Accounting and Business Research, Vol. 30. No, 3. pp. 213-225. Summer 2000 213

Accounting for goodwill: an examination offactors influencing management preferencesPelham Gore, Fauziah M. Taib and Paul A. Taylor*

Abstract—This paper investigates factors that influenced the position of managements of UK-listed companies inthe heated debate that surrrounded proposals for a new standard on goodwill accounting, i.e. the factors influencingwhether managements preferred immediate write-off or capitalisation-based approaches. The factors investigatedare derived from contracting cost theory, and include those associated with debt covenant restrictions and profit-based management schemes. They also include non-agency contracting costs. A key feature of the design is that,compared to prior research, we specify more rigorously circumstances where such contracting cost effects are, orare not, likely to be binding. In addition, the paper investigates the effects on management preferences of theirbeliefs about revisions in market perceptions of their companies resulting from changes in goodwill accounting.Our results support certain contracting cost-based hypotheses, but they also indicate that management beliefs aboutchanges in market perceptions of their companies constitute a strong influence on their preferences.

1. IntroductionThis paper reports the results of an investigationof factors that influenced the positions taken bymanagements of UK-listed companies during theheated debate surrounding proposals for a newstandard on goodwill accounting. This was at atime when the UK Accounting Standards Board(ASB) had not decided upon an appropriatemethod and was canvassing various constituencies,including managements, on how it should proceed.One set of factors examined here to explain whymanagements favoured the mandating by the ASBof capitalisation or immediate write-off based ap-

*Pelham Gore and Paul Taylor are lecturers in Accountingand Finance at Lancaster University Management School.Fauziah Taib is lecturer at the Management School, UniversitySains Malaysia. The authors wish to thank members of theDepartment of Accounting and Finance at Lancaster, in parti-cular Ken Peasnell, Steven Young, T.S. Ho, John O'Hanlonand Peter Pope, and delegates at the British AccountingAssociation Annual Conference, Cardiff, March 1996, the JointLancaster/Manchester University Accounting Seminar June1996 and the Financial Accounting and Auditing ResearchConference, London, July 1996; also two anonymous reviewersand the associate editor, Peter Moizer, for comments duringthe development of this paper; Brian Francis and Damon Ber-ridge for advice on statistical issues; and the Research Boardof the Institute of Chartered Accountants in England andWales for the funding of this research. The views expressed andresponsibility for any errors are, however, those of the authors.Correspondence should be addressed to Paul Taylor, Depart-ment of Accounting and Finance, The Management School,Lancaster University, LAI 4YX, UK. Tel: 01524-65201; Fax:01524-847321; E-mail: P.Taylor@lancaster.ac.ukThe final version of this paper was accepted in February 2000.

proaches is based on contracting cost theory. Thisincludes agency costs associated with debt cove-nants and management compensation schemes andother, non-agency, costs. Also examined areinformation effects deriving from managements'beliefs about the impact goodwill accountingwould have on market perceptions of theircompanies.

Our paper adds to prior UK work on this issueby gathering information that is not publicly avail-able about the nature of the bond covenants andmanagement compensation schemes of large UKlisted companies. Our results indicate that certaincontracting cost factors, such as binding gearing-based debt covenant restrictions and applicableprofit-based management compensation plans, doaffect management preferences. They also showthat such preferences are strongly affected bymanagement beliefs about market perceptions.Our identification of such factors can assist pol-icymakers in understanding what may infiuencemanagement positions during future debates overother reporting standards.

The next section of the paper examines thedevelopment of goodwill accounting in the UK.Related studies are then reviewed and hypothesesdeveloped. After describing the data and the var-iables used to operationalise hypothesis testing, weexplain the model used. Our results and relatedsensitivity analyses are then discussed. Finally, asummary and conclusions are presented and im-plications for standard-setters discussed.

214 ACCOUNTING AND BUSINESS RESEARCH

2. The UK debate on goodwill

Accounting for goodwill has long been controver-sial in the UK. There is evidence that, in the ab-sence of regulation, preparers have been dividedover which approach to use. The Survey of Pub-lished Accounts 1979 (Skerratt, 1980:158) showsthat in 1976/77 38% of companies disclosing agoodwill accounting policy carried goodwill atoriginal cost, 42% used immediate write-off to re-serves and 20% capitalised and amortised it. Sub-sequently, the Companies Act 1981 prohibited thefirst of these methods. The Accounting StandardsCommittee's (ASC) SSAP 22 Accounting for Good-will (ASC, 1984) was unusual in allowing alterna-tive treatments: it stated that goodwill should nor-mally be dealt with by immediate write-off directto reserves but also permitted capitalisation andamortisation. However, the vast majority of UKcompanies chose immediate write-off, with the re-sult that their reported profit figures were notadversely affected by amortisation of purchasedgoodwill, but their gearing (leverage) ratios wereseverely increased due to depletion of accountingequity. Weetman and Gray (1991) found that, fora sample of 41 UK companies in 1986, profits un-der UK GAAP were 10.2% higher than under USGAAP because of such goodwill accountingdifferences.

The often dramatic balance sheet effects of im-mediate write-off elicited considerable attention.Fierce debate arose over the capitalisation of in-tangibles such as brand names so as to amelioratesuch adverse gearing effects (Barwise et al, 1989),and the possible concomitant incentives to revalue(upwards) fixed assets were also examined (Linand Peasnell, forthcoming). Considerable argu-ment was also generated over how goodwill itselfshould be accounted for (see, for example, Arnoldet al; 1992). In particular, concerns were aired thatthe SSAP 22 position was at variance with theinternational norm of capitalisation and amortis-ation. Arguments were advanced for capitalisa-tion, predicated, for example, on the desirabilityof international harmonisation and concerns thatimmediate write-off against reserves might giveBritish companies an advantage in making inter-national take-over bids (Choi and Lee, 1991; ASB,1997). The fact that UK managements had discre-tion in choosing their companies' goodwill ac-counting treatment was also a cause for greatconcern.

Proposals by the ASC (Exposure Drafts 47 and52: ASC, 1990a, 1990b) to impose capitahsationand amortisation of goodwill and other intangiblesmet with overwhelming opposition: 93% of cor-porate submissions and 73% of all submissionswere against the goodwill proposals (ASB,1993:12). In December 1993, the ASB, admittingit could not achieve agreement among its board

members on this topic, issued a Discussion Paper(DP) Goodwill and Other Intangibles (ASB, 1993),outlining six alternatives for accounting for good-will, three involving capitalisation and amortisa-tion, and three immediate write-off.' The DPindicated that the ASB wished to adopt a singleapproach to goodwill accounting.

