introduced by Givoly et al. (1999). We refer to this proxy as Dissim. Exploiting the hierarchical nature of
the SIC coding scheme, we assign Dissim a value of four when a segment’s primary and secondary 4-digit
SIC codes do not share the same first digit. We set Dissim equal to three when the primary and secondary
4-digit SIC codes share the same first digit, two when the SIC codes share the same first 2-digits, one
when the they share the same first 3-digits, and zero if the primary and secondary 4-digit SIC codes are
identical. Accordingly, Dissim is also decreasing in industry relatedness. Since H1 predicts a significant
increase in industry relatedness following the adoption of SFAS No 131, a significant decline in Dissim
provides support for our first hypothesis.
Some researchers charge that the SIC system does not do a particularly good job of distinguishing
among industries, motivating some researchers to rearrange the SIC codes into researcher-defined
“industries”. For example, Fama and French (1997 (FF)), reclassify SIC codes into 48 “industries” that
6
Beginning in 1997 the government supplanted SIC codes with the North American Industrial Classification Codes
(NAICS) for use in its statistical analyses. NAICS codes are based on similarities in production processes. Bhojraj,
Lee and Oler (2003) find that the two coding methods perform similarly in empirical tests, however. Given this, and
consistent with prior research, we generally employ SIC codes, although our measures for vertical relatedness as
well as our measures based on profit margin/asset turnover portfolios use NAICS industry classifications.
13
they argue better reflect common risk factors.7 We also examine a measure of industry relatedness based
on the FF industry groupings. SameFF is an indicator variable that we set equal to one if a segment’s
primary and secondary SIC codes belong to the same FF industry, and zero otherwise. Given its
construction, SameFF also mitigates against the cardinal interpretation problem that plagues Threediff.
SameFF is increasing in industry relatedness. Since H1 predicts a significant increase in industry
relatedness following the adoption of SFAS No 131, a significant increase in SameFF provides support
for our first hypothesis.
We base our final industry relatedness measure on relationships between asset turnover and profit
margin that generally arise from firms’ business strategy choices. Selling and Stickney (1989 (SS)), show
that industries with a high proportion of fixed costs and high entry barriers tend to have the highest profit
margins and lowest asset turnovers. In contrast, industries with low capital intensity and commodity-like
products tend to have the lowest profit margins and highest asset turnovers.
We capitalize on SS’s finding as follows. We compute industry profit margin and asset turnover
ratios using data for 1990 through 1995 for the Compustat population of single-segment firms.8 We
examined several different definitions of “industry” including 1-, 2-, and 3-digit SIC codes, as well as 2-,
3-, and 4-digit NAICS. Since SS document a negative correlation between asset turnover and profit
margin that is linked to underlying business strategy, we decided to “let the data speak” and define
industry based on which definition yields the largest negative correlation between the industry profit
margin and asset turnover ratios. Based on this analysis we define industry in terms of 2-digit NAICS
code.
Based on quintile cutoffs of industry profit margin and asset turnover, respectively, we define five
“profit margin” portfolios, and five “asset turnover” portfolios, with portfolio 5 comprised of the
industries with the highest values. We then allocate industries to twenty-five “strategy” portfolios based
7
See Appendix A for a cross-reference of the Fama and French classification system to SIC codes.
We compute profit margin as the ratio of operating profit after depreciation (Compustat data item #178) to Sales
(data item #12), and asset turnover as the ratio of Sales to average total assets (data item #6). For purposes of these
computations, we eliminate firms with assets less than zero, and sales less than $20 million. We also require a
minimum of 10 firms in each industry category.
8
14
on combinations of profit margin, and asset turnover portfolios. For example, we indentify industries with
the highest profit margin and lowest asset turnover by the strategy portfolio combination (5, 1). We then
define an indicator variable, PMATpf, which we set equal to one if a segment’s primary and secondary 2digit NAICS codes belong to the same strategy portfolio, and zero otherwise. PMATpf is increasing in
industry relatedness. Since H1 predicts a significant increase in industry relatedness following the
adoption of SFAS No 131, a significant increase in PMATpf provides support for our first hypothesis.
Given its construction, PMATpf overcomes several potential concerns with our other measures.
First, it mitigates against the cardinal interpretation problem that troubles Threediff. Second, it combines
2-digit NAICS codes into “industries” based on combinations of ratios, which prior research links to
strategic choices. This is in contrast to the Fama French approach which is based on a researcher
determined definition of industry groupings.
The four measures described above capture the extent to which firms engage in horizontal
integration, but some firms adopt a strategy of vertical integration. Our final industry relatedness measure
addresses this possibility. Following prior research (Fan and Lang, 2000; Schoar, 2002; Shahrur, 2005;
Kale and Shahrur, 2007), we derive our vertical integration measure from the Use tables of the benchmark
input-output accounts for the U.S. economy. The Bureau of Economic Analysis (BEA), a division of the
U.S. department of Commerce, publishes the benchmark input-output accounts including data on the
goods and services produced by each industry and the inputs required for production. The benchmark
input-output accounts are presented in five tables; the main one is the Use table. The BEA publishes the
summary benchmark accounts annually. We use data from the 1997 table because it is closest to our
sample period.
The Use tables report the dollar value of commodity flows between pairs of industries and the
total production of each industry. For each industry pair, i and j, the table reports the dollar value of i’s
output that was used to produce j’s output. This value is denoted aij. Fan and Lang (2000) divide aij by the
dollar value of industry j’s output to obtain a measure of the percent of industry j’s total output that is
supplied by industry i. Similarly, when aji is divided by the value of industry i’s output we get a measure
15
of the percent of industry i’s total output that is supplied by industry j. Conceptually, vertical integration
is more prevalent when a larger percentage of one industry’s output is supplied by another industry.
Following Fan and Lang, we use the average of these two ratios to measure the extent of vertical
integration for pairs of industries where the pairs are the primary and secondary segment SIC. We employ
this variable, which we denote Vrel, to proxy for vertical relatedness. Vrel is increasing in the degree of
vertical integration.
The issue of vertical integration is particularly relevant to our study because, among the many
changes the standard brought to bear, SFAS No. 131 imposed a change in the reporting requirements for
vertically integrated operations. SFAS No. 14 did not require the disaggregation of vertically integrated
operations because the standard defined industry segments in terms of products and services sold to
unaffiliated customers. SFAS No. 131 removed this exception stating, “the definition of an operating
segment should include components of an enterprise that sell primarily or exclusively to other operating
segments of the enterprise if the enterprise is managed that way”.9 Thus, to the extent that firms manage
vertically integrated operations within the same operating unit, this should give rise to an increase in
within-segment vertical relatedness.10 Since if we observe an effect it is likely attributable to a change in
reporting requirements, we do not draw any conclusions regarding our hypotheses from our analysis of
Vrel.
Across-Segment Tests (H4)
H4 involves examining whether there is greater cross-segment industry diversity following the
adoption of SFAS No. 131. To test this hypothesis, we employ the entropy measures introduced in
Jacquemin and Berry (1979), which include a measure of total firm diversification (TotDiv) that is
decomposed into related diversification (RelDiv) and unrelated diversification (UnrDiv). UnrDiv reflects
the extent to which segment sales are distributed across unrelated industries while, RelDiv captures the
9
SFAS No. 131, ¶79.
Alternatively, to the extent that firms manage vertically integrated operations across different operating units, we
expect to observe an increase in across-segment vertical relatedness. Currently, we have left this analysis to a future
draft of the paper.
10
16
extent to which segment sales are distributed within related industry groups. These measures incorporate
the number of segments a firm operates in, the relative importance of each segment’s sales to total firm
sales, and the degree of similarity in product line classification.11
These measures are particularly useful in the context of our study since they distinguish between
related and unrelated diversification. We test H4 by comparing measures of related and unrelated
diversification across segments over the two reporting regimes. We expect to observe an increase in
unrelated diversification (UnrDiv) and a decrease in related diversification (RelDiv), following the
adoption of SFAS No. 131, if the standard results in greater cross-segment industry diversity.
The diversification measures use 2-digit SIC codes to define industry groups, with operations in
the same 2-digit industry deemed related, and those in different 2-digit industry groups deemed unrelated.
We estimate total diversification, TotDiv, using Jacquemin and Berry’s formula, which we present below.
𝑇𝑜𝑡𝐷𝑖𝑣 = ∑𝑁
𝑖=1
Where: Si
TtlS
N
𝑆𝑖
1
𝑙𝑛 (𝑆𝑖
)
𝑇𝑡𝑙𝑆
⁄𝑇𝑡𝑙𝑆
(1)
= segment i’s sales.
= total firm sales.
= total number of segments.
TotDiv captures diversification to the extent the firm reports different segments. On the one hand,
some of these segments may belong to the same industry group (i.e. 2-digit SIC code), in which case the
“diversification” is into a related operation. On the other hand, some of the segments belong to different
industry groups, in which case the “diversification” is into unrelated operations. To partition TotDiv
between related, RelDiv and unrelated UnrDiv diversification, we group segments into S industry groups
based the segments’ primary 2-digit SIC codes. This yields S industry groups comprised of Z segments.
These industry groups are distinct from one another in terms of 2-digit SIC code. We estimate RelDiv and
UnrDiv using Jacquemin and Berry’s formulas, which we present below.
11
Due to data limitations we cannot use the entropy measures for our within-segment analysis. To do so, we would
require sales produced by the primary SIC code operations versus the sales produced by the secondary SIC code
operations. These data are not publicly available.
