13
MANAGERIAL AND DECISION ECONOMICS Manage. Decis. Econ. 22: 251–263 (2001) Modeling the Competitive Process Robert Jacobson* and Gary Hansen† School of Business Administration, Uniersity of Washington, Seattle, WA, USA Understanding the behavior of markets requires understanding not just the level of return but also its dynamics. The speed at which abnormal returns dissipate is one useful indicator of the competitive process. We model for US and Japanese firms the process of competition as reflected in the persistence of abnormal returns. While we find a similar aggregate distribution of firm persistence in both countries, we observe cross-national differences manifesting themselves in terms of inter-country differences in industry persistence. We find that both industry- and firm-specific factors influence firm persistence. Copyright © 2001 John Wiley & Sons, Ltd. DOI: 10.1002/mde.1020 INTRODUCTION The opportunity for profit encourages firms to engage in entrepreneurial discovery. Schumpeter (1942) describes this discovery as ‘to reform or revolutionize the pattern of production by exploit- ing an invention, or more generally, an untried technological possibility for producing a new commodity or producing an old one in a new way, by opening up a new source of supply of materials or a new outlet for products, by reorga- nizing an industry’. Firms must trade in the mar- ketplace to realize the benefits of their discovery. By doing so they provide opportunities for com- petitors to imitate successful practices and so dis- sipate abnormal returns. More generally, profitability, or the lack thereof, signals the direction that firm resources should flow to satisfy customer needs. The adjust- ment of resources and output into areas earning excess profits and away from areas earning below average profits will, in time, tend to bring returns back towards the firm’s cost of capital. This ad- justment is ‘the competitive process’ and the speed at which these abnormal returns dissipate is of fundamental importance to the firm because it impacts the value of any strategic initiative. Yet the issue of the persistence of abnormal returns is downplayed in most theories of the firm. Hayek (1948) noted that perfect competition in neoclassical economics is a market that is al- ready in equilibrium. It is a state of affairs in which there is no opportunity to compete. He felt that this conceptualization assumes away the fun- damental dynamic characteristics of competition, which include, for example, imitation, innovation, entry and exit. Neoclassical economics concen- trates primarily on specifying the effects of com- petition after the process of competition has reached its limits. While describing equilibrium, it does not describe the process that led to this outcome. Current strategic thinking has tended to follow neoclassical traditions and focus on equilibrium concepts of firm performance. In particular, the resource-based theory of the firm emphasizes het- erogeneous resources that allow a firm to achieve sustainable competitive advantage, i.e. an advan- tage that competitors are unable to imitate. 1 By focusing on one type of resource (i.e. inimitable assets that offer advantage until circumvented by a shock to the marketplace), resource-based per- spectives de-emphasize the role of resources that create a dissipating abnormal profit stream. A dichotomous view of resources (i.e. inimitable ver- sus immediately imitable) is inconsistent with the continuous innovation and change that tends to characterize the marketplace. * Correspondence to: School of Business Administration, Uni- versity of Washington, Seattle, WA, USA. Tel.: +1 206 5434780; e-mail: [email protected] † E-mail: [email protected] Copyright © 2001 John Wiley & Sons, Ltd.

Modeling the competitive process

Embed Size (px)

Citation preview

MANAGERIAL AND DECISION ECONOMICS

Manage. Decis. Econ. 22: 251–263 (2001)

Modeling the Competitive ProcessRobert Jacobson* and Gary Hansen†

School of Business Administration, Uni�ersity of Washington, Seattle, WA, USA

Understanding the behavior of markets requires understanding not just the level of return butalso its dynamics. The speed at which abnormal returns dissipate is one useful indicator ofthe competitive process. We model for US and Japanese firms the process of competition asreflected in the persistence of abnormal returns. While we find a similar aggregatedistribution of firm persistence in both countries, we observe cross-national differencesmanifesting themselves in terms of inter-country differences in industry persistence. We findthat both industry- and firm-specific factors influence firm persistence. Copyright © 2001John Wiley & Sons, Ltd.

DOI: 10.1002/mde.1020

INTRODUCTION

The opportunity for profit encourages firms toengage in entrepreneurial discovery. Schumpeter(1942) describes this discovery as ‘to reform orrevolutionize the pattern of production by exploit-ing an invention, or more generally, an untriedtechnological possibility for producing a newcommodity or producing an old one in a newway, by opening up a new source of supply ofmaterials or a new outlet for products, by reorga-nizing an industry’. Firms must trade in the mar-ketplace to realize the benefits of their discovery.By doing so they provide opportunities for com-petitors to imitate successful practices and so dis-sipate abnormal returns.

More generally, profitability, or the lackthereof, signals the direction that firm resourcesshould flow to satisfy customer needs. The adjust-ment of resources and output into areas earningexcess profits and away from areas earning belowaverage profits will, in time, tend to bring returnsback towards the firm’s cost of capital. This ad-justment is ‘the competitive process’ and the speedat which these abnormal returns dissipate is offundamental importance to the firm because itimpacts the value of any strategic initiative.

Yet the issue of the persistence of abnormalreturns is downplayed in most theories of thefirm. Hayek (1948) noted that perfect competitionin neoclassical economics is a market that is al-ready in equilibrium. It is a state of affairs inwhich there is no opportunity to compete. He feltthat this conceptualization assumes away the fun-damental dynamic characteristics of competition,which include, for example, imitation, innovation,entry and exit. Neoclassical economics concen-trates primarily on specifying the effects of com-petition after the process of competition hasreached its limits. While describing equilibrium, itdoes not describe the process that led to thisoutcome.

Current strategic thinking has tended to followneoclassical traditions and focus on equilibriumconcepts of firm performance. In particular, theresource-based theory of the firm emphasizes het-erogeneous resources that allow a firm to achievesustainable competitive advantage, i.e. an advan-tage that competitors are unable to imitate.1 Byfocusing on one type of resource (i.e. inimitableassets that offer advantage until circumvented bya shock to the marketplace), resource-based per-spectives de-emphasize the role of resources thatcreate a dissipating abnormal profit stream. Adichotomous view of resources (i.e. inimitable ver-sus immediately imitable) is inconsistent with thecontinuous innovation and change that tends tocharacterize the marketplace.

* Correspondence to: School of Business Administration, Uni-versity of Washington, Seattle, WA, USA. Tel.: +1 2065434780; e-mail: [email protected]† E-mail: [email protected]

Copyright © 2001 John Wiley & Sons, Ltd.

252 R. JACOBSON AND G. HANSEN

Penrose (1959), widely acknowledged as a foun-dation for resource-based theory, noted a limita-tion in equilibrium models since ‘new productiveservices are continuously being created’. Givenmarketplace dynamics, strategy and tactics thatgenerate these dissipating profits are likely to beas important, if not more so, to business successthan activities that cannot be duplicated. Allow-ing for a broader classification of value-generatingresources, consistent with a continuous process ofinnovation, offers the potential to strengthen thecontribution of resource-based theory. Resource-based perspectives need not be limited to assetsthat create an inimitable advantage. Rather, thetheory can address a continuum of assets that giverise to advantages ranging from inimitable toimmediately imitable and the effect of which isreflected in the speed of the competitive process,i.e. in the persistence of return.

