determinants of residual income

Upload: niazahmed27

Post on 03-Apr-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/28/2019 Determinants of Residual Income

    1/29

    What Determines Residual Income?

    Author(s): Qiang ChengReviewed work(s):Source: The Accounting Review, Vol. 80, No. 1 (Jan., 2005), pp. 85-112Published by: American Accounting AssociationStable URL: http://www.jstor.org/stable/4093162 .

    Accessed: 27/10/2012 03:22

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

    .

    JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    American Accounting Association is collaborating with JSTOR to digitize, preserve and extend access to The

    Accounting Review.

    http://www.jstor.org

    http://www.jstor.org/action/showPublisher?publisherCode=aaasochttp://www.jstor.org/stable/4093162?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/4093162?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=aaasoc
  • 7/28/2019 Determinants of Residual Income

    2/29

    THE ACCOUNTING REVIEWVol.80,No. 12005pp. 85-112

    W h a t Determines Res idua l I n c o m e ?Qiang ChengThe Universityof British Columbia

    ABSTRACT:This paper investigates the determinantsof residualincome scaled bybook value of equity,i.e., abnormalreturnon equity(ROE),by analyzing he impactofvalue-creation(economic rents)and value-recording conservativeaccounting)proc-esses on abnormalROE.I relyon economic theories to characterizeeconomic rentsand develop an empiricalmeasure-the conservativeaccounting factor-to capturethe effect of conservativeaccounting.As expected, industryabnormalROE ncreaseswith industryconcentration, ndustry-levelbarriers o entry,and industryconservativeaccounting factors. Also as expected, the differencebetween firmand industryab-normalROE ncreases withmarket hare, firmsize, firm-levelbarrierso entry,andfirmconservativeaccounting factors. Integratinghese determinants nto the residualin-come valuationmodel significantlyncreases its explanatorypowerforthe variationnthe market-to-bookratio.Keywords:equityvaluation;he residual ncome valuationmodel;economic rents;con-servativeaccounting.Data Availability:The data used in this study are availablefromsources identified nthe text.JELClassification: D4; G12;M4.

    I. INTRODUCTIONheresidualncomevaluationmodel RIM)provides parsimoniousrameworkink-ing accountingnformationo firmvalue.It has beenwidelyusedto estimateirmvalueand the cost of equity.'TheRIMmodelstates hatfirmvalue s the sumofbookvalue and the presentvalue of expected uture esidualncome.Thus,forecastingfuture esidual ncome s critical o RIM mplementations.riorresearch ften relieson1 See Lee (1999) and Lo andLys (2000) for reviewsof RIM-based tudies.Thispaper s based on my dissertation inishedat the Universityof Wisconsin-Madison. thank wo anonymousreferees,Holly Ashbaugh,Sid Balachandran,oy Begley, WaltBlacconiere,Bob Bowen, LarryBrown,DavidBurgstahler,andraChamberlain,WernerDe Bondt,PeterEaston,RichardFrankel,BruceHansen,ElizabethKeat-ing, MarkKohlbeck,Per Olsson, StephenPenman,MortonPincus,PervinShroff,ShyamSunder,Jim Wahlen,JohnWild,PeterWilson,workshopparticipantst the followinguniversities:BostonCollege,Columbia,Cornell,GeorgiaState, Indiana, owa,Michigan,Minnesota,MIT,Northwestern, niversity f California,Berkeley,Wash-ington(Seattle),WashingtonSt. Louis),Wisconsin-Madison, ndYale,conferenceparticipantst the 2001 AAAAnnualMeeting n Atlanta, he 2001 SummerSymposiumon AccountingResearchat Hong KongUniversityofScience & Technology, ndthe 2001 AccountingSymposium tLondonBusinessSchool,andespeciallyXiaChen,S. P. Kothari,CharlesM. C. Lee, TerryShevlin,andTerryWarfield adviser), or valuablecommentson earlierversionsof this paper.Editor'snote:Thispaperwas acceptedby PatriciaDechow,Editor.

    Submitted January 2003Accepted May 200485

  • 7/28/2019 Determinants of Residual Income

    3/29

    86 Chengthe linear nformation ynamicsproposedby Ohlson(1995) andFelthamandOhlson(1995)for this purpose.However,Dechow et al. (1999) and Myers (1999) find that these infor-mationdynamicscombinedwith the RIM fail to capture irmvalueeffectively.This failurecalls for a betterunderstandingf the determinants f residual ncome (Beaver1999).To this end, this paperfirstinvestigates he impactof value-creationeconomicrents)andvalue-recordingconservative ccounting)processeson residual ncome scaledby bookvalue of equity, .e., abnormal eturnon equity(ROE).It then examineswhether ntegratingeconomic and accountingdeterminants f abnormalROE into the RIM can improveitsabilityto explainfirmvalue.This paperhypothesizes hatabnormalROEincreaseswith economicrentproxiesandconservativeaccounting actors. It relies on theoriesof competition o identifyproxiesforeconomic rents. These proxiesmeasure the extentof imperfectcompetition,underwhichfirms can price productsabove marginalcosts and earnpositive abnormalROE.Conser-vative accounting,such as expensinginvestmentswith futurebenefits,can cause recordedearningsand book value of equityto differfrom economic measuresand lead to nonzeroabnormalROE. To capturethe impact of conservatismon abnormalROE, I develop anempiricalmeasure, he conservativeaccounting actor.This factortakesinto considerationthe impactof conservatismon both net income andbook valueof equity.This paperassumes that abnormalROE follows a first-orderautoregressive AR(1))process and allows the AR(l) parameter i.e., persistence)to vary with economic rentproxiesand conservativeaccountingfactors.Because industryabnormalROEis likely tohavedifferentpersistence rom firmdifferential bnormalROE(thedifferencebetweenfirmandindustryabnormalROE), heempiricalanalysesare conducted eparately t theindustrylevel and the firm evel. The industry-level nalysis suggeststhatthe persistenceof industryabnormalROE increases with industryconcentration,ndustry-levelbarriers o entry,andindustryconservativeaccounting actors.The firm-levelanalysis suggests thatthe persist-ence of firmdifferentialabnormalROEincreaseswith marketshare,firmsize, firm-levelbarrierso entry,and firmconservativeaccounting actors.Likethepersistenceof abnormalROE, ts permanentevel (i.e., the unconditionalmean)might also vary systematicallyacross industriesand firms,especiallyover finitehorizonsanalyzedin empiricalstudies (Mueller 1977; Zhang2000). The abnormalROE analysesare then extendedto explain the permanentevel of abnormalROE. The results suggestthatthe permanentevel increaseswith economicrentproxiesandconservativeaccountingfactors as well.

    Last, this paperconducts market-to-book atio analyses to investigatethe extent towhich economic and accountingdeterminants f abnormalROEimproveRIMimplemen-tations.Integratinghese determinantsnto a model with currentabnormalROE andotherfactors examinedin priorresearch ncreasesthe explanatorypower (i.e., the adjustedR2)by 40 percent and 33 percent for industryand firm market-to-book atio, respectively.Additionalanalyses indicatethat these determinantshave incrementalexplanatorypowerover models analyzed n priorstudies(Fairfieldet al. 1996;BhojrajandLee 2002).This study contributes o the valuationliterature n three ways. First, it provides aframework or forecastingfuture residual income and improvesour understanding f thelinkbetweencurrentandfutureaccountingnformation.t decomposesabnormalROE ntoan economicrentcomponentand a conservativeaccountingcomponent,and finds thatbothcomponentssignificantly mproveforecastsof abnormalROEand the abilityof the RIM

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    4/29

    WhatDeterminesResidualIncome? 87to explain firmvalue.2Such evidence is important or both valuationstudiesand studiesthatrely on the RIMvaluation ramework, uch as value relevancestudies,by suggestinga list of economic and accountingvariablesthatcan capture he cross-sectionalvariationin firmvalue.Second, it provides insights into the impact of economic rents on abnormalROE.Extendingpriorresearch,ncludingLev (1983) andBaginskiet al. (1999), thispaperstudiesboththepersistenceof abnormalROEand its permanentevel. This paperalso uses a morecomprehensiveet of economicrentproxies.This is important iventhatdifferent conomicrent proxies likely correlatewith each other (Porter1980). The analyses suggest that in-dustryconcentration,irmsize, market hare,andbarrierso entryallpositivelyaffectfutureabnormalROE.Third,while priorresearchrecognizesthe impactof conservatismon accountingnum-bers(e.g.,Zhang2000), there s limitedresearchon how to empirically apture uchimpact.This paperdevelops an empiricalmeasure to capturethe impact of conservatismon ab-normalROE for individualfirms. This measurereflects the notion that the conservativeaccountingeffect is contingenton the growthof investmentsrecordedconservatively: heeffect is negativewhen the growth s high andpositive when the growth s low. Empiricalresultsindicate that this measurehelpsforecast future abnormalROEof individual irmsandindustries.Thus,thispaperextendspriorempiricalstudieson conservativeaccounting,such as Lev et al. (1999) and Monahan(1999), which use a portfolio approach o dem-onstrate heeffect of capitalizingR&Dexpenditures n contemporaneouserformancemea-suresandreturns-earningsssociations,amongotherthings.The remainder f the paper s organizedas follows. SectionII analyzeseconomic andaccounting actorsthataffectabnormalROE andpresentshypotheses.SectionIIIdescribessampleselection andvariablemeasurement. ectionIV reports he abnormalROEanalysesand Section V reports he market-to-bookatioanalyses.SectionVI concludes.