17
∑𝑍
𝑗=1 𝑆𝑗
𝑆𝑗
𝐸𝑤 = ∑𝑍𝑗=1 (∑𝑍
𝑗=1 𝑆𝑗
𝑅𝑒𝑙𝐷𝑖𝑣 = ∑𝑆 (𝐸𝑤
𝑈𝑛𝑟𝐷𝑖𝑣 = ∑𝑆 (
Where: Sj
Z
S
𝑙𝑛 (
∑𝑍
𝑗=1 𝑆𝑗
𝑇𝑡𝑙𝑆
∑𝑍
𝑗=1 𝑆𝑗
𝑇𝑡𝑙𝑆
𝑆𝑗
))
(2)
)
(3)
𝑇𝑡𝑙𝑆
𝑙𝑛 (∑𝑍
𝑗=1 𝑆𝑗
))
(4)
= sales of segment j, member of industry group S.
= total number of segments in industry group S.
= total number of industry groups.
Ew is the weighted average of the percent of firm sales accounted for by the Z segments in an
industry group, the weight for each segment is the natural log of the inverse of its share of firm sales.
Thus, both the number of segments in the group and the importance of each segment, based on sales, is
reflected in this measure. RelDiv is the weighted average of the related diversity of each group, Ew, across
all S industry groups in the firm with each group weighted by its share of firm sales. UnrDiv is the
weighted of each industry group’s share of firm sales across all groups. TotDiv is the sum of RelDiv and
UnrDiv. Appendix B provides a numerical example of these diversification measures taken from Palepu
(1985), p. 253.
Proxy for Competitive Environment
Within-Segment Tests (H2)
We use the Harris (1998) speed of positive abnormal profit adjustment as a measure of industry
competitive advantage. The speed of abnormal profit adjustment measures the persistence of return on
assets (ROA)12 above the industry norm. Firms maintain a competitive advantage longer when other
firms, (i.e. potential entrants), cannot precisely identify the source of excess profits. Under SFAS No. 14,
it was argued that multi-segment firms operating in industries with differing competitive environments
aggregated results of their diverse operations in such a way as to delay new entry and sustain abnormal
profitability, thereby minimizing the proprietary costs of disclosure.
12
ROA is calculated as the ratio of net income (Compustat data # 18) to total assets (data #6).
18
We estimate Harris’ measure of the rate of positive abnormal profit adjustment with the following
regression:
Xijt =  0j + 1j(DnXijt-1) + 2j(DpXijt-1) + eijt
where:
Xijt = the difference between firm i’s return on assets and the mean return on assets for
industry j in year t;
Dn = 1 if Xijt-1 is less than or equal to 0, and 0 otherwise; and
Dp = 1 if Xijt-1 is greater than 0, and 0 otherwise.
We estimate this equation for each industry using pooled cross-sectional time-series data from the
Compustat population of firms with sales from 1996 – 1998. We define industries based on 3-digit SIC
codes. The coefficient estimate  2j measures the persistence of positive abnormal ROA within industry j.
A significant positive 2j implies some firms can consistently earn above average rates of return
suggesting that other firms or new entrants cannot mimic this performance. A larger coefficient therefore
implies less competition.
We calculate a segment level variable (Compdiff) that is the absolute value of difference between
the abnormal profitability measures for the segment’s primary and secondary SIC codes. Prior research
suggests that under SFAS No. 14 firms tended to “hide” operations with a slower rate of positive
abnormal profit adjustment (Harris 1998) by grouping such operations with unrelated activities. H2
predicts that under SFAS No. 131 firms are more likely to group these operations with operations that
face a similar competitive environment. If the business units comprising a segment become more alike in
competitive environment following SFAS No. 131, Compdiff will decrease.
Across-Segment Tests (H5)
At the same time that within-segment similarity in competitive environment increases, H5
predicts a decrease in across-segment similarity in competitive environment. To capture this, we compute
Comp_range as the absolute value of the maximum less the minimum speed of positive abnormal profit
adjustment assigned to the firm’s segments. For purposes of this analysis we assign speed of profit
19
adjustment estimates to the segments based on primary SIC code. An increase in this metric is consistent
with an increase in across-segment diversity in competitive environment, which would support H5.
Proxy for Growth Opportunities
Within-segment Tests (H3)
Growth opportunities, or options to generate returns above the required rate of return, represent
an important facet of the business environment and therefore affect firm strategy and organizational
design. Chen and Zhang (2003) find that the usefulness of segment data is associated with differences in
investment opportunities caused by differences in growth potential across segments. We estimate two
measures of investment opportunities commonly seen in related literature: market-to-book equity ratio
and market-to-book assets ratio (e.g. Smith and Watts 1992, Gaver and Gaver 1993).
We define the market-to-book equity ratio as the market value of equity (closing price x shares
outstanding) divided by common equity. The numerator represents the present value of all future cash
flows while the denominator measures the accumulated value of existing assets only. Growth
opportunities are captured to the extent the market believes the firm’s prospects surpass its return on
assets in place.
The amount of leverage in a firm’s capital structure can complicate the interpretation of this
measure. Accordingly, we also calculate the market-to-book assets ratio. Research evaluating various
proxies for investment opportunities (including the market-to-book equity ratio) finds that, regarding
growth opportunities, the market-to-book assets measure offers the greatest information content among
four commonly used proxies (Adam and Goyal 2006).
The market-to-book assets ratio is closely related to Tobin’s q. Both illustrate a relation between
assets in place and assets in place plus the market’s beliefs regarding the potential return on future assets.
We follow Gaver and Gaver’s (2003) formula and compute the measure as the sum of total assets less
common equity plus market value of equity, (essentially the market value of common equity plus the
book value of debt and other preferred claims) all over total assets.
20
We estimate the mean and median of both ratios for each industry using annual cross-sectional
data for the Compustat population of firms over the years 1996-1998, with industries defined in terms of
3-digit SIC code. We then assign these industry measures of growth opportunities to each segment based
on the segment’s primary and secondary 3-digit SIC codes. We define the segment-level proxy, MBEdiff
as the absolute value of the difference between the industry market-to-book equity ratios assigned to a
segment’s primary and secondary 3-digit SIC codes. Likewise, we define MBAdiff as the absolute value
of the difference between the industry market-to-book assets ratios assigned to a segment’s primary and
secondary 3-digit SIC codes.
Our test of H3 examines whether the mean (median) measures of within segment growth
opportunities changed post SFAS No. 131. Observing a decrease in the difference of our growth
opportunity measures, MBEdiff and MBAdiff at the segment-level indicates support for H3.
Across-segment Tests (H6)
Our final hypothesis looks at the across segment diversity of growth opportunities following the
adoption of SFAS No. 131. Prior research suggests that abundant growth opportunities are a source of
persistent above average returns (Gaver and Gaver 1993). Based on this and the findings of Harris (1998),
we assume that under SFAS No. 14 firms tended to “hide” operations with greater growth opportunities.
If under SFAS No. 131, these operations are grouped with operations that face similar growth
opportunities, the firm level weighted-average value should increase. For purposes of this analysis, we
focus on each segment’s primary SIC code. We compute MBE_range and MBA_range as the absolute
value of the maximum less the minimum respective growth opportunity variables assigned to a firm’s
segments based on primary SIC codes. An increase in this metric supports H6. This would also be
consistent with improved segment reporting as Chen and Zhang (2003) find that the usefulness of
segment reporting increases as differences in growth rates across segments increases.
21
IV.
SAMPLE AND DESCRIPTIVE STATISTICS
Sample Construction
We collect data for all firms included on the Compustat segment file in the year prior to and the
year of SFAS No. 131 adoption. Due to the adoption provisions of SFAS No. 131, December year-end
firms adopted the standard in 1998. For these firms, 1997 is the final year they applied SFAS No. 14 (our
pre-SFAS No. 131 year), and 1998 is the year they first applied SFAS No. 131 (our post-SFAS No. 131
year). In contrast, non-December year-end firms adopted SFAS No. 131 in 1999. Accordingly, we draw
data for the pre- and post-SFAS No. 131 periods from 1998, and 1999, respectively.
We begin with 8,041 firms reporting 24,885 business segments in the year before and the year of
mandatory adoption of SFAS No. 131. To eliminate firms that altered their segment definitions for
reasons other than SFAS No. 131, we remove firms undergoing mergers, acquisitions, or divestitures in
the adoption year.13 This reduces our sample by 995 firms and 5,827 segments14.
We also remove 631 firms reporting 1,568 segments because the firm has negative equity or no
assets. We eliminate these firms from our analyses because measures of market-to-book equity and
market-to-book assets are ill defined for firms with negative equity and zero assets. We also eliminate
segments (though not necessarily firms) with negative sales or no primary SIC code since these
“segments” often represent corporate transfers or eliminations. This reduces our sample by 25 firms and
487 segments. Finally, for inclusion in our sample we require firms to have at least one segment
observation in the pre- and post-SFAS No. 131 periods, resulting in the elimination of 2,619 firms
reporting 5,578 segments. Our final sample consists of 3,771 firms and 11,425 segment observations.
Table 1 summarizes our sample selection process.
13
We use the Compustat footnote for Sales (data #12) to identify firms with mergers and acquisitions. If the
footnote indicates that the current year sales data includes a merger or acquisition, we remove the firm from the
sample. To identify divestitures, we review the notes of annual 10-k reports of firms reporting a decrease in the
number of segments in the period after the adoption of SFAS No. 131.
14
Although this follows convention in related literature (Street et al. 2000; Herrmann and Thomas, 2000; Berger and
Hann, 2003 and 2007), this screening mechanism is criticized for its potential elimination of firms whose actual
structure changed precisely because of the new regulation (Piotroski, 2003). For example, firms with poorly
performing business units might divest such units to avoid potential monitoring following the exposure of poor
performing operations by SFAS No. 131. In the context of our research question, eliminating these firms biases
against finding a significant change in segment definitions that is associated with organizational design.
22
Insert Table 1 here.
Of the 3,771 firms in our final sample, 1,081 firms changed their segment definitions in
conjunction with the adoption of SFAS No. 131, whereas 2,690 firms did not.15 We refer to the former
group as change firms and the latter group as no-change firms. The 1,081 (2,690) change firms (nochange firms) comprise 29% (71%) of the full sample, while their 4,939 (6,486) segments comprise 43%
(57%) of the total number of segments.
