Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
JOURNALOF
Journal of Financial Economics 37 (1995) 39-65
Diversification’s effect on firm value
Philip G. Berger*Va, Eli Ofekb
a Wharton School, University
of
Penns.vlvania. Philadelphia, PA 19104-6365, USA
bStern School qf Business. New York University, New York, NY 10003, USA
(Received March 1993; final version received June 1994)
Abstract
We estimate diversification’s effect on firm value by imputing stand-alone values for
individual business segments. Comparing the sum of these stand-alone values to the
firm’s actual value implies a 13% to 15% average value loss from diversification during
1986-1991. The value loss is smaller when the segments of the diversified firm are in the
same two-digit SIC code. We find that overinvestment and cross-subsidization contrib-
ute to the value loss. The loss is reduced modestly by tax benefits of diversification.
Key words: Diversification; Focus; Overinvestment; Cross-subsidization; Organizational
structure
.IEL classijrarion: G32; G34
1. Introduction
During the 1950s and ’60s many corporations undertook massive diversifica-
tion programs. This process reached its climax with the merger wave of the late
1960s and the accompanying rise to prominence of huge conglomerate firms. In
the last 15 years the trend has reversed, with recent studies by Comment and
*Corresponding author.
Berger acknowledges the financial support of Coopers & Lybrand and the University Research
Foundation of the University of Pennsylvania. We appreciate the helpful comments of Patty
Dechow, Ned Elton, Bob Holthausen, Richard Sloan, and especially the many constructive sugges-
tions of Wayne Mikkelson (the editor) and John Byrd (the referee). We have also benefited from the
comments of workshop participants at New York University, Rutgers University, Tel Aviv Univer-
sity, and University of Pennsylvania.
0304405X/95/$07.00 @ 1995 Elsevier Science S.A. All rights reserved
SSDI 0304405X9400798 6
40
P.G. Berger, E. OjhklJournal
of
Financial Economics 37 (1995) 3665
Jarrell(1994) and Liebeskind and Opler (1993) documenting a return to special-
ization. This push toward focus apparently resulted from the view that unrelated
diversification decreases firm value.
Theoretical arguments suggest that diversification has both value-enhancing
and value-reducing effects. The potential benefits of operating different lines of
business within one firm include greater operating efficiency, less incentive to
forego positive net present value projects, greater debt capacity, and lower taxes.
The potential costs of diversification include the use of increased discretionary
resources to undertake value-decreasing investments, cross-subsidies that allow
poor segments to drain resources from better-performing segments, and mis-
alignment of incentives between central and divisional managers. There is no
clear prediction about the overall value effect of diversification.
We use segment-level data to estimate the valuation effect of diversification
and to examine the potential sources of value gains or losses. We compare the
sum of the imputed stand-alone values of the segments of diversified companies
to the actual values of those companies. We document that diversified firms
have values that average, during 1986-91, 13% to 15% below the sum of the
imputed values of their segments. The loss in value is, however, considerably less
for related diversifications. We identify overinvestment in segments from indus-
tries with limited investment opportunities, as measured by a low Tobin’s
q-ratio,
as one source of the value loss. An additional source of loss in value
is cross-subsidization of poorly performing divisions by better-performing
divisions. The value loss is reduced by a modest decrease in taxes.
Section 2 reviews the related literature and details the predictions resulting
from prior theoretical work. Section 3 describes the sample and explains our
empirical approach. Section 4 assesses the overall value effect of diversification
using imputed segment values and a comparison of profitability between diversi-
fied and single-segment firms. Section 5 explores the individual sources of value
loss and gain, and Section 6 concludes. The appendix provides additional details
on our empirical approach and variable construction.
2. Hypo&eses
Consistent with observed trends in diversification activity, theoretical argu-
ments developed during the late 1960s and early ’70s generally address the
benefits of diversification, whereas more recent papers address the costs.
Chandler (1977) argues that, because multidivision firms create a level of
management concerned with coordination of specialized divisions, they are
inherently more efficient and thus more profitable than their lines of business
would be separately. Weston (1970) states that resource allocation is more
efficient in internal than in external capital markets. He therefore contends that
diversified firms allocate resources more efficiently because they create a larger
P.G. Berger, E. OfeklJournal
qf
Financial Economics 37 11995) 3965
41
internal capital market. A version of this argument raised by Stulz (1990) is that
diversified firms, by creating a larger internal capital market, reduce the
underinvestment problem described by Myers (1977). These internal
capital market arguments predict that diversified companies make more
positive net present value investments than their segments would make as
separate firms.
Another potential benefit of diversification arises from combining businesses
with imperfectly correlated earnings streams. This coinsurance effect gives
diversified firms greater debt capacity than single-line businesses of similar size
(Lewellen, 1971). One way in which increased debt capacity creates value is by
increasing interest tax shields. Thus, diversified firms are predicted to have
higher leverage and lower tax payments than their businesses would show if
operated separately. A further tax advantage arises from the tax code’s asym-
metric treatment of gains and losses. Majd and Myers (1987) note that undiver-
sified firms are at a significant tax disadvantage because tax is paid to the
government when income is positive, but the government does not pay the firm
when income is negative. This disadvantage is reduced, but not eliminated, by
the tax code’s carryback and carryforward provisions. The Majd and Myers
analysis predicts that, as long as one or more segments of a conglomerate
experience losses in some years, a conglomerate pays less in taxes than its
segments would pay separately.
Diversification can create several costs. Stultz (1990) argues that diversified
firms will invest too much in lines of business with poor investment opportuni-
ties. Jensen’s (1986) assertion that managers of firms with unused borrowing
power and large free cash flows are more likely to undertake value-decreasing
investments has a similar implication. To the extent that lines of business have
access to more free cash flow as part of a diversified firm than on their own,
Jensen’s argument predicts that diversified firms invest more in negative net
present value projects than their segments would if operated independently.
Meyer, Milgrom, and Roberts (1992) make a related argument regarding the
cross-subsidization of failing business segments. Since a failing business cannot
have a value below zero if operated on its own, but can have a negative value if it
is part of a conglomerate that provides cross-subsidies, Meyer, Milgrom, and
Roberts predict that unprofitable lines of business create greater value losses in
conglomerates than they would as stand-alone firms. For example, when
Michael Walsh was preparing to take over as CEO of Tenneco, he discovered
that ‘Tenneco’s profitable auto-parts and chemicals divisions didn’t strive as
hard as they might for higher earnings because their surplus was routinely
dumped into the company’s money-losing farm-equipment operation’
(Wall
Street Journal,
March 29, 1993). Finally, Myerson (1982) and Harris, Kriebel,
and Raviv (1982) discuss the information asymmetry costs that arise between
central management and divisional managers in decentralized firms. These costs
are higher in conglomerates than in focused firms to the extent information is
42
P.G. Berger, E. OfeklJournal @‘Financial Economics 37 (1995) 39-65
more dispersed within the firm, leading to the prediction that diversified firms
are less profitable than their lines of business would be separately.
