THE PARADOX OF SUCCESS: AN ARCHIVAL AND A LABORATORY STUDY OF STRATEGIC … 12 The paradox of...

© Academy of Management fournal2000, Vol. 43, No. 5. 837-853.

THE PARADOX OF SUCCESS: AN ARCHIVAL AND ALABORATORY STUDY OF STRATEGIC PERSISTENCEFOLLOWING RADICAL ENVIRONMENTAL CHANCiE

PINO G. AUDIALondon Business School

EDWIN A. LOCKEKEN G. SMITH

University of Maryland

An archival study of the airline and trucking industries over a ten-year period and alaboratory study revealed that greater past success led to greater strategic persistenceafter a radical environmental change, and such persistence induced performancedeclines. The laboratory study also demonstrated that dysfunctional persistence is dueto greater satisfaction with past performance, more confidence in the correctness ofcurrent strategies, higher goals and self-efficacy, and less seeking of information ixomcritics.

Previous research has shown that past organiza-tional success leads to strategic persistence—a ten-dency for firms to stick with strategies that haveworked in the past (e.g., Lant, Milliken, & Batra,1992; Miller & Chen, 1994). Typically, such persis-tence is beneficial. Success goes to those who de-velop, refine, and enhance key competencies thatlead to sustainable competitive advantage (March,1991). But when the environment changes, thesuccess-persistence relationship may prove detri-mental. Major shifts in competitive, technological,social, and legal conditions may render prior strat-egies ineffective (Haveman, 1992; Smith & Crimm,1987). To ensure alignment with the new environ-mental context, organizations must anticipate ordetect such changes and initiate strategic transfor-mations. Yet, surprisingly, an evolving literaturesuggests that managers often do not respond toenvironmental signals that indicate the need forstrategic change (Hedberg, 1981; Tushman & Ro-

We thank Susan Taylor, Jerry Hage, and Rick Guzzo fortheir useful suggestions during the development of thisresearch. Thanks also to Chris Earley, Dennis Gioia, PeterMoran, Anand Narasimhan, Henry Tosi, Anne Tsui, andthe three anonymous reviewers for their comments onearlier versions of this article. Special thanks to AmirErez for statistical advice. We are also grateful for accessto the airline database assembled by Min-Jer Ghen, Mar-tin Gannon, and Gurtis Grimm and to the trucking data-base prepared by Tom Gorsi. Financial support for thecompletion of this research was kindly provided by Boc-coni University and the University of Maryland.

manelli, 1985). Instead, managers often fall into apattern of dysfunctional strategic persistence.

Past success itself may set the stage for dysfunc-tional persistence. It is easy to assume that a strat-egy that worked in the past will be the most effec-tive strategy in the future (Prahalad & Bettis, 1986).Of course, managers are also exposed to forces driv-ing toward strategic change, such as when dramaticenvironmental change occurs and past strategiesbegin to fail (e.g., Kiesler & Sproull, 1982). Thequestion is. How do managers deal with these com-peting forces? After a period of success, do theystill have the ability to recognize when it is time tochange? Or is the pressure tov/̂ ard persistence cre-ated by past success so strong that it blinds them toearly signals that past strategi(3s may fail?

We addressed these questions in two studies. Inthe first study, we examined the airline and truck-ing industries in the United States over a ten-yearperiod surrounding a discrete and radical environ-mental change, a deregulation. This study adds toprevious research in two important ways. First,although previous research Itias established thatpast success leads to strategic persistence (Boeker,1997; Lant et al., 1992; Meyer, Coes, & Brooks,1993; Miller & Chen, 1994), it has not examinedthis relationship in the context of a discrete andradical environmental change. Second, there havebeen only limited efforts to document the dysfunc-tional consequences of strategic persistence after anenvironmental change.

In the second study, we took the analysis to adeeper level by examining a set of factors that maymediate and thus explain the effect of past success

837

838 Academy of Management Journal October

on Strategic persistence. In contrast to prevalentaccounts of inertia, which call into consideration avariety of structural constraints (Hannan & Free-man, 1984), our focus here is on psychological me-diators. The idea that the psychological processesof organizational decision makers mediate strategicrigidity has not gone unrecognized (Hambrick &Finkelstein, 1987; Staw & Ross, 1987). For example,a number of investigators have argued that pastsuccess leads to complacency and the formation ofrigid cause-and-effect beliefs (Kiesler & Sproull,1982; Prahalad & Bettis, 1986). Others have pointedto the importance of external monitoring activities(Dutton & Duncan, 1987; Miller & Chen, 1994) andaspirations (Creve, 1998; Lant & Montgomery,1987). However, no study to date has examined theactual psychological processes that underlie strate-gic persistence. Our second study was an attempt tofill this gap in the literature.