    II. HYPOTHESIS DEVELOPMENTAssumingthataccountingsatisfies the clean surplusrelation thechangein bookvalueof equity equals net income minus net dividends)and that the intrinsicvalue of equityequals the presentvalue of future net dividends,one can represent he intrinsic value ofequity(V,)by the residual ncome valuationmodel (RIM):

    V,=BVX

    (1)=(1+r)whereBV,is book value of equity,E,(.) is the expectationat time t, Xa is residual ncome,and r is the cost of equity,which is used to discountall futurepayoffsto equityholders.Residual income denotes the difference between net income (X,) and the requiredreturnfrombeginning-of-period ook value of equity,i.e., Xa = Xt - rBV_1.Basically,the RIMmodel statesthat firmvalue is the sum of book value and the presentvalue of expectedfutureresidual ncome.2 Priorresearchgenerally ocuses either on economic rentsor on conservative ccounting.One exception s aconcurrent tudy,Jansen(2002), which also investigatesboth effects. Comparedo Jansen 2002), this paperuses a morecomprehensiveet of economic rentproxiesand allows both the persistenceand the permanentlevel of firmperformanceo varywith economicrentsand conservative ccountingactors, nsteadof attributingthe permanentevel to conservative ccounting ffects and thepersistenceo economicrents,as done in Jansen(2002).

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    5/29

    88 Cheng

    Accordingly,forecastingresidualincome is critical to implementationsof the RIMmodel. Toimprove heforecasts, hispaper nvestigateshe determinantsf residual ncome.Since analysis of residualincome is potentially subjectto scale problems (Brown et al.1999), this paperfocuses on abnormalROE(ROE"), esidual ncome scaledby book valueof equity,or the differencebetweenROE and the cost of equity.Intermsof abnormalROE,the RIMmodel can be rewrittenas:

    Vt=BVt?+EtLXROTBVtT-1 +

    (1')T=l 0(1+)TThe RIM model holds underboth unbiasedaccountingand conservativeaccounting.Underunbiasedaccounting,a firm'sROEequalsits cost of equityif the firmoperatesunderperfectcompetition.However,if the competition s imperfect,the firmcan chargepriceshigherthan its costs; accordingly,economic rents arise and abnormalROE is no longerzero.Underconservativeaccounting,accountingmeasuresdeviatefromeconomicmeasures.As a result,accountingROE s different rom economicrateof return,anda firm'sabnormalROE can be nonzeroeven if the firmoperatesunderperfectcompetition.Thatis, conser-vativeaccountings also animportant eterminant f abnormalROE(Ohlson1995;Felthamand Ohlson 1995, 1996; Zhang 2000). Below, I first decompose abnormalROE underconservativeaccounting nto a conservativeaccountingcomponentand an economic rentcomponent,andthendiscussthese two components n detail.

    Economic Rents, Conservative Accounting, and Abnormal ROESupposethat the clean surplusrelations under unbiasedand conservativeaccountingare:BVy; BV;_+ X-BV,= BV,_,+ X,- d,,

    where d is net dividends,which are assumedto be the same under the two accountingsystems, i.e., net dividendsare not affectedby accountingpolicies. Accountingnumbersunder unbiasedaccountingare denoted by " ' "Subtractinghe secondequation rom the first one yields:

    BV;- BV,= BV'-, - BV,_ + X; - ,.The difference in book value of equity betweenunbiasedand conservativeaccounting sdefined as the estimatedreserve (ER), as in Penman and Zhang (2002). The estimatedreservecan be interpreted s a measureof how conservativean accountingsystem is. Asin FelthamandOhlson(1996) andGjesdal(1999), conservativeaccountings definedsuchthatthe carryingvalueassignedto investmentsyields anexpected accountingrate of returngreater hanthe internalrateof return.Thus, the estimatedreserve is positive under con-servativeaccountingand the higherthe estimatedreserve,the more conservative he ac-countingsystem.In termsof the estimatedreserve,the above equationcan be writtenas:The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    6/29

    WhatDeterminesResidualIncome? 89

    X - X, = ER, - ER,_1. (2)That is, the differencein net income underthe two accounting systems is the change inthe estimatedreserve. Net income underconservativeaccounting s lower than thatunderunbiasedaccounting f the estimatedreserveincreases,and vice versa.Dividingboth sides of Equation 2) by beginning-of-periodookvalueof equityunderconservativeaccountingand aftersome algebra,one can obtainthe relationbetweencon-servativeaccounting,economicrents,andabnormalROE:

    S(1 + r)ER,_1 - ER, ( ER,ROEa + - + ROEta' 1 + BV~R/1=CAF,ROE 1+B-, . (3)

    See AppendixA for the derivation.This equationsuggests thatabnormalROEundercon-servativeaccountingconsists of two components.The firstcomponent,referred o as theconservativeaccountingfactor (CAF), captures he impact of conservatismon abnormalROE.The secondcomponentcaptures he impactof economicrents-abnormal ROEunderunbiasedaccounting.Conservative Accounting FactorConservatisms "a prudent eaction o uncertaintyo tryto ensurethatuncertainty ndrisksinherentn business situationsareadequately onsidered FASB 1980,2)." Oneprimeexample of conservatismunderU.S. accountingis the expensingof R&D expenditures.Because the benefitsfromR&D expendituresareuncertain, hese investmentsare not rec-ognized as assets and are required o be expensed.3Expensingan investmentwith futurebenefitshas a negative impacton abnormalROEin the investmentperiodand a positiveimpactfor the rest of its useful life. If the growthof investmentss high, the negativeim-pact of contemporaneousnvestmentson currentabnormalROE dominatesthe positiveimpact of past investments,and if the growthis low, the positive impact dominatesthenegative mpact.Thequantitativempactof conservatism n abnormalROE canbe capturedby the conservativeaccounting actordevelopedabove (see Equation 3)):4

    (1 + r)ERt_1 - ER,(ERt_1

    - ER,) + rER,_1CAFB=

    BVV,,- BV,_Basically, CAFis the inversechange in the estimatedreserve(i.e., the net income effect)afteradjusting or the opportunity ost of the beginning-of-periodstimatedreserve(i.e.,the book valueeffect).53 An exceptionto this rule is that firms can capitalizeR&D expenditures ssociatedwith computer oftwaredevelopmentncurred fter he establishment f technological easibility.4 Equation 3) suggeststhat conservative ccountingalso affects abnormalROEthrough he lagged estimatedreserve thesecondterm).Giventhaton average conomicrents ROE

  • 7/28/2019 Determinants of Residual Income

    7/29

    90 ChengThe above discussion andEquation 3) lead to the firsthypothesis(in alternativeorm):H1: AbnormalROEincreases with conservativeaccounting actors.

    Economic RentsCompetitionwithin an industrycan drivepricesdownto marginal osts so thatthe netpresentvalue of an investmentandindustryabnormalROEunderunbiasedaccountingarezero. However, mperfectcompetitioncan lead to economicrents,andthe less competitivean environments, the highereconomic rents are.Neo-classical economic theoriessuggest that the level of competitionand economicrentsdependon industrystructure, .e., industryconcentration nd industry evel barriersto entry.Concentrationan yield economic rents becausefirmsin a concentratedndustrycan collude and set prices above marginalcosts so that they can have positive abnormalROE. As predictedby oligopoly theories, the effectiveness of collusion increases with

    concentrationStrickland ndWeiss 1976). Barriers o entry,on the otherhand,caninduceand sustaineconomicrentsby reducingthreats rom outsidecompetition.Product nnova-tions (temporarymonopolypower),productdifferentiationbuyer nertiaandloyalty),cap-ital intensity(minimumrequired apital),scarceresources,patents,and immobilemanage-ment talents all could result in high barriers o entry (Eatonand Lipsey 1981; Mueller1977).These discussionslead to the second hypothesis(in alternativeorm):H2: AbnormalROE increases with industryconcentrationand industry-levelbarriersto entry.If firms in an industryare not homogeneousas assumedin neo-classical economictheories,then firmcharacteristics,uch as marketshare,firmsize, or firm-levelbarriers oentry,can affectprofitability s well (Brozen1971;Martin1983). Firms with largemarketshareenjoy economic rents because large marketshare leads to economies of scale andincreasedbargainingpowerin an oligopoly. Similarly, argefirmscan enjoy economiesofscale in both the productmarketand the financingmarket.Furthermore,Mueller(1986)arguesthat firm-level barriers o entryare more importanthanindustry-levelbarriers oentryin protecting irms fromoutsidecompetition.These discussionslead to the last hypothesis(in alternativeorm):H3: AbnormalROE increases with marketshare,firmsize, and firm-levelbarriers oentry.One can expressfirmabnormalROE(ROE,,) as the sum of industryabnormalROE(ROEj,)andfirmdifferentialabnormalROE(DROE ):

    ROE= ROEt + DROE ,,wherej stands for the industry o which firm i belongs. Because industryabnormalROEvaries with industrycharacteristics nd firmdifferentialabnormalROE varies with firmcharacteristics,hese two componentsmight have differentpersistence.6Accordingly, his6 This is empirically confirmed based on 219 industries and 1,052 firms with at least 25 consecutive years of dataon abnormal ROE. The difference in abnormal ROE persistence (estimated from AR(1) processes) between afirm and the industry to which it belongs has a wide range. For an industry with a persistence of 0.55 (thesample average), half of the firms in the industry have a persistence higher than 0.65 or lower than 0.33.The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    8/29

    WhatDeterminesResidualIncome? 91

    paperanalyzes them separately.The industry-levelanalysis examines how industrychar-acteristicsaffect industryabnormalROE and the firm-levelanalysis examines how firmcharacteristics ffect firmdifferentialabnormalROE.7III. SAMPLE SELECTIONAND VARIABLEMEASUREMENTSample SelectionAll firm-yearobservations n the period 1976-1997 with available dataon abnormalROE and firm characteristics rom the 1998 Compustatdatabase are used in the ab-normalROEanalyses.The sample periodbegins with 1976, when there areenoughobser-vations to estimate the predicted uture ndustryabnormalROE(which is used in the firmabnormalROE analyses, and the estimationof which is describedin Section IV). Thesampleperiodendswith 1997due to the need for one-year-aheadbnormalROE.Regulatedfirms, ncludingfinancial nstitutions SICcodesbetween 6000 and6999) and utilities(SICcodesbetween4900 and4999), areexcludedbecausetheiroperations remarkedlydifferentfrom otherfirms.8Observationswith book value of equity less than$1 million and thosewith abnormalROEgreater han 100 percentor less than -100 percent(2.2 percentof theoriginalsample)are dropped o reduce the influence of extremevalues. The final sampleconsists of 3,270 observations or the industryabnormalROEanalysesand 22,536 obser-vations for the firm abnormalROE analyses. The sample for the market-to-book atioanalyses is described n Section V.