The theoretical arguments discussed above do not distinguish between related
and unrelated diversification. Many authors argue, however, that related diver-
sified firms perform better than conglomerates. Rumelt (1974) argues that
related diversification affects value more positively than unrelated diversifica-
tion because skills and resources can be used in related markets. Others discuss
the effects of reputation and economies of scope, which arise when the joint cost
of producing two or more outputs is less than the sum of the costs of producing
each output by itself. Nayyar (1993), for example, argues that benefits from
a positive reputation in an existing business and from economies of scope are
available from related, but not from unrelated, diversification. The resulting
prediction is that the valuation effect of diversification is more positive for
related than unrelated lines of business.
Empirical studies have generally produced mixed results on diversification’s
overall value effect.’ Copeland and Weston (1979) cite several studies that find
superior stock price performance by conglomerates over mutual funds in the
1960s and early ’70s. Other studies, however, find inferior performance during
this period. Ravenscraft and Scherer (1987) argue that the performance of
a sample of conglomerates becomes noticeably worse if the 1970s are included.
De (1992) however, finds no cross-sectional correlation between the degree of
focus and measures of excess return calculated over the period 1976-85.
Other studies addressing the effect of focus on performance also produce
mixed results. Chatterjee and Wernerfelt (1991) review the literature measuring
performance with accounting numbers and find that no consensus emerges.
Event studies generally attribute a comparative penalty to diversification, espe-
cially in the 1980s although its magnitude is rarely statistically significant.
While these event studies examine the buyers of related and unrelated firms,
John and Ofek (1994) find that increased focus is a significant determinant of
seller gains from asset sales.
Two recent studies provide evidence of a negative relation between diversifi-
cation and value. Comment and Jarrell (1994) find a negative relation
during 1978-89 between abnormal stock returns and several measures of diver-
sification, including the number of segments reported by management and
revenue- and asset-based Herfindahl indexes. Similarly, Lang and Stulz (1994)
present evidence of a negative relation between Tobin’s
q-ratio
and these
diversification measures. Neither study examines the potential sources of these
value losses.
‘Many of these studies are discussed by Comment and Jarrell(1994), and we draw heavily on their
review.
PG. Berger, E. OfekJJournal of Financial Economics 37 (1995) 3945
43
3. Sample selection and estimation of segment values
3.1. Sample selection and description
FASB No. 14 and SEC Regulation S-K require firms to report segment
information for fiscal years ending after December 15, 1977. Firms must report
audited footnote information for segments whose sales, assets, or profits exceed
10% of consolidated totals. The Compustat Industry Segment (CIS) database
reports segment information for all active Compustat firms other than utility
subsidiaries. For approximately 6,500 firms per year, the file contains informa-
tion on five variables by segment [net sales, earnings before interest and taxes
(EBIT), depreciation, capital expenditures, and assets] as well as the number of
reported segments for the firm.
We obtain data for all firms on the CIS tape during the 1986-91 period that
have total sales of at least $20 million and have no segments in the financial
services industry (SIC codes between 6000 and 6999). To avoid distorted
valuation multiples for firms with sales or assets near zero, we require firms to
have sales of at least $20 million. Firms with financial services segments are
removed from consideration because applying the valuation methods we use is
problematic for such firms. Specifically, many firms in the financial services
industry do not have information available on earnings before interest and taxes
(EBIT), because such earnings are not meaningful for financial companies. To be
included in the final sample, diversified corporations must have data on both the
CIS and Compustat data files. Additionally, total capital (measured as market
value of common equity plus book value of debt) for the firm must be available.
Unlike assets and earnings, sales are usually completely allocated among the
reported segments of a diversified firm; therefore, we require that the sum of
segment sales must be within 1% of total sales for the firm. These procedures
result in a sample of 3,659 firms with 16,181 observations, of which 5,233 are
multi-segment. Of the multi-segment observations, 2,473 are two-segment, 1,557
are three-segment, 752 are four-segment, and 451 report five or more segments.
Table 1 describes the single-segment and multi-segment firms, as well as the
differences between the groups. Due to skewness in the distributions, we empha-
size medians rather than means. At the median, multi-segment firms have three
segments and roughly three times the total capital of single-segment firms. Their
median industry-adjusted leverage ratio (book value of debt over total assets) is
2.9% higher than that of focused firms, consistent with segments of diversified
firms borrowing more, and their median industry-adjusted taxes do not differ
from those of single-line firms, inconsistent with tax benefits from diversification.
The segment level characteristics in the bottom panel of Table 1 are based
on the segment data. Segments of diversified companies tend to be slightly
smaller than stand-alone segments, although this difference is due in part to an
incomplete allocation of assets by diversified firms to their segments. When we
Table 1
Descriptive statistics for a sample of 16,181 observations of single-business-segment and multi-segment firms with sales of more than $20 million and
information available from the Compustat Industry Segment (CIS) database
Segments are lines of business for which separate accounting disclosures are made by management in accordance with FASB No. 14 and SEC Regulation
S-K. Single-segment firms are those reporting exactly one segment on the CIS database, whereas multi-segment firms are those reporting two or more
segments. Significance levels are indicated for the difference between multi-segment and single-segment firms. The significance of the difference in medians
is assessed using the nonparametric median test.
Single-segment firms
Med. Mean STD
Multi-segment firms
Med. Mean STD
Difference
Med. Mean
Sample characteristics at the firm level
Number of segments 1.000 1.000 0.000 3.000 2.885 1.088 2.000s 1.885h
Total capital (S millions) 116 878 4086 316 1961 4734 2oOb 1083b
Total debt to assets 0.257 0.278 0.217 0.290 0.305 0.191 0.033h 0.026b
Industry-adjusted leverage 0.000 0.027 0.198 0.029 0.041 0.193 0.029’ 0.014s
Taxes/sales 0.016 0.027 0.044 0.017 0.024 0.039 0.000 - 0.003b
Industry-adjusted taxes 0.000 - 0.001 0.023 0.000 - 0.002 0.015 O.ooo’ - O.OOIb
Observations 10,948 5,233
Sample characteristics at the segment leuel
Segment sales ($ millions)
123
Segment assets (S millions)
106
Capital expenditures/sales
0.044
Negative CF segmentsa
0.000
Observations
10,948
-~
716
4224
116 649
2033 - 7’
- 127b
784
3494
92 549
1662 - 14* - 23Sb
0.087 0.136
0.045 0.101
0.176 0.001
0.014b
0.082
0.274
0.000
0.090 0.286 0.000’ 0.008’
15,093
Total capital
is the sum of book value of debt and market value of equity.