As as result of responses to the DP, the ASBissued a Working Paper (WP), Goodwill and Intan-gible Assets (ASB, 1995). This generally supportedcapitalisation and predetermined life amortisation,with impairment tests for goodwill or intangibles'believed to have indefinitely long lives'. The ASBmade a 'tentative conclusion' that goodwill shouldbe shown as an asset, but still allowed the poss-ibility that it would be shown as a separately dis-closed deduction from reserves. After public hear-ings in September 1995, the ASB issued FRED 12in June 1996 and the FRS 10 in December 1997(ASB, 1996, 1997). Both required capitalisation ofgoodwill and other intangibles, with amortisationover their useful economic lives where these areregarded as of limited duration. These documents,like the WP, included a rebuttable presumptionthat such lives will not exceed 20 years. However,where goodwill or intangibles are regarded as hav-ing an indefinite life and are capable of continuedmeasurement, they are not to be amortised, but areto be written down only if necessary, using annualimpairment tests as a way of recognising the lossof value.

This paper focuses specifically on the goodwillaccounting aspects of the debate and, in particular,on the period between the ASB's 1993 DiscussionPaper and its 1995 Working Paper proposals, atime of uncertainty over which course of action theASB would take. The inclusion by the ASB in theDP of both immediate write-off and capitalisationoptions provides the opportunity to study the pref-erences of UK-listed companies for the accountingtreatment to be adopted in the future standard andalso the factors that were driving their preferences.

3. Development of hypotheses

Studies examining the factors influencing account-ing choice have been difficult to carry out in theUK due to limitations on data availability. Ourwork builds on earlier UK studies that examinedcontracting and other explanations for accountingchoice decisions. Mather and Peasnell (1991), forexample, studied the decision in 1988 by major

'The six alternatives were:i. capitalisation with predetermined life amortisation,ii. capitalisation subject only to impairment tests,iii. a combination of capitalisation approaches i and ii.iv. immediate write-off direct to unspecified reserves.V. immediate write-off to a separate reserve,vi. combination of immediate write-off to reserves with im-pairment tests.

SUMMER 2000 215

British companies to capitalise brands. They foundthat capitalisation was associated with high lever-age and the implications of Stock Exchange Class1 transactions (discussed later). Grinyer et al.(1991) examined factors that influenced the pro-portion of a company's purchase price assigned tonet tangible assets and therefore to goodwill. Theyfound that a contracting cost-related variable, thepost-acquisition gearing ratio, influenced this de-cision. However, such British studies have beenhampered by the lack of publicly availableinformation about many of the variables widelyassumed to drive accounting choices, e.g. companydebt covenant restrictions and proflt-basedmanagement compensation plans. To overcomethis problem we collect data by survey and usethem in conjunction with published flnancial data.

Other UK studies have examined the form ofrestrictions in UK debt covenants, relevant here inour measurement of contracting cost variables.Citron (1992a) examined 25 UK bank loan con-tracts and found that the great majority werebased on rolling (extant) GAAP, and all werebased on consolidated amounts. Non-GAAP ap-proaches were adopted in net worth covenant re-strictions relating to some intangibles and to good-will, but apparently not in the gearing- orproflt-based restrictions that we use in our study.Earlier, Citron (1992b) had found by survey thatbankers claimed they would be understanding indealing with technical breaches caused by changesin GAAP, and that costs imposed on borrowersresulting from such breaches are therefore likely tobe smaller than for other breaches.

Citron (1995) found in 108 UK public debt is-sues (over an earlier period than our study) that,unlike US public debt agreements, UK account-ing-based covenants are mainly afflrmative, i.e. ap-plying at all times while the debt is outstanding.He conjectured (p. 148) that such covenants po-tentially influence managements' accounting policychoices more than where covenants only becomeoperational at the time of specifled actions such asraising new debt. Day and Taylor, in similarly or-ientated research, interviewed 18 officials from fln-ancial institutions (Day and Taylor, 1995), 44 cor-porate treasurers from major UK companies (Dayand Taylor, 1996) and examined a sample of loanagreements. They reported (Day and Taylor, 1994)that gearing and interest cover, used in our paper,were two of the main accounting variables used inbank loan covenants in the UK. Studies, summar-ised in Leuz et al. (1998), found that dividendbased covenants were not widely used in the UK,̂

We now set out the hypotheses to be tested.

^Other non-accounting policy choice research on goodwill,based on US data, has focussed on the relationship betweenthe amount of reported goodwill in that country and the mar-ket capitalisation of equity. For example, Chauvin and Hir-

3.1. Contracting cost hypothesesWe use costly contracting theory to identify fac-

tors, which include both agency and non-agencycosts, expected to underlie management prefer-ences for particular accounting treatments ofgoodwill (Watts and Zimmerman, 1990: 134-5).

3.1.1. Agency contracting costsThe paper focuses on three primary agency re-

lationships discussed in the literature, between 1)management and shareholders, 2) shareholders/managers and debtholders, and 3) the companyand outside regulators ('political costs'), andframes hypotheses in terms of the standard varia-bles used to measure 'opportunistic' managementbehaviour (Watts and Zimmerman, 1990: 138-140).

Based on such a framework, we would predictcompany preferences for goodwill accounting totake the following forms:•Where there are management compensationschemes based on accounting proflts', influencedby the effects of goodwill accounting, there willbe a preference for methods not impactingadversely on reported proflt, i.e. immediate write-off to reserves methods. We therefore predict anegative association between capitalisation-basedapproaches and the existence of such compensa-tion plans.

•Where a company has debt covenant restrictionsbased on balance sheet ratios such as gearing, wehypothesise a positive association between capi-talisation-based preferences and the presence ofsuch restrictions. However, if debt covenant re-

schey (1994), using a simultaneous equations approach, exam-ined interrelationships between goodwill, net income andmarket capitalisation. McCarthy and Schneider (1995) alsoexamined the relationship between equity market capitalisationand goodwill, including income-based explanatory variableswithin an Ohlson-type framework (Ohlson, 1995). They con-cluded that, for US firms, the market does seem to treatacounting goodwill as a value component. Jennings et al. (1996)report a strong cross-sectional association between equity val-ues and recorded goodwill balances for US firms, but only aweak negative relationship between goodwill amortisation andequity values. Barth and Clinch (1996) examine the effects ofinternational accounting differences and their effects on shareprices. Their US findings are consistent with the above studiesand they find that for IJK companies share prices seem to actas if goodwill were an asset, but with a lower coefficient thanother assets.