Better understanding firm behavior and perfor-mance requires a better understanding of thecharacteristics of the persistence of returns. Ourknowledge of these characteristics is limited. Howquickly do abnormal returns dissipate? Howmuch variation in persistence exists among firmsand industries? Why does this variation occur?Our study seeks to extend past research in thearea by analyzing firm-specific estimates of persis-tence. In particular, we seek to (i) obtain andanalyze descriptive statistics relating to empiricalestimates of firm persistence and (ii) assess theextent to which firm persistence depends on econ-omy-wide, industry, and firm-specific factors.2 Wefirst propose a measure of persistence and illus-trate its importance through its influence on firmvalue. We then provide an overview of its deter-minants. We then discuss our methodological ap-proach and the data sample prior to providingour analysis of persistence for US and Japanesefirms.

QUANTIFYING THE COMPETITIVEPROCESS

Although other possible measures can provideinsights into the dynamics of return, e.g. thefrequency that a firm earns profits above somethreshold, the first-order autoregressive coefficientcharacterizing the return series may be the bestdescriptive measure of the competitive process.Consider the model AccRt=�+� �AccRt−1+�t,

where AccR is an accounting return measure suchas return on investment (ROI) or return on equity(ROE). The coefficient � tracks the speed atwhich returns dissipate; specifically abnormalprofits will dissipate at a rate of (1−�) perperiod.

The autoregressive representation models re-turn as a function of a constant, contemporane-ous shocks, as well as factors that are manifestedin the returns from previous periods. Past re-search has focused on mean return, i.e. �/(1−�),and on the conditions under which this level ofreturn will differ from the cost of capital, e.g. therelative importance of industry-wide versus firm-specific factors. While of value, this approach haslimitations in that it overlooks differences in firmdynamics. The firm’s mean return or return in agiven period is an inadequate descriptor of firmperformance. Further, because of continuingshocks, it may be most appropriate to view thecost of capital or some other long-term rate not asa characteristic of the firm’s profitability, butrather as an ‘attractor’ (Nelson, 1995).

Differing values of the parameter � correspondto differing dynamics of the competitive processand reflect the degree to which a firm, whether bydesign or chance, has insulated itself from com-petitive forces. Values of �=1.00 imply that firmprofits do not dissipate and are consistent withthe notion of sustainable competitive advantage.Values of �=0.00, which indicate that abnormalprofits dissipate immediately, are consistent withthe notion of perfect competition or contempora-neous reversion to some underlying return. Theconcept of the competitive process, implying thatprofits dissipate over time, suggests 0.00���1.00.

PERSISTENCE AND THE VALUE OF THEFIRM

To appreciate the importance of persistence ininfluencing the behavior of firms and markets, itis useful to consider its impact on firm value.3 Westart with the classical valuation model where thestock market value of the firm (M) equals thepresent value of expected future cash flows, i.e.

Mt= ��

s=1

� 11+k

�s

Et [CashFlowt+s ]. (1)

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

253MODELING THE COMPETITIVE PROCESS

We make the simplifying assumption that thepresent value of expected future cash flowsequals the present value of expected future earn-ings, which gives

Mt= ��

s=1

� 11+k

�s

Et [Earningst+s ]. (2)

We model expected accounting return (AccR)to follow an AR(1) process that decays to thecost of capital (k), i.e. AccRt+s=k+� �(AccRt+s−1−k). By definition Earnings=B �AccR, where B is the book value of size (e.g.assets or equity). With initial conditions AccRt

and Bt, we can rewrite Equation (2) taking intoaccount the autoregressive structure of account-ing return as

Mt= ��

s=1

� 11+k

�s

Bt(� s(AccRt−k)+k) (3)

or

Mt=Bt��(AccRt−k)

(1+k−�)+1

n. (4)

Dividing through by book value gives

Mt

Bt

=�(AccRt−k)(1+k−�)

+1 (5)

Equation (5) depicts the market-to-book ratioas depending on three components. One, it de-pends directly on the accounting return versuscost of capital differential, i.e. spread. All elseequal, firms with larger spreads have greatermarket value. Two, it depends on the cost-of-capital. The greater the cost-of-capital, the lessimportant future-term earnings are in influenc-ing firm value. Three, the market-to-book ratiodepends on persistence. Persistence has an ‘inter-active’ effect on the market-to-book ratio. Thatis, its effect depends on the firm’s spread.Higher persistence magnifies value enhancementfor positive spread firms; it magnifies value de-struction for negative spread firms.

While it is still commonplace for competitiveadvantage (monopoly power) to be equated withthe level of return, we see from Equation (5)that the effect of return on firm value cannot beunderstood independent of persistence. Thegreater the persistence, the greater the effect ofreturn on market value. With � equal to 1.00,Equation (5) simplifies to M/B=AccR/k, i.e.the market-to-book ratio is equal to the ratio ofaccounting return to cost-of-capital. Under this

condition, positive (negative) spread firms reapsignificant long-term rewards (penalties) fromtheir competitive advantage (disadvantage) andthis is reflected in a higher (lower) market-to-book ratio. With � equal to 0.00, Equation (5)simplifies to M/B=1. With no persistence, theinitial-condition accounting return supplies noinformation about future-term returns. Accord-ingly, the firm’s market value will equal itsbook value. Given its effect on firm value and,therefore, on managerial decisions, understand-ing the behavior of markets requires understand-ing the persistence of returns.4

DETERMINANTS OF PERSISTENCE

In the absence of market imperfections, com-petitor response would be immediate and com-plete. Abnormal profits would dissipateinstantaneously. There are, however, market-place imperfections (i.e. information and produc-tion asymmetries) that provide barriers toimitation, which Rumelt (1987) labels ‘isolatingmechanisms.’ Isolating mechanisms, such as in-formation impactedness (tacitness) that createsambiguity on the part of competitors that pre-vents competitor response, organizational struc-tures and incentives that make competitors slowto respond, buyer switching costs that createloyalty to the brand, the degree of innovative-ness on the part of the firm5 and its competi-tors, and the manner in which the firm choosesto exploit its advantage, interact to determinethe persistence of return.

Given the complexity of this interaction, it isnot surprising that Geroski and Jacquemin(1988) conclude that ‘it remains difficult to findfactors which are systematically associated withthe persistence of profits.’ Mueller (1990a) sug-gests that industry characteristics are relativelymore important than firm characteristics in in-fluencing the persistence of profits. Geroski andJacquemin (1988), in turn, suggest that country-wide factors turn out to be more discriminatingthan firm- or industry-specific ones. While as-sessing the relative importance of these factors iscomplicated by the complexity of their inter-re-lationship, theory and past empirical work sug-gest that firm persistence will depend oncountry-, industry- and firm-specific factors.