Industry-adjusted
leverage is the difference between a firm’s actual leverage,
defined as the ratio of total debt to total assets, and its imputed leverage. Imputed leverage for each of the firm’s segments is calculated as the segment’s
assets multiplied by the industry’s median ratio of debt to assets.
Industry-adjusted taxes
paid is the difference between the firm’s actual taxes paid and its
imputed taxes, all standardized by total sales. Imputed taxes for each of the firm’s segments are calculated as the segment’s EBIT multiplied by the
industry’s median ratio of taxes paid to EBIT.
“Indicator that equals 1 if a segment has negative cash flow (EBITD) and 0 otherwise.
‘Significant at the 1% level.
‘Significant at the 5% level.
dSignificant at the 10% level.
46
P.G. Berger, E. OfeklJournal
qf
Financial Economics 37 (1995) 3945
allocate all assets of a firm to its segments, the difference between the median
assets of segments of diversified firms and stand-alone firms from the same
industry is insignificant (result not reported). Segments do, however, differ
somewhat from single-line firms in their investment levels and their likelihood of
having negative cash flow. The investment level, as measured by the ratio of
capital expenditures to sales, is higher for the segments of diversified companies,
although only the difference in means is significant. In addition, the 9% of
segments of diversified firms with negative cash flow is higher than the 8.2% for
single-line businesses. The larger presence of negative cash flow among segments
of diversified firms is consistent with diversified firms keeping poorly performing
divisions in business beyond the point where they would fail if independent.
3.2. Estimating segment values using multipliers
To examine whether diversification enhances or decreases corporate value, we
measure the percentage difference between a firm’s total value and the sum of
imputed values for its segments as stand-alone entities (see the Appendix for
additional details not described below). We calculate the imputed value of each
segment by multiplying the median ratio, for single-segment firms in the same
industry, of total capital to one of three accounting items (assets, sales, or
earnings) by the segment’s level of the accounting item. The industry median
ratios are based on the narrowest SIC grouping that includes at least five
single-line businesses with at least $20 million of sales and sufficient data for
computing the ratios. Using this algorithm, the imputed value for 44.6% of all
segments of diversified companies is based on four-digit SIC code industries,
25.4% on three-digit industries, and 30% on two-digit industries.
The sum of the imputed values of a company’s segments estimates the value of
the firm if all of its segments are operated as stand-alone businesses. The natural
log of the ratio of a firm’s actual value to its imputed value is our measure of
excess value, or the gain or loss in value from diversification. Positive excess
value indicates that diversification enhances the value of segments beyond that
of their stand-alone counterparts. Negative excess value indicates that diversifi-
cation reduces value.
The validity of the multiplier approach depends on management disclosure
policies. Theoretical models of managerial disclosure decisions suggest that
managers may have incentives to misstate segment data to both providers of
capital and product market competitors.2 Their ability to misstate depends on
the discretion managers have to allocate dollars between segments. Since
segment assets must be specifically identifiable with the segment for which they
*See Darrough and Stoughton (1990), Wagenhofer (1990), Feltham and Xie (1992), Feltham, Gigler,
and Hughes (1992), and Newman and Sansing (1993).
P.G. Berger, E. OfekJJournal of Financial Economies 37
(1995J
3945
41
are reported, there is little discretion to misstate them. Managers do have some
ability to allocate sales and greater discretion to allocate expenses, so segment
sales and earnings (sales less expenses) are vulnerable to manipulation. Givoly,
Hayn, and D’Souza (1993) assess the quality of segment reporting, with quality
defined as the difference between the correlation of accounting measures from
segments with those of their industries and the correlation of measures from
single-line firms in the same industries with the aggregate industry measures. They
find a marginally significant difference in correlation coefficients of 0.061 for sales
and a significant difference of 0.146 for earnings. Thus, there is some evidence that
the segment earnings numbers may not be as reliable as segment sales and asset
figures. The earnings multiplier has the advantage, however, of imputing value
directly from current profitability, which may be more directly linked to firm
value than sales or assets. Therefore, we report results for all three multipliers.
Examining diversification using the industry multiplier approach on indi-
vidual business segments has several advantages over other methods. For
example, an event study requires a clearly defined event date. It is difficult to
identify and precisely date an event that unambiguously conveys information
about diversification. The stock price response to takeover announcements may
reflect the terms of the offer, the probability of success, or information signalled
about opportunities in the bidder’s core line of business. Thus, it is difficult to
clearly identify investors’ attitudes about diversification by examining an an-
nouncement-date stock price response.
Tobin’s
q-ratio
is also widely used in studies examining how the level of firm
value varies with firm structure. Calculation of
q
requires assumptions about
rates of depreciation and inflation to estimate the firm’s replacement value. In
addition, valuation studies do not generally industry-adjust
q
despite its large
variation across industries. Attempting to industry-adjust
q
is, however, prob-
lematic when studying diversification, because neither segment market values
nor segment replacement values can be computed directly from available data.
Additionally, event studies and studies that assess value effects using Tobin’s
q-ratio
provide only limited opportunities to examine the potential sources of
gains or losses from diversification. The industry multiplier approach not only
provides a direct estimate of the excess value associated with diversification, but
also allows further investigation at the segment level of the sources of any
overall value effect.
4. The overall value effect of diversification
4.1. The excess value measure of overall efect
Table 2 describes the excess value measures obtained using each of the three
multipliers. We report the single-segment firms’ values to evaluate whether the
4x
P.G. Berger, E. OjbklJournai of Financial Economics 37 (199.5) 39-65
Table 2
Descriptive statistics for the excess value measures using asset, sales, and EBIT multiples
Excess value is the natural logarithm of the ratio of a firm’s actual value to its imputed value. A firm’s
imputed value is the sum of the imputed values of its segments, with each segment’s imputed value
equal to the segment’s assets, sales, or EBIT multiplied by its industry median ratio ofcapital to that
accounting item. The sample includes 16,181 observations between 1986 and 1991. The significance
of median values is based on the Wilcoxon signed-rank test. The significance of the difference in
medians is assessed using the nonparametric median test.
Quartiles
Actual/imputed value
Using asset multiples
Single-segment firms
Multi-segment firms
Using sales multiples
Single-segment firms
Multi-segment firms
Using EBIT multiples
Single-segment firms
Multi-segment firms
Med. Mean 1st
3rd
- o.ooo 0.014”
- 0.266
0.265
- 0.162”. - 0.122asb - 0.371 0.098
0.000 0.001 - 0.370 0.366
- 0.1068*b
- 0.097”,b
- 0.47 1 0.278
0.009” 0.073” - 0.225 0.342
- 0.079a*b - 0.061”.b - 0.323
- 0.177
10,664
3,884
STD
Obs.
0.463
0.408
0.561
0.542
0.486
0.414
10,373
5,015
10,047
2,973
Using asset multiples:
The natural logarithm of actual value/imputed value where: actual value is
total book value of debt plus market value of equity, and imputed value is the sum of the imputed
values of the firm’s segments. Each segment’s imputed value is the segment’s assets multiplied by its
industry median capital-to-assets ratio.