' in our questionnaire we asked about the existence of ac-counting profit-based management compensation plans and ac-counting-hased debt covenant restrictions, in the full knowlegethat such contracts also contain a wide range of other non-accounting indicators (e.g. see Smith and Warner, 1979 andCitron, 1992b and 1995). We make the assumption that othercontracting provisions do not mask such first order effects inour study. We also specifically adduced information about thetreatment of goodwill in such contracts. Unfortunately, as inmany other studies, we are not able to examine the upper andlower bounds of management compensation contracts dis-cussed by Healy (1985).

216 ACCOUNTING AND BUSINESS RESEARCH

strictions are based on income statement-basedratios, companies will prefer methods that do notreduce profits, i.e., immediate write-off to re-serves methods. Thus we hypothesise a negativeassociation between capitalisation-based prefer-ences and the presence of interest cover-basedrestrictions.

•In politicially visible companies, we use size toproxy for political costs (Watts and Zimmerman,1990: 139) and hypothesise that capitalisation-based approaches would be preferred by largecompanies since they reduce reported profit com-pared to immediate write-off ones (i.e., a positiveassociation between capitalisation-based ap-proaches and size). Size variables are ambiguousto interpret, and can be characterised as controlvariables. We also employ a dummy variable inour models to assess effects unique to utilities,which are more heavily regulated; however, asregulation is complex, we do not predict thedirection of this relationship.We predict that company preferences would

only be affected where such accounting-basedmanagement compensation contracts and debtcovenant restrictions are near contract limits('binding') and based on accounting numbers cal-culated using rolling GAAP; no effect is predictedif they are based on UK GAAP 'fixed' at the timethe agreement was signed.

3.1.2. Non-agency contracting costsThese take many forms, for example transaction

costs (e.g. brokerage costs), renegotiation costsand other differential costs. We examine two po-tentially differential costs incurred by UK-incor-porated listed companies: whether companies havea US quote and whether they have had to considera (UK) Stock Exchange Class 1 transaction. SinceUS GAAP requires capitalisation and amortisa-tion, companies having a US quote already haveto produce accounts on this basis (or a reconcili-ation between US GAAP and their accounts pro-duced under UK GAAP) regardless of the methodthat they choose for UK reporting purposes. If theASB required capitalisation, such companieswould incur fewer extra compliance costs thancompanies not quoted in the US. We would thuspredict a positive association between preferencefor capitahsation-based approaches and having aUS quote.

Listed companies incur transaction costs, suchas obtaining shareholder approval or preparingand circulating certain prospectus-type informa-tion, if they engage as bidders in take-over dealsclassified by the Stock Exchange as Class 1 trans-actions. In periods leading up to the study. Class1 status was attributed to a take over deal if thefair value of the assets being acquired (or disposedof) exceeded the book value of the acquiring (dis-

posing) company by 15% or more (Yellow Book,Section 6). Furthermore, the book value of the ac-quiring company included goodwill as an assetonly if capitalisation was the accounting pohcyused in its financial statements. We would thuspredict that companies planning takeover bidswhich had to consider Class 1 transactions wouldprefer capitalisation of goodwill so as to increasetheir book value relative to the fair value of (prob-ably less acquisitive) target companies and thusminimise the chance of such transactions fallingwithin the relevant Class 1 criteria, i.e. a positiveassociation between a preference for capitalisationand the fact that the company has needed to con-sider Class 1 transactions in planning bids.

3.2. Information effect ('IE') hypothesesAs noted earlier, many of the concerns about

possible changes to goodwill accounting require-ments in the practitioner-orientated accountingliterature seem to be directly related to balancesheet and income statement impacts of goodwillaccounting (e.g. Pearce, 1994). T'o test such alter-native explanations for company preferences, weexamine whether such preferences are infiuencedby finance directors' beliefs about how analystswill interpret any change in reported ratios andprofits, particularly if they have naive beliefs abouthow analysts will react. For example, would com-panies have a preference for the ASB to choosecapitalisation-based alternatives if they believedthat analysts would treat increased gearing as asign of increased riskiness? Also, would they preferimmediate write-off methods if they believed thatreduced reported profits resulting from goodwillamortisation would cause analysts to downgradetheir assessment of the company?

The existence of a strong relationship betweencorporate management preferences and their be-liefs about how analysts will react could be inter-preted as suggestive of a naive view of how themarket operates—see for example the discussionin CoUinson et al. (1993). Sometimes, however,there may be more complex linkages between be-hefs and preferences (O'Keefe and Soloman,1985). For example, suppose companies preferredthe ASB to choose immediate write-off as the soleapproach and expressed the belief that analystsconsider that the immediate write-off of goodwillgives UK firms an international competitive ad-vantage. Such a behef could stem from finance di-rectors having a naive view of the effects of ac-counting numbers and how analysts interpretthem, and preferring immediate write-off becauseof higher reported profits under that approach. Onthe other hand, it could be that they believe thatsuch competitive advantage arises because of thedifferences in profit-based management compen-

SUMMER 2000 217

sation in different countries owing to the differentreported profit figures (Lee and Choi, 1992).

We also examine a number of possible other,information effect (IE)-related, beliefs. We hypoth-esise a positive association between the preferencefor capitalisation-based approaches and finance di-rectors' beliefs that:•capitalising goodwill would enable analysts andother users better to assess their group size relativeto other groups than not doing so;•analysts would consider the company less risky ifgoodwill were capitalised and gearing ratios con-sequently improved;•the increased gearing ratios caused by immediatewrite-off compared to capitalisation and amortis-ation has a negative effect on the way that analystsperceive the company;•immediate write-off distorts the inferences thatanalysts draw in making comparisons betweencompanies that have grown by acquisition andthose that have grown by internal growth.

We hypothesise a negative association betweenthe preference for capitalisation-based approacheswhere finance directors believed that:•analysts would downgrade their rating if theirprofits were lower due to enforced goodwill amor-tisation;•that the widely used UK practice of immediatewrite-off to reserves, which results in no profit andloss account charge, gives UK companies an inter-national competitive advantage compared tosimilar non-UK companies.

4. Sample, data and variables4.1. Data sources and sample selection

Two main data sources were used for our study:Extel's Company Research CD-ROM databaseand the results of a postal questionnaire survey.We utilised the former to compile a dataset giving1994 year-end financial statement information thatenabled the calculation of variables such as gear-ing levels and interest cover. The survey providedcompany data relating to contracting cost hypoth-eses, such as the existence and nature of debt cov-enant restrictions and senior management profit-based compensation contracts. Certain interactionvariables, discussed later, make use of bothsources.