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

254 R. JACOBSON AND G. HANSEN

Country Effects

To the extent that markets continue to becomemore global, we would expect similarities in per-sistence across nations. However, national mar-kets differ in important ways that affect firmsoperating within their boundaries. Porter (1990)notes that national attributes establish the ‘play-ing field’ that operates for its industries. He high-lights factor conditions, demand conditions,related and supporting industry conditions, andfirm strategy, structure and rivalry as the fourcentral attributes of the playing field. As nationsdiffer with respect to these attributes, we wouldexpect differing behavior on the part firms andcustomers cross-nationally. And, therefore, wewould expect cross-national differences in the per-sistence of profits.

Consider that in Japan there has traditionallybeen little price competition, and it is considered‘impolite’ to mention your competitor or to di-rectly compare your products in advertisements.In this context, fewer tools are available for mar-ket entry, and consumers have increased searchcosts for product attribute information. In itself,this would limit the dissipation of abnormal re-turns. But, it also encourages increased non-pricecompetition and alternative information sources.In the US, technology-based companies receiveprotection from imitation through patent protec-tion. In Hong Kong and Malaysia, the absence ofprotected intellectual property rights retards inno-vation yet stimulates ‘the art’ of imitation and therapid diffusion of new technologies and processes.In Germany, and to a lesser degree elsewhere inEurope, employment is considered more of a fixedcost of doing business. In the US, firms have agreater inclination to restructure and downsize theworkforce. This flexibility allows US firms greaterability to retain profits during downturns. Thecost of this flexibility is that the firm can lose tacitknowledge and, in addition, competitors can moreeasily acquire it. As the above examples illustrate,nations create different competitive contexts forfirms operating within their boundaries.

Industry Effects

Traditionally, industrial organization has focusedon the industry as the relevant unit of analysis. Inparticular, the emphasis has been on the role ofindustry structure (e.g. concentration) in influenc-ing conduct and performance. The host of con-

flicting theories and empirical results has madethe effect of concentration on innovation and itsdiffusion one of the most widely researched topicsin industrial organization. That is, concentrationcan be posited to both retard and provide addedincentives for innovation and market reaction.

While recent work suggests concentration mayhave little association with persistence (Waring,1994) or, for that matter, profitability, a host ofother factors may affect not just a single firm butrather all firms in the industry. That is, industryeffects involve much more than information re-flected in industry concentration. Since, for exam-ple, opportunities for innovation and imitationdiffer by industry, so should persistence. Whilefirm strategy is obviously critical, firms cannotcompletely isolate themselves from industry-wideforces. Elements of customer demand and techno-logical change, for example, are outside the con-trol of managers. Whether it be the impact of theInternet, or customer preference for genericbrands, firms in the same industry may face simi-lar shifts in demand. Each industry faces a differ-ent combination of forces that evolve over timeand that manifest themselves in the nature ofimitation and innovation.

Firm-Specific Effects

Idiosyncratic capabilities allow a firm to partiallyisolate itself from economy-wide and industry-specific competitive forces. Within an industry,firms compete by building, acquiring and com-bining assets and resources. While the industrysets the context for firm behavior, recent work instrategy has emphasized the importance of firm-specific factors as determinants of business perfor-mance. Rumelt (1987), for example, found thatthe variance in profits within industries is three tofive times larger than the variance across indus-tries.

The resource-based theory of the firm, in par-ticular, highlights that companies can consciouslyaccumulate unique combinations of resourcesover time that provide advantage relative to otherfirms facing a similar competitive context(Barney, 1991; Peteraf, 1993). These resourcesinclude visible assets, such as physical assets,patents and property, as well as invisible assets,such as know how, routines, processes and skills.Both the level and persistence of profits depends

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

255MODELING THE COMPETITIVE PROCESS

on the distinctiveness and imitability of these keyresources.

Differing capabilities allow for the utilization ofdiffering isolating mechanisms. Some firms, e.g.Coca-Cola, seek to leverage brand equity (reputa-tion). Others seek to exploit buyer switching costs,e.g. Microsoft. Sony uses its skills in new productdevelopment to create a constant flow of newproducts, thus influencing customer behavior toincrease demand. Yet, since it lacks the low costproduction skills of its rival, Matsushita, itsbrands encounter dissipation of profits from lowcost imitators. Matsushita’s low cost productioncapability precludes it from achieving the samelevel of product innovativeness as Sony. Thus,each firm develops differing resources to generateand sustain profit and to overcome its rivals’advantages.

MODELING THE TIME SERIESPROPERTIES OF RETURN

Although notable exceptions exist, e.g. Mueller(1990b), relatively little empirical research hasbeen undertaken to investigate the competitiveprocess. The absence of quantitative informationhas, to some extent, limited discussions of thecompetitive process. This is not to suggest thatsome ‘equilibrium’ speed of adjustment exists andthat empirical methods can uncover it. Rather,empirical analysis provides a historical perspec-tive, in addition to other insights, that can be ofvalue in understanding the current and futurebehavior of markets. We know that adjustmentrates differ. We do not, however, have an ade-quate understanding of the extent of the variationor why it occurs. We seek to assess these issuesempirically.

Past research assessing the competitive processhas primarily used three methodological variantsto estimate the persistence parameter �.6 Thethree different approaches correspond to differingassumptions about the inter-firm differences inthe properties of the return series. One approachinvolves using ordinary least squares to estimatean autoregressive model of the form

AccRit=�+� �AccRit−1+�it (6)

This approach assumes that there are no firm-specific differences either in the level of return orin the persistence of return (i.e. �i=�j and �i=

�j). Of particular concern is that, to the extentthat firms differ in their level of return, not ac-counting for this difference will result in an up-ward bias in the estimated persistence coefficient.

To circumvent this potential bias, a secondapproach has used a model of the form

AccRit=�i+� �AccRit−1+�it (7)

Equation (7) differs from Equation (6) in that itallows for a firm-specific constant, i.e. it allowsthat rates of return differ across firms. Estimationof Equation (7) can be achieved following, forexample, the procedure suggested by Andersonand Hsiao (1981, 1982), i.e. a generalized methodsof moments estimator with a diagonal set ofinstruments. The Anderson and Hsiao (1981) ap-proach involves taking first differences of Equa-tion (7) to remove the fixed effect and then usinglagged values (of order two or higher) of ROI asinstrumental variables for the lagged first differ-ence in ROI. The limitation of Equation (7) isthat it constrains the persistence parameter to bethe same across firms.

A third approach is a model of the form

AccRit=�i+�i �AccRit−1+�it (8)

Equation (8), the most general of the threeapproaches, allows for both firm-specific differ-ences in the constant and the persistence of re-turn. This generality, however, has some costs.Unlike the other models, the assumption of afirm-specific persistence parameter restricts theefficiency gains associated with the pooling oftime series and cross-sectional data.