Using sales multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s sales multiplied by its
industry median capital-to-sales ratio.
Using EBIT multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s EBIT multiplied by
its industry median capital-to-EBIT ratio.
“Significant at the 1% level.
bThe difference between the single-segment and multi-segment medians or means is significant at the
1% level.
excess value measures are well-behaved, and report the multi-segment values to
provide a preliminary indication of diversification’s effect on value. By using
multipliers for the median single-segment firm in each industry, the excess values
are constructed with median values of approximately zero for the single-segment
firms. The median values differ slightly from zero due to the elimination of
extreme excess values, and adjustments to the EBIT multiplier for segments with
negative earnings (as described in the Appendix). The distributions of excess
value for single-segment firms are quite symmetric around zero for the valuation
multiples of assets and sales. For the multiples of EBIT, the distribution is
positively skewed. Our tests are based on mean differences between the excess
P.G. Berger, E. OfeklJournal of Financial Economics 37 (1995) 3945
49
values of diversified and single-segment firms, however, so a mean excess value
of zero for the single-segment firms is not required in order to measure diversifi-
cation’s value effect.
Table 2 reports negative differences in mean and median excess values
between stand-alone and multi-segment firms, indicating that diversification
reduces value. Additional evidence of an association between value loss and
diversification is provided in panel A of Table 3, which reports regressions of
excess value on various controls and an indicator variable that equals one if the
firm is multi-segment3 The multi-segment indicator variable captures the per-
centage difference in average excess value between focused and diversified firms.
We control for factors that could affect excess value and whose magnitudes are
not entirely determined by whether or not the firm is diversified. Firm size,
profitability, and growth opportunities are controlled for by using the firm’s
natural log of total assets, its EBIT-to-sales ratio, and its ratio of capital
expenditures to sales. The results are consistent across the three multipliers, with
the lost value from diversification ranging from 12.7%, using the asset multi-
plier, to 15.2%, using the EBIT multiplier. Since these losses are based on total
capital, the value loss to equity holders is even larger than these measures
suggest. Using the average book leverage of about 30% for multi-segment firms,
and assuming no effect on debt value, we calculate the value loss to equity
holders as ranging from 18.1% to 21.7%.
We also calculate average dollar losses from diversification as the mean
difference between imputed and actual value. Using the asset multiplier, the
mean dollar loss per firm during 1986-91 is $235.1 million, implying a total loss
in value for the approximately 850 multi-segment sample firms of $200 billion.
We also compare the value losses of multi-segment firms with differing numbers
of segments by comparing the value losses among diversified firms with two,
three, four, and five or more segments. The (unreported) results show the value
loss increases with the number of segments.4
3The White test rejects the null of homoskedasticity at the 0.01 level for most reported regressions.
Therefore, reported significance levels are calculated using White (1980) heteroskedasticity-consis-
tent standard errors.
4All reportedpercentage changes in value represent logarithmic percentages. The results presented
in panel A of Table 3 are unchanged when (1) extreme values are included in the sample, (2)
influential observations are omitted (see Belseley, Kuh, and Welsch, 1980, for the method
used
to
identify influential observations), and (3) the effect of the fineness of industry partitioning is
addressed by using just two-digit SIC codes to define industries. We also test the sensitivity of the
results to our diversification metric by replacing the multi-segment indicator variable with a rev-
enue-based Herfindahl index, whose calculation is detailed in the Appendix. The regression coeffi-
cient on the Herfindahl index is significantly positive, consistent with a larger loss of value as
diversification increases.
Table 3
Coefficient estimates from regressions of excess value on: a multi-segment indicator and control variables (in panel A) and the number of segments,
a related segments measure, and control variables (in panel B)
Excess value is the natural logarithm of the ratio of a firm’s actual value to its imputed value. A firm’s imputed value is the sum of the imputed values of its
segments, with each segment’s imputed value equal to the segment’s assets, sales, or EBIT multiplied by its industry median ratio of capital to that
accounting item. The sample in panel A includes 16,181 observations between 1986 and 1991. The sample in panel B includes 5,233 observations between
1986 and 1991 that report two or more segments. P-values (in parentheses) are based on the White-adjusted standard errors.
-
Panel A: Value lossfiom diversification
Dependent variable: Obs.
Actual/imputed value
R2
Intercept
Multi-segment
indicate?
Log of
assets EBlT/sales
Capexjsales
Using asset multiples 14,547
- 0.004 - 0.127d 0.01 id
0.935d
0.051’
0.086 (0.730)
~O.ooo)
ww ww
(0.064)
Using sales multiples
Using EBIT multiples
15,287 - 0.329* - 0.144d
0.046d
1.038” 0.320d
0.114
(O.@w
wm ww uw (O.ow
12,952 0.003 - 0.152d 0.01 Id 0.174d
0.021 (0.833)
~O.ooo)
(0.~)
(O.@W
Panel B: Relatedness and excess value in diversijied firms
Dependent variable: Obs.
Actual/imputed value R2 Intercept
Using asset multiples
3870 - 0.102d
0.070
VW
Using sales multiples 4901 - 0.332“
0.109
(O.@m
Using EBIT multiples
2893 - 0.087’
0.012 (0.011)
No. of
segment?
- 0.027d
(0.002)
- 0.0776
(OJW
- 0.022’
(0.033)
Related
Log of
segments’
assets
0.026d
(0.006)
0.037d
(0.001)
- 0.002
(0.843)
- 0.010’
(0.017)
0.050d
(0.~)
0.010
(0.028)
EBIT/sales Capex/sales
1.126d - 0.009
ww
(0.866)
0.937d
0.509d
KJ.ow ww
0.273’
mw
Using asset multiples:
The natural logarithm of actual value/imputed value where: actual value is total book value of debt plus market value ofequity, and
imputed value is the sum of the imputed values of the firm’s segments. Each segment’s imputed value is the segment’s assets multiplied by its industry
median capital-to-assets ratio. Using sales multiples: The natural logarithm of actual value/imputed value with each segment’s imputed value equal to the
segment’s sales multiplied by its industry median capital-to-sales ratio. Using EBIT multiples: The natural logarithm of actual value/imputed value with
each segment’s imputed value equal to the segment’s EBIT multiplied by its industry median capital-to-EBIT ratio.
Vndicator that equals 1 if the firm has two or more reported segments, and 0 if the firm reports only one line of business.
bathe number of segments reported by the multi-segment firm.
‘The difference between the total number of segments reported by the diversified firm and the number of segments with different main two-digit SIC codes.
dSignificant at the 1% level.
‘Significant at the 5% level.
‘Significant at the 10% level.