In the survey, we asked for and obtained specificand detailed information on how goodwill wasexplicitly or implicitly treated in contracts—whether its treatment was based on UK GAAP,another form of GAAP or a non-GAAP basis, andfor GAAP treatments, whether they were based onrolling (extant) or frozen (fixed at the time of thecovenant) GAAP. As far as we are aware, this sur-vey, sent out in December 1994, was the first ofthis type to be sent to all large UK hsted com-panies. This information constitutes the only data

source on the treatment of goodwill in debt cove-nant restrictions and management compensationcontracts extensive enough for our purposes.Compared to previous UK studies (e.g.. Citron,1992a, 1995; Day and Taylor, 1995, 1996), basedon smaller samples, we identified a smaller pro-portion of companies having non-GAAP goodwillaccounting methods specified in their contracts.

Standard survey design procedures, includingpre-testing, feedback meetings, and a pilot surveywere utilised. To reduce response bias due to lackof knowledge of the proposals, we sent with thesurvey instrument a one-page summary of theASB's Discussion Paper on goodwill accounting.We also asked about prior familiarity with theproposals and included control questions to checkanswer consistency. Our sampling frame com-prised finance directors of those UK-incorporatedcompanies within The Times 1000 (1994 edition)having a stock market listing. Of the 502 suchcompanies, 21 were used for the pilot study and481 companies in the main survey. Prima faciethese companies were more likely than others tobe involved in take-over activity and hence havematerial levels of goodwill, to have formaliseddebt covenants and management compensationschemes, and to have an active interest in new ac-counting proposals. The survey instrument wasmailed early in December 1994, with a second re-quest four weeks later.'*

The overall response rate of 58.4% (281 com-panies) is high for this type of survey, as is theusable response rate of 44.1% (212 companies).Analysis of the 69 unusable responses reveals nosystematic biasing factors: 52% of these stated thatit was company policy not to take part in surveys,16% did not wish to take part in this particularsurvey, 13% cited time pressures, and the remain-ing 19% comprised a variety of other reasons.Respondent companies were from a broad spreadof business sectors, as shown in Table 1. A Z-testshowed that in 14 out of the 15 industry categoriesthere was no significant difference between theproportion of companies responding and the pro-portion in the population.

Further, to assess response bias we comparedrespondents and non-respondents by industrygrouping, gearing ratio, interest cover ratio, totalassets and turnover. Kruskal-Wallis and Mann-Whitney tests showed no significant difference be-tween the two groups. We also compared re-sponses to first and second mailings usingChi-squared and Mann-Whitney tests. Theseshowed that differences in responses between thegroups were in line with what might be expectedby chance. Financial statement characteristics, in

''A copy of the survey instrument is available from theauthors on request.

218 ACCOUNTING AND BUSINESS RESEARCH

Table 1Industry profile of sample and overall sampling frame

Industry

EngineeringConsumer goodsLight industrialBuilding materials and constructionLeisureChemicals, metals and othermaterialsAgriculture and foodUtilitiesTransport servicesOils, gas, mines and nuclearBusiness servicesFinancialCommunicationsPropertyMiscellaneousTotal

Respondents

Number

422423231922

19129634213

212

Based on The Times 1000 (1994 Edition), Times Books 1994.

%

19.811.310.910.99.0

10.4

9.05.74.22.81.41.90.90.51.4

100.0

Total UK-incorporated listedcompanies in Times 1000

Number

847166555243

2625211211744

21502

%

16.714.113.111.010.48.6

5.25.04.22.42.21.40.80.84.2

100.0

terms of quartiles, of respondents and of the sam-pling frame as a whole are shown in Table 2.

Of individual respondents, 52% were financialdirectors, 23% financial controllers and 23% otheraccountants or managers: only 2% were compara-tively junior or undeclared. Results of tests withand without such junior or undeclared respondentsare very similar: those reported include this lastgroup. All respondents were either professionallyqualified and/or had an extensive background inaccounting.

We are confident that preferences expressed tous reflect a corporate rather than a personal view.

The questionnaire objectives made clear that wesought 'the views and preferences of large com-panies...'. We deliberately targeted finance direc-tors and financial controllers so that responseswould reflect appropriate seniority and decision-making power. Our analysis of respondents illus-trates that our targeting strategy was successful.

5. The model and variables used

Because of the non-normality of the explanatoryvariables and the binary nature of the dependentvariable, PREF, we used logistic regression to

Table 2Characteristics of respondents and

Characteristic

In £m

SalesTotal assetsProfit before interest and taxGearingInterest cover

Source: Extel Company Research

sampling frame

Total (usable)respondents (212)

Quartiles

25%

152.08110.59

5.8317.46%3.28

1994 data.

50% 75%

327.95 1,258.58285.77 1,134.70

22.23 92.3331.07% 48.59%6.17 12.27

Total UK Listed companiesin The Times 1000 (502)

Quartiles

25%

149.42111.63

6.1316.64%3.35

50%

316.57275.64

20.0929.18%

6.56

75%

1,102.41999.45

75.8545.38%12.95

SUMMER 2000 219

Table 3Description of variables used in logistic regression model

Variable

PREF .

GEAR

Description

= 1 if company expresses a first preference for the ASB to adopt capitalisation of goodwill,and = 0 if preference is for immediate write-off.High gearing dummy = 1 if gearing level > upper quartile gearing of sampling framecompanies. The gearing ratio is calculated from Extel definitions as (Totalborrowings+Preference capital)/(Total borrowings+Preference capital+Ordinarycapital+Reserves+Minorities — Intangibles), winsorised at 100%. The Extel definition usesthe immediate write-off basis for measuring gearing. This is justified in this study on theassumption that lenders assess the relative tightness of debt covenant restrictions betweencompanies by using a common UK comparative framework, and only then do theyreformulate such restrictions in terms of the actual goodwill accounting approach used bythe companyHigh gearing x Restriction. Restriction dummy = 1 if company has gearing-based debtcovenant restriction'High gearing x Restriction x Rolling GAAP. Rolling GAAP dummy = 1 if company has arolling GAAP restriction^Low Interest cover = 1 if < lower quartile interest cover of sampling frame companies.Interest cover ratio defined as (Profit before interest and taxation/Interest charge),winsorised at 50 timesLow Interest cover x Restriction. Restriction dummy = 1 if company has interest cover-based debt covenant restriction'Low Interest cover x Restriction x Rolling GAAP. Rolling GAAP dummy = 1 if companyhas a rolling GAAP restriction^= 1 if profit-based management compensation scheme exists'Profit-based management compensation scheme x Rolling GAAP. Rolling GAAPdummy = 1 if company compensation scheme has a rolling GAAP restriction^= 1 if US quote'= 1 if company needed to consider the implications of London Stock Exchange Class 1transactions in the previous five years'= 1 if company is in utilities sectorNatural logarithm of sales= 1 if company felt that capitalising goodwill would enable analysts and other users betterto assess their group size relative to other groups'= 1 if company felt that analysts would downgrade their rating if their profits were lowerdue to enforced goodwill amortisation'= 1 if company felt that analysts would consider the company less risky if goodwill werecapitalised and gearing ratios consequently improved'= 1 if company felt that the UK practice of immediate write-off to reserves, resulting in noprofit and loss account charge, gives UK companies an international competitive advantageover similar US companies'= 1 if company felt that the increased gearing ratios caused by immediate write-offcompared to capitalise and amortise has a negative effect on the way that analysts wouldperceive the company'= 1 if company felt that immediate write-off distorts analysts' inferences when makingcomparisons between companies that have grown by acquisition and those that have grownby internal growth'Eight industry categories based on the London Business School database

