Page 1

Journal of Business Finance & Accounting, 33(9) & (10), 1508–1534, November/December 2006, 0306-686x

doi: 10.1111/j.1468-5957.2006.00641.x

The Diversification Puzzle: Revisiting

the Value Impact of Diversification

for UK Firms

Edel Barnes and Gael Hardie-Brown*

Abstract:

and has examined industrial rather than geographic diversification. This study exploits the Lang

and Stulz (1994), Berger and Ofek (1995) and Bodnar et al. (1999) methodologies and controls

for form of diversification in assessing value impact for a sample of UK firms, for the period

1996-2000. Using an adjusted value metric that controls for industry effects, we report a significant

diversification, value impact, internationalization, industry, risk

1. INTRODUCTION

Corporatediversificationinbothitsgeographicandindustrialdimensions,hasreceived

substantial attention from both the academic and business communities. However,

the literature has been skewed towards an examination of the industrial form of

diversification and has primarily focused on US firms. The extant theoretical literature

suggests a diversification puzzle. On the one hand the presence of diversification

along either dimension potentially provides a company with increased opportunity

and flexibility, plausible increments to value. On the other hand, again for both forms,

diversification beyond domestic, single activity operations introduces a complexity

to organisational structure that potentially increases agency-related concerns, with a

consequent diminution in value.

Empirical research in the area has identified a negative relationship between firm

value and US corporate industrial diversification (Lang and Stulz, 1994; and Berger

and Ofek, 1995) while Lins and Servaes (1999) document a similar relation across

a range of jurisdictions including Germany, Japan and the UK. A number of studies

*The authors are from the Department of Accounting, Finance and Information Systems, University College

Cork, Ireland. All errors and omissions remain the authors’ sole responsibility. (Paper received March 2006,

revised version accepted May 2006. Online publication August 2006)

Address for Correspondence: Edel Barnes, Department of Accounting, Finance, and Information Systems,

University College Cork, O’Rahilly Building, College Rd. Cork, Ireland.

e-mail: e.barnes@ucc.ie

C ?2006 The Authors

Journal compilationC ?2006 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK

and 350 Main Street, Malden, MA 02148, USA.

VALUE IMPACT OF DIVERSIFICATION FOR UK FIRMS

1509

have focused on identification of the drivers of this association (Rajan, Servaes and

Zingales, 2000; and Brusco and Panunzi, 2000), as well as evaluating its sensitivity to

industrial diversification type (Graham, Lemmon and Wolf, 1999). The value impact

of geographic diversification has received less attention, and again research has been

primarily restricted to the US environment. Such studies have identified a positive

relationship between firm value and geographic diversification for US firms (Errunza

and Senbet, 1981; and Morck and Yeung, 1991), while Reeb, Kwok and Baek (1998)

provideempiricalevidenceofapositiverelationshipbetweengeographicdiversification

and systematic risk. As many firms are both geographically and industrially diversified,

recent US research has examined the impact of geographic and industrial diversifica-

tion in a framework that controls for both forms of diversification (Bodnar, Tang and

Weintrop, 1999). Bodnar et al. provide further US evidence supporting the generally

held view that geographical (industrial) diversification is associated with a positive

(negative) value impact respectively. More recently the existence of a diversification

discount has been called into question (Martin and Sayrak, 2003), with a number of

studies now suggesting that industrial diversification may in fact create shareholder

value and to the extent that diversified firms trade at a discount, this may be due not

to the diversification decision per se, rather to characteristics of firms that make the

diversification choice.