    Variable MeasurementFirm Abnormal ROEFirmabnormalROEis measuredas the differencebetweenROEandthe industrycostof equity,where ROE is measuredas net income before extraordinarytems available orcommonequitydeflatedby beginning-of-periodook valueof equity.The industrycost ofequityis the sum of the annualizedone-monthT-billyield andtheindustryequitypremium.This latterterm is estimatedfrom the conditionalthree-factormodel as studiedin FamaandFrench(1997). This paperuses the industrycost of equitybecauseit containssmallermeasurement rrors hanthe firmcost of equity(Famaand French1997).Firm CharacteristicsMarket hare s measuredas the ratio of a firm's sales to totalindustry ales. Firmsizeis measuredas thenaturalogarithmof totalassets.Thispaperuses threeproxiesforbarriersto entry:R&D intensity(the ratio of R&D expenditures o sales) for product nnovation,advertisingntensity(the ratio of advertisingexpenditureso sales) for productdifferenti-ation,andcapitalintensity(theratioof depreciation,depletion,andamortization xpensesto sales) for minimumrequired apital.However,high capital ntensitycould also result nlow abnormalROEdue to high capacityadjustment osts (Lev 1983). Thus, the directionof the impactof capital ntensity s unclear.Conservativeaccounting actorsare calculated or R&D and advertising xpenditures.The R&Dreserveis the unamortized ortionof R&D assetsgeneratedby currentandpastR&D expenditures if these expenditures were capitalized. Similarly for the advertisingreserve. R&D assets are amortized using the coefficients reported in Lev and Sougiannis7 Combiningheindustrynd irmanalysesogether, hilestillallowinghetwocomponentso havedifferentpersistence,does not change he inferencesbut makesresultpresentationsumbersome.8 Althoughcertain ndustry haracteristicsreused to explain ndustryperformance,pecialfeaturesof regulatedindustries, uch as regulations n the rate of return, till warrant eparateanalyses.See Nwaeze (2000) for adetaileddiscussionof electricutilities.

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    9/29

    92 Cheng

    (1996). Advertisingassets are amortizedusing an acceleratedmethod(sum-of-yeardigitsover two years),as in PenmanandZhang(2002).Note that economic rents and conservativeaccountingeffects might be related.Forexample, successful R&D investmentscan affect abnormalROE throughboth economicrentsand conservativeaccounting actors.These two effects are conceptuallydistinguish-able though.While abnormalROE due to economicrentsincreaseswith the level of R&Dexpenditures, bnormalROE due to conservativeaccountingdependson both the level andthe growth of R&D expenditures.These effects are empiricallydistinguishablef theirempiricalproxies, R&D intensityand the R&D conservativeaccountingfactor,are nothighly correlated.The correlationbetween these two variables s low (-0.05) in the sam-ple.9Similarargumentsandcorrelation videnceapplyto advertisingexpenditures.'0IndustryAbnormalROE and IndustryCharacteristicsIndustriesareclassified on the basis of three-digitSICcodes. IndustryabnormalROE,economic rent proxies (except concentration),and conservativeaccountingfactors areweightedaveragesof accompanying irmmeasures.The weight for an industrymeasure sthe denominator sedto calculate he accompanying irmmeasure.Forexample,theweightfor industryabnormalROE is beginning-of-period ook value of equity. Industryconcen-tration s proxiedby the Herfindahlndex,calculatedas the sum of squaredmarket haresof all firmsin an industry-the higherthe Herfindahl ndex, the more concentrated heindustry.OtherFactorsPriorresearchsuggests that special items, accruals,and the magnitudeof abnormalROEmightnegativelyaffect the persistenceof firmperformanceBrooksand Buckmaster1976; Fairfieldet al. 1996; Sloan 1996). Accordingly, he empiricalanalyses ncludethesevariables,referred o as otherfactors,as controlvariableswhen analyzingthe incrementalexplanatorypower of economic rent proxies and conservativeaccountingfactors. Thesevariablesare measured n absolutevalues, and special items and total accrualsare scaledby beginning-of-period ook value of equity,as done in priorresearch.Variablemeasurement s summarizedn AppendixB.DescriptiveStatisticsPanel A of Table1 reportsvariablemeans andmediansat the industryandfirm evels.IndustryabnormalROE has a mean of -0.017 and firm abnormalROE has a mean of-0.053." The cost of equity is around0.14. On average,a firmenjoys 5 percentof theproductmarket and has about $130 million of assets (the table reports log values.) TheHerfindahl ndex for an industry s around0.24. R&Dexpenditures rearound1.5 percent9 This low correlationreflects the underlying ow associationbetween the level and the growth of R&Dexpenditures.0oNevertheless, eaders hould nterprethe results with caution.Using R&D and advertising ntensity o proxyfor economic rents is somewhatarbitrary nd might captureconservative ccounting ffects. Also, the CAFmeasure s sensitive o the amortizationmethodand the cost of equityestimation." The reason for the negativemeanabnormalROE couldbe that:(1) the equity premiums overestimated;2)bankruptirms ncluded n theCompustatResearch ile have low ornegativeROE;or (3) firmswithhigh growthin R&Dor advertising xpendituresn the sampleperiodhavenegativeabnormalROEbecauseof conservative

    accounting.Firms n the last two scenariosaregenerally malland havelowerweightin calculating he meanindustry bnormalROE(book-value-weightedverageof firmabnormalROE) han n calculatinghe mean irmabnormalROE (equal-weighted verage).This explainswhy the mean firm abnormalROEis more negativethanthe meanindustryabnormalROE.

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    10/29

    TABLE 1Descriptive Statistics

    Descriptivestatisticson all variables,except the market-to-bookatio andestimatedreserves,arebasedon thanalysis:22,536 firm-year bservationsn the period 1976-1997 for firm level analysesand3,270 industry-industryevel analyses.Descriptivestatisticson the market-to-bookatioand estimatedreservesare based onratioanalysis:19,898 firm-year bservations or firmlevel analysesand3,396 industry-yearbservationsorB for variablemeasurement.Panel A: Descriptive Statisticson Industry and Firm Characteristics

    Industry LevelVariable Mean MedianAbnormalROE -0.017 -0.010Firm differentialabnormalROE NA NACost of equity 0.136 0.140Market-to-bookatio 1.947 1.848Marketshare NA NAFirm size NA NAConcentration 0.236 0.208Barriers o entryR&Dintensity 0.015 0.007Advertising ntensity 0.020 0.012Capital ntensity 0.040 0.032Estimated eserves(scaledby book value of equity)R&D 0.088 0.080Advertising 0.025 0.021Conservative ccounting actorsR&D 0.005 0.001

    Advertising 0.000 0.000Other actorsSpecialitems 0.025 0.012Total accruals 0.174 0.152Magnitudeof industryor firm differentialabnormalROE 0.069 0.050

    h3

    OE:;S;SClg3C~ch

    C;SXaivoot~

  • 7/28/2019 Determinants of Residual Income

    11/29

    TABLE 1 (continued)Panel B: Correlationsbetween VariablesThe upper (lower) triangle reports Spearman correlations between variables at the industry (firm) level. AbnoROE in the upper triangle and firm differential abnormal ROE in the lower triangle. Unless noted, the correl

    Abnormal Market Firm R&D Advertising CapitalROE Share Size Concentration Intensity Intensity IntensityAbnormalROE NA NA 0.03# -0.01# 0.15 -0.14Market hare 0.15 NA NA NA NA NAFirm size 0.18 0.72 NA NA NA NAConcentration NA NA NA -0.02# -0.07 -0.04R&D intensity -0.03 -0.21 0.04 NA -0.10 0.36Advertising intensity 0.03 0.01 0.02 NA 0.05 -0.21Capital intensity -0.18 -0.09 0.16 NA 0.29 -0.12Conservative accounting 0.22 0.05 0.00# NA -0.05 -0.08 0.03factor-R&DConservative accounting 0.15 0.02 0.00# NA -0.02 0.01# 0.01factor-advertisingSpecial items -0.24 0.03 0.12 NA 0.08 0.03 0.16Total accruals -0.09 -0.05 -0.02 NA -0.05 -0.03 0.12#Thecorrelations insignificant t the 0.05 level.