The two studies complement each other. Thearchival study explores the effects of past successon persistence and the effects of persistence onperformance in real organizations experiencing adramatic environmental change. However, this re-search was carried out without direct observationof the managerial decision-making process. In con-trast, the laboratory study, while replicating salientfeatures of the archival study, focuses directly onsome of these decision processes in a controlled,simulated environment. Therefore, the archivalstudy provides external validity and generalizabil-ity, whereas the laboratory study provides internalvalidity and control.

MACRO ANALYSIS OF THE EFFECT OF PASTSUCCESS ON DYSFUNCTIONAL STRATEGIC

PERSISTENCE

Key Questions

Two questions have guided previous research onsuccess and persistence. The first is. Does successaffect persistence? Empirical evidence clearlylends support to a success-persistence causal link.In a study of the furniture and software industries,Lant, Milliken and Batra (1992) found that organi-zations with performance above the industry aver-age over the period 1980-84 were less likely thanother industry firms to reorient their strategies inthe two-year period 1984-86. In a study of 450California hospitals over a period of 11 years,Meyer, Goes, and Brooks (1993) found that highperformance led to fewer strategic reorientations.Miller and Chen's (1994) study of the airline indus-try in the postderegulation period between 1979and 1986 showed that a company's previous per-

formance was associated with higher levels of com-petitive inertia, measured as the number of changesin competitive practices compared to rivals' num-bers of such changes. Boeker's (1997) study of 67semiconductor producers over a 14-year periodshowed that high past performance made strategicchanges less likely. Consistent with this pattern offindings, in a study of the radio broadcasting indus-try, Greve (1998) showed that, as an organization'sperformance increased, the probability of changedecreased.

It must be noted that past performance is likely tocause persistence over a period of time only iforganizations are consistently successful. If organi-zations experience a mixed series of increases anddeclines in performance, strategic persistencewould not necessarily be expected. For this reason,when we use terms such as "success," "past suc-cess," or "history of success" in this article, werefer to a series of past, positive performance out-comes.

The second question that has guided relevantpast research is. Why does success foster persis-tence? Previous research has proposed several ex-planations. One account, rooted in the reinforce-ment-expectancy model of learning, is thatorganizations tend to repeat actions that are asso-ciated with positive outcomes (Cyert & March,1963; Prahalad & Bettis, 1986). A second explana-tion suggests that organizations become committedto retaining proven competencies, because doing sois more efficient than trying to develop new ones(Levitt & March, 1988). A third thesis is that organ-izations are not motivated to change when theirperformance meets or exceeds their aspiration lev-els (Greve, 1998; Lant & Montgomery, 1987). Obvi-ously, these three explanations have much in com-mon in that they all rest on an assumption thatpersistence stems from the thinking processes ofstrategic decision makers. A fourth account stressesinstead the role of structural constraints. In thisview, past success favors the development of rigidorganizational structures and increases the pres-sure for stability coming from external stakeholders(Hannan & Freeman, 1984). Such constraints inturn stifle decision makers' ability to alter a currentstrategy.

Research on the success-persistence effect hassomewhat neglected a third question, namely.What are the performance consequences of strate-gic persistence? Theory suggests that strategicpersistence can be a double-edged sword for orga-nizations (March, 1991). Under conditions of envi-ronmental stability, persistence can be highly ad-vantageous. It facilitates the development ofcompetencies that have proven valuable in the

2000. Audia, Locke, and Smith 839

past, increases efficiency and quality and, as a con-sequence, builds legitimacy with external stake-holders (Hannan & Freeman, 1984). Persistencealso facilitates learning because in the absence ofchange, cause-effect relationships are clearer-cut(e.g., Levine, 1971). Furthermore, persistence re-duces risks, both economic and social, stemmingfrom low cohesion (Shaw, 1976), discouragement(Bandura, 1986), or distrust (Hollander & Julian,1969).

Nonetheless, a persistence-success pattern of be-havior may become self-destructive if it leads topersistence in the face of major environmentalshifts, such as technological breakthroughs, regula-tory changes, or alterations in trade barriers (Have-man, 1992; March, 1991; Smith & Grimm, 1987).After a period of prior success and persistence,organizations may be slow to recognize that it istime to change. Driven by the natural tendency tocontinue exploiting previously effective strategies,successful organizations may ignore the implica-tions of major environmental changes until drasticperformance declines compel new strategies (Tush^man & Romanelli, 1985).