'Established by survey question. The information effect variables included an 'unsure' category, which wasincluded in our tests, but not reported in Table 5 to preserve clarity of presentation.^We established through the questionnaire that the relevant debt covenant included goodwill in the restrictioncalculation and did not contain a special treatment for goodwill.

GEARI(l)

GEARI(2)

INTCOV

INTCOVI(l)

INTC0VI(2)

MCOMPMCOMPI

USQCLl

UTILLOGSALESBSZ

DRATE

LRISK

INTCADV

WGEAR

DCOMP

220 ACCOUNTING AND BUSINESS RESEARCH

evaluate the validity of our hypotheses, PREFtakes the value of one if management's first pref-erence is for capitalisation-based approaches (61respondents) and zero for immediate write-off ap-proaches (151 respondents),^ Our model containedboth contracting cost variables and informationeffect (IE) variables.

p p GEARI(l)GEARI(2)+P4 INTCOV+P5 INTCOVI(l)

+p6 INTCOVI(2)+P7 MC0MP+p8 MCOMPI+P5 USQ+pio CLl+Pi, UTIL-i-p,2 LOGSALES+P,3 BSZ+P,4 DRATE+p,5 LRISK-i-p,6 INTCADV-i-p,, WGEAR+P,8 DCOMP+1

The variables are described in Table 3, In testingthe contracting cost issues discussed above we usea model based on categorical variables. In thismodel, main effect variables include the existenceof profit-based management compensation plans(MCOMP); and debt covenant restrictions (forwhich we used one type of balance sheet restric-tion, gearing level (GEAR), which takes the valueone for high-geared firms and zero otherwise, andone income statement-based restriction, interest-cover (INTCOV), which takes the value one forlow interest cover firms and zero otherwise); thepresence of a US quote (USQ); the active consider-ation of Class 1 Transactions (CLl); whether ornot the company is a utility (UTIL); and size(LOGSALES), The variables used to test theinformation effect hypotheses directly relate to thehypotheses outlined earlier and so are not dis-cussed further here,*

The earliest studies on contracting costs and ac-counting choice used only continuous main effectexplanatory variables, such as gearing and interestcover ratios, to measure the effects of debt cove-nant restrictions. Such an approximation formeasuring 'binding' restrictions was criticised asbeing inexact (Duke and Hunt, 1990; Press andWeintrop, 1990), Later studies used simple inter-action effects to refine measurement, by measuring

'The former are alternatives i, ii, and iii in the ASB's 1993DP, as detailed in footnote 1, and the latter alternatives iv, vand vi. Alternative specifications were also tested: see Section6.

'Many of the survey questions upon which our variables arebased allowed responses of 'yes', 'no' or 'unsure'. The 'unsure'category was included in our tests for information effect vari-ables. While this reduces the power of the tests somewhat andtherefore makes our results conservative, it avoids artificiallyconstraining the responses. We only report results for 'yes' ver-sus 'no' in our tables because reporting the 'unsure' resultswould overload with detail an already complicated presenta-tion, and we do not have formal hypotheses for the compari-sons with 'unsure'. (As might be expected though, the resultsfor 'yes' versus 'unsure' were generally in a similar direction tothose for 'yes' versus 'no' but showed less significance.) We didnot include an 'unsure' category for the contracting variablesbecause to do so would have exponentially increased modelcomplexity. There were also fewer such answers.

the effects of gearing or interest cover designated'high' or 'low' relative to medians (e,g, Ayres,1986; Dunne, 1990), Our study contains further re-finements: we use quartiles to reflect more plausi-bly the point at which gearing and interest coverrestrictions become binding, and categorical vari-ables to challenge the assumption implicit in thecontinuous model that the degree of 'bindingness'(and therefore the consequences of breakage ofdebt covenant restrictions) increases in a linearfashion beyond the upper or lower quartile point.We argue that it is more reasonable to assume thatdebt covenant restrictions only become binding be-yond a particular point and not before (though wefind similar results, discussed later, when con-tinuous variables are used). In addition, we focuson higher level interaction terms to target moreexactly firms having binding restrictions of the rel-evant type, i,e, based on rolling GAAP (Mohr-man, 1996),

Our hypotheses and modelling are thus based oninteraction and not main effect terms. For exam-ple, GEAR, with its coefficient, p,, is the incre-mental effect in log-odds ratio terms of high-geared firms over low-geared firms, GEARI(2), thehighest level interaction variable, with coefficientP3, can be interpreted as the incremental effect ofthe simultaneous presence of high gearing and roll-ing GAAP debt covenant restrictions over merelythe presence of high gearing and debt covenantrestrictions which are not based on rolling GAAP,Thus, P1+P2+P3 measures the contracting cost ef-fect for high geared firms with gearing restrictionsbased on rolling GAAP, p,+P2 measures the effectfor the remaining highly geared firms with gearingrestrictions, but not based on rolling GAAP, P,measures the effect for the rest of the firms whichare high geared but do not have gearing debt cov-enant restrictions. We do not model all possiblemain effects and interaction terms, for examplemeasuring the existence of a debt covenant restric-tion per se, as we are only interested in incrementaleffects relating to gearing.

Our main contracting cost hypotheses relateonly to highest-level interaction terms such asGEARI(2), where we predict that P3 will be signifi-cant and positive, i,e, a positive association be-tween management preference for capitalisation-based approaches and gearing level for high gearedcompanies (above the top quartile) where there isalso a rolling-GAAP debt covenant restriction. Ifany of these conditions were not met, for exampleif GAAP were non-rolling, contracting cost theorywould not predict any particular association. Sim-ilarly we predict that the coefficient of INT-C0VI(2) will be negative, i,e, the incremental in-terest cover effect for companies with bindinginterest cover restrictions, having rolling GAAP-based interest cover restrictions and with interestcover below the bottom quartile. We also hypoth-

SUMMER 2000 221

esise that only where there are rolling GAAP-based profit-based management compensationplans will this influence management preferences,so we predict that Pg the regression coefficient ofMCOMPI, will be negative.