The Stein Estimator

One means of lessening this limitation is throughthe use of empirical Bayes procedures. Stein(1955) showed that it is possible to uniformlyimprove maximum likelihood estimates (MLEs)in terms of total square error in estimates. Ineffect, Stein estimators take a weighted average ofthe MLE and a fixed-point estimate, usually theaverage value of an entire sample of MLEs. Con-structed in this way, Stein estimators are oftencalled ‘shrinkage estimators,’ as they operate byshrinking each parameter estimate towards a fixedpoint. Any specific Stein estimate cannot be ex-pected to outperform any specific maximumlikelihood estimate, but the set of Stein estimatesshould outperform the set of maximum likelihoodestimates.

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

256 R. JACOBSON AND G. HANSEN

A number of studies have extended and gener-alized Stein’s results (e.g. Efron and Morris, 1973,1975). The most relevant extension for this studyinvolves a Stein estimator that allows the vari-ances of persistence estimates for different firmsto be different. Assuming that

�i ��i�N(�i, Di), i=1, . . . , k

and

�i�N(�, A), i=1, . . . , k

a Stein estimator of the underlying persistenceparameter of a particular firm �� i* can be calcu-lated as

�� i*=�� +c�

�� i−�� �where �� i is the least squares estimate of �i, (�� ) isan estimate of �, and c=1− [D� i/(A� +D� i)], withD� i being the least squares estimate of the varianceof �i, and A� being an estimate of the variance of�i about the mean value �.

The Stein estimate is composed of the maxi-mum likelihood estimator and the mean of theentire sample of MLEs. The value c has the effectof shrinking the estimator to the sample estimateof the mean value of persistence. The shrinkage isgreatest as c approaches 0 and least as c ap-proaches 1. By allowing for different variances,Di, the procedure shrinks the estimate closer tothe mean the larger the variance of the estimate,and it allows for less shrinkage for estimates withsmaller variances.

Although Stein’s results were written as a statis-tical paradox, they have a straightforward empiri-cal Bayes interpretation. Robbins (1955) showedthat it is possible to achieve the same minimumrisk associated with Bayes’ rule without knowl-edge of a prior distribution as long as the numberof means being estimated is very large.

We make use of the Stein estimator to estimatethe persistence parameter in Equation (8) for abroad sample of US and Japanese firms. Ourmeasure of accounting return is ROI, defined asnet income divided by total assets at the beginningof the period.7 To improve stationarity in the timeseries, we subtract the yearly mean ROI fromeach observation. Thus, our analysis is based ondeviations of firm ROI from the annual economy-wide (sample) average. We first look at descriptivestatistics for the persistence estimates and thenassess the role of country, industry and firm-

specific factors in explaining differences in persis-tence among firms.

DATA

For the US, the 1992 Standard and Poor’s Com-pustat annual industrial file provides us with ac-counting information for companies listed in theNYSE, AMEX and NASDAQ stock exchanges.For a firm to be included in our study, we re-quired that it reported its net income and assetsfor the entire 20-year period 1973–1992, be anon-financial firm, and that there be at least threeother firms represented from that same industry.A total of 1039 firms met this selection criterionand were included in the analysis.

The twenty year time frame was used to allow asufficient time series for estimation. However,particularly since we make use of Stein estima-tion, firms with fewer observations could alsohave been included in the analysis. However, thesame (20) year criteria allow for greater compara-bility across firms. We limited our sample tonon-financial firms speculating that the financialfirms may have different characteristics, e.g. thenature of the asset base, than non-financial firmsand because our Japanese data sample is limitedto non-financial firms. Using only firms that comefrom industries with a minimum of four firms inthe sample gives us the ability to test for industry-wide effects.

For Japan, annual balance sheet and incomestatement items come from the Japan Develop-ment Bank. Firms in the sample are part of theFirst Section of the Tokyo Stock Exchange (TSE).The TSE has two sections, but the First Sectiondominates the Second Section, e.g. the size of theSecond is less than 10% of the First in terms ofmarket value. Although there are seven otherstock exchanges in Japan, they are very smallrelative to the TSE. In addition to the samesampling criteria for the US sample (e.g. a 20-yearreporting criteria), data availability allows us tolook only at Japanese firms with a March fiscalyear, the most common. A total of 271 firms wereincluded in the analysis covering the periodMarch 1974–March 1993.

We should note that the selection process re-duces the generality of our results. Both for theUS and for Japan, the sample is not representa-tive of all firms as, for instance, it does not

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

257MODELING THE COMPETITIVE PROCESS

include young firms or firms that have failed.While differing perspectives suggest the possibilityof this sample being unrepresentative of the popu-lation of firms as a whole, both the sign of thebias and its magnitude is speculative. Nonetheless,some generalization can be expected and, perhapsmost importantly, the firms included in the sam-ple are important in their own right.

RESULTS

Table 1 provides descriptive statistics for the esti-mates of �i for the 1039 US and 271 Japanesefirms in our sample. The mean persistence, �� forthe US sample suggests a persistence parameter ofapproximately 60% per year, i.e. about 40% ofabnormal returns dissipate per year. The estimateof mean persistence for Japanese firms is in closecorrespondence (0.569). The median value of per-sistence is also very similar for the two samples,i.e. 0.595 for the US and 0.584 for Japan.

As a sensitivity check, we also estimated aggre-gate persistence making use of the fixed-effectsmodel. Using the Anderson and Hsiao (1981)approach and deleting the upper 1% and lower1% of the sample based on the change in ROI, weestimated a firm-specific constant model (Equa-tion (7)) for the US and Japanese samples.8 Theestimated value of persistence obtained from thisestimation process was 0.56 with a standard errorof 0.02 for the US and 0.58 with a standard errorof 0.03 for Japan. This provides additional sup-port for our Stein-based estimate of persistence ofapproximately 0.60.

Given the differences in, for example, social,economic and market structures between thecountries, the similarity in aggregate persistence issurprising. While Japan appears to have a more

heavily skewed distribution of low persistencefirms, e.g. 5% of US firms have persistence below0.304 while 5% of Japanese firms have persistencebelow 0.162, the distribution of persistence acrossthe two countries is in close correspondence.Mechanisms are in place in both countries thatallow firms to earn abnormal profits but whichcause these profits to dissipate over time.

The estimates of persistence reported in Table 1are larger than those found in previous work. Forexample, Mueller (1990a), based on a sample of551 manufacturing firms for the period 1950–1972, reports a mean persistence parameter of0.183, i.e. about one-third of the estimate wereport in Table 1. While it is unclear what ac-counts for the significant difference, two possibili-ties are (i) difference in the type of firms includedin the sample (i.e. our sample includes firms out-side of manufacturing) and (ii) differences in thetime period of study. This second explanation issupported by the findings of Odagiri and Ya-mawaki (1990b), who report lower persistenceestimates for the period 1950–1972 as comparedwith 1964–1980, i.e. 0.19 versus 0.47 for USfirms. Indeed, our estimates are more in line withthe findings of studies making use of more recentdata. For example, Geroski and Jacquemin (1988)report estimates of persistence for France, Ger-many and the UK to be 0.459, 0.461 and 0.520,respectively. Odagiri and Yamawaki (1990a) re-port a mean persistence of 0.46 for their 1964–1982 sample of Japanese firms.