    (hcsx;s~ZC.;scro

    Ca?rOO~I

  • 7/28/2019 Determinants of Residual Income

    12/29

    WhatDeterminesResidualIncome? 95of sales for average industries and 3 percentfor averagefirms; advertising ntensityisaround2 percentandcapitalintensity s around4 percent.At the firmlevel, unrecognizedassets from R&D and advertisingexpenditures qual16 percentand2 percentof book valueof equity,respectively.The ratios are9 percentand3 percentat the industryevel. As discussed n SectionII, conservativeaccountingpoliciesincreaseabnormalROEfor firmswith high growth n investmentsand decreaseabnormalROEfor those with low growth; husthe averageconservativeaccounting actor s aroundzero.While the magnitudeof specialitems is relativelysmall, about2 percentof bookvalueof equity,the absolute value of totalaccruals s about 18 percentof book valueof equity.The absolutevalue of industryabnormalROE and firmdifferentialabnormalROEare onaverage0.07 and0.12, respectively.Panel B of Table 1 reportsthe correlationsbetweenvariables,the uppertriangleforindustrycharacteristics ndthe lower trianglefor firmcharacteristics.Althoughmost cor-relationsare significantlydifferent romzero, they aregenerallysmall and the regressionsarenot subjectto multicollinearity ased on the condition ndex (Belsley et al. 1980).12

    IV. ABNORMAL ROE ANALYSESAssumingthatindustryandfirmdifferentialabnormalROEfollows anAR(1) process,this sectionexamineswhether he AR(1) parameters arywith economicrentproxiesandconservativeaccountingfactorsas hypothesizedabove.13 I first reportresults when eco-nomic rentproxiesandconservativeaccounting actorsareused to explainthe persistenceof industryabnormalROEandfirmdifferentialabnormalROE.I thenextendthe analysesby allowingthe permanentevel of abnormalROEto varyas well. A discussion of sensi-tivitytests concludesthis section.The Industry Abnormal ROE AnalysisThis subsection nvestigateswhether ndustryeconomic rentproxiesand conservativeaccounting actorshelp predictfutureabnormalROEusingthe following regression:

    ROE~+1= o + arROEa + a(xROE X Concentration,+ a2ROE x Barriers to entry,+ a3ROEt x Conservative accounting factors, + ot4ROEtx Otherfactors, + e +*. (4)

    The industrysubscript s dropped rom all variables or ease of notation.Industrycharac-teristicsare mean-adjustedwithineach year so that %o epresents he averagepersistenceof industryabnormalROE.See AppendixB for variablemeasurement.Hypotheses1 and2 imply thatcoefficients on concentration,barriers o entry(except capital intensity),andconservativeaccounting actorsarepositive.12 As reportedn PanelB, firmsize is highly correlatedwith market hare.This high correlation ould lead toinsignificant oefficientson firmsize and market hareeven if theyjointlyhave significant xplanatory ower.

    Accordingly,heregression nalysesuse residualsromyearlyregressions f firmsize on market hare o capturethe incrementalmpactof firmsize beyondmarket hare.The underlying ssumption f this approachs thatmarket hare s a morefundamentalconomicconstruct.13 Unlikethispaper,priorresearchgenerallyassumes heparameterso be constantacrossfirmsand overtime.

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    13/29

    96 ChengTable 2 reports yearly regression results. t-statistics of average coefficients are used totest significance levels (Fama and MacBeth 1973). As in the regressions presented below,all independent variables are winsorized at the 1st and 99th percentiles to reduce the influ-

    ence of extreme values. For comparison purposes, Column A reports results when only thecurrent industry abnormal ROE and other factors are included. The average persistence ofindustry abnormal ROE is 0.79. Consistent with prior research, special items, total accruals,and the magnitude of industry abnormal ROE have negative coefficients.Column B reports the results from the full model. Although the impact of concentrationis insignificant, the impact of barriers to entry is significantly positive,, as expected. Alsoas expected, the R&D conservative accounting factor has a significantly positive impact.The advertising conservative accounting factor has an insignificant coefficient, probably dueto the short useful life of advertising investments (Kothari et al. 2002). Overall, includingindustry economic rent proxies and conservative accounting factors significantly increases

    TABLE2The Industry Abnormal ROE AnalysisROE',, = N, + %oROE" oROEI X Concentration, + o2ROE, x Barriers to entry,+ a3ROE x Conservative ccountingactors, + oa4ROEt Otherfactorst+ Et+ (4)

    ROE'is industryabnormalROE.The industry ubscripts dropped rom all variables or ease ofnotation.See AppendixB for variablemeasurement.The analysis s based on 3,270 industry-yearobservations n the period1976-1997. Regressionsare estimatedyearlyandindustrycharacteristicsaremean-adjusted ithin each year. ReportedcoefficientsandadjustedR2 areyearlyaverages.t-statistics or the averagecoefficientsarereported FamaandMacBeth1973).Waldtests are usedto test whether he explanatory owerof the model specificationn ColumnB is higherthanthatinColumnA. The medianp-valueof yearlyWald tests is reported.Predicted ColumnA Column BSigns Coefficient t Coefficient t

    Intercept ? -0.00 -3.06 -0.00 -4.88ROEa + 0.79 27.29 0.79 25.40ROEaX Concentration + 0.05 0.48ROEax Barriers to entryR&D intensity + 0.62 1.69Advertisingntensity + 2.75 3.04Capital ntensity ? 1.17 2.28ROEaX Conservative ccountingfactorsR&D + 0.69 1.88Advertising + -2.49 0.74ROEaX Other actorsSpecialitems - -3.21 -2.33 -3.45 -2.85Total accruals - -0.31 -1.96 -0.21 -1.66Magnitudeof industryabnormalROE - -0.14 -0.78 -0.07 -0.29AdjustedR2 0.61 0.64p-valueof Wald test 0.0257The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    14/29

    WhatDeterminesResidualIncome? 97theexplanatory ower.TheadjustedR2 increases rom 0.61 in ColumnA to 0.64 in ColumnB, significantat the 0.0257 level based on Waldtests.4',15The Firm Abnormal ROE AnalysisThis subsection nvestigateswhetherfirmeconomic rentproxiesand conservativeac-countingfactorshelp predictfirmabnormalROEusing the following regression:

    ROE+', = o + ROE +I,,++ %oDROE'+ axDROE x Market sharet+ 2DROEa x Firm size, + a3DROEa x Barriers to entry,+ t4DROEa x Conservative accounting factors,

    + ctDROEa x Otherfactors, + e,,,. (5)The firm subscript s dropped rom all variablesexcept the predictedfutureindustryab-normal ROE (ROEj,,+), for which subscript represents ndustry to which the firmbelongs.ROEj,a+ is used to capture he industrycomponentof future irmabnormalROE.16Firmcharacteristics re mean-adjustedwithin each year so that %oepresents he averagepersistenceof firmdifferentialabnormalROE. See AppendixB for variablemeasurement.Hypotheses1 and 3 suggest thatcoefficients on marketshare,firmsize, barriers o entry(exceptcapitalintensity),andconservativeaccounting actorsarepositive.Table3 reports yearly regressionresults. For comparisonpurposes,ColumnA onlyincludespredicted uture ndustryabnormalROE,current irm differentialabnormalROE,andotherfactors,and ColumnB reports he resultsfromthe full model. As expected,theimpactof marketshare,firmsize, and barriers o entryis significantlypositive.The coef-ficient on the R&Dconservativeaccounting actor s significantlypositive as predicted,butthat on the advertising onservativeaccounting actor s insignificant.Overall, ncorporatingfirmeconomiccharacteristicsndconservativeaccounting actorssignificantlyncreases heexplanatorypower.The adjustedR2 increases from 0.30 in ColumnA to 0.34 in ColumnB, significantat the 0.0003 level based on Wald tests.In sum, the results in this and the previoussubsectionsuggest thatas expected,eco-nomic rentproxies and conservativeaccounting actors are positively associatedwith thepersistenceof abnormalROE. Allowing the persistenceto vary with these determinantsincreasesthe explanatorypowerby 5 percentand 13 percentfor industryandfirm differ-entialabnormalROE,respectively.Explaining the Permanent Level of Abnormal ROEWhileallowingthepersistenceof the AR(1)processto vary, he aboveanalysesassumeits intercept,which captures he mean abnormalROE,to be constantacross industriesandfirms.Priorresearchsuggests thatthis assumptionmight not hold. Mueller(1977) argues14 ThispaperusesWald ests to compare xplanatory owerof alternativemodelspecifications ndreportsmedianp-valuesof yearlyWald ests.15 This relativelysmall improvement artlyresults from not controlling or the variationn the permanentevel,as discussed ater.16 Usingrealized uture ndustry bnormalROE s inconsistentwithforecasting urposes.PredicteduturendustryabnormalROE is estimated rom rolling AR(1) regressionsof abnormalROE for individual ndustries.Forexample, o predictanindustry's bnormalROE n 1990,all availableabnormalROEof theindustry pto 1989(since 1963) are used to estimatean AR(1)process.The estimatedparameters ndabnormalROE n 1989 arethenused to predict ndustryabnormalROE n 1990. To ensure he accuracyof estimatingAR(1), at least 12time-seriesobservations re used.

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    15/29

    98 Cheng

    TABLE3The Firm Abnormal ROE AnalysisROE,+1 = 0o+ IROE, +1+ %oDROEa aoDROE' x Market share, + a2DROE'x Firm size, + o3DROEt x Barriers to entry, + a4DROEt

    X Conservative accounting factors, + %aDROEt x Otherfactors, + E,,1 (5)ROEa is firmabnormalROE,ROEf,,+ s predicted uture ndustryabnormalROE,andDROEais firmdifferential bnormalROE. The firmsubscripts dropped rom all variablesexceptROEt+,,for which subscript representsndustry to which the firmbelongs.See AppendixB for variablemeasurement.The analysis s basedon 22,536 firm-year bservationsn the period1976-1997.Regressionsareestimatedyearlyand firmcharacteristics remean-adjustedwithin eachyear.ReportedcoefficientsandadjustedR2areyearly averages. -statistics or the averagecoefficientsarereported FamaandMacBeth1973).Wald tests are used to test whether he explanatory owerofthe model specificationn ColumnB is higherthanthat in ColumnA. The medianp-valueofyearlyWaldtests is reported.