We further distinguish between the change and no-change firms based on whether they are singlesegment or multi-segment firms in the pre-SFAS No. 131 year. Of the 1,081 firms that changed their
segment definitions, 785 (73%) reported a single segment under SFAS No. 14, but multiple segments
under SFAS No. 131. We refer to this subset of firms as the change firm single_multi group. The
remaining 296 (27%) of the change firms are multi-segment firms in both years. We refer to the firms in
this group as change firm multi_multi. Similarly, 2,344 (87%) of the no-change subsample report as
single segment firms under both reporting regimes, whereas the remaining 346 (13%) report multiple
segments before and after the adoption of SFAS No. 131. We refer to these subsamples of no-change
firms as no-change single segment and no-change multi segment firms, respectively.
We conduct all of our analyses on these subsamples of firms since they might differ in terms of
their reporting incentives and other firm characteristics. For example, Botosan and Stanford (2005) focus
on the change firm single_multi subset because “by virtue of their decision to withhold industry segment
information pre-SFAS No. 131, they utilized the flexibility in SFAS No. 14 to the greatest extent
possible, presumably because of perceived high disclosure costs.”16 By separately examining the segment
definitions of each subsample before and after SFAS No. 131, we gain insight into differences in the
We identify change firms using Compustat’s segment identifier (SID). When, for any reason, a listed segment is
not comparable to segment listed in the prior year, Compustat assigns a new SID to the segment. We initially
classify as a change firm each firm with a new SID, but as noted in our sample selection procedures, we remove
firms with mergers, acquisitions, or divestitures to isolate firms for whom SFAS No. 131 unequivocally motivated
the change in segment disclosure.
16
Botosan and Stanford, 2005, pg. 752.
15
23
firms’ reporting behavior under SFAS No. 14, and upon the adoption of SFAS No. 131, which could help
us better understand reporting incentives.
Descriptive Statistics
Table 2 presents the distribution of our sample firms across SIC “divisions”. These high-level
categories, to an extent, are analogous to economic sectors. We assign the sample firms to divisions based
on their primary SIC code in 1998. Relative to the Compustat population of firms (also reported in the
table), our sample contains a slightly lower proportion of firms in the finance, insurance and real estate
industries, and a slightly greater proportion of firms in the manufacturing industry. This effect is driven
by our subsample of no-change single segment firms. This group is significantly less (more) populated by
firms in the finance, insurance, and real estate (manufacturing) industries that the other subsamples or the
Compustat population of firms are. Otherwise, the industry breakdown of our total sample and the
subsamples roughly resembles the Compustat population.
Insert Table 2 here.
Table 3 shows the distribution of the firms based on the number of segments they report pre- and
post-SFAS No. 131. Consistent with prior research, we document an increase in the number of segments
reported by firms altering their disclosures in response to SFAS No. 131. For example, 48.7% (33.6%) of
the change firms reporting a single segment under SFAS No. 14, report two (three) segments under SFAS
No. 131. Similarly, of the change firms reporting multiple segments before and after the adoption of
SFAS No. 131, only 15.9% reported two segments under SFAS No. 131, as compared to 48% under
SFAS No. 14. Interestingly, the no-change multi-segment sample tend to report fewer segments than the
change firm multi_multi sample both before and after the adoption of SFAS No. 131. Just over 60% of
the no-change multi segment firms report two segments as compared to 48% (15.9%) of the change firm
multi_multi sample before (after) the adoption of SFAS No. 131. This suggests that the multi segment
firms that changed their disclosure in response to SFAS No. 131 are systematically different from the
multi segment firms that did not.
Insert Table 3 here.
24
Table 4 provides firm-level descriptive statistics for the four subsamples of firms, before and after
the adoption of SFAS No. 131. Panels A and B provide data for the change firm single_multi, and
multi_multi subsamples, respectively, while panels C and D provide data for the no-change firm single,
and multi samples. Consistent with findings in table 3, the median number of segments increased from
one to three for the change firms that reported a single segment under SFAS No. 14, and three to 3.5 for
the change firms that reported multiple segments under both reporting regimes.
Insert Table 4 here.
Not unexpectedly, all of our firm size variables (total assets, total sales and MVE) indicate that
the subsamples are skewed toward larger firms. Additionally, our sample firms tend to be larger than the
population of firms on Compustat. In 1998 median total sales for the Compustat population is $79
million. Median sales in the SFAS No. 131 adoption year (1998 for the majority of our firms) exceeds
this figure for all subsamples except the no-change single segment firms where median total sales is
$42.32Million. Using any of our three firm size measures we find that firms in the change firm
multi_multi segment sample are larger than the firms in any other subsample. Moreover, change firms
reporting a single segment under SFAS No. 14, or no-change firms reporting a single segment under both
reporting regimes, tend to be smaller than firms reporting multiple segments pre- and post-SFAS No. 131,
regardless of whether they changed their segment definitions.
Profitability, as measured by ROA, is stable over the sample period. Median ROA for both
subsamples of change firms is constant at 3%. The no-change single segment firms tend to be slightly less
profitable with an ROA of 2%, whereas the no-change multi-segment firms tend to be slightly more
profitable at 4%.
To summarize the main differences between the subsamples of firms, we find that the no-change
multi-segment firms disclose fewer segments and are more profitable than the firms in the other
subsamples. We also find that firms in the change firm muli_multi sample are much larger than the firms
in the other subsamples. Finally, the firms in the no-change single segment subsample are smaller, with
25
proportionately fewer firms in financial industries and proportionately more firms in the manufacturing
sector.
V.
EMPIRICAL RESULTS
Correlation Analysis
Table 5 provides Pearson correlation coefficients and p-values for our explanatory variables
measured at the segment-level. The statistics we present are based on all observations without regard for
subsample membership. We limit our presentation to the pooled results because the subsample results are
not significantly different. We present the correlation among the variables measured in the final SFAS
No. 14 year below the diagonal, while the statistics above the diagonal are based on data measured in the
year SFAS No. 131 is adopted.
Insert Table 5 here.
Table 5 shows a high correlation between our industry relatedness variables that focus on
horizontal integration (Threediff, SameFF, and Dissim). The correlation among the three measures
exceeds 60%.17 This suggests that these measures capture the same underlying construct. Moreover, it
suggests that the cardinal interpretation problem with Threediff, which is not shared by SameFF and
Dissim, is not a major concern. In contrast, the correlation between these variables and our measure of
industry relatedness arising from vertical integration (Vrel) is less than 30% across the board. This
suggests that Vrel captures an aspect of industry relatedness not captured by the other three measures.
Our proxy for within segment similarity in competitive environment (Compdiff) is not highly
correlated with any of our other proxies, indicating that this variable captures a distinct aspect of firms’
operating environment. Our two measures of similarity in growth opportunities (MBEdiff and MBAdiff)
are not highly correlated with one another. Both before and after the adoption of SFAS No. 131 the
correlation is 4%. This suggests that these measures capture different aspects of growth. Moreover, the
correlation between MBAdiff and the industry relatedness measures based on horizontal integration
17
Since Threediff, Dissim move in the opposite direction than SameFF, a negative correlation is expected.
26
(Threediff, SameFF, and Dissim) exceed 25%, while the correlation between MBEdiff and these measures
ranges from 10% to 20%. This suggests that the aspect of growth captured by MBAdiff tends to be more
related to industry than that of MBEdiff.
Univariate Results
Within-segment Results (H1 through H3)
Table 6 presents the results of our test of H1 through H3 based on data for firms that changed their
segment definitions in response to SFAS No. 131. H1 predicts that following the adoption of SFAS No.
131 there is a significant increase in industry relatedness among the operations grouped as segments. We
find results consistent with this prediction both for firms that reported a single segment under SFAS No.
14, and for those that reported multiple segments under both regimes. All of our industry relatedness
measures (Threediff, SameFF, Dissim, and Vrel) move in the expected direction, although the Vrel result
is weak for the sample of multi_multi change firms. These results suggest that when firms changed their
segment definitions following the adoption of SFAS No. 131 they defined segments more in line with
their internal organizational structures than they did under SFAS No. 14.
Insert Table 6 here.
Another interesting finding is the difference between the two subsamples of change firms before
the adoption of SFAS No. 131. Under SFAS No. 14, the single_multi change firms tended to group much
more diverse operations into segments than the multi_multi change firms. For example, under SFAS No.
14, 54% of the operations combined into the same segment by the single_multi change firms belonged to
the same Fama-French industry group (mean SameFF = 0.54). This compares to 64% for the multi_multi
change firms. Panel C indicates that these differences are significant with t-statistic of 4.15, indicating
that single_multi firms combine more dissimilar operations in the SAFS 14 era. The comparable figures
for the post-SFAS No. 131 year are 76% and 72% for the single_multi and multi_multi change firms,
respectively; panel C indicates that these percentages are significantly different. These results are
consistent with the single_muti change firms taking greater advantage of the flexibility in SFAS No. 14 to
27
“hide” certain operations, and SFAS No. 131 successfully forcing more comparable reporting across
firms.
H2 predicts that following the adoption of SFAS No. 131, there is a significant increase in the
similarity of competitive environment among the operations grouped as segments. The results presented
in Table 6 are generally consistent with this hypothesis. For example, the mean (median) difference in the
speed of profit adjustment between segments’ primary and secondary industries (Compdiff) declined from
1.11 (0.05) to 0.44 (0.00) among change firms that began reporting multiple segments following the
adoption of SFAS No. 131. We document a similar, albeit weaker result, for the change firms that
reported multiple segments under both regimes (0.89 versus 0.73 at the mean). In contrast to our result
pertaining to industry relatedness, our Compdiff results suggest that the single_multi and multi_multi
change firms both took advantage of the flexibility in SFAS No. 14 to “hide” certain operations by
combining opterations facing different competitive environments in a single segment. After adoption of
SFAS No. 131, Compdiff is smaller for the single_multi firms than the multi_multi firms suggesting
multi_multi firms continue to combine operations facing different levels of competition.