Surprisingly, most previous empirical researchon the success-persistence relationship has over-looked the distinction between the beneficial ef-fects of persistence under conditions of environ-mental stability and the detrimental effects ofpersistence under conditions of radical environ-mental change. Nonetheless, three studies haveshown negative effects of strategic persistence onperformance following a radical environmentalchange, though they did not examine the relation-ship between success and persistence. Smith andGrimm (1987) studied the effects of railroad dereg-ulation on the strategy-performance relationship.They found that railroad companies that did notchange their strategies after deregulation performedworse than those that did. Haveman (1992) foundthat persistence with past strategies on the part ofsavings and loan organizations after a radical envi-ronmental change decreased performance. Finally,Zajac and Kraatz (1993) found that strategic persis-tence in the educational programs of American lib-eral arts colleges in response to the cumulativeeffect of environmental changes decreased perfor-mance.

Hypothesis, Study 1

In sum, our review suggests that, although thereis evidence showing that success leads to persis-tence and that persistence, in the context of a rad-ical environmental change, has dysfunctional per-formance consequences, past research has not

examined the whole sequence, success-persis-tence-performance. To address this deficiency, wetested the following hypotheses in study 1 of thisresearch:

Hypothesis la. Following a discrete and radi- .cal environmental change, organizations witha greater history of success are more likely topersist with their past strategies than thosewith a lesser history of success.

Hypothesis lb. Following a discrete and radi-cal environmental change, organizations witha high level of persistence with their past strat-egies are more likely to experience a reductionin their performance than those with less per-sistence.

MICRO ANALYSIS OF THE PROCESSESMEDIATING THE EFFECT OF PAST SUCCESS

ON DYSFUNCTIONAL STRATEGICPERSISTENCE

We noted above several accounts of the reasonswhy past success may lead to persistence. We nowconsider the mediating role of the individual psy-chological processes and behaviors of organiza-tional decision makers, a causal link often over-looked in previous research on inertia. Someresearchers consider information seeking the criti-cal factor. They have suggested that success de-creases information seeking and that less seeking ofinformation reduces the ability of managers to rec-ognize environmental changes (e.g.. Miller & Chen,1994). Others, however, are not entirely convincedthat information seeking is sufficient to motivatechange in strategy and have emphasized insteadthe role of cognitions. Adopting this view, someauthors have suggested that people may gather in-formation about environmental changes and stillmiss their impact on future performance becausetheir rigid beliefs act as distorting filters (Kiesler &Sproull, 1982), whereas yet others have proposedthat success facilitates the development of rigidmental maps that lock people into patterns of ac-tion that are hard to modify (Prahalad & Bettis,1986).

Although these micro explanations give promi-nence to particular processes, some scholars haveproposed broader arguments that comprise all themicro processes that seem to be involved in thesuccess-persistence relationship, including thosementioned above and others such as complacencyand overconfidence (Miller, 1993; Milliken & Lant,1991). Surprisingly, however, as Milliken and Lant(1991) noted, very little research has examined thevalidity of any of these theoretical explanations. In


our second study, we addressed this gap. Ground-ing our analysis in previous research, we focusedon concepts that are closely related to past perfor-mance and. strategic rigidity. These concepts aresatisfaction, beliefs, self-efficacy, goals, and infor-mation seeking.

When people succeed in attaining their objec-tives or values, they experience satisfaction, whichwe view, drawing on Locke and Latham (1990), asan affective response based on a subconsciousvalue appraisal. Following success, individualsalso develop beliefs regarding cause-effect relation-ships—that is, the connection between their ac-tions and the outcomes achieved. Such beliefs in-clude conclusions about what particular taskstrategies are effective in performance. People alsodevelop beliefs about their own competence or self-efficacy. Past success enhances efficacy, and effi-cacy motivates them to set higher goals for thefuture (Bandura, 1986; Locke & Latham, 1990). It isnot the case, however, that individuals focus onlyon the past when planning future actions. Peoplehave the power of forethought (Bandura, 1986,1997). They can anticipate what will happen, forexample, by seeking information about the futureand projecting its implications. We now considerhow past success can make these processes operatein a detrimental fashion.

Mediating Processes

Satisfaction with performance. It is well knownthat success in attaining one's goals leads to satis-faction [e.g., Bandura, 1986; Locke & Latham,1990). The more the standard is surpassed, thestronger the feeling of satisfaction the person expe-riences. Satisfaction is pleasurable and can giveindividuals fuel for further action, but the danger isthat these positive feelings reduce individuals' mo-tivation to initiate new adaptations. Further, strongfeelings of satisfaction can lead individuals to in-terpret warning signs as benign or positive, thuseliminating the possibility of questioning whetherthey are taking the correct actions (Isen & Baron,1991). Complacency, a common correlate of satis-faction, is often mentioned as an explanation of thelink between success and dysfunctional persis-tence. For example. Miller and Chen posited that"success can make managers so complacent, socontent with the status quo, that they resist change"(1994: 3).