    Predicted Column A Column BSigns Coefficient t Coefficient tIntercept ? -0.04 -5.40 -0.03 -5.27Predicted uture ndustryabnormalROE + 0.78 18.27 0.79 18.28DROEa + 0.62 34.77 0.61 29.44DROEa x Market share + 0.17 2.00DROEa x Firm size + 0.01 1.98DROEa x Barriers to entry

    R&Dintensity + 0.45 2.07Advertisingntensity + 1.07 2.66Capital ntensity ? 0.82 2.41DROEa x Conservative accounting factorsR&D + 0.31 2.30Advertising + -0.38 -0.42DROEa x OtherfactorsSpecialitems - -0.84 -10.33 -0.94 -10.94Total accruals - -0.19 -4.07 -0.15 -3.16Magnitudeof firmdifferential bnormal - -0.20 -2.37 -0.26 -3.01ROEAdjustedR2 0.30 0.34p-valueof Waldtest 0.0003thatbecause firmsspendresources o maintainbarrierso entryand economicrents,theremightbe permanentdifferences n profitability crossindustriesandfirms.Consistentwiththis argument,Mueller(1977) finds that firmprofitabilitydoes not convergeto the com-petitiverate of return.Mueller(1986) furtherarguesthatthese differencesvary systemati-cally with industryand firmcharacteristics. imilarly,Zhang(2000) models the impactofconservativeaccountingon earnings n a steadystateand shows thatthe impactcould benonzeroandvaries with industryandfirmcharacteristicss well.This subsectionextends the above analyses by allowing the mean abnormalROE,re-ferredto as the permanent evel, to vary with economic rent proxies and conservative

    The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    16/29

    WhatDeterminesResidualIncome? 99

    accounting actors. Since the interceptof an AR(1) processis the productof the meanand1 minusthe persistence, he regressionequation or industryabnormalROEbecomes:17ROEt+, = ROEa(1 - p) + pROEt + 4+1,

    where boththe permanentevel (ROEa) and the persistence(p) areexpressedas functionsof economic rentproxiesand conservativeaccounting actors:ROEa = 0 +f3IConcentration, + 32Barriers o entryt

    + 33Conservativeaccounting factorstp = o0+ oaConcentrationt + %c2Barrierso entryt+ o3Conservative accounting factors, + oO40therfactors,.Coefficientsa's capture he impacton the persistence,as in Equation 4), and coefficientsp's capture he impacton the permanentevel.Table 4 reportsyearlyregressionresults.Since the regressionequation s nonlinear nthe coefficients,the nonlinearOLSregressionmethod s used.'8The resultsforpersistenceare similar to those reported n Table2. The impactof industryconcentration n the per-manent evel is significantlypositive, suggestingthat concentratedndustriescan earnpos-itive abnormalROEin the long run.R&D intensityand advertising ntensityhave a sig-nificantlypositive impact,as expected.Capital ntensityhas a negative mpact(marginallysignificant).Conservative ccounting actorshave a positiveimpacton thepermanentevel,as expected. Overall, controllingfor the impact of industrycharacteristicson both thepersistenceand thepermanentevel increases he adjustedR2 from0.61 in ColumnA, whenonly currentndustryabnormalROE andotherfactors are included,to 0.67 in ColumnB,significantat the 0.0012 level.The regressionequationfor firmabnormalROEis constructed imilarlyand Table5reportsregressionresults.The results for persistenceare similarto thosereportedn Table3. The impactof both market shareand firmsize on the permanentevel is significantlypositive. The impact of barriers o entry is mixed-marginally significantlypositive forR&D intensity, nsignificant or advertising ntensity,andsignificantlynegativefor capitalintensity.The impactof the R&D conservativeaccountingfactor is significantlypositive,but thatof the advertisingconservativeaccounting actoris insignificant.Overall,the ad-justedR2 increasesfrom0.31 in ColumnA to 0.37 in ColumnB, significantat the 0.0001level.

    Comparingheresults n Table4 with thosein Table2, orTable5 with Table3, suggeststhatcontrolling or the cross-sectionalvariation n the permanentevel further ncreases heexplanatorypower.Untabulated esultssuggest that the increases are significant.In sum,compared o a model with currentabnormalROE and otherfactors,using economic rentproxiesandconservativeaccounting actorsto explainthe variation n both the persistence17 To see this, one can writean AR(1) processas: ROEt+I= 4o + pROE

  • 7/28/2019 Determinants of Residual Income

    17/29

    100 ChengTABLE4The Industry Abnormal ROE Analysis:ExplainingBoth the Persistenceand the PermanentLevel

    This tablereports he results fromthe following regression hatallowsboth the persistenceofindustryabnormalROEandits permanentevel to varywith economicrentproxiesandconservativeaccounting actors:ROE'+, = ROEa(1 - p) + pROEt + c,+1,

    where the permanent level (ROEF)and the persistence (p) are expressed as:ROE" = 0 + f3,Concentration, + 32Barriers o entry,

    + 33Conservativeaccounting factors,p = to + otConcentration, + a2Barriers to entry,

    + o3Conservative accounting factors, + 4,Other factors,ROE' is industry abnormal ROE. The industry subscript is dropped from all variables for ease ofnotation. See Appendix B for variable measurement. The analysis is based on 3,270 industry-yearobservations in the period 1976-1997. Regressions are estimated yearly using the nonlinear OLSregression method and industry characteristics are mean-adjusted within each year. Reportedcoefficients and adjusted R2 are yearly averages. t-statistics for the average coefficients are reported(Fama and MacBeth 1973). Wald tests are used to test whether the explanatory power of the modelspecification in Column B is higher than that in Column A. The median p-value of yearly Waldtests is reported.

    Predicted Column A Column BSigns Coefficient t Coefficient t

    Explaining the persistence (%, ,,......c,)Intercept + 0.82 27.80 0.82 22.80Concentration + 0.02 0.17Barriers to entryR&D intensity + 2.12 2.84Advertising intensity + 1.92 2.85Capital intensity ? 0.78 2.21Conservative accounting factorsR&D + 0.94 1.97

    Advertising + 1.60 0.46OtherfactorsSpecial items - -2.99 -2.03 -2.77 -5.87Total accruals - -0.25 -1.42 -0.34 -2.84

    Magnitude of industry abnormal ROE - -0.03 -0.92 -0.12 -2.54Explaining the permanent level (Po,Pr1,....3)ntercept ? 0.06 1.23 0.03 1.99Concentration + 0.06 2.25

    Barriers to entryR&D intensity + 2.47 2.06Advertising intensity + 0.92 2.55Capital intensity ? -1.48 -1.58Conservative accounting factorsR&D + 2.61 1.73Advertising + 7.27 2.97

    Adjusted R2 0.61 0.67P-value of Wald test 0.0012The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    18/29

    What Determines Residual Income? 101

    of abnormal ROE and its permanent level increases the adjusted R2 by 10 percent and 20percent for industry and firm abnormal ROE analyses, respectively.Sensitivity TestsIndustryClassificationIndustry classification involves a trade-off between homogeneity of firms in the sameindustry and appropriateclassification for diversified firms. The more detailed the industryclassification, the more homogenous the nondiversified firms in an industry are, but themore problematic is the classification for diversified firms. Unfortunately, there is no agree-ment on appropriateindustry classification. Using four-digit SIC industries or the industryclassification in Fama and French (1997) yields similar results.VariableMeasurement

    First, using total market share of the four largest firms in an industry, instead of theHerfindahl ndex, to proxy for industryconcentrationyields similar results. Second, theresultsbased on alternative stimates of the industryequitypremiumare similar:(1) theaverage historical equity premium for all industries (0.0516, as reportedin Fama and French1997), (2) Capital Asset Pricing Model (CAPM), and (3) the three-factor model. Using thefirm cost of equity estimated from CAPM or the three-factor model also yields similarresults. Third, the impact of market share and firm size could be confounded by diversifi-cation because market share and firm size might be correlated with diversification anddiversification can result in a deviation of firm performance from the industry level (Mueller

    TABLE 5The Firm Abnormal ROE Analysis:ExplainingBoth the Persistence and the PermanentLevel

    This tablereports he resultsfromthe following regression hat allowsboth the persistenceof firmdifferentialabnormalROEand its permanentevel to varywitheconomic rentproxiesandconservativeaccounting actors:ROEa1 = IROE,,+ + DROEa(1 - p) +DROEl + E,+1,wherethe permanentevel (DROEa)and the persistence p) areexpressedas:DROEa = P, + P,1Market hare, + 32Firm ize, + 33Barrierso entry,

    + 34Conservativeaccounting factors,p = %o+ e1Market share, + O2Firmsize, + c3Barriers to entry,+ ao4Conservative ccounting factors, + '5Other factors,

    ROE~ s firmabnormalROE,ROEat+I s predicted uture ndustryabnormalROE,andDROEais firmdifferential bnormalROE. The firmsubscripts dropped romall variablesexceptROE7,,+1,or which subscript representsndustry to which the firmbelongs.See AppendixB for variablemeasurement.The analysis s based on 22,536 firm-year bservationsn the period1976-1997.Regressionsareestimatedyearly using the nonlinearOLS regressionmethodandfirmcharacteristicsremean-adjustedwithineach year. ReportedcoefficientsandadjustedR2areyearlyaverages. -statistics or the averagecoefficientsarereported Famaand MacBeth1973).Waldtests

    are used to test whether he explanatory owerof the modelspecificationn ColumnB is higherthanthat in ColumnA. The medianp-valueof yearlyWaldtests is reported.(continued on next page)