Our use of non-interaction (GEAR, INTCOVand MCOMP) and lower level interaction varia-bles (e.g. GEAR (1)) is merely to proxy for omit-ted variables. We also use industry to control foromitted variables correlated with industry effects,e.g. capital intensity and differential financingregimes.

A correlation matrix for the main effect varia-bles is shown in Table 4. A number of significantcorrelations are apparent but the following pointsmay be made. The correlation between interestcover and gearing level is dealt with in our sensi-tivity analyses where we fit models with each in-dividually. As might be expected, industry classi-fication (which includes utilities) is correlated withtwo contracting cost variables. This may lead tosome understatement of the strength of our results,and thus they should be seen as conservative.LOGSALES is correlated with our transactioncost variables. We hypothesise that they operate inthe same direction, and so would not expect theeffect of each to mask the others. We specificallyaddress correlations between contracting cost andinformation effect variables by also fitting separatemodels for each category of variables (see Section6). The correlation between some of the informa-tion effect variables is significant, but this does notprevent many of these variables proving individ-ually significant in the tests reported later, thoughthe level of significance of individual variables maytherefore again be understated.

6. ResultsThe main results are shown in Table 5. The modelhas very good overall explanatory power as shownby its log-likelihood ratio (model chi-square),overall fit and goodness-of-fit statistic. Individualvariables act generally in the directions hypothes-ised and a number of them are highly significant.

Results for individual variables provide con-siderable support for contracting-cost effects.Binding highest level interaction gearing andmanagement compensation contracting effects arein the directions hypothesised, GEARI(2) beingpositive and MCOMPI negative, and are highlysignificant. The binding interest cover contractingeffect, INTCOVI(2) is not significant, although itssign is as predicted. Other, non-agency, contract-ing variables are of the hypothesised sign, exceptthe consideration of Class 1 transactions, CLl,which has the opposite sign to that predicted, but

none are significant.' Certain control variables aresignificant (INTCOV) or marginally significant(INTCOVI(l) and MCOMP).

Four IE variables are also significant at the 5%level or better, and five have the predicted sign.BSZ, the belief that capitalisation and amortisa-tion would enable analysts better to assess the sizeof the company, is very highly significant. Signifi-cant at the 5% level are DRATE, the belief thatcapitalisation and amortisation leads to analystsdown-rating the company, INTCADV, the beliefthat the practice of immediate write-off to reservesgives UK companies an international competitiveadvantage, and DCOMP, the belief that themethod of goodwill accounting distorts compari-sons between companies that have grown byacquisition and those which have grown by inter-nal means.

Sensitivity analysis included fitting a modelwhere gearing and interest cover levels weremeasured as continuous variables consistent, asdiscussed earlier, with previous studies: othersensitivity analyses on the model were conductedas follows:

i) Measuring the tightness of debt covenant re-strictions both in terms of population medians andindustry quartiles.*

ii) Fitting models with debt covenant restrictionsbased on gearing variables only, and then with in-terest cover variables only.

iii) Fitting models based on contracting varia-bles only and information effect variables only, foreach of the sensitivity runs in i) and ii).

v) Restricting contracting and information effectonly models to the same cases as the main model.

iv) Remeasuring the dependent variable PREFusing different groupings of the six goodwill ac-counting alternatives given in the ASB's 1993 DP.'

'One reviewer has pointed out that our results for the US-quote variable, USQ, may have been contaminated to a certainextent by our survey questions not distinguishing between a fullUS quotation and where company securities are traded asAmerican Depository Receipts. With hindsight it would havebeen better to distinguish these. We address this ambiguity byinstead re-running the model using a variable based on anothersurvey question, whether the company produces a reconcilia-tion to US GAAP, such as the SEC form 20-F. This use of theredefined variable produces a similar result to that reported,indicating that not distinguishing between the two types of USlisting is unlikely to have affected the results.

^This was to capture any tendency for lenders to set the val-ues of individual covenants to industry levels (even though theywould also recognise individual company circumstances aswell). This also addresses industry-specific accounting effects.

'The six alternatives are given in footnote 1. Differentiatingcapitalisation (alternatives i, ii and iii) and immediate write-offapproaches (alternatives iv, v and vi), as in our main tests, isundeniably best for capturing balance sheet effects. It also cap-tures income statement effects to a large extent, since alterna-tives i, ii and iii all have income statement effects. However,case vi also has some income statement effect. Therefore thesecond specification used differentiated alternatives i, ii, iii andvi from alternatives iv and v (which have no income statement

222 ACCOUNTING AND BUSINESS RESEARCH

I

I

ase

es

ao

(U <U

s t

o

uz

£2

PQ

t/3

w

OO

IND

u

a

PL(

ou

8

•s o

Var

iabh

oo OS

p —d d

i n —4 —O •^ ro

d d d

t-~ -^ "n CTvOS i-~ >n -H(N — fN —'di d d di

^̂ in ^̂ in r--t-~ ^ -^ <N ooo r<̂ -H ( s csd d d d d

OS O <N - ^ O\I ^ - ,-^ O ^ - O

123

o

*

O O

160

o

057,

o

079

o .

010

o

062

o

099

o

047

o

050^

o

072

o

110^

o

070

o

063

o

. OS I> O '

I 00 •1 O '

I in OS O\. o o oI O O O O O O

VO t^ 00 t^^^ CN fS tnp p —I pd d I

o 00 ̂ • ^ O\t^ (N vi m op p p p p

d d d d d d d

-̂'Osoosooooomr--r-iTfos'*cSTrmm^-O'^^<N'npppp—'pppp-^pddddddddddd

rO'-^Tj-fn(N^OsO^^'^(NV^r ^ os fN ^^ OS r*̂ I*™* f"̂ f*̂ i n o s r*"i(-^^p__p_p__pppdddddddddddd

Q l

OKSDUCJPQQJH^Q

13

—

8cS

igni

V3

13

s?"n

cS

CC3

a60

13

ific

•1

o

lat

goulU

egat

i

a

are

C/5U

3SO

•2

ket

o

PQ

SUMMER 2000 223

Table 5Results of logistic regression of company preference for desired ASB goodwill treatment0 = immediate write-ofQ against contracting

Model:PREE = Po+P, GEAR+P2 GEARI(1)+P3 G

and information effect variables

EARI(2)-i-P4 INTCOV+P5 INTCOVI(1)+P,P, MCOMP+Ps MCOMPI-l-p, USQ+P,o CLl+P,, UTIL+p,j LOGSALES+pp BSZ+PPi, INTCADV+P,, WGEAR-hp.g DCOMP+Z P, IND;