Still, our estimates are higher than those previ-ously reported. This difference among estimatesraises the question: how sensitive are persistenceestimates to sample and time period? To investi-gate this question, we split our sample into threeperiods: 1976–1981, 1982–1987 and 1988–1992.9

We then estimated mean persistence for each pe-riod using the fixed-effects model (with outliersdeleted). For Japan, estimated persistence is indis-tinguishable among the three time periods, i.e.0.58, 0.59 and 0.57. This stability of parameters isconsistent with the results of Odagiri and Ya-mawaki (1986), who find no structural change inJapan following the 1973 oil shock. For the US,however, we find substantial difference among theperiods. For the two early periods the estimatedpersistence is 0.63 and 0.68. For the later period,it is 0.37. Thus, the similarity in our persistenceestimates between the US and Japan would notseem to reflect uniformity in the competitive

Table 1. Descriptive Statistics by Firm: Esti-mates of Persistence

US Japan

1039cobs 271Mean 0.5690.603

0.187S.D. 0.2330.918 0.94395%0.731 0.71475%

0.5840.595Median25% 0.473 0.423

0.1620.3045%

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

258 R. JACOBSON AND G. HANSEN

process across global markets. Rather, by merehappenstance, inter-temporal differences in persis-tence produced a similar overall measure of per-sistence for the particular period of study. Thisreinforces the view that � should not be viewed asa constant or as exogenous. Rather, it is a dy-namic factor that reflects firm-, industry- andeconomy-wide conditions.

Within-Country and Cross-National Differences inIndustry Persistence

Indeed, as evidenced by the distribution of persis-tence reported in Table 1, differences in persis-tence exist among firms. To give more insight intothe nature of the distribution, Table 2 provides anascending order listing of industry persistence (i.e.the mean of the firm persistence for firms in thesame four-digit industry classification) for the twocountries.

An examination of Table 2 indicates both simi-larities and differences across countries and indus-tries, which reflects similarities and differencesacross firms. For example, we observe a substan-tial difference cross-nationally between electricservices. In the US, these utilities have well-aboveaverage persistence, i.e. approximately 0.75 forboth electric service providers and electric andother service providers.10 In Japan, persistence forelectric power providers is well below average(0.26). Pacific Gas and Electric, for example, hasa persistence of 0.70. Tokyo Electric is at 0.31.Railroads provide another area of difference. Inthe US, both railroads and railroad equipmentmanufacturers have below average persistence, i.e.0.44 and 0.45, respectively. With few barriers toimitation and with numerous substitutes, abnor-mal returns in the US railroad industry dissipatemore quickly. The Japanese railroad industry,with substantially fewer transportation alterna-tives, i.e. a less developed highway and airportsystem, and with passenger rail differentiated byservice, schedule and price, has average persis-tence (0.59). But similarities also exist. For exam-ple, high persistence characterizes the pharma-ceutical industry in both countries, i.e. 0.76 inthe US and 0.70 in Japan. Merck and FujisawaPharmaceutical both have among the very highestestimated values of persistence, i.e. 1.00. Sim-ilarities exist in the steel industry as well. Theestimated persistence is 0.52 for the US and 0.48for Japan. Both Bethlehem and Nippon Steel

Table 2. Industry Persistence

Industry name PersistenceObservation

US Industries1 Abrasive, asbestos, misc. minerals 0.408312 Railroad equipment 0.438973 Metalworking machinery & eq 0.452494 0.45377Railroads5 Aircraft engine, engine parts 0.45885

0.459896 Elec apparatus & equip-whsl0.46757 Gold and silver ores0.47297Petroleum refining80.480329 Air cond, heating, refrig eq

10 Hotels, motels, tourist courts 0.481990.48724Bolt, nut, screw, rivet, washrs11

12 Ortho, prosth, surg appl, suply 0.49096Ship & boat bldg & repairing13 0.49277

14 Machine tools, metal cutting 0.4943215 Metals service centers—whsl 0.49509

0.49611Engineering services16Semiconductor, related device 0.4964817

18 Natural gas transmission 0.4968519 Motor vehicle part, accessory 0.49876

0.4989320 General industrial mach & eq21 Household audio & video eq 0.4990722 Electronic comp, accessories 0.5056423 Water transportation 0.5086724 General indl mach & eq, NEC 0.5107225 Engr, acc, resh, mgmt, rel svcs 0.5135926 Cmp integrated sys design 0.5151

0.51523Misc fabricated metal prods27Trucking, except local 0.5153228

29 Periodical: pubg, pubg & print 0.5199830 Crude petroleum & natural gs 0.52023

0.5204831 Books: pubg, pubg & printingSteel works & blast furnaces 0.5220632

33 Construction machinery & eq 0.52602Soap, detergent, toilet preps34 0.52703

0.527235 Radio, TV broadcast, comm eq36 Special industry machy, NEC 0.52838

0.53334Metal mining3738 Hardwr, plumb, heat eq—whsl 0.5339239 Misc elec machy, eq, supplies 0.5360940 Engines & turbines 0.537841 Indl coml fans, blowrs, oth eq 0.5388642 Prim production of aluminum 0.53955

0.54227Photographic equip & suppl4344 Electric lighting,, wiring eq 0.5443445 0.54438Rolling & draw nonfer metal46 Paperboard mills 0.5479247 Textile mill products 0.54897

Plastics, resins, elastomers 0.554564849 Newspaper: pubg, pubg & print 0.554750 Brdwoven fabric, mill, cotton 0.5551951 Cutlery, handtools, gen hrdwr 0.55638

Elec meas & test instruments 0.557935253 0.55862Operative Builders54 Electronic computers 0.5607755 Eating places 0.56218

0.5649756 Misc. chemical productsPaper mills57 0.56812Footwear, except rubber58 0.57595

0.57703Fabricated rubber pds, NEC590.5785660 Perfume, cosmetic, toilet prep

61 Advertising agencies 0.578570.57984Sugar & confectionery prods62

63 Srch, det, nav, guid, aero sys 0.58021

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

259MODELING THE COMPETITIVE PROCESS

Table 2. (Continued)

Industry name PersistenceObservation

0.5802264 Commercial printingFabricated plate work 0.5803565

0.5812766 Electronic components, NEC0.58788Household furniture670.5922168 Men, yth, boys, frnsh, wrk clthg0.5936Apparel and accessory stores69

Heavy constr—not bldg constr70 0.598180.60094Beverages71

Motors and generators72 0.601020.6034473 Surgical, med instr, apparatus0.60364Chemicals & allied prods740.6046575 Air transport, scheduled0.60624Help supply services76