H3 predicts that following the adoption of SFAS No. 131, there is a significant increase in the
similarity of growth opportunities among the operations grouped as segments. Our results with respect to
MBEdiff and MBAdiff are generally consistent with this prediction; similarity in the growth opportunities
of operations grouped as segments increased (at least for the median firm) for change firms in both
subsamples. Differences in the industry mean market to book equity and assets between the primary and
secondary segment industries declined in the post SFAS No. 131 period. Panel C indicates that singlemulti change firms combined operations with greater differences in growth opportunities in the SFAS 14
period (MBAdiff 1.76 versus 1.25) but changed their segment reporting choices in response to SFAS 131
in a manner consistent with that of the multi_multi change firms.
Summary of Results
In summary, with few exceptions, our univariate findings based on data measured at the segmentlevel suggest that post-SFAS No. 131, business activities grouped into the same segment are significantly
28
more related, have more similar competitive environments and more similar growth opportunities than the
business activities grouped together under SFAS No. 14. These results support the conclusion that SFAS
No. 131 is successful in persuading firms to disclose segment information that corresponds more with
their internal organizational structure than disclosure under SFAS No. 14. Based on these results, we draw
the following conclusions. First, SFAS No. 131 is successful in forcing firms to define segments more in
line with their internal organizational structures than SFAS No. 14 was. Second, change firms that
reported a single segment under SFAS No. 14 took greater advantage of the flexibility inherent in that
standard to define segments that deviated from their internal organizational structure than firms that
reported multiple segments under SFAS No. 14. Thus, the single_multi change firms started out with
what appears to be a much different grouping strategy as compared to the multi_multi firms, but upon
adoption moved closer to the strategy employed by the multi_multi firms. This finding is consistent with
the former group using the flexibility inherent in SFAS No. 14 to “hide” certain operations to mitigate
proprietary, agency or other costs of disclosure.
These conclusions are based on an analysis of operations grouped within segments. Next we
examine the extent to which across segment diversity in operations increased following the adoption of
SFAS No. 131. Our across segment analysis is limited to the change firm multi_multi sample, since by
virtue of their reporting of a single segment under SFAS No. 14 and multiple segments under SFAS No.
131, across segment diversity must increase for all firms in the change firm single_multi sample.
Across-segment Results (H4 through H6)
We present the results of our across-segment tests in Table 7. H4 predicts that following the
adoption of SFAS No. 131, there is a significant increase in the diversity of industry relatedness across
segments. Consequently, we expect to observe an increase in unrelated diversification across segments
and a decrease in related diversification across segments. Our results are only partially consistent with our
expectations. As predicted we observe a significant increase in unrelated diversification across segments,
0.41 versus 0.48 (t-statistic 2.16). In contrast to our expectations we also document a significant increase
29
in related diversification, 0.27 versus 0.34 (t-statistic 4.66). We are currently undertaking additional
analysis to further our understanding of this result.
Insert Table 7 here.
H5 predicts that following the adoptions of SFAS No. 131, there is a significant increase in the
diversity of competitive environments across segments. We find that the range in speed of profit
adjustment across segments (Comp_range) increases for the median firm following the adoption of SFAS
No. 131. This result is weak, however, with a p-value of 10% and insignificant at the mean.
Finally, H6 predicts that following the adoption of SFAS No. 131, there is a significant increase in
the diversity of growth opportunities across segments. Our results are consistent with this prediction.
Specifically, we document an increase in the range of MBE and MBA (i.e. MBE_range and MBA_range)
following the adoption of SFAS No. 131 for the median firm. The median value of MBE_range increases
from 2.3 to 3.2, while the median value of MBA_range increases from 1.0 to 1.37. However, results for
the mean are insignificant.
In untabulated analyses, we also observe statistically significant changes, in the predicted
direction, for both the mean (median) values of our competition and growth opportunity variables at a
weighted average firm-level where the weightings are proportionate to the contribution of a given
segment’s sales to total segment sales. These findings support the conclusions drawn above based on
ranges versus weighted average values.
Summary of Results
Our analysis of the change in the across-segment diversity of industry relatedness, competitive
environment, and growth opportunities produces results that are consistent with our hypotheses for the
median firm. Thus, our across segment analyses provide moderate support for the conclusion that SFAS
No. 131 is successful in forcing firms to define segments more in line with their internal organizational
structures than SFAS No. 14 was. Overall, we conclude that there is strong evidence to support the
conclusion that as the operations grouped within segments became more similar following the adoption of
SFAS No. 131 and weak evidence that the segments themselves became more diverse.
30
Comparison of Change to No-change Firms
In an attempt to further understand firms’ reporting strategies with respect to their segment
definitions, we compare univariate statistics for our within-segment variables across the change and nochange subsamples of firms. We present these data in Table 8.
Insert Table 8 here.
Panel A presents statistics comparing the change firm single_multi subsample to the no-change
multi-segment, and no-change single segment subsamples, in the SFAS No. 14 period. The results
indicate that the firms reporting a single-segment under SFAS No. 14, but multiple segments under SFAS
No. 131, made significantly different decisions when grouping operations into segments in the SFAS No.
14 reporting period, than the firms that did not alter their segment definitions in response to SFAS No.
131. The single_multi change firms tended to combine more diverse operations into segments in the
SFAS No. 14 period than either the single segment or multiple-segment no-change firms. This result
holds across all of our industry relatedness measures, as well as our measures of similarity in competitive
environment and growth opportunities.
For example, under SFAS No. 14, 54% of the operations combined into the same segment by the
single_multi change firms belonged to the same Fama-French industry group (mean SameFF = 0.54).
This compares to 66% for the no-change multi-segment firms and 59% for the no-change single segment
firms. These results are consistent with the single_muti change firms taking greater advantage of the
flexibility in SFAS No. 14 to “hide” certain operations perhaps because these firms face greater
proprietary, agency or other costs of disclosure.
Panel B presents statistics comparing the change firm single_multi subsample to the no-change
multi-segment, and no-change single-segment subsamples, in the SFAS No. 131 period. The results
indicate that the firms reporting a single segment under SFAS No. 14, but multiple segments under SFAS
No. 131, continue to make significantly different segment definition decisions following the adoption of
SFAS No. 131 than the firms not altering their segment definitions in response to SFAS No. 131.
Notably, however, the direction of the effect is opposite across all variables. That is, the results suggest
31
that after the adoption of SFAS No. 131, the single_multi firms that changed their segment definitions
began grouping more similar operations into segments than the firms that did not change their segment
definitions.
For example, following the adoption of SFAS No. 131, 76% of the operations combined into the
same segment by the single_multi change firms belonged to the same Fama-French industry group (mean
SameFF = 0.76), while these statistics remain at 66% and 59% for the no-change multi-segment and
single segment firms respectively. These results might suggest that there is continued reticence on the part
of some firms (those included in the no-change subsamples) to define segments consistent with the firm’s
presumed internal organizational structure.
This latter conclusion is further bolstered by the results presented in Table 8, Panel C. Panel C
presents statistics comparing the change firm multi_multi subsample to the no-change multi-segment, and
no-change single segment subsamples, in the SFAS No. 14 period. The evidence indicates that the
multi_multi change firms made segment definition decisions that were very similar to the no-change
multi-segment firms in the SFAS No. 14 period. Except for Vrel, none of measures are significantly
different across the two subsamples. Yet, the results in Panel D, which present the same comparison for
the SFAS No. 131 year, indicate that after adopting SFAS No. 131, the multi_multi change firms appear
significantly different from the no-change multi-segment firms. Moreover, the direction of the difference
is consistent with the multi_multi change (no-change) firms choosing (not choosing) segment definitions
that reflect their internal organization structure.
VI.
SUMMARY AND CONCLUSION
SFAS No. 131 governs disclosure about segments of an organization. Critics charged that the
latitude extended to managers under SFAS No. 14 led to segment disaggregation in which unrelated or
vaguely related business activities were combined under broad industry headings. Consistent with this,
prior research finds that the economic consequences of disclosure, such as proprietary and/or agency
costs, motivated firms’ segment reporting decisions under SFAS No. 14. By prescribing a management
32
approach to segment disaggregation, SFAS No. 131 calls for segmentation that more closely reflects the
firms’ organizational design. The purpose of our study is to evaluate the success of SFAS No. 131 in
inducing firms to disclose more about their organizational design in their segment disclosures.
Our measures include relatedness in industry, competitive environment and growth opportunities.
We find an increase in within segment relatedness of operations and some evidence of an increase across
segment diversity of operations with the implementation of SFAS No. 131. Moreover, we find that firms’
reportable segments include business divisions that share more similar competitive environments and
growth opportunities after SFAS No. 131 than before. Overall, our results suggest that firms that changed
their segment definitions following the adoption of SFAS No. 131 did so to increase the alignment of
segment disclosure with their presumed internal organizational structures.
Our results also indicated that firms reporting a single segment under SFAS No. 14, but multiple
segments under SFAS No. 131 tended to take greater advantage of the flexibility in SFAS No. 14 to
“hide” certain operations. This is consistent with such firms facing greater proprietary, agency or other
costs of disclosure. Finally, we provide evidence that firms that did not change their segment definitions
in response to SFAS No. 131, use post-SFAS No. 131 segment definitions that deviate from their
presumed internal organizational structures to a greater extent than firms that changed their segment
definitions in response to SFAS No. 131.