Confidence in the effectiveness of current strat-egies. Successful individuals, even if their pastachievements were due to accidents of timing, canbecome confident about the effectiveness of theiractions (March & Olsen, 1976) and so treat their

beliefs about strategy-performance links as fact.Past research suggests that high confidence incause-and-effect beliefs increases individuals' ef-fort and persistence (Vroom, 1964) and also leadsthem to persist with strategies that were successfulin the past (Schwartz, 1982). Such confidence inthe continued efficacy of previously successfulstrategies is beneficial if the conditions that pro-duced success do not change but detrimental ifconditions do change. The same holds for cause-and-effect beliefs about organizational strategies;cause-and-effect beliefs based on the past can in-hibit future change because they can be retainedeven when they no longer apply (Kiesler & Sproull,1982; Prahalad & Bettis, 1986).

Self-efficacy. In social cognitive theory, task-specific confidence is known as self-efficacy, theperception of one's capabilities to attain perfor-mance outcomes. Self-efficacy is a natural conse-quence of "enactive mastery" (Bandura, 1986,1997). Thus, the better people perform on a taskover a period of time, the higher their confidence inbeing able to perform well in the future. Previousresearch shows that higher self-efficacy has severalbeneficial effects: it facilitates high future perfor-mance, encourages the setting of high performancegoals, strengthens commitment, fosters the selec-tion of effective task strategies, and motivates pos-itive responses to negative feedback (Bandura,1997).

It remains unclear, however, whether the benefi-cial effects of self-efficacy apply to dynamic taskenvironments. To be an accurate predictor of futureperformance, self-efficacy must be based on accu-rate feedback regarding past performance on thesame task, performed under the same conditions.When the task changes, efficacy beliefs based ex-clusively on past performance may become an in-accurate guide to the future. If individuals antici-pated changes in the task and reassessed theirabilities in light of new task requirements, thenthey would avoid forming inaccurate efficacy be-liefs. However, this may not happen because, asconsecutive successes accumulate, past achieve-ments become the primary influence in the forma-tion of efficacy beliefs, and possibly counteractinginfluences lose their potency (e.g., Lindsley, Brass,& Thomas, 1995). Falsely assuming that what theydid in the past will continue to work, individualsare likely to overestimate their ability to perform inthe new situation.

Goals. Previous research indicates that, follow-ing success, people tend to raise their goals (Locke& Latham, 1990). This occurs because individualsassume that they have learned from successful ex-periences and, as a result, perceive themselves as

2000, Audia, Locke, and Smith 841

capable of additional improvement. Less is knownabout the effect of high goals on strategic persis-tence following a radical environmental change.Earley and Perry (1987) showed that hard goalsenhanced the use of "primed" strategies, whetheror not these were effective. A strategy is primedwhen the person who is choosing a strategy is giveninformation that induces a certain mental set. Sub-jects with hard goals did better than subjects with"do-your-best" goals when the primed strategy wassuitable. However, when the primed strategy wasinappropriate to the task, subjects with hard goalsdid worse than subjects with do-your-best goals.Thus, successful individuals may stick to outdatedstrategies because past success has primed them touse what worked in the past. When higher futuregoals are linked with incorrect strategies, this lethalcombination can cause performance to drop morerapidly than it would have if high goals had notbeen set.

Amount of information sought Miller and Chen(1994) postulated that success may be interpretedas a sign that less vigilance and less environmentalscanning are required. They hypothesized that re-duced scanning decreases the motivation to under-take corrective adjustments. This argument is con-sistent with the literature on information seeking.Ashford and Cummings (1983) proposed that indi-viduals' motivation to seek information depends onthe value they place on additional information.This is affected by the importance placed on attain-ing a given goal as well as the degree of uncertaintyover the behaviors appropriate for attaining a goal.Enduring success is likely to lead individuals toseek less information because success increases in-dividuals' certainty that they are already doing theright thing. Because noticing changes in an envi-ronment depends, in part, on the amount of infor-mation sought (Daft & Weick, 1984; Kiesler &Sproull, 1982), it is expected that the less strategicdecision makers seek information, the higher theirpersistence with past strategies following a radicalenvironmental change will be.