The aim of this paper is initially to exploit an identified gap in the literature, by

exploring the UK value impact of geographic and industrial diversification in a model

that controls for both forms, and to revisit the diversification debate in light of recent

trends in the financial literature that call into question the perceived wisdom that

industrially diversified firms trade at a discount. In addition, evidence to suggest that

value creation occurs through geographic diversification may also be in doubt, in light

of the apparent failure of certain high profile UK firms to successfully implement a

globalexpansionstrategy.Thefirststageofourstudyappliesthemarkettobook(MTB)

methodology supported by Lang and Stulz (1994), updating it to control for both

forms of diversification (Bodnar et al. 1999). However, given evidence that anomalies

(‘puzzles’) with respect to valuation implications of other corporate financial decisions

may be rationalized by inadequate risk measurement (Eckbo, Masulis and Norli, 2000,

among others), our analysis is further refined by the addition of a control variable

for systematic risk (Reeb, Kwok and Baek, 1998). In this broad MTB analysis of the UK

market,consistentwithpriorUSstudies,wereportasignificantgeographic(industrial)

diversification premium (discount) of approximately 12% and 10% respectively.

Notwithstanding much anecdotal evidence that geographic diversification has often

been an ‘unhappy’ experience for UK firms, this stage of our analysis suggests there is

nothing fundamentally different in the broad UK market as compared to the US that

couldexplainsuchfailure.Perhapsitisjustalimitednumberof(possiblyskewing)firms

thatbuckthetrendandprovideevidenceoffailurethatiscontrarytotheUSexperience

1510

BARNES AND HARDIE-BROWN

measure of value impact. Here our findings differ interestingly from the extant AVM

literature in respect of the US market and suggest a UK geographic diversification

discount of 14% and no significant UK industrial diversification value impact.

This subsection of the UK population is focused on a small number of industries

and it seems likely that ‘Industry’ type drives this apparent within-sample conflict,

which accords with our priors that the information content of industry has been vastly

under-exploited in the literature to date. We suggest that in the UK environment,

certain industries may be disproportionately sensitive to demographic and cultural

issues, which become more acute in the case of geographic diversification, for example

the ‘retail’ sector. This may account for the perverse geographic discount that appears

inconsistent with the extant literature that reflects a positive value impact of corporate

geographic diversification, and may also explain the apparent UK anomaly regarding

geographic diversification. It is possible that certain industries are disproportionately

represented in the failing globalisation strategies of certain UK firms so highly

publicisedinthefinancialpress.Thissectorinfluenceisanareanotpreviouslyexploited

in the geographic diversification literature. In consequence our second aim is to speak

directly to this theme by developing a decision choice model, conditional on the

diversification decision, which explains and potentially predicts diversification form

withahighdegreeofaccuracy.Specificallywefindsectoralcharacteristics,togetherwith

Size and contemporaneous Sales to be highly significant predictors of the dimension

of diversification choice for our sample of UK firms.

The remainder of this paper is organised as follows. Section 2 provides a review of

extantliteratureasitrelatestogeographicandindustrialdiversification,whileSection3

discussestheadjustedvaluemethodologyadoptedandspeakstosampleselectionissues.

Section 4 presents the results of our value impact assessment and Section 5 outlines,

implements and discusses our decision choice model and results thereof. Our final

section summarizes and concludes.

2. LITERATURE REVIEW

(i) Geographic Diversification

The history of empirical research into the value impact of geographic diversification

canbetracedbacktotheworkofErrunzaandSenbet(1981),whowerefirsttoexamine

empirically the implications of geographic diversification for firm value. These authors

documentasignificantpositiverelationbetweeninternationalinvolvement(asproxied

by the foreign sales ratio) and excess value, which increases in barriers to capital flow,

suggesting that value impacts may be period specific and associated with changes in

economic indicators over time. Specifically there may be periods when multinational

firms provide ‘below market cost diversification services’ (Bodnar et al. 1999, p. 6).

VALUE IMPACT OF DIVERSIFICATION FOR UK FIRMS

1511

there is little shareholder protection in domestic markets and external financing

is difficult. In such cases, there can frequently be a diversification premium as the

benefits of corporate diversification outweigh the agency-related costs, which contrasts

with evidence of a diversification discount among high-income countries where capital

markets are integrated and well developed.