52
P.G. Berger, E. OjhkJJournal of Financial Economies 37 (1995) 39-C
In panel B of Table 3, we examine how the value loss varies between related
and unrelated diversified firms. We use an SIC code algorithm to classify
segments as related. Therefore, we cannot identify vertically integrated seg-
ments, which get classified as unrelated. Specifically, we classify segments within
a firm as unrelated if they have different SIC codes at the two-digit level. We
then measure the difference between the total number of segments reported by
a firm and its number of unrelated segments. This related-segments variable
varies between zero, when no segments are related, and the number of firm
segments minus one, when no segments differ.
The excess value measure is regressed on the number of segments reported by
the diversified firm, the related-segments variable, and the controls for size,
profitability, and growth opportunities. The regressions use only the 5,233
multi-segment observations. The significantly negative coefficient estimates on
the number of segments indicate that, all else being equal, multi-segment firms
lose more value as they become more diversified. This result is consistent with
the nearly monotonic relation we found between the magnitude of the value loss
and the number of segments. The significantly positive coefficient estimates,
using the asset and sales multipliers, on the related-segments variable show that
relatedness mitigates the value loss from diversification. All the coefficient
estimates remain positive when the pooled regressions reported in panel B of
Table 3 are performed by year, with the magnitudes significant at better than the
0.10 level in five of 12 cases.
Panel A of Table 4 reports the coefficient estimates on the multi-segment
indicator when the regressions reported in panel A of Table 3 are performed by
year. These results are of interest because the t-statistics for the pooled results of
Table 3 are overstated if the inclusion of the same firm for multiple years results
in observations that are not independent. Using all three multipliers, the coeffi-
cient estimates are significant at the 0.01 level in all years. There is some
variation in the value loss during the period, with losses ranging from 17.7% to
9.2% for the asset multiplier, from 14.9% to 13.1% for the sales multiplier, and
from 17.6% to 12.7% for the EBIT multiplier.
Panel B of Table 4 reports the coefficient estimates on the multi-segment
indicator when the regressions reported in panel A of Table 3 are re-estimated
(absent the size control) on four size-based subsamples with approximately the
same number of observations. The results show that the value loss occurs for all
firm sizes, with the greatest percentage loss occurring in the smallest firms.
4.2. Profitability as an alternative measure of overaN efect
Table 5 examines whether segments of diversified companies have lower
operating profitability than their counterpart single-line firms. We use two
industry-adjusted measures of profitability, operating margin (EBIT/sales) and
P.G. Berger, E. OfeklJournal of Financial Economics 37 (1995) 39-65
53
Table 4
Measures of the percentage value loss from diversification by year and by firm size
The value loss measure is the coefficient estimate of the multi-segment indicator from regressions of
the logarithm of the ratio of a firm’s actual value to its imputed value on a multi-segment indicator
variable, the natural logarithm of assets (in panel A regressions only), EBIT/sales, and capital
expenditures/sales. All of the reported coefficients are significant at the 0.01 level. The sample
includes 16,181 observations between 1986 and 1991.
Panel
A: Loss from diversijcation by year
Actual/imputed value
1986
1987 1988
1989 1990
1991
Using asset multiples - 0.177 - 0.137 - 0.092
- 0.103 - 0.120 - 0.133
Using sales multiples - 0.149 - 0.138 - 0.131
- 0.143 - 0.131 - 0.146
Using EBIT multiples - 0.135 - 0.164 - 0.140
- 0.176 - 0.127 - 0.166
Number of observationsa 2,143 2,401 2,547
2,643 2,742 2,806
Panel B: Loss from diversification by firm size (TA = total assets in millions of dollars)
Actual/imputed value
TA < 50 50< TA<
150
I50 <
TA < 500
TA > 500
Using asset multiples
Using sales multiples
Using EBIT multiples
Number of observation?
-____-
- 0.161 -0.117
- 0.142 - 0.120
- 0.167 -0.112
- 0.144 - 0.141
- 0.201 - 0.116
- 0.159 - 0.146
3,964 3,849
3,482 3,989
Using asset multiples:
The natural logarithm of actual value/imputed value where: actual value is
total book value of debt plus market value of equity and imputed value is the sum of the imputed
values of the firm’s segments. Each segment’s imputed value is the segment’s assets multiplied by its
industry median capital-to-assets ratio.
Using
sales
multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s sales multiplied by its
industry median capital-to-sales ratio.
Using EBIT multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s EBIT multiplied by
its industry median capital-to-EBIT ratio.
“Number of observations in the sales multiplier regressions.
the return on assets (EBIT/assets), or ROA (see the Appendix for additional
details). To correct for unallocated EBIT, we gross up or down the EBIT of each
segment in a firm by the percentage deviation, if any, between the sum of its
segments’ EBITs and the total firm EBIT. We correct for unallocated assets in
the same way. In addition, to examine whether profitability differences depend
on segment size, we report the differences in profitability for three size sub-
samples.
To increase the power of the remaining tests, we exclude multi-segment firms
with all segments in the same two-digit SIC code (completely related multi-
segment firms). The profitability results, using both the operating margin and
54
P.G. Berger, E. Ofek/Journal qf’Financia1 Economics 37 (1995) 3945
Table 5
Mean and median differences in industry-adjusted profitability, as a function of sales or assets,
between single-segment firms and segments of diversified firms
Multi-segment firms that do not have two or more segments with different two-digit SIC codes are
eliminated from the sample. The first number in the square brackets is the number of single-segment
observations and the second number is the number of observations from multi-segment firms. The
sample period is 1986-91. The significance of the difference in medians is assessed using the
nonparametric median test.
___-
Segment sales (SS) in millions of dollars
Total sample
ss < 50
Profitability measured by industry-adjusted EBIT/sales
Mean - 0.020
- 0.037”
Median
- 0.009”
- 0.024”
Observations [10922, 84221 [2767, 27651
50 < ss < 250 SS > 250
- 0.011” - 0.008”
- 0.009” - 0.003b
[4614,2459] [3541,3198]
Projitability measured by industry-adjusted EBIT/total assets
Mean - 0.015”
- 0.030” - 0.006b - 0.005b
Median - 0.019”
- 0.039” - 0.017”
- 0.004
Observations
[10922, 82271
I-2767, 27261 [4614, 24311 [3541,3070]
EBZT-to-sales ratio
is the ratio of the segment minus the industry median EBIT-to-sales ratio. If the
sum of the segments’ EBIT does not match the total firm EBIT then each segment’s EBIT for that
firm is grossed up (or down) by the percentage deviation.
EBIT-to-asset
is the ratio of the segment
minus the industry median EBIT-to-assets ratio. If the sum of the segments’ EBIT does not match
the total firm EBIT then each segment’s EBIT for that firm is grossed up (or down) by the percentage
deviation. A similar procedure is performed for assets.
“Significant at the 1% level.