Electronic parts, eq—whsl, NEC77 0.607580.60866Industrial organic chemicals78

Prim smelt, refin nonfer metl79 0.609230.6145880 Grocery stores0.61553Natural gas transmis & distr810.6172582 Pumps and pumping equipment0.61775Plastics products, NEC83

Aircraft84 0.620350.62157Metal cans85

Television broadcast station86 0.6246387 Grain mill products 0.62593

0.62611Mng, quarry nonmtl minerals880.6320789 Metal Forgings and Stampings0.63296Computer communication equip90

Food and kindred products91 0.634910.63532Department stores92

Misc amusement & rec service93 0.64550.6540294 Farm machinery and equipment0.65982Tele & telegraph apparatus950.6611596 Women’s clothing stores0.66207Drug & proprietary Stores97

Convrt papr, paprbrd, ex boxes98 0.666140.67239Manifold business forms99

Guided missiles & space vehc100 0.67593Variety stores101 0.67607

0.67629Can, froznpreserv fruit & veg1020.68192103 Paints, varnishes, lacquers0.6824Motor vehicles & car bodies104

Cement, hydraulic105 0.683740.6871Lumber & oth bldg matl—retl106

Indl trucks, tractors, trailrs107 0.68834Water supply108 0.69031

0.69807Groceries & related pds—whsl109Industrial measurement instr110 0.70572

0.7133111 Concrete, gypsum and plaster0.71715Aircraft parts, aux eq, NEC112

Computer & office equipment113 0.73678Apparel & other finished pds 0.74055114

0.7485Electric services1150.75567116 Electric & other serv comb0.75868Pharmaceutical preparations117

118 Drilling oil and gas wells 0.763250.76329Phone comm ex radiotelephone119

120 Natural gas distribution 0.78206

Japanese IndustriesBridges and steel frames1 0.257

2 Electric power 0.2623 Oil refining 0.284

0.370Shipping-tramps45 Petrochemicals 0.434

0.4656 Chemicals-soda based

Table 2. (Continued)

PersistenceIndustry nameObservation

7 Steel rolling 0.4688 0.481Trucking9 Steel-blast furnace 0.484

10 Ink and pigment 0.49011 Paperboard 0.49312 Automobile parts 0.503

0.50413 Conveying machineryElectronic parts14 0.505

15 Electric wire/cable 0.51716 Warehousing 0.53217 0.540Copper, lead & zinc18 Synthetic resin 0.552

0.568Real estate1920 Heavy duty electrical equipment 0.57221 Railroad transportation 0.58622 Automobiles 0.58823 Construction-dredging & 0.605

reclamation0.60724 Wholesale-trading companies

25 Specialized trading companies 0.60726 Cameras 0.607

0.608Confectionery & bakery2728 Household machinery 0.63429 Shipbuilding 0.637

Civil engineering 0.650300.657Road pavement31

Synthetic fibers 0.69332Pharmaceuticals 0.69633

0.73334 General constructionInstallation work 0.74035

36 0.749Communications equipment37 Plastic processing 0.751

Paper38 0.7540.835Household electrical appliances39

are at 0.4. These cross-national similarities anddifferences suggest that the primary role of na-tional factors is that of having differential impactson industry persistence, as opposed to inducinguniformly different levels of persistence across allindustries.

Inter-firm Differences in Persistence

Within an industry, we also observe similaritiesand differences in firm persistence. Are these in-ter-firm differences random? Or are there system-atic differences related to other factors? Toaddress this issue we seek to assess the differentialimpact of industry-wide and firm-specific determi-nants of persistence. In particular, we focus onthree factors (industry persistence, relative firmprofitability and relative firm size) as potentiallyassociated with inter-firm differences in persis-tence. To assess the extent to which firm per-sistence varies systematically we regress the

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

260 R. JACOBSON AND G. HANSEN

estimates of firm persistence obtained from Steinestimation on (i) the mean persistence of otherfirms in the same four digit industry, (ii) relativefirm profitability, and (iii) relative firm size. Thatis, we estimate the model

�� i*=�0+�1��� sic+�2�ROIi−ROIsic

+�3�Assetsi/Assetssic+�it. (9)

Many of the factors influencing persistence arenot firm-specific but rather are common to allfirms in the industry. To the extent this is true, thepersistence of other firms in the industry shouldbe positively related to firm persistence. Firms,however, differ in their ability to cope with andinfluence industry forces. Thus, firm-specific fac-tors should also affect persistence. One such fac-tor is relative firm profitability. We posit that firmprofitability reflects the extent to which the firmpossesses and utilizes unique and valuable re-sources. While not emphasized in previous empiri-cal work, since ROI and persistence jointlydetermine market value, firms with unique re-sources will seek to influence both. By doing so,they will induce a positive correlation between thetwo. Finally, we would suspect that larger firms inthe industry should have more persistent returnsthan smaller firms. Again using the resource-baseperspective, we would expect that some of thesame unique resources that allow a firm to getbigger also allow it more persistent returns, i.e. avariation of the Demsetz (1973) hypothesis.

Table 3 provides the results of the estimation ofEquation 9 for the US and for Japan. For boththe US and the Japanese sample, we observe that

firm persistence has a highly significant industrycomponent. Some of the same factors that affectthe firm also affect other firms in the same indus-try. For the US, 59% of the industry persistence isreflected in the firm-specific persistence level. ForJapan, approximately 50% of firm persistence isindustry-specific. Industry characteristics affectnot only the level of the earnings stream but alsothe duration of that stream. Markets do matter.

We find that, for both samples, firm persistencedepends on relative ROI. This effect of relativeROI reflects the combined effect of two factors,i.e. the positive role of firm ROI and the negativerole of industry ROI. The positive effect of firmROI is consistent with resource-based perspec-tives. Some of the same factors that allow a firmto achieve high ROI allow it to achieve highpersistence. The negative effect of industry ROI isconsistent with, for example, higher profitabilityencouraging increased entry into the market andimitation.11 That is, the larger potential rewardsprovide added incentive for activities that induceless persistence.

For the US, the estimated effect of relative ROIis 0.88. The estimated coefficient is higher forJapanese firms (3.71). This difference in magni-tude is somewhat misleading, however. The vari-ance in US ROI is greater than the variance inJapanese ROI. In terms of standardized regres-sion coefficients, the effect of ROI on persistenceis about the same. A one standardization changein ROI for the US is associated with a 0.21standard deviation change in persistence. ForJapan, we find that a one standardization changein ROI is associated with a 0.24 standard devia-tion change in persistence. Therefore, in terms ofoverall impact on persistence, relative return has asimilar effect in both countries.

We observe a difference in the role of relativesize. The effect is small and insignificant for theUS. It is positive (0.02) and statistically significantfor Japan. This difference supports the argumentthat the differences in economy-wide and marketstructures result in differential relationships influ-encing persistence. A variety of possible explana-tions may account for this cross-nationaldifference. For example, US firms may choose toexploit their unique resources more in terms ofROI than market share. Relative size for Japanesefirms may be more strongly associated with theunique resources that allow for maintaining com-parative advantage because they place a greater

Table 3. Why Persistence Differs Across Firms

Dependent variable: �� i*

JapanUS

Coefficient S.E. Coefficient S.E.