Reeb, Kwok and Baek (1998) provide evidence of a very significant positive relation-

ship between systematic risk and international diversification, which suggests that the

internationaldiversificationbenefitofreducedcashflowcorrelationmaybedominated

by other negative influences that increase overall firm volatility of returns and in

consequence increase systematic risk. This suggests a value discount, ceteris paribus,

for internationally diversifying firms. Foreign exchange and political risk, increased

asymmetricinformationandmanagementself-fulfillingpropheciesrepresentpotential

negative influences that may enhance systematic risk of internationalised firms. This is

consistent with anecdotal evidence that managers often use a higher hurdle rate when

assessing international projects, in consequence frequently accepting riskier projects,

suchthattheconceptofincreasinglyriskyinternationalprojectsbecomesaself-fulfilling

prophecy.Reebetal.concedehowever,thatanincreaseinsystematicriskassociatedwith

international diversification may equally be due to comparatively higher indigenous

betas within the international countries under review.

(ii) Industrial Diversification

Empiricalresearchintoindustrialdiversificationdatesbacktothe1930s(Coase,1937),

althoughtheimplicationsofdecreasedfocusforcorporatevaluehaveremainedlargely

unexplored until the 1990s. Lang and Stulz (1994) are the first to focus on value, as

distinct from the performance implications of industrial diversification, and identify a

negative relationship between US industrial diversification and value as measured by

both Tobin’s q and Market to Book, which persists even when an industry adjusted

value metric (AVM or chop shop approach) is utilised. Berger and Ofek (1995)

examine the potential source of these value losses and provide evidence that a greater

propensity to over-invest is associated with a lower value for industrially diversified

firms. They also find that subsidization of poorly performing segments contributes to

this diversification value loss, and identify a reduction in the diversification discount

in related, as opposed to unrelated diversification, which supports the relevance of

economies of scale and leveraging of intangible assets such as managerial expertise to

achieve operating efficiency. They conclude that an industrial diversification strategy

could produce small benefits in the form of increased debt capacity and tax savings

given proper controls for over-investment and cross-subsidization.

Rajan,ServaesandZingales(2000)confronttheargumentthatindustrialdiversifica-

tion destroys value through over-investment in value-destroying projects, by exploring

1512

BARNES AND HARDIE-BROWN

and the bright side of internal capital markets’ (Brusco and Panunzi, 2000, p.19).

In the context of US-based studies, Graham, Lemmon and Wolf (1999) explore the

sensitivityoftheindustrialdiscounttodiversificationtype,onthedimensionsoforganic

versus acquired investment and provide empirical evidence to suggest that organic

diversification does not result in a decline in value.

Iforganicdiversificationisrelatedtocorefirmactivities,thesefindingsareconsistent

withBergerandOfek(1995)andwithRumelt(1974),whoarguethatrelatedindustrial

diversification is beneficial from a skills and resource utilisation perspective. However,

Graham et al. (1999) find evidence that diversification through acquisition does result

in such a discount, and show that about half of this value loss derives from the fact that

the acquired firms are already priced at a discount to their industry prior to merger.

This is an interesting point as the majority of prior studies have dismissed self-selection

inindustrialdiversificationwhenfocusingonthecharacteristicsoftheacquiringfirms.1

Infocusingonacquiredfirms,theirstudyidentifiesanalternativeformofself-selection

that can account for some but not all, of the discount.

LinsandServaes(1999)examinetheimpactofindustrialdiversificationaspartofan

internationalstudyencompassingtheUK,GermanyandJapan,anddocumentevidence

of an industrial diversification discount of approximately 15% in the UK for the years

1992and1994acrossasampleofapproximately700UKfirms.2Incontrasttothis paper

their analysis fails to control for geographic diversification in assessing diversification

value impact. This study reports a discount of 10% in a UK model controlling for both

forms of diversification.