‘Significant at the 10% level.
ROA measures, indicate that segments of diversified companies are significantly
less profitable than stand-alone companies for the total sample and across all
three size groups. The - 0.009 lower operating margin of the full sample of
segments is significant at the 0.01 level, as is the - 0.019 median difkence in
ROA. Under both measures, the reduction in operating profitability is greatest
for the smallest segments, which was also true of the reduction in value
documented in Table 4. When the single-segment companies are compared to
the excluded group of completely related multi-segment firms, the (unreported)
median differences in profitability continue to be significantly negative, but with
magnitudes only about 60% of those reported in Table 5. Thus, related diversifi-
cation has a less negative effect than unrelated diversification on operating
profitability. Overall, the operating profitability results are consistent with those
found using the excess value measures.
P.G. Berger, E. Ofek/Journal
ef
Financial Economics 37 (1995) 3945
55
5. Sources of gains and losses from diversification
5.1. Overinvestment and the value loss in diversljiedfirms
Jensen (1986) and Stulz (1990) argue that overinvestment is a potential source
of value loss from diversification. We examine whether overinvestment is asso-
ciated with value loss in diversified firms by regressing excess value on a measure
of overinvestment and controls for size, profitability, and capital expenditures.
We restrict the sample to unrelated multi-segment firms. We measure a firm’s
overinvestment as the sum of the depreciation-adjusted capital expenditures of
all its segments operating in industries whose median Tobin’s 4 is in the lowest
quartile (below 0.76), scaled by total sales (see the Appendix for details of the
Tobin’s 4 calculation). Thus, higher values of the overinvestment variable
represent more unprofitable investment.
Table 6 shows that the coefficient estimates range from - 0.399 to - 0.924
on the overinvestment variable and are significant at the 0.10 level or better for
all three methods of calculating excess value. Thus, more overinvestment is
associated with less excess value for multisegment firms with unrelated seg-
ments. An increase in depreciation-adjusted capital expenditures of 1% of sales
in
low-q
industries is associated with an average decrease of 0.4% to 0.9% in
excess value. The results continue to hold when the regressions are performed
using data for separate years.
The overall value loss we documented in Section 4 was a mean difference
between diversified and single-line firms. Therefore, the amount of value loss
attributable to overinvestment can be calculated as the product oh (1) the rate of
value loss per unit of overinvestment (estimated in Table 6) and (2) the mean
difference in overinvestment between segments of diversified firms and single-
segment businesses. Using a regression analysis controlling for profitability and
size, we find (but do not report in a table) that, on average, overinvestment by
unrelated segments of multi-segment firms exceeds that of single-line companies
by 3.6% of sales. When segments are related, the difference in overinvestment is
2.6% of sales. Given the coefficient estimates on the overinvestment variable in
Table 6, the difference in overinvestment of 3.6% of sales implies an excess value
loss of 1.4% to 3.3% for low investment opportunity segments of diversified
firms.
5.2. Cross-subsidization and the value loss in diverstjiedfirms
Another theorized source of the value loss in diversified firms is the subsid-
ization of failing segments. For example, Jensen (1989, 1991, 1993) argues that
the constraints against cross-subsidization in LB0 associations are one source
of their value gains. Moreover, Meyer, Milgrom, and Roberts (1992) contend
that cross-subsidization results in unprofitable lines of business creating greater
56 P.G. Burger, E. Ofek/Journal qf Financial Economies 37 (1995) 3945
Table 6
Regression estimates of the relation between overinvestment and excess value in multi-segment firms
Overinvestment is the level of capital expenditure in excess of depreciation in industries with median
Tobin’s 4 in the bottom quartile. Multi-segment firms that do not have two or more segments with
different two-digit SIC codes are eliminated from the sample. The sample includes multi-segment
firm observations between 1986 and 1991. P-values (in parentheses) are based on the White-adjusted
standard errors.
Dependent variable: Obs.
Actual/imputed value
R2
Intercept
Over- Log of EBIT/
Cwx/
investmenta assets sales
sales
Using asset multiples
1282 - 0.108b - 0.399' - 0.022b 1.211b 0.180
0.074 (0.006)
(0.056)
(O.ow ww
(0.115)
Using sales multiples 1535 - 0.409b - 0.924' 0.025b 0.766b 1.083b
0.077 (0.000) (0.008) (0.001)
v-w
VW
Using EBIT multiples 966
- 0.075
- 0.736b - 0.005
0.601b
0.016 (0.211)
mw
(0.528)
woo)
Using asset multiples:
The natural logarithm of actual value/imputed value where: actual value is
total book value of debt plus market value of equity and imputed value is the sum of the imputed
values of the firm’s segments. Each segment’s imputed value is the segment’s assets multiplied by its
industry median capital-to-assets ratio. Using
sales multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s sales multiplied by its
industry median capital-to-sales ratio.
Using EBIT multiples:
The natural logarithm of actual
value/imputed value with each segment’s imputed value equal to the segment’s EBIT multiplied by
its industry median capital-to-EBIT ratio.
“The sum, for a firm, of capital expenditures in excess of depreciation in segments operating in
industries with median Tobin’s
q
less than 0.76, scaled by total firm sales.
bSignificant at the 1% level.
‘significant at the 10% level.
value losses in conglomerates than they would as stand-alone firms. To investi-
gate this proposition, we use negative cash flow [as measured by EBIT plus
depreciation (EBITD)] as a proxy for poor performance, noting that it will be
a noisy measure of poorly performing segments if managers use discretion in
disclosure to disguise poorly performing units. We examine whether the pres-
ence of negative cash flow in one or more segments has a more negative effect on
diversified firm value than the presence of negative cash flow has on focused firm
value. Such an effect is consistent with poorly performing segments of diversified
firms draining value from other lines of business.
We construct a conditional excess value measure, which uses separate multi-
pliers to impute values for positive and negative EBIT observations (see the
Appendix for additional details). The multipliers are partitioned on the basis of
EBIT rather than cash flow (EBITD) because the small number of negative cash
flow segments makes it very difficult to obtain enough observations within an
P.G. Berger, E. OfeklJournal of Financial Economics 37 (199s) 394.5
57
industry to calculate a median. Because the multipliers are partitioned based on
EBIT, we redo the reported tests based on the presence of negative EBIT (rather
than negative cash flow) in one or more segments. The inferences from this
sensitivity test are identical to those reported below.
Excess value conditional on whether EBIT is negative is used because the
cross-subsidization test assesses whether multi-segment firms with negative cash
flow segments have lower values than multi-segment firms without such seg-
ments. If we conducted this test with the unconditional measure of excess value
used in the rest of the paper, we would expect actual values to be less than
imputed values for the negative cash flow observations if market prices are
a function of discounted cash flows. Conditioning the excess value measure on
the sign of EBIT allows the measure to reflect the effect of the negative cash flow
itself on firm value. The imputed value of segments with positive EBIT is
calculated using the industry median multiple of single-segment firms with
positive EBIT and the imputed value of segments with negative EBIT is
calculated using the industry median multiple of single-segment firms with
negative EBIT. Segments are matched with industry multipliers at the two-digit
level, if possible, and industry medians require at least three single-segment
firms.