33
TABLE 1
Sample Selection
Change Firms
Single_Multi
Firms
Seg
Population of observations in
the Compustat segment tape
with data before and after
SFAS No. 131
Change Firms
Multi_Multi
Firms
Seg
No-change
Single
Segment
Firms
Seg
No-change
Multi Segment
Firms
Seg
1,745
6867
646
4,505
4,987
9,974
663
3,539
Less:
Observations with a merger,
acquisition, or divestiture in
the adoption year
301
2.334
112
1,565
490
980
92
948
Observations with negative
equity or no assets
84
306
16
86
492
984
39
Observations with negative
segment sales, no primary
SIC or classified as ‘no
operations’
1
179
0
174
21
42
Observations lacking
sufficient data in both the pre
and post SFAS No. 131
periods
574
1,074
222
715
1,640
Total
Percent of Sample
785
21%
2974
26%
296
8%
1965
17%
Pre SFAS No. 131
Post SFAS No.131
785
785
785
2,189
2,974
296
296
857
1,108
1,965
34
Total
Firms
Seg
8,041
24,885
995
5,827
192
631
1,568
3
92
25
487
3,280
183
509
2,619
5,578
2,344
62%
4,688
41%
346
9%
1,798
16%
3,771
100%
11,425
100%
2,344
2,344
2,344
2,344
4,688
346
346
899
899
1,798
3,771
3,771
7,542
4,885
6,540
11,425
TABLE 2
Industry Distribution of Sample Firms
SIC Division
Agric, forestry, fishing
Construction
Fin, Insur, Real Estate
Manufacturing
Mining
Retail Trade
Services
Trans, Comm, Utilities
Wholesale Trade
Total
Change Firms
Single_Multi
Freq
%
3
0.4
11
1.4
161
20.5
295
37.6
11
1.4
50
6.4
147
18.7
55
7.0
52
6.6
785
100.0
Change
Firms
Multi_Multi
Freq
%
3
1.0
5
1.7
65
22.0
119
40.2
15
5.1
13
4.4
35
11.8
17
5.7
24
8.1
296 100.0
No-change
Single Segment
Freq
%
6
0.3
19
0.8
293
12.5
1048
44.7
97
4.1
232
9.9
389
16.6
181
7.7
79
3.4
2,344
100.0
No-change
Multi
Segment
Freq
%
5
1.5
7
2.0
74
21.4
135
39.0
20
5.8
13
3.8
40
11.5
31
8.9
21
6.1
346 100.0
Total
Sample
Freq
%
17
0.5
42
1.1
593
15.7
1597
42.3
143
3.8
308
8.2
611
16.2
284
7.5
176
4.7
3771 100.0
1998
Compustat
Population
Freq
%
144
0.3
476
1.1
10,712
24.0
14,846
33.3
2,360
5.3
2,438
5.5
7,960
17.8
4,092
9.2
1,562
3.5
44,608 100.0
SIC divisions are based on the primary SIC coded at the 2 digit level as reported under SFAS No. 131.Change firms are those that changed
the number or composition of segments based on SID (segment identification) in the Compustat business segment file during the year of the
adoption of the new standard. Single_Multi indicates change firms that reported a single segment under SFAS No.14 and multiple segments
under SFAS No.131. Multi_Multi firms reported multiple segments under SFAS No. 14 and increased the number of segments reported
under SFAS No. 131. No-change firms report the same grouping of business activities, either a single segment or multiple segments, under
both SFAS No. 14 and SFAS No. 131.
35
TABLE 3
Number of Segments Reported
Pre and Post adoption of SFAS No. 131
Panel A: Change Firms Single_Multi
Under SFAS No. 14
# of
Cumulative Cumulative
Seg Firms
Freq
%
%
1
785
100.0
785
100.0
2
0
0.0
785
100.0
3
0
0.0
785
100.0
>4
0
0.0
785
100.0
Total Firms
785
Total Segments
785
# of
Seg Firms
1
0
2
382
3
264
>4
139
Total Firms
Total Segments
Under SFAS No. 131
Cumulative
Freq
%
0
0
48.7
382
33.6
646
17.7
785
785
2,189
Panel B: Change Firms Multi_Multi
Under SFAS No. 14
# of
Cumulative
Seg Firms
Freq
%
1
0
0.0
0
2
142
48.0
142
3
89
30.1
231
>4
65
22.0
296
Total Firms
296
Total Segments
857
# of
Seg Firms
1
0
2
47
3
98
>4
151
Total Firms
Total Segments
Under SFAS No. 131
Cumulative
Freq
%
0.0
0
15.9
47
33.1
145
51.0
296
296
1,108
Cumulative
%
0.00
48.0
78.0
100.0
Panel C: No_Change Firms: Single Segment
Under SFAS No. 14
# of
Cumulative Cumulative
Seg Firms
Freq
%
%
1
2,344
100
2,344
100
Total Firms
2,344
100
Total Segments
2,344
Under SFAS No. 131
# of
Cumulative
Seg Firms
Freq
%
1
2,344
100
2,344
Total Firms
2,344
Total Segments
2,344
Panel D: No_Change Firms: Multi Segment
Under SFAS No. 14
# of
Cumulative Cumulative
Seg Firms
Freq
%
%
1
0
0.0
0
0.0
2
209
60.4
209
60.4
3
91
26.3
300
86.7
>4
46
13.3
346
100.0
Total Firms
346
Total Segments
899
# of
Seg Firms
1
0
2
209
3
91
>4
46
Total Firms
Total Segments
36
Under SFAS No. 131
Cumulative
Freq
%
0.0
0
60.4
209
26.3
300
13.3
346
346
899
Cumulative
%
0.0
48.7
82.3
100.0
Cumulative
%
0.0
15.9
49.0
100.0
Cumulative
%
100
100
Cumulative
%
0.0
60.4
86.7
100.0
Grand Total Firms
Grand Total
Segments
3,771
4,885
Grand Total Firms
Grand Total
Segments
3,771
6,540
The number of segments is as reported in the Compustat Business Segment file for the fiscal year-end prior to and during the year of the
mandatory adoption of SFAS No. 131. Change firms are those that changed the number or composition of segments based on SID
(segment identification) in the Compustat business segment file during the year of the adoption of the new standard. Single_Multi
indicates change firms that reported a single segment under SFAS No.14 and multiple segments under SFAS No.131. Multi_Multi firms
reported multiple segments under SFAS No. 14 and increased the number of segments reported under SFAS No. 131. No-change firms
report the same grouping of business activities, either a single segment or multiple segments, under both SFAS No. 14 and SFAS No.
131.
37
TABLE 4
Firm Level Descriptive Statistics
Panel A: Change Firms Single_Multi
N = 785
Under SFAS No. 14
Variable
Mean Median
Min
Max
Std Dev
Segments
1.00
1.00
1.00
1.00
0.00
Total Assets 2,332
143
2.01
217,318 14,265
Total Sales
1,008
121
0.02
47,061
3,821
MVE
134
112
0.86
126,526
6,615
ROA
(0.01)
0.03
(2.39)
0.44
0.21
N = 785
Panel B: Change Firms Multi_Multi
N = 296
Under SFAS No. 14
Variable
Mean Median
Min
Max
Std Dev
Segments
3.00
2.00
10.00
1.18
2.90
Total Assets
735
1.59
304,012 30.412
8,448
Total Sales
617
0.86
153,627 13,685
4,089
MVE
437
0.82
239,536 20,688
5,628
ROA
0.03
(2.39)
0.20
0.20
(0.01)
N = 296
Panel C: No-change Firms Single Segment
N = 2,344
Under SFAS No. 14
Variable
Mean Median
Min
Max
Std Dev
Segments
1.00
1.00
1.00
1.00
0.00
Total Assets
700
47.89
0.12
391,673
8,349
Total Sales
449
38.20
0.00
32,183
1,892
MVE
612
63.30
0.00
164,758
4,605
ROA
(0.07)
0.02
(5.82)
4.84
0.41
N = 2,344
Panel D: No-change Firms Multi Segment
N = 346
Under SFAS No. 14
Variable
Mean Median
Min
Max
Std Dev
Segments
2.60
2.00
2.00
6.00
0.91
Total Assets 2,401
212
0.14
80,918
7,489
Total Sales
2,080
208
0.03
117,958
8,116
MVE
2,545
146
0.62
89,219
9,634
ROA
0.02
0.04
(1.62)
0.53
0.17
N = 346
Mean
2.78
2,712
1,136
1,755
(0.02)
Mean
3.72
8,887
3,939
6,754
(0.01)
Mean
1.00
788.75
481.74
780.06
(0.11)
Mean
2.60
2,553
2,133
3,264
0.01
Under SFAS No. 131
Median
Min
Max
3.00
2.00
9.00
149
1.61
321,421
134
0.11
42,370
101
0.25
174,083
0.03
(2.91)
0.48
Std Dev
1.01
18,130
3,774
9,451
0.23
Under SFAS No. 131
Median
Min
Max
3.50
2.00
9.00
708
0.59
355,935
621
0.47
144,416
412
0.75
333,672
0.03
(4.68)
0.21
Std Dev
1.39
32,323
12,920
27,267
0.32
Under SFAS No. 131
Median
Min
Max
1.00
1.00
1.00
51.66
0.08
485,014
42.32
0.00
33,674
50.30
0.18
165,189
0.02
(7.48)
4.81
Std Dev
0.00
10,251
2,058
5,955
0.48
Under SFAS No. 131
Median
Min
Max
2.00
2.00
6.00
242
0.14
87,691
198
0.02
137,634
145
0.27
191,264
0.04
(1.51)
0.50
Std Dev
0.91
8,000
8,753
15,117
0.16
This table presents firm level descriptive statistics under both reporting regimes for the four subsamples in the study. Change firms
are those that changed the number or composition of segments based on SID (segment identification) in the Compustat business
segment file during the year of the adoption of the new standard. Single_Multi indicates change firms that reported a single segment
under SFAS No.14 and multiple segments under SFAS No.131. Multi_Multi firms reported multiple segments under SFAS No. 14
and increased the number of segments reported under SFAS No. 131. No_Change firms report the same grouping of business
activities, either a single segment or multiple segments, under both SFAS No. 14 and SFAS No. 131. Segments are the number of
segments reported in the Compustat business segment file. Total Assets are the total dollar amount in millions reported per
Compustat data item #6. Total Sales are the total dollar amount in millions reported per Compustat data item #12. Market Value of
38
Equity (MVE) is calculated as closing price (data#199) x common shares out (data#25). Return on Assets (ROA) is income before
extraordinary items (data#18) divided by total assets (data #6).