    The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    19/29

    102 ChengTABLE 5 (continued)

    Predicted Column A Column BSigns Coefficient t Coefficient tPredicted uture ndustryabnormalROE(dx) + 0.77 18.87 0.76 19.37Explaining he persistence %o,,...cs)Intercept + 0.60 30.14 0.60 24.24Market share + 0.17 2.30

    Firm size + 0.02 2.00Barriers to entryR&Dintensity + 0.15 1.82Advertisingntensity + 0.14 2.16Capital ntensity ? 0.17 2.35Conservative accounting factorsR&D + 0.58 2.40Advertising + -0.89 -0.79OtherfactorsSpecialitems - -0.84 -9.03 -1.06 -11.17Total accruals - -0.16 -3.73 -0.16 -3.65Magnitudeof firmdifferentialabnormalROE - -0.24 -3.22 -0.25 -4.08

    Explaining he permanentevel (3,o. 134)Intercept ? -0.11 -5.28 -0.11 -4.40Market share + 0.48 4.67Firm size + 0.02 3.45Barriers to entryR&Dintensity + 0.32 1.67Advertising ntensity + -0.31 -0.96Capital ntensity ? -0.92 -2.95Conservative accounting factorsR&D + 0.28 1.66Advertising + 0.72 0.63

    AdjustedR2 0.31 0.37p-valueof Wald test 0.0001

    1977). Controlling or diversificationmeasuredas the sum of squaredsales shares of allindustrysegmentsin the firm)does not changethe inferences.Finally,because the measurement f certainvariables capital ntensity,firmsize, spe-cial items,and totalaccruals) s affectedby accountingchoices, theircoefficientsmightbeconfoundedby conservativeaccountingeffects. Using alternativemeasurementshat areless likely to be affectedby accounting hoices-the ratio of plant,property, ndequipmentto sales as a proxyfor capital ntensity,naturalogarithmof sales for firmsize, andspecialitems andtotalaccrualsscaled by sales-yields similarresults.Model EstimationThis paperdoes not analyzethe impactof financial everageon abnormalROE basedon the notion thatfinancialactivities do not create abnormalprofitsand are accounted orunbiasedly(FelthamandOhlson 1995). Controlling or financial everagein the empiricalanalysesby addinga main effect and an interactionwith currentabnormalROE does notaffect the inferences.Also, estimatingpooled regressionsyields similarresults.The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    20/29

    WhatDeterminesResidualIncome? 103V. MARKET-TO-BOOKRATIO ANALYSESThe abnormalROEanalyses suggest thatthe determinants f abnormalROE can im-proveforecasts of one-period-aheadbnormalROE.However, f the informationdynamic

    of abnormalROEdiffersfrom an AR(1) process,investigatingonly one-period-aheadb-normalROEis insufficient or valuationpurposes.Forinstance,untabulated nalyses sug-gest thatan AR(2) processis moredescriptive han an AR(1) processfor about 35 percentof the industrieswith at least 25 yearsof consecutivedata.Thus,this sectiondirectly inksthe determinants f abnormalROE to firmvalue,i.e., the market-to-bookatio,to examinethe extentto which integratinghese determinantsmprovesvaluation mplementations.Like the abnormalROEanalyses, the market-to-book atioanalyses are conductedatboth the industryandfirmlevels. Marketvalue is measured ourmonthsafterfiscal-year-ends so thatthe capitalmarketscan integrate he informationn the currentyear'sfinancialstatements.It is calculatedas marketvalue at fiscal-year-endsmultipliedby cumulatedreturnsn the fourmonthsafterwards.Overall,3,396 industry-yearbservations nd 19,898firm-yearobservations n the period 1976-1997 are used in the industryand the firmmarket-to-bookatioanalyses,respectively.19 s reportedn PanelA of Table1, the meanof industrymarket-to-bookatios is 1.947 and the mean of firmmarket-to-book atiosis1.944.The Industry Market-to-BookRatio AnalysisThe regressionspecification or the market-to-book atioanalyses is derived fromtheRIMmodel. Scalingboth sides of Equation 1') underunbiasedaccountingby book valueof equity underconservativeaccountingindicatesthat the market-to-book atio: (1) in-creases with the difference in book value of equity between unbiased and conservativeaccounting, .e., the estimatedreserve,(2) decreases with the cost of equity,(3) increaseswiththe growthof book value of equity,and(4) increaseswithfutureabnormalROEunderunbiasedaccounting,which in turnincreases with currentabnormalROE and economicrentproxies.20Thus,the regression or the industrymarket-to-book atioanalysisis:

    MV ? ERB o + 41r + l,Cost of equity, + 32Growth,+ f3Concentration,V, Bvt+ 34Barriersto entry, + aoROEa + caROEa Cost of equityt+ ot2ROEa x Growth, + 3ROEa x Concentrationt + oa4ROEax Barriers to entry, + uotROEaX Otherfactors, + E. (6)

    The industrysubscripts dropped romall variables or ease of notation.The main effectsof industrycharacteristicsare includedto capturethe permanent evel effect and theirinteractionswith abnormalROE are includedto capture he persistenceeffect. Growth sdefinedas the percentagechange in book value of equity.Other variablesare defined asbefore (see AppendixB for variablemeasurement.)19 Themarket-to-bookatioanalyseshaveslightlydifferent amples rom the abnormalROEanalysesbecause heformerrequiredataon marketvalue andthe latterrequiredata on futureabnormalROE.20 See Cheng(2002) for a detaileddiscussionof the link fromdeterminantsf abnormalROEto the market-to-book ratio.Theoretically,onservative ccounting ffects should also be controlled or by adjusting bnormalROE for CAE For simplicity, he analyseshere do not make such adjustments; oing so does not affect theinferences.

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    21/29

    104 ChengTable 6 reportsthe yearly regressionresultsfrom Equation(6) and, for comparisonpurposes,the results from two simpler specifications.21 ColumnA includes only currentindustryabnormalROE and other factors. ColumnB controlsfor conservativeaccounting

    effects by addingestimated reserves. As expected,coefficientson estimatedreserves aresignificantlypositive. The adjustedR2 increasesfrom 0.25 to 0.29, significantat the 0.01level. This suggests that incorporating onservativeaccountingeffects helps explain thevariation n the industrymarket-to-bookatio.ColumnC reports he results from the full model. Althoughcoefficientson estimatedreservesbecomeinsignificantaftercontrolling or other ndustry haracteristics,oefficientson most otherindustrycharacteristics re significant n the predicteddirections.The ad-justed R2 increasesfrom0.29 in ColumnB to 0.35 in ColumnC, significantat the 0.0036level. Overall,these resultssuggestthatincorporatingconomicandaccountingcharacter-istics helps explainthe variation n the industrymarket-to-bookatio.The Firm Market-to-Book Ratio AnalysisThe regression or the firmmarket-to-bookatioanalysisis constructed imilarly:

    MV, MV ERB= + ++ 2ROE B ,Cost of equityt + f2Growth,B V B V I B V+ P3Market share, + P4Firmsize, + PBarries to entry, + %DROE+ arDROEa x Cost of equity, + o2DROEOx Growth, + c3DROE]x Market share, + c4DROE x Firm size, + X5DROEx Barriers to entry, + ct6DROEax Otherfactors, + E,. (7)

    The firm subscript s droppedfrom all variablesexcept lagged industrymarket-to-bookratio (MVj,,_,/BVj,,_,)nd the estimatedindustry permanentabnormalROE (ROEj), forwhich subscript represents ndustry to whichthe firmbelongs. These two industryvar-iables are included to capturethe industry evel market-to-book ariation.22 he industrypermanentabnormalROE is estimatedfrom rollingAR(1) regressionsof abnormalROEfor individual ndustries,as explained n SectionIV. Othervariablesare defined as before(see AppendixB for variablemeasurement.)Table 7 reports he yearly regressionresultsfrom Equation 7) and from two simplerspecificationsto test the incrementalexplanatorypower of conservativeaccountingandother firm characteristicseparately.ColumnA includes lagged industrymarket-to-bookratio, the estimated ndustrypermanentabnormalROE,current irm differentialabnormalROE,and other factors. Column B controlsfor conservativeaccountingeffects by addingestimatedreserves.As expected,coefficientson estimated eservesaresignificantlypositive.The adjustedR2 increases from0.33 to 0.38, significantat the 0.0001 level.ColumnC reports he resultsfromthe full model.As expected,coefficientson the costof equityand on its interactionwith firm differentialabnormalROEaresignificantlyneg-ative.Also as expected,coefficientson growth,market hare,firmsize, R&Dintensity,andtheirinteractionswith firmdifferentialabnormalROEaresignificantlypositive, exceptthat21 All observationswith an absolutevalue of R-studentmeasuregreater han 3 are identifiedas outliersandexcluded Belsleyet al. 1980).Note that he abnormalROEanalysesdo not identifyorexcludeoutliersbecausedoing so requiresnformation bout utureabnormalROE andis inconsistentwith forecasting urposes.22 Using contemporaneousndustrymarket-to-bookatio s inconsistentwithvaluationpurposes.TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    22/29

    What Determines Residual Income? 105

    TABLE6The Industry Market-to-bookRatio AnalysisMVt ER=-ok + C-, + ,3Cost of equity, + 32Growth,+ 33Concentration,B V B V+ 34Barriersto entry, + %aROEw acROEt x Cost of equity, + ot2ROEa

    x Growth, + x3ROEax Concentration, + o4ROEt x Barriers to entry,+ a5ROEt x Otherfactorst + F, (6)MV is marketvalue,BV is book valueof equity,ERis the estimatedreserve,and ROEa s industryabnormalROE.The industry ubscripts dropped rom all variables or ease of notation.SeeAppendixB for variablemeasurement.The analysis s basedon 3,396 industry-yearbservationsnthe period1976-1997. Regressionsareestimatedyearlyandindustrycharacteristicsremean-adjustedwithineach year.Reported oefficientsandadjustedR2 areyearlyaverages. -statistics orthe averagecoefficientsarereported Famaand MacBeth1973).Waldtests areused to test whetherthe explanatory owerof the model specificationn one columnis higherthan thatin the previouscolumn. The medianp-valueof yearlyWaldtests is reported.