In Table 7, the coefficient estimate on the indicator for firms with one or more
negative cash flow segments (for regressions using both the asset and sales
multiplier as dependent variables) is insignificantly different from zero for
focused businesses, but negative and statistically significant for the multi-seg-
ment firms. The insignificant difference from zero for single-segment firms shows
that conditioning the excess value measure on the sign of EBIT results in excess
values that do not diverge significantly from zero based on the sign of the
segment’s cash flow. The coefficient estimates on the negative cash flow indi-
cator for the multi-segment firms are - 11.3% using the conditional asset
multiplier and
- 11.5% using the conditional sales multiplier. Unlike the
single-segment firms, the unrelated multi-segment firms with negative cash flow
segments have a large and significant value loss. We find that, in an average year,
26% of unrelated multi-segment firms have at least one negative cash flow
segment (result not reported). Interpreting this as a 26% probability of engaging
in cross-subsidization in a given year, and multiplying this probability by the
average value loss we document when cross-subsidization does take place, we
estimate the value lost from cross-subsidization by the average unrelated multi-
segment firm at 2.9% to 3%. When the Table 7 tests are repeated using the
completely related multi-segment firms in place of the unrelated multi-segment
firms used to generate the reported results, the coefficient estimates on the
indicator variables are - 4.2% and - 4.5%, respectively. The value lost from
cross-subsidization is thus smaller in related diversifications than in unrelated
diversifications. Finally, the results shown in Table 7 continue to hold when the
pooled regressions are performed separately by year.
58
P.G. Berger, E. Ofek JJournal qf Financial Economics 37 (199.5) 39-6.5
Table 7
Regression estimates of the effect of cross-subsidies from good to bad segments on diversified firm
value
The regressions of excess value on a negative cash flow indicator and control variables are
performed separately on single-segment and multi-segment firms. Multi-segment firms that do not
have two or more segments with different two-digit SIC codes are eliminated from the sample.
P-values (in parentheses) are based on the White-adjusted standard errors,
Obs.
RZ
Intercept
Negative
CF
indicator”
Log of
assets
Capex/
sales
Conditional excess
value using
asset multiples
Single-segment firms 10580
0.061b
0.001
(O.o(w
Multi-segment firms 1973 - 0.1 lob
0.024
(0.001)
Conditional excess value using sales multiples
- 0.013 - O.OOgb 0.07gb
(0.384)
(0.004)
(0.001)
- 0.113b
- 0.004b
0.334b
ww
0.447
(O.ooo)
Single-segment firms 10241 - 0.333b 0.017 0.060b 0.440b
0.054 (0.000) (0.400)
(O.ow ww
Multi-segment firms 2493 - 0.422b
- 0.1 15b
0.041b
l.Oolb
0.096
ww
(0.000)
WJW (O.ow
Conditional excess value using asset multiples:
The natural logarithm of actual value/imputed value
where: actual value is the firm’s total book value of debt plus market value of equity and imputed
value is the sum of the imputed values of the firm’s segments. Each segment’s imputed value is the
segment’s assets multiplied by the relevant industry median capital-to-assets ratio. The imputed
value of segments with positive EBIT is calculated using the industry median multiple of single-
segment firms with positive EBIT, and the imputed value of segments with negative EBIT is
calculated using the industry median multiple of single-segment firms with negative EBIT.
Condi-
tional excess value using sales mtdtiples:
Similar to the calculation explained above, except that the
segment’s sales are multiplied by the industry median ratio of capital to sales.
a An indicator variable that equals one if the firm has at least one segment with negative cash flow
(EBITD).
‘Significant at the 1% level.
The Table 7 results are consistent with cross-subsidization explaining part of
the value loss from diversification. We find that diversified firms with negative
cash flow segments have significantly lower excess values than diversified firms
without such segments, but only after imputing value in a manner that does not
result in a difference in excess value between stand-alone segments with positive
cash flow and stand-alone segments with negative cash flow. Although these
results are consistent with cross-subsidization, they may also result from a sig-
nalling explanation in which the presence of one or more negative cash flow
segments signals the market that the diversified firm’s management is of low
quality.
P.G. Berger, E. OfeklJournal qf Financial Economics 37 (1995) 3945
59
5.3. Debt and taxes as advantages of diversijication
Theory suggests the value loss from diversification may be mitigated by
increased debt capacity and reduced tax payments. Combining businesses with
imperfectly correlated earnings streams increases debt capacity (Lewellen, 1971),
and any increased borrowing that results increases interest tax shields. Diversifi-
cation creates a further tax advantage by allowing the losses of some segments to
be offset contemporaneously against the gains of others, rather than merely
carried forward to future tax years.
Table l’s fourth row shows that, after adjusting for industry differences (see
the Appendix for details on the industry adjustment), multi-segment firms have
mean borrowings that are 1.4% more of their assets than those of single-segment
firms. To examine whether the univariate difference in industry-adjusted lever-
age is driven by factors other than Lewellen’s (1971) coinsurance argument, we
regress the industry-adjusted leverage measure on the multi-segment indicator
and on size, profitability, and growth opportunity controls. We find that
multi-segment firms borrow 1% more of their assets than their segments would
as separate entities. This result suggests that the majority (1%/1.4%) of the
higher borrowing by diversified firms arises from coinsurance. Increasing debt
by 1% of assets does not appear economically significant, suggesting that
diversified firms are unlikely to achieve major savings from higher interest tax
shields.
Table 1 also shows that the difference in median industry-adjusted taxes (see
the Appendix for details of the industry adjustment) between focused and
diversified firms is zero. The mean difference of a 0.1% reduction in taxes paid as
a percentage of sales is significant statistically, although unlikely to be econom-
ically significant. Regressing the industry-adjusted taxes paid on the multi-
segment indicator and the control variables produces a significant (0.10 level)
coefficient estimate (not reported) of - 0.1% on the multi-segment indicator,
consistent with the univariate result. This result indicates that the higher interest
tax shields arising from diversified firms’ higher leverage are not large. In
addition, the results provide little evidence that multi-segment firms save signifi-
cant taxes by offsetting losses from some segments against the profits of the rest
of the firm.
6. Conclusions
We study the effects of diversification on firm value by estimating the value of
a diversified firm’s segments as if they were operated as separate firms. In doing
so, we find that diversification reduces value. We estimate that this value loss
averages 13% to 15% over the 1986-91 sample period, occurs for firms of all
sizes, and is mitigated when the diversification is within related industries. We
60
P.G. Berger, E. Ofek/Journai qf‘Financia1 Economics 37 (1995) 394.5
find additional support for the conclusion that diversification reduces value by
documenting that the segments of diversified firms have lower operating profit-
ability than single-line businesses.