39
TABLE 5
Pearson correlations (p-values) Pooled Segment Level Variables
Under SFAS No. 14 (below the diagonal) and Under SFAS No. 131 (above the diagonal)
Threediff
Threediff
Dissim
SameFF
PMATpf
Vrel
Compdiff
MBEdiff
MBAdiff
0.68
(.0001)
-0.64
(.0001)
-0.65
(.0001)
-0.27
(.0001)
0.09
(.0001)
0.10
(.0001)
0.35
(.0001)
Dissim
SameFF
PMATpf
Vrel
0.69
(.0001)
-0.66
(.0001)
-0.83
(.0001)
-0.67
(.0001)
-0.71
(.0001)
0.68
(.0001)
-0.25
(.0001)
-0.22
(.0001)
0.33
(.0001)
0.31
(.0001)
-0.82
(.0001)
-0.70
(.0001)
-0.25
(.0001)
0.12
(.0001)
0.19
(.0001)
0.32
(.0001)
0.66
(.0001)
0.37
(.0001)
-0.10
(.0001)
-0.17
(.0001)
-0.26
(.0001)
0.35
(.0001)
-0.10
(.0001)
-0.10
(.0006)
-0.29
(.0006)
-0.06
(.0001)
-0.09
(.0001)
-0.14
(.0001)
Compdiff
MBEdiff
MBAdiff
0.11
(.0001)
0.14
(.0001)
-0.10
(.0001)
-0.11
(.0001)
-0.06
(.0001)
0.10
(.0001)
0.20
(.0001)
-0.18
(.0001)
-0.12
(.0001)
-0.09
(.0001)
0.07
(.0001)
0.35
(.0001)
0.35
(.0001)
-0.27
(.0001)
-0.30
(.0001)
-0.13
(.0001)
0.24
(.0001)
0.04
(.0006)
0.06
(.0001)
0.28
(.0001)
0.04
(.0080)
Threediff is the absolute value of the segments primary 3- digit SIC code less the segment’s secondary 3-digit SIC code. Dissim
is discrete random variable assigned a value of 0 when the segments primary and secondary SIC codes are identical, a value 1
when the primary and secondary SIC codes have the same first 3 digits, a value of 2 when the primary and secondary SIC codes
have the same first 2 digits, a value of 3 is assigned when the primary and secondary SIC codes have the same first digit and
finally, a value of 4 when the segment’s primary and secondary SIC codes have no common digits. SameFF is an indicator
variable assigned a 1 if the segment’s primary and secondary SIC codes are in the same Fama-French industry and zero
otherwise. PMATpf is an indicator variable assigned a 1 if the segment’s primary and secondary 2 digit NAICS codes are in the
same strategy portfolio as determined by profit margin and asset turnover quintiles and zero otherwise. Vertical relatedness (Vrel)
represents the average input transfers between the segment’s primary and secondary industries based on the Input-Output
accounts for the US economy. Compdiff is the absolute value of the estimate of the speed of abnormal profit adjustment (Harris
1998) for the segment’s primary SIC industry less the speed of abnormal profit adjustment for the segment’s secondary SIC
industry. MBEdiff is absolute value of the difference between growth opportunities measured as market-to-book equity of the
segment’s primary SIC and segment’s secondary SIC. MBE is measured as MVE (closing price (data#19) x shares out (data
#25))/common equity (data #60). MBAdiff is the absolute value of the difference between growth opportunities measured as
market-to-book assets of the segment’s primary SIC and segment’s secondary SIC. MBA is calculated as (total assets less
common equity plus MVE)/total assets.
40
TABLE 6
Univariate Tests of Relatedness of Operations
Within Segment Analysis
Panel A: Change Firms: Single_Multi
N = 785
Variable
Under SFAS No. 14
Three diff
96.58
7.00
Dissim
2.32
2.00
SameFF
0.54
1.00
PMATpf
0.61
1.00
Vrel
0.05
0.01
Compdiff
1.11
0.05
MBEdiff
3.80
0.69
MBAdiff
1.76
0.34
Panel B : Change Firms: Multi_Multi
N = 857
1.00
58.36
Three diff
2.00
1.87
Dissim
1.00
0.64
SameFF
1.00
0.72
PMATpf
0.03
0.08
Vrel
0.01
0.89
Compdiff
0.30
3.88
MBEdiff
0.12
1.25
MBAdiff
N = 2,189
Under SFAS No.131
47.65
0.00
1.20
0.00
0.76
1.00
0.80
1.00
0.06
0.02
0.44
0.00
2.73
0.00
1.04
0.00
N = 1,108
46.38
1.52
0.72
0.78
0.08
0.73
3.53
1.07
0.00
1.00
1.00
1.00
0.02
0.00
0.00
0.00
+
+
+
-
Mean
Test
-8.95***
-17.60***
11.60***
11.16***
2.88***
-3.11***
-1.44
-3.91***
Median
Test
-15.40***
-17.86***
11.35***
10.94***
3.16***
-14.62***
-14.60***
-14.82***
+
+
+
-
-2.16**
-4.99***
3.84***
3.32***
1.17
-0.63
-0.38
-0.88
-4.57***
-5.09***
3.83***
3.31***
1.43*
-3.80***
-3.92***
-3.87***
Panel C : Comparison of Multi_Multi Firms (Panel B) and Single_Multi Firms (Panel A)
Mean
Median
Mean
Median
Test
Test
Test
Test
-0.29
5.65***
-5.39*** -5.73***
Three diff
-6.14***
5.67***
6.55***
-6.18***
Dissim
4.14***
-2.14**
-2.17**
4.15***
SameFF
4.68***
-1.48
-1.50*
4.71***
PMATpf
7.35***
8.03***
7.79***
6.83***
Vrel
-5.18***
1.88*
5.55***
-0.65
Compdiff
-4.10***
1.19
6.16***
0.08
MBEdiff
-5.18***
0.20
5.75***
-2.24**
MBAdiff
*, **, *** indicates significance levels at the 10%, 5% and 1% levels, respectively. Change firms are those that changed the
number or composition of segments based on SID (segment identification) in the business segment file during the year of the adoption
of the new standard. The Single_Multi sample includes firms that report a single segment under SFAS No. 14 and multiple segments
under SFAS No. 131. Multi_Multi change firms report more than one segment under SFAS No.14 and increased the number segments
reported under SFAS No.131. Threediff is the absolute value of the segments primary 3- digit SIC code less the segment’s secondary
3-digit SIC code. Dissim is discrete random variable assigned a value of 0 when the segments primary and secondary SIC codes are
identical, a value 1 when the primary and secondary SIC codes have the same first 3 digits, a value of 2 when the primary and
secondary SIC codes have the same first 2 digits, a value of 3 is assigned when the primary and secondary SIC codes have the same
41
first digit and finally, a value of 4 when the segment’s primary and secondary SIC codes have no common digits, as described more
fully in section III. SameFF is an indicator variable assigned a 1 if the segment’s primary and secondary SIC codes are in the same
Fama-French industry and zero otherwise. PMATpf is an indicator variable assigned a 1 if the segment’s primary and secondary 2
digit NAICS codes are in the same strategy portfolio as determined by profit margin and asset turnover quintiles and zero otherwise.
Vertical relatedness (Vrel) represents the average input transfers between the segment’s primary and secondary industries based on the
Input-Output accounts for the US economy. Compdiff is the absolute value of the estimate of the speed of abnormal profit adjustment
(Harris 1998) for the segment’s primary SIC industry less the speed of abnormal profit adjustment for the segment’s secondary SIC
industry. MBEdiff is absolute value of the difference between growth opportunities measured as market-to-book equity of the
segment’s primary SIC and segment’s secondary SIC. MBE is measured as MVE (closing price (data#19) x shares out (data
#25))/common equity (data #60). MBAdiff is the absolute value of the difference between growth opportunities measured as marketto-book assets of the segment’s primary SIC and segment’s secondary SIC. MBA is calculated as (total assets less common equity
plus MVE)/total assets.
42
TABLE 7
Univariate Tests of Relatedness of Operations
Across Segment Analysis
Change Firm Multi_Multi Sample
Variable
RelDiv
UnrDiv
Comp_range
MBE_range
MBA_range
Sample Size
Under SFAS No. 14
Mean
Median
0.27
0.07
0.41
0.43
1.69
0.11
7.27
2.30
2.45
1.00
N = 296
Under SFAS No. 131
Mean
Median
0.41
0.34
0.48
0.53
2.09
0.14
8.42
3.20
3.66
1.37
N = 296
Expect
+
+
+
+
Mean
Test
4.66***
2.16**
0.64
0.57
1.71
Median
Test
4.91***
2.07**
1.51*
2.03**
2.18**
*, **, *** indicates significance levels at the 10%, 5% and 1% levels, respectively. Change firms are those that changed the
number or composition of segments based on SID (segment identification) in the Compustat business segment file during
the year of the adoption of the new standard. Multi_Multi firms reported multiple segments under SFAS No. 14 and
increased the number of segments reported under SFAS No. 131. Single_Multi indicates change firms that reported a single
segment under SFAS No.14 and multiple segments under SFAS No.131; these firms are not tabulated above as across
segment variables equal zero when a firm reports a single segment. The entropy measures of diversification are calculated
as follows: Total diversification = Pi ln (1/Pi) where Pi is defined as the proportion of segment i’s sales in the total sales of
the firm. To calculate the additive related and unrelated components of the total diversification number, segment sales at the
4-digit SIC level are consolidated into industry groups which share the same 2-digit SIC code. Let Ps equal the proportion
of total firm sales within a 2-digit industry group. An entropy measure within 2-digit industry groups is Ew= (Pi/Ps) ln
(Ps/Pi). Related diversification = Ps Ew. Unrelated diversification is then the weighted average of segment sales occurring
across 2-digit SIC industry groups or DU =  Ps ln (1/Ps). Comp_range is the absolute value of the difference between the
maximum and minimum measures of the abnormal speed of profit adjustment (Harris 1998) based on the primary SIC codes
assigned to firm’s segments. MBE_range is the absolute value of the difference between the maximum and minimum
measures of the market-to-book equity measure of growth opportunities based on the primary SIC codes assigned to firms’
segments. MBE is measured as MVE (closing price (data#19) x shares out (data #25))/common equity (data #60).