Intercept 0.25 0.03 0.25 0.060.500.050.59�� sic 0.10

0.88 0.863.710.12ROIi−ROIsic

−0.003Assetsi/Assetssic 0.002 0.02 0.01R2 0.14 0.15cobs 1039 271

Where �� i*= firm-specific estimate of persistence, �� sic=per-sistence of return of the other firms in the same industry,ROIi−ROIsic=deviation of firm ROI from mean ROI ofother firms in the industry, Assetsi/Assetssic=size of firmrelative to average size of other firms in the same industry.

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

261MODELING THE COMPETITIVE PROCESS

emphasis on market share. Another possible ex-planation relates to the historically strong roleplayed by keiretsu industry groups, which devel-oped around dominant firms. Relative size maybe acting as a proxy for the strength of thekeiretsu.

The explanatory power of the models (as re-flected in, for example, the R2 values of 0.14 forthe US and 0.15 for Japan) raises three issues ofnote. One, it suggests that the mean (economy-wide) value is a relatively good predictor of firmpersistence. While firms in every industry havediffering opportunities for competition, we ob-serve that the impact for competition is somewhatuniform across firms and markets. Economy-widepersistence can approximate the dynamics of re-turn in highly disparate firm and industry struc-tures. Managers are limited in their ability toaffect the competitive process. Two, it is difficultto obtain observable measures that explain firmpersistence. Indeed, our choice of explanatory fac-tors in Equation (9) reflects this perspective, since,rather than making use of explicit industry-specific measures (such as concentration), we usedan outcome-based measure (mean industry persis-tence). Many of the factors influencing persistenceare unobservable. Indeed, Winter (1987) suggeststhat, because of their tacitness, it is unobservablefactors that are most likely to have the greatestand longest-lasting impact on performance.Three, the importance of context (heterogeneity)cannot be overemphasized. Factors having a givenimpact in one context, may have a different im-pact in another context. For example, in somemarkets advertising may facilitate differentiationand, therefore, higher persistence. In other mar-kets, advertising may have the opposite effect byleading to increased information and price compe-tition. This suggests, for example, the potentialusefulness of industry specific research in order tomore fully understand the determinants of firmpersistence.

SUMMARY

We find the mean estimates of persistence for theperiod 1975–1992 approximately the same for theUS and for Japan (about 0.6). These estimates,however, mask both inter-temporal differences inthe US and cross-national differences in industrypersistence. The cross-national differences in in-

dustry persistence suggest that country-wide fac-tors may matter as much, if not more, in shapingthe context in which industries operate as in influ-encing the aggregate level of persistence. Theinter-temporal heterogeneity in US meanpersistence suggests that a useful direction forfuture research would be to assess whether thelower persistence observed for the US since 1988reflects a structural change in the US economy orwhether it is a transitory phenomenon.

At a more disaggregate level, we observe, forboth the US and Japan, differences in industrypersistence and differences in persistence amongfirms in the same industry. We find that marketsdo matter in influencing firm persistence. Firmscannot isolate themselves completely from in-dustry-wide competitive factors. We find thatcommon industry characteristics affect firmpersistence, and that higher industry returnsprovide incentives that accelerate the dissipationof abnormal returns.

Assessing the longevity of persistence remainsan area for future research. For example, in theUS and Japan, as well as other industrializedcountries, high persistence has characterized thepharmaceutical industry. This has been attributedto, for example, high historical levels of productdifferentiation, the absence of buyer pressure,high barriers to entry, and patent protection.With changes in health care policy and rapidtechnological innovation, some of the factors in-ducing high persistence in the past may no longerbe operative. Technological change has alreadyhad a major impact on the telecommunicationsindustries and we would expect to observe inter-temporal changes in industry persistence in a vari-ety of other markets as well.

While industry-wide factors matter, firms dohave the ability to influence the persistence ofreturns by their strategic choices. In particular, wefind that some of the same factors having apositive impact on the level of return also posi-tively impact its persistence. The role of size ismore ambiguous. We find a positive associationin Japan, but no association in the US.

Our estimates of persistence reinforce the per-spective that viewing business performance as anequilibrium concept is inadequate. A significantamount of firm profitability is, in fact, transitory.As such, understanding business success requiresunderstanding the factors influencing the dynam-ics of return. As these national, industry and

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

262 R. JACOBSON AND G. HANSEN

firm-specific factors vary, so will persistence. Byunderstanding how these factors interact to influ-ence the persistence of return we can come tobetter appreciate the functioning of the competi-tive process.

NOTES

1. See Foss (1994) for a more detailed discussion ofthe use of equilibrium concepts in the resource-based theory of the firm.

2. A focus of past research has been to compare thestructure-conduct-performance paradigm (whichemphasizes industry effects) with the resource-based paradigm (which emphasizes idiosyncraticfirm effects) in explaining differences in the level ofreturn. We compare the paradigms’ implicationswith respect to persistence.

3. See also, for example, Miller and Rock (1985),Kormendi and Lipe (1987) and Jacobson andAaker (1993) for a discussion of how persistenceeffects the stock market reaction to a shock inearnings or size adjusted earnings (i.e. accountingreturn). Namely, persistence magnifies the earningsshock in that it effects the present value of therevision in expected future earnings. If �=0, thenthe earnings shock is purely transitory and has noeffect on future term earnings. Conversely, if �=1.00 then the earnings shock is permanent and allfuture term earnings will be impacted.

4. While it is a fundamental determinant of firmvalue, we do not place judgments as to the desir-ability of high versus low values of persistence.First, as evidenced from Equation (5), persistencedoes not have a direct effect on the market-to-bookratio. High persistence increases M/B for positivespread firms. It decreases M/B for negative spreadfirms. As such, from the firm’s standpoint, increas-ing persistence may be undesirable as it may lockthe firm into an unprofitable position. Indeed, theeffect of persistence on the firm may change overtime. Some of the same barriers that allowed verti-cally integrated steel mills to preclude entry in the1960s, precluded exit and expansion in the 1980sand 1990s. Second, we are sympathetic to the posi-tion that profitability provides firms the incentiveto innovate. In the absence of persistence, firmswould be unable to reap the rewards of their inno-vation. As such, less innovation would occur andsociety would be worse off. So while persistencemay allow for marketplace inefficiencies, it alsoallows for discovery and entrepreneurial behavior.

5. Innovation on the part of the firm can both in-crease and decrease persistence. A firm may chooseto dissipate its own profits rather than have com-petitors do so. However, to the extent the firmengages in innovation correlated with its existingbrands, it can increase persistence (Roberts, 1996).

6. A less widely used fourth approach involves usingnon-linear least squares to estimate the persistenceparameter based on a stock market valuationmodel, Jacobson (1988, p. 422).