We find that overinvestment is associated with lower value for diversified
firms, and that segments of diversified firms overinvest more than single-line
businesses do. These results are consistent with one source of the value loss
being the greater propensity of multi-segment firms to overinvest. We also find
evidence that suggests the subsidization of poorly performing segments contrib-
utes to the value loss from diversification.
Two potential benefits of diversification are increased interest tax shields
resulting from higher debt capacity and the ability of multi-segment firms to
immediately realize tax savings by offsetting losses in some segments against
profits in others. Our estimate of the tax saving, however, is only 0.1% of sales,
far too small to offset the documented value loss.
Our study confirms recent evidence documenting a significant loss of value in
corporations that followed a diversification strategy during the 1980s. In addi-
tion, we provide evidence regarding possible sources of this value loss. The
evidence that diversification represented a suboptimal managerial strategy
raises questions about the effectiveness of the corporate control and monitoring
mechanisms in place during this period. In addition, the evidence suggests that if
overinvestment and cross-subsidization are properly controlled, a diversifica-
tion strategy can produce small benefits in the form of increased debt capacity
and tax savings.
Appendix
Multiplier estimation of imputed value and
excess value: Eqs. (1) and (2)
illustrate the approach:
(1)
EX VAL = ln( V/Z( V)) ,
(2)
where
I(v)
AZi
= imputed value of the sum of a firm’s segments as stand-alone
firms,
= segment
i’s
value of the accounting item (sales, assets, or EMT)
used in the valuation multiple,
Zndi(
V/AZ),/
= multiple of total capital to an accounting item (sales, assets, or
EMT) for the median single-segment firm in segment i’s industry,
EX VAL
= firm’s excess value,
P.G. Berger, E. OfeklJournal
of
Financial Economics 37 (1995) 3945
61
V
= firm’s total capital (market value of common equity plus book
value of debt),
n
= total number of segments in segment i’s firm.
Eq. (1) shows that the firm’s imputed value is the sum of segment-imputed
values, which are obtained by multiplying an industry median multiplier of total
capital to an accounting item by the segment’s level of the accounting item. Eq.
(2) shows that the firm’s excess value measure is the natural logarithm of the
ratio of the firm’s actual value to its imputed value.
To compute excess value using the sales multiplier, we multiply the industry
median multiple of capital-to-sales for the stand-alone firms in the segment’s
industry by the segment’s sales to obtain the imputed capital of the segment. We
repeat this process for each of the firm’s segments and then sum to obtain the
firm’s imputed value. Finally, we find the firm’s excess value by taking the
natural logarithm of the ratio of actual to imputed value. Extreme excess values
are excluded from the analysis, resulting in the loss of 790 observations (4.9%)
for the sales multiplier. ‘Extreme’ is defined for all three multipliers as natural
logarithms of actual to imputed value above 1.386 or below - 1.386 (i.e., actual
values either more than four times imputed, or less than one-fourth imputed).
The asset multiple imputed values are found in an analogous manner. An-
other issue that arises is that it is much more common for the segment asset
figures from the CIS tape to disagree with the Compustat firm totals than is the
case with sales. The segment sum is usually less than the firm figure, indicating
that the problem arises from unallocated assets. We deal with this problem in
one of two ways: If the sum of the segment asset figures for a firm deviates from
the firm’s asset figure by more than 25%, we exclude the observation from all
analyses using the asset multiples. This results in the exclusion of 1,309 asset
multiplier observations (8.1%). If the deviation is within 25%, we adjust the
firm’s imputed value to reflect the fact that the multipliers have been multiplied
by segment asset figures that are too small or too large. Specifically, the firm’s
imputed value is grossed up or down by the percentage deviation between the
sum of its segments’ assets and total firm assets. The excess value measure based
on asset multiples is then found in the same way as the measure using sales
multiples, with the exclusion of extreme values reducing the asset multiplier
sample by 321 observations (2%).
The earnings before interest and taxes (EBIT) multiple imputed values use the
same adjustment procedures as the asset multiple imputed values, resulting in
the exclusion of 2,189 EBIT multiplier observations (13.6%). One additional
issue that arises with the EBIT measure is how to treat segments with negative
EBITs, since multiplier approaches do not typically assign negative values to
firms with negative earnings. We address this issue by replacing the EBIT
multiplier imputed value with either an EBIT-plus-depreciation (EBITD) multi-
plier imputed value, if positive, or with the segment’s sales multiplier imputed
62 P.G. Berger, E. OfeklJournal of Financial Economics 37 (1995) 3945
value. The exclusion of extreme values reduces the EBIT multiplier sample by
1,031 observations (6.4%).
Revenue-based Herfindahl index:
The Herfindahl index,
H,
is calculated
across n business segments as the sum of the squares of each segment
i’s
sales, Si,
as a proportion of total sales:
H=iSf
i=l
(3)
Thus, the closer
H
is to one, the more the firm’s sales are concentrated within
a few of its segments.
Industry-adjusted profitability measures:
Only single-segment firms are used
in calculating the industry profitability ratios. In addition, to avoid outliers
having an impact on mean profitability, if the profitability measure has a value
above one it is truncated to one, and if it has a value below minus one it is
truncated to minus one. Very few observations require truncation. The profitab-
ility results reported in Section 4.2 are not altered if these observations are
excluded.
Depreciation-adjusted capital expenditures:
Segments are excluded from the
capital spending analysis if the sum of the segments’ capital expenditures for
a firm is not within 3% of the firm’s capital expenditures, or if the segment’s sales
are less than $1 million.
Tobin’s
4: The numerator of Tobin’s
q,
market value of total capital, is
calculated as market value of common equity plus book value of debt. The
denominator, replacement value of assets, is estimated using a modification of
the Lindenberg and Ross (1981) algorithm. Plant and equipment are valued by
setting up an acquisition schedule and adjusting for price-level changes and
depreciation, as suggested by Lindenberg and Ross, while the technological
change parameter of Lindenberg and Ross is, following Smirlock, Gilligan, and
Marshall (1984), assumed to be zero. Specifically, we assume that the value of
plant in 1970 (or the first year with available Compustat data) is equal to book
value. Following Smirlock, Gilligan, and Marshall, we reduce the value of plant
and equipment by 5% each year to compensate for depreciation, and then adjust
it for inflation using the GNP deflator. We then apply the Lindenberg and Ross
formula. For inventories, we apply the Lindenberg and Ross algorithm directly.
Conditional excess value:
The conditional excess values are calculated separ-
ately for positive- and negative-EBIT firms by using separate multipliers. Only
single-segment firms are used in calculating the multipliers. We attempt to