MBA_range is the absolute value of the difference between the maximum and minimum measures of the market-to-book
assets measure of growth opportunities based on the primary SIC codes assigned to firms’ segments.
43
I.
TABLE 8
Univariate Tests of the Relatedness of Operations Within Segments
Comparing Change Firms to No-change Firms in the pre and post period
Panel A: Segment Level Analysis comparing Single_Multi Change Sample to No_Change Segment Samples Under SFAS No.14
Single_Multi – Pre N = 787
No_Change Multi Segment –Pre N = 899
No_Change Single Segment- Pre
N = 2344
Under SFAS No. 14
Under SFAS No. 14
Mean
Median
Under SFAS No. 14
Mean
Median
Variable
Mean
Median
Mean
Median
Testa
Testa
Mean
Median
Testa
Testa
Threediff
96.58
7.00
62.91
1.00
-4.66*** -6.36***
106.87
1.00
1.45
-3.87***
Dissim
2.32
2.00
1.78
2.00
-7.22*** -7.16***
1.92
2.00
-5.59***
-5.94***
SameFF
0.54
1.00
0.66
1.00
4.82***
4.79***
0.59
1.00
2.28**
2.29**
PMATpf
0.61
1.00
0.72
1.00
5.00***
4.96***
0.63
1.00
1.26
1.27
Vrel
0.05
0.01
0.07
0.03
4.16***
5.23***
0.04
0.01
-1.83*
0.46
Compdiff
1.11
0.05
0.99
0.01
-0.36
-5.21***
0.49
0.01
-2.73***
-5.65***
MBEdiff
3.80
0.69
4.58
0.16
0.79
-5.43***
3.30
0.19
-.075
-4.62***
MBAdiff
1.76
0.34
1.56
0.06
-0.78
-5.68***
1.78
0.08
0.08
-3.83***
Panel B: Segment Level Analysis comparing Single_Multi Sample to No_Change Segment Samples Under SFAS No. 131
Single_Multi Sample – Post N=2,187
No_Change Multi Segment –Post N = 899
No_Change Single Segment– Post N = 2,344
Under SFAS No. 131
Under SFAS No. 131
Mean
Median
Under SFAS No. 131
Mean
Median
Variable
Mean
Median
Mean
Median
Testa
Testa
Mean
Median
Testa
Testa
Threediff
47.65
0.00
60.14
1.00
2.55**
8.38***
109.00
1.00
13.51***
13.96***
Dissim
1.20
0.00
1.76
2.00
8.98***
9.59***
1.93
2.00
15.18***
15.27***
SameFF
0.76
1.00
0.66
1.00
-5.46*** -5.43***
0.59
1.00
-12.55*** -12.34***
PMATpf
0.80
1.00
0.74
1.00
-3.74*** -3.74***
0.64
1.00
-12.89*** -12.65***
Vrel
0.06
0.02
0.07
0.03
2.79***
3.81***
0.04
0.01
-6.50***
-4.03***
Compdiff
0.44
0.00
0.85
0.01
2.51***
8.40***
0.50
0.01
0.55
11.87***
MBEdiff
2.73
0.00
4.54
0.12
2.39***
8.38***
3.32
0.19
1.21
12.60***
MBAdiff
1.04
0.00
1.62
0.06
3.03***
8.34***
1.78
0.10
5.11***
13.44***
a
test reflects comparison to the single_multi sample
44
TABLE 8 (continued)
Panel C: Segment Level Analysis comparing Multi_Multi Change Sample to No_Change Segment Samples Under SFAS No. 14
Multi_Multi Change–Pre N=857
No_Change Multi Segment –Pre N = 899
No_Change Single Segment –Pre N = 2,344
Under SFAS No. 14
Under SFAS No. 14
Mean
Median
Under SFAS No. 14
Mean
Median
b
b
b
Variable
Mean
Median
Mean
Median
Test
Test
Mean
Median
Test
Testb
Threediff
58.36
1.00
62.91
1.00
0.73
-0.90
106.87
1.00
7.35***
1.85**
Dissim
1.87
2.00
1.78
2.00
-1.09
-1.15
1.92
2.00
0.89
0.89
SameFF
0.64
1.00
0.66
1.00
0.62
0.62
0.59
1.00
-2.79***
-2.75***
PMATpf
0.72
1.00
0.72
1.00
0.24
0.24
0.63
1.00
-4.42***
-4.41***
Vrel
0.08
0.03
0.07
0.03
-3.11**
-2.33**
0.04
0.01
-11.78***
-8.99***
Compdiff
0.89
0.01
0.99
0.01
0.35
-0.39
0.49
0.01
-2.08**
0.14
MBEdiff
3.88
0.30
4.58
0.16
0.71
-1.33
3.30
0.19
-0.87
-0.13
MBAdiff
1.25
0.12
1.56
0.06
1.28
-0.69
1.78
0.08
2.60***
1.72**
Panel D: Segment Level Analysis comparing Multi_Multi Change Sample to No_Change Segment Samples Under SFAS No. 131
Multi_Multi Change–Post N=1,108
No_Change Multi Segment –Post
N = 899
No_Change Single Segment –Post N = 2,344
Under SFAS No. 131
Under SFAS No. 131
Mean
Median
Under SFAS No. 131
Mean
Median
Variable
Mean
Median
Mean
Median
Testb
Testb
Mean
Median
Testb
Testb
Three diff
46.38
0.00
60.14
1.00
2.45**
3.15***
109.00
1.00
10.70***
6.92***
Dissim
1.52
1.00
1.76
2.00
3.44***
3.41***
1.93
2.00
6.94***
6.87***
SameFF
0.72
1.00
0.66
1.00
-2.93*** -2.94***
0.59
1.00
-7.90***
-7.83***
PMATpf
0.78
1.00
0.74
1.00
-2.02**
-2.01**
0.64
1.00
-8.77***
-8.68***
Vrel
0.09
0.03
0.07
0.03
-4.00*** -3.21***
0.04
0.01
14.04*** -11.52***
Compdiff
0.73
0.00
0.85
0.01
0.52
3.00***
0.50
0.01
-1.33
4.72***
MBEdiff
3.53
0.00
4.54
0.12
1.08
2.37***
3.32
0.19
-.034
4.59***
MBAdiff
1.07
0.00
1.62
0.06
2.34**
2.77***
1.78
0.10
3.89***
6.11***
b
test reflects comparison to the multi_multi change sample.
*, **, *** indicates significance levels at the 10%, 5% and 1% levels, respectively. Change firms are those that changed the number or composition of segments
based on SID (segment identification) in the Compustat business segment file during the year of the adoption of the new standard. Single_Multi indicates change
firms that reported a single segment under SFAS No.14 and multiple segments under SFAS No.131. Multi_Multi firms reported multiple segments under SFAS
No. 14 and increased the number of segments reported under SFAS No. 131. No_Change firms report the same grouping of business activities, either a single
segment or multiple segments, under both SFAS No. 14 and SFAS No. 131.
See Table 6 for variable descriptions.
45
Appendix A
Fama-French (1997) Industry Classifications
Four-digit SIC codes are assigned to 48 industries as follows:
Agric
Food
Soda
Beer
Smoke
Toys
Fun
Books
Hshld
Clths
Hlth
MedEq
Drugs
Chems
Rubbr
Txtls
BldMt
Cnstr
Steel
FabPr
Mach
ElcEq
Misc
Autos
Aero
Ships
Guns
Gold
Agriculture
Food
Products
Candy and
Soda
Alcoholic
Beverages
Tobacco
Products
Recreational
Products
Entertainment
Printing and
Publishing
Consumer
Goods
Apparel
Healthcare
Medical
Equipment
Pharmaceutic
al Products
Chemicals
Rubber and
Plastic
Products
Textiles
Construction
Materials
Construction
Steel Works,
Etc.