7. Much discussion has centered around the validityof accounting return in general (Fisher andMcGowan 1983) and the relative usefulness of dif-ferent measures of accounting return. The particu-lar measure we use correlates with stock return (forboth US and Japanese) as well as any accountingreturn measure, (Jacobson and Aaker, 1993). Assuch, despite the inherent limitation in accountingreturn measures, the ROI measure contains infor-mation about the firm’s financial performance.

8. The fixed-effects estimator appears to be highlysensitive to extreme value (i.e. ‘outliers’). Withoutdeleting the extreme values the estimated values off from the fixed effects model were −0.035 for theUS and 0.35 for Japan. To appreciate how differentthis fixed-effects estimate is from the Stein estima-tor, consider that for the US the fixed effectsestimate is smaller than every single one of the 1039firm estimates of persistence obtained using theStein estimation procedure. One of the advantagesof the Stein estimator is that by weighting estimatesbased on estimated variances, the role of extremevalues is minimized.

9. Our sample begins in 1976 to allow for instrumen-tal variable estimation of the fixed effects model,which requires the use of ROI lagged two periodsas an instrument.

10. We find that, in general, regulated industries in theUS have among the highest levels of persistence.

11. We also estimated a model that allowed for sepa-rate effects for firm return and industry return. Wecould not reject the hypothesis that the estimatedcoefficients were of the same magnitude but of theopposite sign, i.e. a positive effect for firm returnand a negative effect for industry return. Thissupports our constrained model based on relativereturn. In addition, we tested for non-linearities inthat resource disadvantaged firms might have morepersistent losses and larger industry losses provideadded incentives for activities that reduce persis-tence; e.g. exit. We did not find evidence of anon-linearity in the relationships. Possible explana-tions include asymmetric behavior on the part ofmanagers and such phenomenon as escalating com-mitment. Another explanation may be sample bias;e.g. a firm with highly persistent negative returnswill be unable to stay in business for the 20 yearsneeded to be in our sample.

REFERENCES

Anderson TW, Hsiao C. 1981. Estimation of dynamicmodels with error components. Journal of the Ameri-can Statistical Association 76: 598–606.

Anderson TW, Hsiao C. 1982. Formulation and esti-mation of dynamic models using panel data. Journalof Econometrics 18: 47–82.

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)

263MODELING THE COMPETITIVE PROCESS

Barney J. 1991. Firm resources and sustained competi-tive advantage. Journal of Management 17: 99–120.

Demsetz H. 1973. Industry structure, market rivalry,and public policy. Journal of Law and Economics 16:1–10.

Efron B, Morris C. 1973. Stein’s estimation rule and itscompetitors—An empirical Bayes approach. Journalof the American Statistical Association 68: 117–130.

Efron B, Morris C. 1975. Data analysis using Stein’sestimator and its generalizations. Journal of theAmerican Statistical Association 70: 311–319.

Fisher F, McGowan J. 1983. On the misuse of account-ing rates of return to infer monopoly profits. Ameri-can Economic Re�iew 73: 82–97.

Foss NJ. 1994. The resource-based perspective andthree economic approaches. Working paper, Instituteof Industrial Economics and Strategy, CopenhagenBusiness School.

Hayek FA. 1948. The meaning of competition. In Indi-�idualism and Economic Order, Hayek FA (ed.). Uni-versity of Chicago Press: Chicago; 92–106.

Geroski PA, Jacquemin A. 1988. The persistence ofprofits: A European comparison. Economic Journal98: 375–389.

Jacobson R. 1988. The persistence of abnormal returns.Strategic Management Journal 9: 415–430.

Jacobson R, Aaker DA. 1993. Myopic managementbehavior with efficient but imperfect financial mar-kets: A comparison of information asymmetries inthe US and Japan. Journal of Accounting and Eco-nomics 16: 383–405.

Kormendi R, Lipe R. 1987. Earnings innovations, earn-ings persistence and stock return. Journal of Business60: 323–345.

Miller M, Rock K. 1985. Dividend policy under asym-metric information. Journal of Finance 40: 1031–1051.

Mueller DC. 1990a. The Dynamics of Company Profits:An International Comparison. Cambridge UniversityPress: Cambridge, UK.

Mueller DC. 1990b. The persistence of profits in theUnited States. In The Dynamics of Company Profits:An International Comparison, Mueller DC (ed.).Cambridge University Press: Cambridge, UK; 35–57.

Nelson RR. 1995. Recent evolutionary theorizing abouteconomic change. Journal of Economic Literature 33:48–90.

Odagiri H, Yamawaki H. 1986. A study of companyprofit-rate time series: Japan and the United States.

International Journal of Industrial Organization 4:1–23.

Odagiri H, Yamawaki H. 1990a. The persistence ofprofits in Japan. In The Dynamics of Company Prof-its: An International Comparison, Mueller DC (ed.).Cambridge University Press: Cambridge, UK; 129–146.

Odagiri H, Yamawaki H. 1990b. The persistence ofprofits: International comparison. In The Dynamicsof Company Profits: An International Comparison,Mueller DC (ed.). Cambridge University Press: Cam-bridge; 169–186.

Penrose ET. 1959. The Theory of the Growth of theFirm. Oxford University Press: Oxford.

Peteraf MA. 1993. The cornerstones of competitiveadvantage: A resource-based view. Strategic Man-agement Journal 14: 179–191.

Porter ME. 1990. The Competiti�e Ad�antage of Na-tions. The Free Press: New York.

Robbins HE. 1955. An empirical Bayes approach tostatistics. In Proceedings of the Third Berkeley Sym-posium on Mathematical Statistics and Probability,vol. 1, Neyman J (ed.). University of CaliforniaPress: Berkeley, CA; 157–163.

Roberts PW. 1996. Innovation and firm-level persistentprofitability: A Schumpeterian framework. Workingpaper, Australian Graduate School of Management,March.

Rumelt R. 1987. Theory, strategy, and entrepreneur-ship. In The Competiti�e Challenge: Strategies forIndustrial Inno�ation and Renewal, Teece DJ (ed.).Ballinger Publishing Company: Cambridge; 137–158.

Schumpeter JA. 1942. Capitalism, Socialism, andDemocracy. Harpers: New York.

Stein C. 1955. Inadmissibility of the usual estimator forthe mean of a multivariate normal distribution. InProceedings of the Third Berkeley Symposium onMathematical Statistics and Probability, vol. 1, Ney-man J (ed.). University of California Press: Berkeley,CA; 197–206.

Waring G. 1994. Industry differences in the persistenceof firm-specific returns. Working Paper, Emory Uni-versity, March.

Winter SG. 1987. Knowledge and competence asstrategic assets. In The Competiti�e Challenge: Strate-gies for Industrial Inno�ation and Renewal, Teece DJ(ed.). Ballinger Publishing Company: Cambridge;159–184.

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 251–263 (2001)