    Predicted Column A Column B Column CSigns Coefficient t Coefficient t Coefficient tIntercept + 1.97 8.34 1.92 9.15 2.25 5.22Estimated reservesR&D + 0.61 2.28 -0.40 -0.88Advertising + 2.75 3.23 -0.27 -0.19Cost of equity -1.30 -1.99Growth + 0.01 0.05Concentration + -0.62 -0.11Barriers to entryR&Dintensity + 5.35 2.37Advertisingntensity + 3.75 2.82Capital ntensity ? -1.88 -2.80ROEa + 5.33 15.61 6.50 17.20 5.05 13.09ROEa x Cost of equity - 1.28 0.17ROEa x Growth + 1.06 2.09ROEa x Concentration + 0.41 1.63ROEa x Barriers to entryR&Dintensity + 24.62 1.93Advertisingntensity + 11.22 1.63Capital ntensity ? -9.74 -1.19ROEa x OtherfactorsSpecialitems - -12.04 -2.66 -10.30 -2.46 -7.09 -1.88Total accruals - -5.34 -4.65 -5.12 -4.76 -4.26 -2.46Magnitudeof industry - -5.91 -3.15 -4.77 -2.58 -7.85 -3.65abnormalROEAdjustedR2 0.25 0.29 0.35p-valueof Waldtest 0.0092 0.0036

    firmsize has an insignificantmain effect andmarketsharehas an insignificantnteractioneffect. The adjustedR2 increases further rom 0.38 in ColumnB to 0.44 in ColumnC,significantat the 0.0001 level.TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    23/29

    106 Cheng

    TABLE7The Firm Market-to-BookRatio AnalysisMV, MV ER= V + 1 Bl ++2ROEj

    + (43 +"lCost of equity, + 32Growth,B VBy>1t RE7 BV+ 33Markethare, + 34Firmsize, + P5Barriers to entry, + o-oDROEa olDROEtx Cost of equity, + o2DROEt x Growth, + c%3DROEtx Market share,+ o4DROEt x Firm size, + otDROE x Barriers to entry,+ a6DROEt x Otherfactors, + e, (7)

    MV is marketvalue,BV is book value of equity,ROEjis the estimated ndustrypermanentabnormalROE,ER is the estimatedreserve,andDROEa is firm differentialabnormalROE.Thefirmsubscripts dropped romall variablesexceptMVt,,_-BVj,_,andROEj,for which subscriptrepresentsndustry to whichthe firmbelongs.See AppendixB for variablemeasurement.Theanalysis s basedon 19,898firm-year bservationsn the period1976-1997. Regressionsareestimatedyearlyand firmcharacteristics remean-adjustedwithineach year. Reportedcoefficientsand adjustedR2 areyearlyaverages. -statistics or the averagecoefficientsarereported FamaandMacBeth1973). Waldtest is used to test whether he explanatory owerof the model specificationin one columnis higherthanthat in the previouscolumn. The medianp-valueof yearlyWaldtestsis reported.Predicted Column A Column B Column CSigns Coefficient t Coefficient t Coefficient t

    Intercept + 1.16 7.70 0.97 7.44 1.06 7.96Lagged industrymarket- + 0.36 14.04 0.35 14.51 0.33 12.85to-book ratioIndustrypermanent + 4.31 7.68 3.33 7.06 2.68 6.11abnormalROEEstimated reservesR&D + 1.33 10.93 0.60 4.60Advertising + 1.23 2.26 1.90 2.95Cost of equity -2.74 -3.01Growth + 0.15 2.35Market share + 0.55 2.14Firm size + 0.02 1.32Barriers to entry

    R&D intensity + 3.77 5.32Advertising ntensity + 0.04 0.06Capital ntensity ? 0.14 0.20DROEa + 5.21 16.01 5.27 15.21 5.10 18.67DROEa x Cost of equity - -26.44 -5.85DROEa X Growth + 0.50 3.27DROE x Market share + -3.03 -0.74DROEa X Firm size + 0.40 4.33DROEa x Barriers toentryR&Dintensity + 3.90 3.40

    Advertisingntensity + 7.17 2.96Capital ntensity ? -10.72 -3.25(continued on next page)

    The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    24/29

    What Determines Residual Income? 107

    TABLE7 (continued)Predicted ColumnA ColumnB ColumnC

    Signs Coefficient t Coefficient t Coefficient tDROEa X OtherfactorsSpecialtems - -1.96 -2.71 -2.11 -3.62 -2.36 -2.64Totalaccruals - 0.07 0.21 -0.22 -0.58 -0.84 -2.14Magnitudef firm - -9.24 -10.65 -8.09 -10.42 -6.98 -9.86differentialabnormal OEAdjustedR2 0.33 0.38 0.44p-value f Wald est 0.0001 0.0001

    In sum, the market-to-book atioanalyses suggest thatintegratingdeterminants f ab-normalROEcan significantly mprove he RIM'sexplanatory owerfor themarket-to-bookratio. Additionalanalyses are conducted to investigatehow well this approachperformsrelativeto models used in priorstudies.For example,Fairfieldet al. (1996) find that dis-aggregating arningscanimprove orecastsof futureROE,andBhojrajand Lee (2002) findthatcertain ndustryandfirmcharacteristics,ncluding ndustrymarket-to-bookatio,firmprofit margin,and analysts' growth forecasts,can help explain the firm market-to-bookratio.Theresults,notreported or the sake of brevity,suggestthatintegratingdeterminantsof abnormalROEprovidessignificant ncremental xplanatorypowerover modelsused inpriorresearch, uch as Fairfieldet al. (1996) in predicting utureROEandBhojrajand Lee(2002) in explainingthe firmmarket-to-bookatio.23VI. CONCLUSIONSThis paper nvestigateswhat determinesabnormalROE.It uses industryandfirmchar-acteristicssuggested by economic theoriesto captureeconomicrentsanddevelopsan em-piricalmeasure-the conservativeaccounting actor-to capture heimpactof conservativeaccounting.The empiricalanalysesexamine whether hese determinantselp predict utureabnormalROE andhelp explainthe market-to-bookatio.As expected,industryabnormalROE increaseswith industryconcentration,ndustry-level barriers o entry,andindustryconservativeaccounting actors.Similarly, irm differ-ential abnormalROE increases with marketshare,firm size, firm-level barriers o entry,

    and firmconservativeaccountingactors.Themarket-to-bookatioanalyses urtherndicatethat integrating hese determinants f abnormalROEinto the RIM significantlyncreasesits explanatorypower for the variation n the market-to-book atio. The adjustedR2 in-creasesby 40 percentand 33 percentat the industryand firm evels, respectively, omparedwith a model with currentabnormalROE and other factorssuggestedby priorresearch(i.e., specialitems,totalaccruals,andthemagnitudeof abnormalROE).Additionalanalysesindicatethatthese determinants rovideincrementalexplanatorypowerover models ana-lyzed in priorstudies(e.g., Bhojrajand Lee 2002; Fairfieldet al. 1996).Overall,this papercontributes o the valuation iteratureby investigatingdeterminantsof abnormalROEandby linkingthesedeterminantso firmvalue.Theanalysesdemonstratethe importanceof thesedeterminantsn predicting utureabnormalROE andin explaining23 TheadjustedR2 after ntegrating eterminants f abnormalROE ntothe modelusedin Fairfield t al. (BhojrajandLee) is 0.06 (0.11) higher,a relative ncreaseof 20 percent 31 percent).Both increasesaresignificant tthe 0.001 level based on Wald ests. See Cheng(2002) for details.

    The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    25/29

    108 Chengthe market-to-book atio. These results are importantnot only for improvingvaluationimplementations,but also for studies relying on the RIM valuationframework,by sug-gesting a list of economic and accountingcontrolvariablesthat can capturethe cross-sectional variation n firmvalue.

    APPENDIX ADerivation of the Decomposition Equation (Equation (3))Equation 2) in the paper:X; - X, = ER, - ER,_,, (2)

    linksthe difference n net income(X)under he two accounting ystems(unbiasedaccount-ing andconservativeaccounting) o the change in the estimatedreserve(ER).Accountingnumbers under unbiased accounting are denoted by " ' "Dividingbothsides of Equation 2) by beginning-of-period ook valueof equityunderconservativeaccountingyields:

    x; X, ER, ER,_,BV,_ BV,_1 BV,_1The first termon the left-handside of Equation 2') can be expressedas:

    X - X; BV;_,BV, BVI, BV,_,X; BV,_, + ER,_1BV;_, BV,_1

    = (ROEa'+). 1+ t ,where ROE`' (= X;/BV-_, - r) is abnormal ROE under unbiased accounting. (Recall thatER is definedas: ER, BV' - BV,.)The second termon the left-handside of Equation(2'), by the definitionof abnormalROE,can be expressedas:X,/BVt_, = ROEa + r.Thus,the left-handside of Equation 2') can be writtenas:

    X' ERX X = (ROE + r) 1 + R -(ROE a + r)B gI_ l B V,tI_1B V ,

    (ER ER=ROE I + BVt ++r. 1+B+ -ROEa-rOBV BV_1

    '( E r.-iEIER=ROE' 1 + BV,_ + r BV,_I ROEBV, BV,_,Substitutinghe above equation nto Equation 2') yields:

    ROEat + ER'-I + r ER,_- ROEa ER, -ER1R V" BV,B _BV,11Afterrearranginghe terms of the aboveequation,one can obtainEquation 3):

    TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    26/29

    WhatDeterminesResidualIncome? 109

    ER-ER-ER ( ERROE =r* t + ROEa . 1 +SBV BV~t-l t BVt-

    (1 + E + ROEa' (1 + ERt-B V,1 B V,_=CAF, +ROE'(?+ BER,(1 + r)ER - ERwhere CAF = ( r)ERt1 -ER referred to as the conservative accounting factor.BVI APPENDIX BVariable Measurement

    MeasurementVariable (Compustat tem numbers in parentheses)Panel A: Firm Abnormal ROEFirm abnormal ROE Net income (#237),/Book value of equity

    (#60),_, - cost of equity,Firm differential abnormal ROE Firm abnormal ROE - industry abnormal ROEPanel B: Firm Market-to-BookRatioMarket-to-bookatio (Marketvalue four monthsafterfiscal-year-

    ends),/(Bookvalue of equity),Panel C: Firm Characteristics-Economic Rent ProxiesMarketshare Sales (#12),/[I(Sales), overthe industry]Firmsize Log(Assets(#6),)Barriers to entryR&D intensity R&Dexpenditures #46),/Sales,Advertising intensity Advertising expenditures (#45),/Sales,Capital ntensity Depreciation,depletion,andamortizationexpenses (#14),/Sales,Panel D: Firm Characteristics-Conservative AccountingFactorsConservative accounting factors are calculated as:[(1 + cost of equity,)(estimated reserve),_, - (estimated reserve),]/(Book value ofequity),_1.Two conservative accounting factors are calculated and the estimated reserve for each factor iscalculated as follows:Estimated reserve-R&D The unamortized portion of R&D assetsgenerated by current and past R&Dexpenditures if these expenditures werecapitalized. R&D assets are amortized using thecoefficients reported in Lev and Sougiannis(1996). To reduce data restriction, theamortization eriod s chosento be five years

    and the coefficients are adjusted proportionallysuch that they add up to 1. Coefficients for"other ndustries" re used for an industry,unless coefficients for the industry are reportedby Lev and Sougiannis.TheAccountingReview,January2005

  • 7/28/2019 Determinants of Residual Income

    27/29

    110 ChengMeasurementVariable (Compustatitem numbers in parentheses)

    Estimated reserve-Advertising The unamortized portion of advertising assetsgenerated by current and past advertisingexpenditures if these expenditures werecapitalized. As in Penman and Zhang (2002),advertising assets are amortized using anaccelerated method (i.e., sum-of-year digits overtwo years).

    Panel E: Firm Characteristics-Other FactorsSpecial items ISpecial items (#17),l/(Book value of equity),,otal accruals ITotalaccruals (#18-#308),I/(Book value ofequity),_ aMagnitude of firm differential abnormal ROE IFirmdifferential abnormal ROEIPanel F: Industry VariableMeasurementIndustry measures, except the Herfindahl Index and the magnitude of industry abnormal ROE, areweighted averages of accompanying firm measures. The weight used for each measure is specifiedbelow:

    Industry abnormal ROE Beginning-of-period book value of equity is theweight.Market-to-book ratio Book value of equity is the weight.Barriers to entry Sales is the weight.Conservative accounting factors Beginning-of-period book value of equity is theweight.Special items and total accruals Beginning-of-period book value of equity is theweight.Herfindahl Index Sum of squared market shares in the industry.Magnitude of industry abnormal ROE Ilndustryabnormal ROEI

    a If operating ash flows (#308) is missingor unavailableprior o 1987),it is estimatedas: funds fromoperations#110) - change n current ssets(#4) + change n current iabilities #5) + change n cashandcash equivalents#1) - change n currentportionof long-termdebt(#34). Current ortionof long-termdebtisset to be zero if it is reported s missingin the Compustat.

    REFERENCESBaginski, S. P., K. S. Lorek, G. L. Wilinger, and B. C. Branson. 1999. The relationship betweeneconomic characteristics and alternative annual earnings persistence measures. The AccountingReview 74 (1): 105-120.Beaver, W. H. 1999. Comments on "An empirical assessment of the residual income valuation model."Journal of Accounting and Economics 26 (1-3): 35-42.Belsley, D. A., E. Kuh, and R. E. Welsch. 1980. Regression Diagnostics: Identifying Influential Dataand Sources of Collinearity. New York, NY: John Wiley & Sons.Bhojraj, S., and C. M. C. Lee. 2002. Who is my peer? A valuation-based approach to the selectionof comparable firms. Journal of Accounting Research 40 (2): 407-439.Brooks, L. D., and D. A. Buckmaster. 1976. Further evidence of the time series properties of ac-counting income. The Journal of Finance 31 (3): 1359-1373.Brown, S., K. Lo, and T. Lys. 1999. Use of R2 in accounting research: Measuring changes in valuerelevance over the last four decades. Journal of Accounting and Economics 28 (2): 83-115.Brozen, Y. 1971. Bain's concentration and rates of return revisited. Journal of Law and Economics14 (2): 351-369.The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    28/29

    What Determines Residual Income? 111

    Cheng,Q. 2002. Essays on the residual ncome valuationmodel. Ph.D. dissertation,UniversityofWisconsin-Madison.Dechow,P.M., A. P. Hutton,andR. G. Sloan. 1999.An empiricalassessmentof the residual ncomevaluation model. Journal of Accounting and Economics 26 (1-3): 1-34.Eaton,B. C., andR. G. Lipsey.1981. Capital,commitment, ndentryequilibrium.The Bell Journalof Economics 12 (2): 593-604.Fairfield,P. M., R. J. Sweeney,and T. L. Yohn. 1996. Accountingclassificationand the predictivecontent of earnings. The Accounting Review 71 (3): 337-355.Fama,E. E, andJ. MacBeth.1973.Risk,returnandequilibrium: mpirical ests. Journalof PoliticalEconomy81 (3): 607-636., and K. R. French.1997. Industry osts of equity.Journalof Financial Economics43 (2):153-193.Feltham,G. A., and J. A. Ohlson. 1995. Valuationand clean surplusaccounting or operatingandfinancial activities. ContemporaryAccounting Research 11 (2): 689-732., and- . 1996.Uncertaintyesolutionandthe theoryof depreciationmeasurement.our-nal of Accounting Research 34 (2): 209-234.Financial Accounting Standards Board (FASB). 1980. Qualitative Characteristics of Accounting In-formation.Statementof FinancialAccountingConceptsNo. 2. Norwalk,CT:FASB.Gjesdal,F. 1999.A steadystategrowthvaluationmodel:A note on accounting ndvaluation.Workingpaper,NorweigianSchoolof Economics.Jansen, . P.2002. Economicandaccountingdeterminants f residualncome.Workingpaper,George-townUniversity.Kothari,S. P., T. E. Laguerre,andA. J. Leone. 2002. Capitalization ersusexpensing:Evidenceonthe uncertaintyof futureearningsfrom capital expendituresversus R&D outlays.ReviewofAccounting Studies 7 (4): 355-382.Lee, C. M. C. 1999.Accounting-basedaluation: mpacton businesspracticesandresearch.Account-ing Horizons 13 (4): 54-68.Lev,B. 1983.Someeconomicdeterminants f time-seriesproperties f earnings. ournalofAccount-ing and Economics 5: 31-48., and T. Sougiannis.1996. The capitalization,amortization,and value-relevanceof R&D.Journal of Accounting and Economics 21 (1): 107-138., B. Sarath,andT. Sougiannis.1999. R&Dreportingbiases and theirconsequences.Workingpaper,New YorkUniversity.Lo, K., andT. Z. Lys. 2000. The OhlsonModel:Contributionso valuation heory, imitations,andempirical applications. Journal of Accounting, Auditing & Finance 15 (3): 337-367.Martin, S. 1983. Market, Firm, and Economic Performance. New York, NY: New York UniversityGraduateSchool of Business,SalomonBrothersCenter or the Studyof Financial nstitutions.Monahan,S. 1999. Conservatism, rowthandthe role of accountingnumbersn the equityvaluationprocess.Workingpaper,Universityof Chicago.Mueller,D. C. 1977. The persistenceof profitsabove the norm.Econometrica 4 (176): 369-380.. 1986. Profits n theLongRun. New York,NY: CambridgeUniversityPress.Myers,J. M. 1999. Implementing esidual ncome valuationwith linearinformationdynamics.TheAccounting Review 74 (1): 1-28.Nwaeze, E. T. 2000. Positiveandnegativeearningssurprises, egulatory limate,andstockreturns.ContemporaryAccounting Research 17 (1): 107-134.Ohlson,J. A. 1995.Earnings,bookvalues anddividends n equityvaluation.Contemporaryccount-ing Research 11 (2): 661-687.Penman,S. H., and X. Zhang.2002. Accountingconservatismand the qualityof earnings.TheAc-counting Review 77 (2): 237-264.

    Porter,M. 1980. Industry tructure ndcompetitive trategy:Keysto profitability. inancialAnalystsJournal36 (4): 30-41.Sloan,R. G. 1996. Do stockpricesfully reflectinformation n accrualsandcash flows aboutfutureearnings? The Accounting Review 71 (3): 289-315.The Accounting Review, January 2005

  • 7/28/2019 Determinants of Residual Income

    29/29

    112 Cheng

    Strickland,A. D., and L. W.Weiss. 1976.Advertising, oncentration,ndprice-costmargins. ournalof Political Economy 84 (5): 1109-1122.Zhang,X. 2000. Conservative ccountingandequityvaluation. ournalofAccountingandEconomics29 (1): 125-149.