Type of information sought. Less attention hasbeen given to the potential mediating role of thetype of information sought. It seems clear that ex-ecutives who have had a long record of success willcombine strong cause-and-effect beliefs with confi-dence that they can attain challenging goals usingthe strategies that worked in the past. Such confi-dence may lead them to seek to maintain theirpositive self-images (e.g., Ashford, 1989). They arethus more likely to disparage and reject those whoquestion their competence and recommend thatnew strategies are in order. After all, great businessleaders often achieve success because they used

their own judgments and ignored others' opinions.Such leaders will be prone to prefer the companyand advice of those who agree with them and sup-port the current strategies (Nystrom & Starbuck,1984). Such an information-seeking pattern wouldreduce the quality of external monitoring activitiesand thereby lower executives' capacity to adjust tochanging circumstances.

Hypotheses, Study 2

If study 1 were to establish that success leads topersistence and that persistence, in the context of aradical environmental change, is dysfunctional interms of performance, then study 2 would need tofirst replicate these findings to allow examinationof the potential psychological explanations. Thus,Hypotheses 2a and 2b replicate Hypotheses la andlb from study 1, and Hypothesis 3 formulates themediating role of the psychological processes dis-cussed above.

Hypothesis 2a. Following a discrete and radi-cal environmental change, individual strategicdecision makers with a greater history of suc-cess are more likely to persist with their paststrategies than those with a lesser history ofsuccess.

Hypothesis 2b. Following a discrete and radi-cal environmental change, individual strategicdecision makers who persist with their paststrategies are more likely to experience a re-duction in their performance than those withless persistence.

Hypothesis 3. The effect of past success onindividual strategic decision makers' persis-

- tence with past strategies following a discreteand radical environmental change will be me-diated by (a] greater satisfaction with the cur-rent level of performance, (b) greater confi-dence in the effectiveness of the currentstrategies, (c) higher self-efficacy, (d) highergoals, (e) a smaller amount of information ac-quired, (f) a greater amount of information ac-quired from favorable sources, and (g) asmaller amount of information acquired fromunfavorable sources.

METHODS, STUDY I

Data

The airline industry. From 1938 to 1978, theCivil Aeronautics Board (CAB) controlled indus-try's entries, exits, and pricing. Deregulation in Oc-tober 1978 removed such controls and was a dis-


continuous environmental change that affected theindustry's competitive environment (Vietor, 1990).Following earlier studies (e.g.. Smith, Grimm, &Gannon, 1992), in our analysis we relied on datafrom the U.S. Department of Transportation con-cerning certified air carriers. To identify differentpatterns of performance history and strategic per-sistence, the interval of time examined covered thefive years preceding the deregulation, 1974-78,and the five years following it, 1979-83. The sam-ple included 25 companies certified during the en-tire period.

Before the deregulation, carriers . competedmainly by adding flights to the routes that theywere currently serving and by increasing the qual-ity of service. After the deregulation, airlines de-vised new strategies to take advantage of the newcompetitive context. New strategies included con-necting flights at particular airports ("hubbing"),which allowed carriers to schedule itineraries withone-hour stopovers, gather traffic from diffusesources, and increase the load factor on previouslythin routes; offering low-cost, limited service for alow fare; and focusing on customers who couldafford higher prices by emphasizing service, trafficcontrol, and distribution.

The trucking industry. From 1935 to 1980, theInterstate Commerce Commission (ICC) regulatedthe motor carrier industry primarily with restric-tive entry policies and controls on prices. By elim-inating these constraints, the Motor Carrier Act of1980 was a radical environmental change that sig-nificantly altered the competitive environment(Corsi & Stowers, 1991). Although the deregulationhad severe consequences for the entire industry,experts suggest that it posed different threats to twotypes of companies: those that handled shipmentsweighing less than 10,000 pounds (less-than-truck-load companies, or LTL) and those that handledshipments of a weight of more than'10,000 pounds(truckload companies, or TL) (Corsi & Stowers,1991).

We examined the impact of deregulation on theLTL segment of the trucking industry. The analysisrelied on data filed with the ICC and publishedannually as the Motor Carrier Annual Report by theAmerican Trucking Association. Data concerned aten-year period comprising the five years before thederegulation, 1976-80, and the five years after it,1981-85. Furthermore, given the large number ofcompanies in the LTL segment (over 2,000), manyof which were very small, we analyzed only LTLcompanies with revenues higher than $1 millionfor which data were available for at least nine of theten years considered in the analysis. This proce-dure yielded a sample of 125 companies.

Before the deregulation, most companies at-tempted to attract customers either by concentrat-ing on the quality of service or by specializing onthe transportation of certain products. After thederegulation, as price competition spread throughthe industry, companies had to find ways to con-tain costs and improve their efficiency. New strat-egies included containing costs by increasing theuse of company drivers, concentrating on high-density, long-haul corridors that allowed higherprices, and reducing labor costs by increasing useof independent truck owners.