Variable name Hypothesised coefficient sign Coefficient

InterceptGEARGEARI(l)GEARI(2)INTCOVINTCOVI(l)INTCOVI(2)MCOMPMCOMPIUSQCLlUTILLOGSALESBSZDRATELRISKINTCADVWGEARDCOMPIND

No. of casesOverall fit(% of correct classifications)Model tdfGoodness-of-fit

Figures provided for each variable are the**significant at 5% level, ***significant atvalue for such directional hypothesis is for

-5.734-2.760-0.104

8.3753.926

-4.057-0.009

2.746-4.445

+ 0.976+ -0.786

-2.423-H 0.224+ 7.158

-2.461+ 1.970

-2.200+ -1.965+ 1.729

all neg

14892.57%

111.60529

83.394

(1 = capitalisation,

INTCOVI(2)+4 DRATE+P,5 LRISK+

Significance level

(0.525)(0.571)(0.984)(0.003)***(0.028)**(0.065)*(0.499)(0.086)*(0.010)***(0.286)(0.212)(0.391)(0.292)(0.000)***(0.037)**(0.150)(0.025)**(0.101)(0.050)**(0.456)

(0.000)***

regression coefficient and the p value (*significant at 10% level.1% level). Where hypothesised coefficient signs are given, the pone-tailed test: otherwise tests are two-tailed.

Variations in the results of the analyses aregreatly outweighed by their cotnmon features. Theeffect of measuring gearing and interest cover ascontinuous variables hardly changes the overallmodel fit. Contracting cost effects in this model arevery similar to the main model, except that themanagement compensation variable is significantat the 5% rather than the 1% level. IE effect var-iables in the continuous model are more highlysignificant than contracting cost variables.

In nearly all the sensitivity analyses the bindingrolling GAAP gearing restriction variable is highly

effect). An even better specification of major income statementeffects might have been to differentiate alternatives ii and iiifrom the rest, but the smaller number of cases in these cate-gories results in a loss of test power. To the extent that thereis any mis-specification, our results for variables defined in in-come statement terms will be too conservative.

significant in the hypothesised direction, while thebinding rolling GAAP management compensationvariable has the predicted sign in all sensitivitytests and is significant at the 5% level in the greatmajority of runs. The binding rolling GAAP in-terest cover variable is rarely significant. Althoughas expected, the explanatory power of the median-based model is lower, it is one of only two to in-dicate any statistical significance for this variable.It also shows lower significance for some informa-tion effect variables, but BSZ retains its pre-emi-nence. In the industry-based quartile model, thebinding rolling GAAP gearing restriction variableis less significant than in all the other runs.

When interest cover variables are run withoutgearing variables, the model %^ falls, which con-firms our impression that management preferenceswere not greatly driven by interest cover contract-

224 ACCOUNTING AND BUSINESS RESEARCH

ing cost variables. Whether a company is in theutilities sector is generally not significant, givingonly very weak support to the political costs hypo-thesis. Contracting cost-only models have less ex-planatory power than information effects onlymodels, and tend not to have a significant fit,whereas the latter are all highly significant. Shiftsin coefficients between partial models and the mainmodel suggest, as might be expected, omittedvariables.

Restricting the overall models to the same casesproduces no major changes. Changing the defini-tional groupings for the dependent variable resultsin lower significance for the binding rolling GAAPgearing variable and higher significance for therolling GAAP management compensation vari-able, as might be expected, while the maininformation effect variables continued to besignificant.

The overall results of the sensitivity analysistherefore suggest a high level of robustness of themodel to alternative specifications.

7. Summary and conclusions

Our study examines empirically factors infiuencingUK corporate managements' preferences forgoodwill accounting. In so doing, it utilises thewindow of opportunity that arose in 1994 whenthe UK Accounting Standards Board was unde-cided on the issue and was canvassing constituentson how it should proceed towards the develop-ment of a standard. The study is novel in that itcombines financial data derived from a commer-cial database with survey data to specify contract-ing cost hypotheses more precisely than in previ-ous studies. We use interaction variables moreexactly to specify binding factors likely to influ-ence management. Our study is also novel in ex-amining non-contracting factors refiectingmanagement's beliefs about market perceptions oftheir companies.

Our results indicate that management's prefer-ences for accounting for goodwill were influencedby factors consistent with a contracting costframework. Both binding gearing-based debt cov-enant restrictions and profit-based managementcompensation schemes were found to be significantin the direction hypothesised, but interest cover-based covenant restrictions much less so.

In addition, our results show that corporatemanagements' goodwill accounting preferences arealso influenced to a great extent by their beliefsabout how the stock market (represented by finan-cial analysts) would respond to the financial state-ment impact of changes in standards. The mostsignificant IE variables relate to management's be-lief that capitalisation of goodwill would allow an-alysts better to assess the size of their company;that analysts would downgrade their rating if their

company's profits were lower due to enforcedgoodwill amortisation; that immediate write-offmethods gives UK companies an advantage ininternational take-over contests; and that imme-diate write-off causes difficulties for analysts whencompairing companies that grew organically andthose that grew by acquisition. However, the im-pacts of these last three are not as strong as thatrelating to beliefs about the assessment of size.This particular variable could be interpreted in anumber of ways:• in terms of managements having 'omitted asset'concerns, believing that the absence of goodwillfrom their balance sheet gives a false view of thecompany and possibly that it is 'better' accountingto include it;• in functional fixation terms, that managementwished the company to be seen as a larger com-pany; or• in 'market' terms, that management intuitedwhat has been found in US empirical studies (seefootnote 2) that there is correlation between re-ported balance sheet goodwill and market values.Overall, our sensitivity analyses indicate that ourresults are robust to differing model specifications.

Two of the novel features of our study are show-ing the importance of utilising high-level interac-tion terms in modelling contracting cost variables,and of defining bindingness more rigorously, e.g.using quartiles rather than medians. We feel thata similar approach could also usefully be appliedto similar empirical studies in other areas of ac-counting. Further research is needed on a numberof issues: into modelling more explicitly the costsof covenant breaches, the relative magnitudes ofdifferent types of contracting costs, and on gaininga better understanding of the sources and develop-ment of management beliefs.

Our work also has policy implications. It indi-cates that the ASB needs to take account of pot-ential contracting costs imposed on companies bychanges in accounting standards. The Boardshould also be sensitive to managements' concernsabout how such changes will affect the way thatthe market perceives their companies. Failure suf-ficiently to appreciate the perceived importance ofboth issues may result in unwelcome adverse re-action to proposed changes and lobbying againstthem. The factors we find driving preparers' think-ing are relevant as a starting point for its deliber-ations on future accounting issues.