Fabricated
Products
Machinery
Electrical
Equipment
Miscellaneous
Automobiles
and Trucks
Aircraft
Shipbuilding,
Railroad Eq
Defense
Precious
0100-0799,2048-2048
2000-2046,2050-2063,2070-2079,
2090-2095,2098-2099
2064-2068,2086-2087,2096-2097
2080-2085
2100-2199 0900-0999,3650-3652,3732-3732, 3930-3949
7800-7841,7900-7999
2700-2749,2770-2799
2047-2047,2391-2392,2510-2519,2590-2599,2840-2844,316
0-3199, 3229-3231, 3260-3260, 3262-3263,
3269-3269,3630-3639,3750-3751, 3800-3800,3860-3879,
3910-3919, 3960-3961,3991-3991,3995-3995
2300-2390,3020-3021,3100-3111, 3130-3159,3965-3965
8000-8099
3693-3693,3840-3851
2830-2836
2800-2829,2850-2899
3000-3000,3050-3099
2200-2295, 2297-2299, 2393-2395, 2397-2399
0800-0899,2400-2439,2450-2459,2490-2499,2950-2952,320
0-3219,
3240-3259,3261-3261,3264-3264,3270-3299,3420-3442,344
6-3452, 3490-3499,3996-3996
1500-1549,1600-1699,1700-1799
3300-3369,3390-3399
3400-3400,3443-3444,3460-3479
3510-3536,3540-3569,3580-3599
3600-3621,3623-3629,3640-3646,
3648-3649,3660-3660,3691-3692, 3699-3699
3900-3900,3990-3990,3999-3999, 9900-9999
2296-2296,2396-2396~3010-3011,
3537-3537,3647-3647,3694-3694, 3700-3716,
3790-3792,3799-3799
3720-3729
3730-3731,3740-3743
3480-3489,3760-3769,3795-3795
1040-1049
Mines
Coal
Enrgy
Util
Telcm
PerSv
Metals
Nonmetallic
Mining
Coal
Petroleum
and Natural
Gas
Utilities
Telecommuni
cations
Personal
Services
BusSv
Business
Services
Comps
Chips
Computers
Electronic
Equipment
Measuring
and Control
Equip
Business
Supplies
Shipping
Containers
Transportatio
n
Wholesale
Retail
LabEq
Paper
Boxes
Trans
Whlsl
Rtail
Meals
Banks
Insur
RlEst
Fin
Restaurants,
Hotel, Motel
Banking
Insurance
Real Estate
Trading
1000-1039,1060-1099,1400-1499
1200-1299
1310-1389,2900-2911,2990-2999
4900-4999
4800-4899
7020-7021,7030-7039,7200-7212,
7215-7299,7395-7395,7500-7500, 7520-7549,
7600-7699,8100-8199, 8200-8299,8300-8399,8400-8499,
8600-8699,8800-8899
2750-2759,3993-3993,7300-7372,
7374-7394,7397-7397,7399-7399, 7510-7519, 8700-8748,
8900-8999
3570-3579,3680-3689,3695-3695, 7373-7373
3622-3622,3661-3679,3810-3810, 3812-3812
3811-3811,3820-3830
2520-2549,2600-2639,2670-2699, 2760-2761,3950-3955
2440-2449,2640-2659,3210-3221, 3410-3412
4000-4099, 4100-4199, 4200-4299,
4400-4499,4500-4599,4600-4699, 4700-4799
5000-5099,5100-5199
5200-5299,5300-5399,5400-5499,
5500-5599,5600-5699,5700-5736, 5900-5999
5800-5813,5890-5890,7000-7019,
7040-7049,7213-7213
6000-6099,6100-6199
6300-6399,6400-6411
6500-6553
6200-6299,6700-6799
47
Appendix B
Numerical Example of Related and Unrelated Diversification
Based on Palepu ((1985), p. 253
Group 1
SIC 10
Total
Sales
100
100
100
100
100
100
100
Group 2
SIC 20
Diversification
Segment 1
1010
Segment 2
1020
Segment 1
2010
Segment 2
2015
Segment 3
2020
Total
Related
Unrelated
100
95
90
80
70
60
20
0
5
10
10
20
10
20
0
0
0
10
10
10
20
0
0
0
0
0
10
20
0
0
0
0
0
10
20
0
0.20
0.32
0.64
0.80
1.23
1.61
0
0.20
0.32
0.32
0.48
0.62
0.94
0
0
0
0.32
0.32
0.61
0.67
Consistent with Palepu (1985) we define industry groups using two-digit SIC and industry segments
using four-digit SIC.
48
REFERENCES
Adam, T. and V. Goyal. 2006. The investment opportunity set and its proxy variables: Theory
and evidence. Working paper. MIT.
Adner, R. and P. Zemsky. 2006. Diversification and performance: linking relatedness, market
structure and the decision to diversify. Working paper. INSEAD.
Baiman, S., D. Larker and V. Rajan. 1995. Organizational design for business units. Journal of
Accounting Research 33: 205-229.
Bens, D. and S. Monahan. 2004. Disclosure quality and the excess value of diversification.
Journal of Accounting Research 42: 691-730.
Berger, P. and E. Ofek. 1995. Diversification’s effect on firm value. Journal of Financial
Economics 37: 39-65.
______. and R. Hann. 2003. The Impact of SFAS No. 131 on information and monitoring.
Journal of Accounting Research 41: 163-223.
Berger, P. and R. Hann. 2007. Segment profitability and the proprietary and agency costs of
disclosure. The Accounting Review 82: 969-906
Bhojraj, S., C. Lee and D. Oler. 2003 . What’s my line? A comparison of industry classification
schemes for capital market research. Journal of Accounting Research 41: 745-774.
Botosan, C. and M. Stanford. 2005. Managers’ motives to withhold segment disclosures and the
effect of SFAS No. 131 on analysts’ information environment. The Accounting review 80:751771.
Brickley, J., C. Smith and J. Zimmerman. 2004. Managerial Economics and Organizational
Architecture. New York, NY. McGraw-Hill.
Bushman, R., R. Indjejikian and A. Smith. 1995. Aggregate performance measures in business
unit manager compensation: the role of intrafirm interdependenices. Journal of Accounting
Research 33:101-128.
Chen, P. and G. Zhang. 2003. Heterogeneous investment opportunities in multiple-segment firms
and the incremental value relevance of segment accounting data. The Accounting Review 78: 397428.
Comment, R. and G. Jarrell. 1995. Corporate focus and stock returns. Journal of Financial
Economics 37: 67-87.
Ettredge, M., S. Kwon, and D. Smith. 2002. Competitive harm and companies’ positions on
SFAS No. 131. Journal of Accounting Auditing and Finance: 93-109.
Ettredge, M., S. Kwon, D. Smith and M. Stone. 2006. The effect of SFAS No. 131 on crosssegment variability of profits reported by multiple segment firms. Review of Accounting Studies.
17: 91-117.
Ettredge, M., S. Kwon, D. Smith and P. Zarowin. 2005. The impact of SFAS No. 131 business
segment data on the market’s ability to anticipate future earnings. The Accounting Review 80:
773-804.
49
Fan, J. and L. Lang. 2000. The measurement of relatedness: an application to corporate
diversification. Journal of Business 73: 629-660.
Fama, E. and K. French. (1997). “Industry costs of equity ”. Journal of Financial Economics 43:
153-193.
Financial Accounting Standards Board (FASB). 1975. Financial Reporting for Segments of a
Business Enterprise. Statement of Financial Accounting Standards No.14. Stamford, CT: FASB.
Financial Accounting Standards Board (FASB). 2008. Disclosures about Segments of an
Enterprise and Related Information. Statement of Financial Accounting Standards No.131 (As
Amended). Stamford, CT: FASB.
Financial Accounting Standards Board (FASB). 2008. Conceptual Framework for Financial
Reporting: The Objectives of Financial Reporting and Qualitative Characteristics and
Constraints of Decision-Useful Financial Reporting Information. Exposure Draft. Stamford, CT:
FASB.
Gaver, J and K. Gaver. 1993. Additional evidence on the association between the investment
opportunity set and corporate financing, dividend, and compensation policies. Journal of
Accounting and Economics 16: 125-160.
Givoly, D., C. Hayn, and J. D’Souza. 1999. Measurement errors and information content of
segment reporting. Review of Accounting Studies.4:15-43.
Harris, M. 1998. The association between competition and manager’ business segment reporting
choices. Journal of Accounting Research: 111-128.
Harris, M., Kriebel, C.H., Raviv, A. 1982. Asymmetric information, incentives and intra-firm
resource allocation. Management Science 28, 604-620.
Hayes, R., and R. Lundholm. 1996. Segment Reporting to the capital market in the presence of a
competitor. Journal of Accounting Research: 261-279.
Healy, P., K. Palepu and R. Ruback. 1992. Does corporate performance improve after mergers?
Journal of Financial Economics 31: 135-175.
Herrmann D., and W. Thomas. 2000. An analysis of segment disclosures under SFAS No. 131
and SFAS No. 14. Accounting Horizons 14: 287-302.
Jacquemin, A. and C. Berry. 1979. Entropy measure of diversification and corporate growth.
Journal of Industrial Economics. June: 359-69.
Jensen, M.C., Meckling, W.H., 1992. Specific and general knowledge and organizational
structure. In: Werin, L., Wijkander, H. (Eds.), Main Currents in Contract Economics. Blackwell,
Oxford, U.K.
Kaplan, S. and M. Weisbach. 1992. The success of acquisitions: evidence from divestitures.
Journal of Finance 47: 107-138.
Krishnan, J., and E. Press. 2003. The North American industry classification system and its
implications for accounting research. Contemporary Accounting Research 20: 685-717.
50
Nayyar, P. 1993. Stock market reactions to related diversification moves by service firms seeking
benefits from information asymmetry and economies of scope. Strategic Management Journal,
14: 569-591.
Palepu, K. 1985. Diversification strategy, profit performance and the entropy measure. Strategic
Management Journal 6: 239-255.
Piotroski, J. 2003. Discussion of the impact of SFAS No. 131 on information and monitoring.
Journal of Accounting Research 41:225-234.
Rajan, R. H. Servaes and L. Zingales. 2000. The cost of diversity: the diversification discount and
inefficient Investment. Journal of Finance 55: 35-80.
Rumelt, R. 1974. Structure and Economic Performance, Harvard University Press, Cambridge,
MA.
Securities and Exchange Commission. Study Pursuant to Section 108(d) of the Sarbanes-Oxley
Act of 2002 on the Adoption by the United States Financial Reporting System of a PrinciplesBased Accounting System. SEC: July 25, 2003.
Selling, T. I. and C. P. Stickney. The effects of business environment and strategy on a firm’s rate
of return on assets. 1989. Financial Analysts Journal (January-February): 43-68.
Smith, C. and R. Watts. 1992. The investment opportunity set and corporate financing, dividend,
and compensation policies. Journal of Financial Economics 32: 263-292.
Street, D., N. Nichols, and S. Gray 2000. Segment disclosures under SFAS No. 131: Has business
segment reporting improved? Accounting Horizons 14:259-285.
Villalonga, B.2002. Diversification discount or Premium? New evidence from the business
information tracking series. Journal of Finance 59: 479-506.
51