Measures

Strategic persistence. Following Finkelsteinand Hambrick (1990), we defined strategic persis-tence as the extent to which a firm's strategic pro-file remained stable over time and measured it byexamining the stability of financial and operationalratios that express the strategic position of a com-pany, on specific issues. For example, R&D expen-diture divided by total revenues is a classic indica-tor of a company's R&D intensity.

For both industries, we selected indicators thatwere expected to change in response to the dereg-ulation. Strategic indicators used for the airlineswere marketing expenses per mile, general ex-penses per mile, equipment expenses per mile, per-centage of scheduled aircraft miles completed, firstversus economy class, first-class revenue-passen-ger load factor, and coach-plus-economy revenue-passenger load factor. Strategic indicators for thetrucking industry were less-than-truckload reve-nues versus other revenues, average revenue perton mile, fuel expenses per mile, amount of ser-vices bought from independent truck owners as apercentage of operating expenses, average load, av-erage length of haul, and total number of trucksdivided by the total number in operation.

The measure of strategic persistence was thencomputed as follows: (1) for each strategic indica-tor, the variance was calculated over the five-yearperiod following the deregulation; (2) variancescores were standardized and multiplied by - 1 sothat positive scores indicate greater persistence;and (3) the standardized indicators were summedto yield an overall measure.

Past performance. This study uses return onsales (ROS) as a measure of performance. Becausesales are very visible to managers, they are likely tobe used as an indicator of performance in small aswell as in large companies. As a result, ROS is oftenused to evaluate a company's performance overtime, to compare it with other companies, and toset future goals. For the trucking industry, we were

2000 Audia, Locke, and Smith 843

also able to compute a second measure of perfor-mance, using return on assets. We computed pastperformance using a procedure similar to that oneadopted by Lant and colleagues (1992). First, pastperformance was calculated for the five years pre-ceding the deregulation. Second, it was assumedthat managers used the average performance of theindustry as reference point to which they comparedtheir own companies' performance. Thus, a firm'sdeviation from the industry median was computedin each of the five years and then averaged.

Change in performance after the environmen-tal change. We adopted the same measure used forpast performance to calculate performance in thefive years after the deregulation. We then computeda difference measure by subtracting performancebefore deregulation from performance afterward.

Control variables. A critical control variable wasorganizational size. Larger organizations are ex-pected to change their strategies less frequentlybecause they are encumbered by greater structuralinertia (e.g., Hannan & Freeman, 1989). We mea-sured size using, for the trucking industry, the nat-ural logarithm of total tons transported (Corsi &Stowers, 1991) and, for the airline industry, thenatural logarithm of total assets (Kelly & Amburgey,1991).

We also controlled for market diversity becauseorganizations serving a variety of customers inter-act with a more heterogeneous environment and

thus collect more information that may enhancetheir capacity to react to external shifts (e.g.. Miller& Chen, 1994). Given that airlines can carry anycombination of passengers, mail, freight, and cargo,we measured market diversity by calculating thepercentage that the largest product line contributedto the total product mix (Kelly & Amburgey, 1991).For the trucking industry, market diversity wasmeasured as the percentage of revenues that de-rived from shipments of a weight of more than1,000 pounds.

Finally, because changes in top management typ-ically facilitate strategic change (e.g., Finkelstein &Hambrick, 1996), we included in our analyseschange in chief executive officer in the year inwhich deregulation took place, that is, between1977 and 1978 for the airline industry and between1979 and 1980 for the trucking industry. There wasone instance of CEO change in the airline industry,and there were seven cases in the trucking indus-try. We obtained this data for the airline industryfrom Moody's Transportation Manual and, for thetrucking industry, from the Executive and Owner-ship Report published by the American TruckingAssociation.

RESULTS, STUDY I

Table 1 presents descriptive statistics and corre-lations, and Table 2 reports regression analyses

TABLE 1Descriptive Statistics and Correlations, Study 1

(la) Airline Industry"

Variable

1. Market diversity2. Size3. CEO change, 1977-784. Strategic persistence, 1979-835. Past performance, 1974-786. Change in performance

(lb) Trucking Industry*"

Variable

1. Market diversity2. Size3. CEO change, 1979-804. Strategic persistence, 1981-855. Past performance, 1976-806. Change in performance

Mean

0.1213.190.040.000.000.00

Mean

0.0812.330.050.000.000.00

s.d.

0.061.880.204.260.080.18

s.d.