In arriving at these conclusions, the usual ca-veats about using survey data should be noted,though such an approach was the only way to col-lect iniportant non-publicly available data.

ReferencesAccounting Standards Board (1993). Goodwill and Other Intan-gible Assets. Discussion paper. London: ASB

SUMMER 2000 225

Accounting Standards Board (1995). Goodwill and IntangibleAssets. Working paper. London: ASB.

Accounting Standards Board (1996). Goodwill and IntangibleAssets. Financial Reporting Exposure Draft 12 (June). Lon-don: ASB.

Accounting Standards Board (1997). Goodwill and IntangibleAssets. Financial Reporting Exposure Standard 10 (Decem-ber). London: ASB.

Accounting Standards Committee (1984). Aceounting for Good-will SSAP 22 (December). London: ASC.

Accounting Standards Committee (1990a). Accounting forGoodwill. Exposure Draft 47 (February). London: ASC.

Accounting Standards Committee (1990b). Consolidated Ac-counts Exposure Draft 52 (June). London: ASC.

Arnold, J., Eggington, D., Kirkham, L., Macve, R. and Peas-nell, K. (1992). Goodwill and Other Intangibles: TheoreticalConsiderations and Policy Issues. London: ICAEW.

Ayres, F. L. (1986). 'Characteristics of firms electing earlyadoption of SFAS 52'. Journal of Accounting and Economics,8: 143-158.

Barwise, P,. Higson, C , Likierman, A. and Marsh, P. (1989).Accounting For Brands. London: ICAEW.

Barth, M. and Clinch, G. (1996). 'International accounting dif-ferences and their relation to share prices: evidence from UK,Australian, and Canadian firms'. Contemporary AccountingResearch, 13 (1) (Spring): 135-170.

Chauvin, K. W. and Hirschey, M. (1994). 'Goodwill, profita-bility, and the market value of the firm'. Journal of Accountingand Public Policy, 13: 159-180.

Choi, F. D. S. and Lee, C. (1991). 'Merger premia and nationaldifferences in accounting for goodwill'. Journal of Inter-national Financial Management and Accounting, 3: 219-240.

Citron, D. (1992a). 'Accounting measurement rules in UKbank loan contracts and accounting policy choice'. Accountingand Business Research, 23 (Winter): 21-30.

Citron, D. (1992b). 'Financial ratio covenants in UK bank loancontracts and accounting policy choice'. Accounting and Busi-ness Research, 22 (Autumn): 322-335.

Citron, D. (1995). 'The incidence of accounting-based cove-nants in UK public debt contracts: an empirical analysis'. Ac-counting and Business Research, 25 (Summer): 139-150.

Collinson, D. J., Grinyer, J. R. and Russell, A. (1993). Manage-ment's Economic Decisions and Financial Reporting. London:ICAEW.

Day, J. F. S. and Taylor, P. J. (1994). 'Accounting aspects:room for improvement'. Accountancy (June): 129.

Day, J. F. S. and Taylor, P. J. (1995). 'Evidence on the prac-tices of UK bankers in contracting for medium-term debt'.Journal of International Banking Law (September): 394-^01.

Day, J. F. S. and Taylor, P. J. (1996). 'Loan contracting byUK corporate borrrowers'. Journal of International BankingZ-aiv (August): 318-325.

Duke, J. C , and Hunt, III H. G. (1990). 'An empirical analysisof debt covenant restrictions and accounting related debtproxies'. Journal of Accounting and Economics, 12: 45-65.

Dunne, K. M. (1990). 'An empirical analysis of management'schoice of accounting treatment of business combinations'.Journal of Accounting and Public Policy, 9: 111—133.

Grinyer, J. R. Russell, A. and Walker M. (1991). 'Managerialchoices in the valuation of acquired goodwill in the UK'. Ac-counting and Business Research, 22 (Winter): 51-55.

Healy, P. M. (1985). 'The effects of bonus schemes on account-ing decisions'. Journal of Accounting and Economics, 7:85-107.

Jennings, R., Robinson, J., Thompson, III R. B. and Duvall,L. (1996). 'The relationship between accounting goodwillnumbers and equity values'. Journal of Business Finance andAccounting, 23(4) (June): 513-533.

Lee, C. and Choi, F. D. S. (1992). 'Effects of alternative good-will treatments on merger premia: further empirical evidence'.Journal of International Financial Management and Account-ing, (Autumn): 220-236.

Leuz, C , Deller, D. and Stubenrath, M. (1998). 'An inter-national comparison of accounting-based payout restrictionsin the United States, United Kingdom and Germany'. Ac-counting and Business Research, 28(2) (Spring): 111-129.

Lin, Y. C. and Peasnell, K. V. 'Fixed asset revaluation andequity depletion in the UK'. Journal of Business Finance andAccounting. Forthcoming.

McCarthy, M. G. and Schneider, D. K. (1995). 'Market per-ception of goodwill: some empirical evidence'. Accounting andBusiness Research, 26(1) (Winter): 69-81.

Mather, P. R. and Peasnell, K. V. (1991). 'An examination ofthe economic circumstances surrounding decisions to capital-ise brands'. British Journal of Management, 2: 151-164.

Mohrman, M. B. (1996). 'The use of fixed GAAP provisionsin debt contracts'. Accounting Horizons, 10 (3) (September):78-91.

Ohlson, J. A. (1995). 'Earnings book value and dividends insecurity valuation'. Contemporary Accounting Research, 11(2)(Spring): 661-687.

O'Keefe, T. B. and Soloman, S. Y. (1985). 'Do managers be-lieve the efficient markets hypothesis? Additional evidence'.Accounting and Business Research, 15 (Spring): 67-80.

Pearce, C. (1994). 'Wrong in theory and wrong in practice'.Accountancy (June): 97.

Press, E. and Weintrop, J. (1990) 'Accounting-based con-straints in public and private debt agreements: the associationwith leverage and impact on accounting choice'. Journal ofAccounting and Economics, 12: 65-97.

Skerratt, L. (1980). Survey of published Accounts 1979. London:ICAEW.

Smith, C. and Warner J. (1979). 'On financial contracting: ananalysis of bond covenants'. Journal of Financial Economics,7: 117-161.

Times Books (1994). The Times 1000 — 1994. London.Watts, R. L. and Zimmerman, J. (1990). 'Positive accountingtheory: a ten year perspective'. The Accounting Review, 65(1)(January): 131-156.

Weetman, P. and Gray, S. (1991). 'A comparative analysis ofthe impact of accounting principles on profits: the USA versusthe UK, Sweden and the Netherlands'. Accounting and Busi-ness Research, 21(84) (Autumn): 363-380.