(iii) Combined View

Similar theoretical arguments pertain to the relationship between both geographic

and industrial forms of diversification and firm value. Both point to potential agency-

related value losses even though industrial diversification literature also explores the

concept of internal power struggles, which might be considered a further aspect to

agency conflicts. Both literatures identify potential value gains resulting from less well-

correlated cash flows, related tax savings, increased debt capacity or the provision of

a shareholder diversification service. Exploitation of intangibles is also identified as

a potential benefit in both forms of diversification. Geographic diversification speaks

to the theory of internalisation, while industrial diversification speaks to the idea of

related rather than unrelated diversification. Agency and corporate control concerns

appear to be dominant negative influences in industrial diversification, whereas the

literature on geographic diversification identifies exploitation of intangible assets

(internalisation) as the dominant and positive influence on value. Neither literature

centrally considers the relevance of systematic risk however, which is likely to be

positively related to geographic diversification (Reeb et al. 1998). This suggests that

the relationship between value and geographic diversification may be negative if risk-

adjusted internationalising projects fail to generate sufficient incremental cash flow.

1 Campa and Kedia (2001) address centrally the endogeneity of the diversification decision, and provide

evidence of self-selection and of a negative correlation between firm value and a firm’s choice to diversify.

2 This international study includes Japan, where firms primarily report financial results for the year ended

31 March, unlike the UK where firms primarily use a 31 December year-end. Though not directly addressed

by Lins and Servaes (1999), it appears that 1993 was omitted to avoid cross over periods, 31 March being

used as the reporting date for the 1992 Japanese sample.

C ?2006 The Authors

Journal compilationC ?Blackwell Publishing Ltd. 2006

Page 6

VALUE IMPACT OF DIVERSIFICATION FOR UK FIRMS

1513

Bodnar,TangandWeintrop(1999)examinethevalueimpactofdiversificationusing

a combined framework that controls for both forms of corporate diversification, using

the basic models of Errunza and Senbet (1981) and Lang and Stulz (1994). Controls

for both diversification forms are critical in identifying the direction and extent of the

impact of the omitted geographic diversification variable on industrial diversification

discountestimates.Bodnaretal.examineasampleofapproximately7,000USfirmsfor

the period 1984 to 1997, results for which are consistent in direction but different in

magnitude to earlier studies. Specifically they report evidence of a 2.7% value premium

for geographic diversification and a 6% value discount for industrial diversification.

(iv) Recent Trends

Martin and Sayrak (2003) note the conventional wisdom that corporate industrial

diversificationdestroysshareholderwealth,thatdiversifiedfirmssellatavaluediscount

and that these effects are particularly pronounced for conglomerates, yet make the

point that many US corporations both choose to diversify and to remain highly

diversified. This observation is equally valid for the UK where large multinationals

tend to dominate any sample of listed firms. Recent refinements in methodology

have called into question the perceived wisdom that industrial diversification destroys

corporate value. These authors cite Graham et al. (1999), Hyland (1999), Campa and

Kedia (1999) and Villalonga (1999a) all of whom argue that industrially diversified

firms may be inherently different from domestic, single-activity firms before choosing

to diversify, so that subsequent value losses may be associated with characteristics of

diversifying firms rather than to the diversification decision per se. Controlling for

3. METHODOLOGY AND DESCRIPTION OF DATA

(i) Data

Segmental reporting is our basis for identifying those firms that are geographically

(multi-national) and/or industrially diversified (multi-activity) for the purposes of this

study. Geographic and industrial segmental information as mandated by SSAP No. 25

BARNES AND HARDIE-BROWN

from foreign sources. As we focus centrally on the presence rather than extent of

diversification, lack of consistency in annual reporting of business segment data is

not expected to qualitatively impact on either analysis or results.3We are constrained

to examining a relatively short sample period here essentially because segmental

information before 1996 is patchy at best, and we recognize the limitations of this

sampling approach.