0.091.160.232.700.040.04

1

.16-.34

.05

- . 41*-.23

1

-.27*-.13^- .23*

.03-.08

2

-.16.74*.15

-.22

2

-.09.21*.23*

-.16*

3

.10

.51*-.10

3

.09-.09-.13^

4

.68*-.39*

4

.25*-.12'^

5

-.36*

5

-.44*

n = 25.'11 = 125.

^p < .10* p < .05

844 Academy of Management Journal Octoher

TABLE 2Results of Regression Analyses for Strategic Persistence and Change in Performance over the Five Years

after Deregulation, Study 1

Variable

Market diversitySizeCEO changePast performanceStrategic persistence

+ p < .10* p < .05

Strategic

-.05.62*.18

.40*

Airline

Persistence

.11

.43*-.24+

.83*

.82*

.42*

Industry

Change in

-.42*-.42*-.27+

.25*

Performance

-.46*.09

-.13

-.82*

.64*

.39*

Strategic

-.17*.18*.09

.09*

Trucking

Persistence

-.19*.12+.10.24*

.14*

.05*

Industry

Cbange in

-.12+-.18*

.10

.05*

Performance

-.14+- . 1 6 *

.11

-.13+

.07*

.02

conducted to test Hypotheses la and lb. For bothindustries, results indicate that the greater the per-formance over the five years prior to deregulation,the more organizations persisted with the paststrategies in the five years following the deregula-tion (airline industry: /3 = .83, p < .05; truckingindustry: fi = .24, p < .01). This pattern of findingssupports Hypothesis la. In the case of the truckingindustry, we also conducted a supplemental regres-sion analysis using ROA as a measure of perfor-mance. Results confirmed those obtained usingROS.

Results also reveal that the greater the strategicpersistence afrer deregulation, the greater the de-cline in performance after deregulation, though thiseffect was only of borderline significance in thetrucking industry (airline industry: /3 = -.82, p <.05; trucking industry: )3 = - .13 , p < .10). Overall,these findings support Hypothesis lb. The onlycontrol variable producing findings largely consis-tent with expectations was size. In both industries,size increased persistence and led to greater dropsin performance after the deregulation, though theeffect on change in performance became weakerafrer persistence entered the analysis. Market diver-sity decreased persistence in the trucking industrybut not in the airline industry. Finally, after pastperformance was added to the analysis, CEOchange had a weak, negative effect on persistence,but only in the airline industry.

METHODS, STUDY 2

The Business Simulation

To facilitate conducting a laboratory investiga-tion of strategic persistence, the first author createdan interactive, computer-based simulation, the Cel-

lular Industry Business Game, in collaborationwith a sofrware company. Perspective Visuals Inc.This game draws on events that occurred in theU.S. cellular telephone industry. Before playing thesimulation, participants read an introductory case,Douglas Cellular Inc. The case provided back-ground information concerning the cellular phoneindustry and the three-year history of Douglas Cel-lular Communication Inc. Participants played therole of MI. Douglas, the founder/CEO and sole de-cision maker for the company, for 13 decision pe-riods, each corresponding to a year of activity. Eachparticipant acted as a separate company. Partici-pants were told that their aspiration was to becomemarket share leader in the northeast region byachieving a 25 percent market share; each startedthe simulation with a 7 percent market share. Al-though participants were told that the simulationwas going to last 15 time periods, the experimenter(the first author) interrupted the game at the end ofthe 13th period to prevent endgame effects.

During each decision period, participants wererequired to make strategic decisions concerning thefollowing areas of activity: pricing, research anddevelopment, advertising, cost containment, salesforce, radio wave capacity, additional products,geographic scope, finance, and alliances with othercompanies. Within each area, participants were al-lowed to take various strategic actions. For exam-ple, in the sales force strategic area they could (1)specify the number of employees in the direct salesforce as well as the number of dealers, (2) indicatethe salary of the direct sales force as well as thecommission for the dealers, and (3) allocate bothdirect sales force and dealers by market.

For each decision period, participants first im-plemented their strategic decisions and then ob-


tained performance feedback. Performance feed-back was the result of previous strategic decisions.The business simulation included a complex set offormulas linking strategic actions to performanceconsequences. These formulas varied over time toreflect changes in the industry. For example, theimpact of advertising expenditures on sales foreach decision period was determined by a coeffi-cient that varied depending on the stage of theindustry.