Breakdown of company sales (turnover), by geographic and industrial segment for

the five corporate reporting year-ends within the period 31 December, 1996 to 31

December,2000,inclusive,wasobtainedfromDatastream?foroursample.TheSEDOL

number allocated by the LSE was chosen as a unique firm identifier to facilitate data

matching across firms, while for core industry matching, key to the Adjusted Value

Measure (AVM) methodology as applied by Lang and Stulz (1994), the three-digit

FTSE industry sub-sector code was utilized. Consistent with Lins and Servaes (1999)

we confine our sample to those non-financial firms listed on the FTSE All Share Index

at March 2001, breadth of index serving to minimize survivorship bias concerns. We

furtherexcludefirmswithsalesoflessthanstg£30mperannum,theseselectioncriteria

being applied annually so that firms are not necessarily included in our sample every

year.

Our‘broadsample’usedfortheMTBregression,thuscomprises495firms,andwhen

selected from the five-year period 1996 to 2000, contains 1,628 firm year observations,

implying an attrition rate of 35% (34%) for individual firms (firm year observations

overall) respectively.4For the second restricted regression, the AVM methodology

requires that each firm year observation have an industry benchmark within the

single-activity domestic category (NN), a difficult requirement to meet within the

context of the UK market. In any year, only those observations with at least four

firm year observations in the NN category in that year formed the ‘restricted sample’,

further reducing our original population of 1,628 to 337 firm year observations or

approximately 21%. Total sales, earnings before interest and tax (EBIT), long term

debt, total assets, net fixed assets, capital expenditure, research and development

expenditure (DS code 119), cash flow, market value, market to book ratio, volatility,

shareholders funds, number of industry firms and SEDOL codes were obtained from

Datastream?for this ‘restricted sample’ for 1996 to 2000 inclusive. We also obtain a

measure of company age since incorporation for all firms in our sample.5

3 Rajan et al. (2000) highlight the problem of inconsistency in annual reporting of business segment data

and SSAP No. 25 provides scope for strategic segmental changes that are a matter of reporting rather than

fundamental changes in business focus. However, directors are more likely to make strategic changes to the

degree of diversification rather than to report diversification (none) where it does not (does) exist.

4 An included firm may have less than five firm year observations where either the SSAP25 10% threshold

was not met, directors exercised discretion in disclosure on grounds of potential prejudice to firm interests,

the firm was not publicly listed for the 5-year period, the turnover or financial services test was failed or the

firm had not yet filed financial statements for a reporting date falling within the calendar year 2000.

5 Computing variables based on calendar year end data rather than on firm specific year end data can give

rise to a distorting mis-match between market values and accounting data (Bodnar et al. 1999) however, it

ensures comparability in firm value measurement. Bodnar notes that use of firm-specific year-end market

values has no qualitative impact on their findings.

C ?2006 The Authors

Journal compilationC ?Blackwell Publishing Ltd. 2006

Page 8

VALUE IMPACT OF DIVERSIFICATION FOR UK FIRMS

1515

(ii) Market to Book Value Measure (MTB)

Consistent with Lang and Stulz (1994) Market to Book (MTB) is defined as follows:

where MTBi,tis the Market to Book for firm i at time t, MVEi,tis the Market Value of

(4)

where Mf ,tis the Industry Multiplier for FTSE industry code f. at time t, MedianNN

is the median of the observations, in the NN category, MVEi,f ,tis the Market Value of

Equity for NN category firm i, within FTSE industry code f. at time t, BVLi,f ,tis the

Book Value of Total Liabilities for NN category firm i within FTSE industry code f. at

1516

BARNES AND HARDIE-BROWN

range of industry classifications, which is feasible within the limitations of our UK

sample of NN firms. On the one hand this sacrifices the high degree of granularity

achieved for US AVM studies, however, on the other hand it facilitates establishment

of a restricted sample, which allows a focused review of a subset of UK industries.

For both the broad and restricted population of firm year observations each is

split into four sub-groups of domestic single activity (NN), multinational single activity

(YN), domestic multi-activity (NY) and multinational multi-activity (YY) observations.

The MTB and AVM metrics are then computed according to equation (1) and (2)