In addition to making strategic decisions, in eachdecision period except the first one, participantswere allowed to request five types of information:information from the Cellular Industry Associationon topics like communications from the FederalCommunications Commission (FCC) and the size ofthe market; information concerning the industry,about topics like new technologies and competi-tors' strategic actions; information on customers,such as their criteria for selecting cell phone ser-vice and hours spent using such services; confiden-tial information and opinions on the effectivenessof specific strategies from industry executives whohad been supportive of Mr. Douglas's ideas in thepast; and confidential information and opinionsconcerning the effectiveness of specific strategiesfrom executives who had questioned Mr. Douglas'sjudgment in the past. The first type of informationcovering any event considered essential for the in-dustry was provided to each participant free ofcharge. The other four types of information wereprovided upon request, each at a cost of $25,000.This information was the same for each participantand, over time, portrayed the different stagesthrough which the industry evolved.

Evolution of the Cellular Service Industry

The evolution of the industry was predeterminedin the game. As in the two industries examined instudy 1, the discrete and radical environmentalchange occurring in the industry in the simulationwas a deregulation. Participants were given severalmessages warning that deregulation was likely tooccur. They received official communication aboutit from the FCC one decision period before it wasimplemented. Participants who were more activeinformation seekers were exposed to alternativeviews about the potential impact of the environ-mental change.

During the first eight decision periods of the sim-ulation, competition was restricted in the four re-gions (that is, the government awarded 20 licensesin each region), and the market was characterizedby steady growth (a 20-30 percent increase in sub-scribers per year). Douglas Cellular Communica-

tions, like the other 19 companies operating in thenortheast region, benefited from the growth of theindustry by gaining new subscribers as well as byincreasing revenues. However, to increase its mar-ket share, it had to achieve rates of growth largerthan those of the industry (for instance, an annual50 percent increase in subscribers). The effectivestrategies were buying licenses to operate in all fivemarkets, acquiring additional radio wave capacityallowing it to carry more calls and avoid networkjams, raising capital to finance those new invest-ments, vigorously increasing sales force and adver-tising expenditures, and concentrating advertisingefforts on business users rather than on privateusers.

In the last five decision periods, two importantenvironmental changes altered the competitivecontext: the U.S. government eliminated the re-gional barriers and allowed competition across re-gions, and the rate of growth of the northeast regiongradually declined from 30 percent to 12-14 per-cent. Because of these changes, competition be-came more intense and was focused on "stealing"subscribers, rather than on acquiring new ones. In-creases in sales force and advertising, which in theprevious stage of the industry were the basic ingre-dients of success, resulted in smaller increases insubscribers, losses in market share, and large re-ductions in operating profits. Price cuts, a strategythat had been ineffective in the previous stage, andmaking alliances to ensure wider geographic cov-erage were the most effective strategies. In conclu-sion, owing to the combined effects of the deregu-lation and the slower growth of the industry, thestrategies that had worked in the past were nolonger effective.

Design and Participants

The study employed a one-by-three design withthree levels of past performance (low, moderate,and high success). Participants, who were ran-domly assigned to the three conditions of the sim-ulation, were 168 graduating seniors [77 womenand 91 men) enrolled in a strategic managementcourse at a large U.S. university. The experimentwas conducted in a computer laboratory in sevensessions with 20-25 individuals in each. Each per-son worked individually in a work area separatedfrom the others by partitions. Participants in thesame session were randomly assigned to differentperformance conditions. The experimenter was notaware of the condition assigned to each participant.

The treatment consisted of providing partici-pants at the beginning of the simulation with dif-ferent levels of task knowledge, which led to dif-


ferent levels of success. Before beginning to playthe role of Mr. Douglas, participants were givensupplemental information consisting of histhoughts about the strategies that v̂ fere most likelyto be effective in the future. All participants weretold to pay attention to advertising, sales force, andfinance as crucial strategic areas for increasing theirmarket shares in the following years. Participantsassigned to the moderate and high success condi-tions also received other detailed suggestions. (In-terested readers can obtain additional informationon the experimental procedure from the first au-thor.)

Although the first author provided tips abouteffective strategies in the first two decision periods,participants had considerable time to make theirchoices and to revise their strategies over time. Atthe end of time period 9, they were asked questionson satisfaction with performance, self-set goals,self-efficacy, and confidence in strategy-perfor-mance relationships. At the end of time period 13,which was also the end of the simulation, partici-pants were asked to answer a question on theirperception of environmental changes.

Measures

Experimental condition. The experimental treat-ment consisted of three levels of task knowledgeprovided at the beginning of the simulation that ledto different levels of success. The three conditionswere coded as follows: 1, low success; 2, moderatesuccess; and 3, high success.

Persistence with past strategy. Although instudy 1 our measure of strategic persistence re-ferred to the overall strategic profile, in study 2 wewere able to use a finer-grained measure. Becausethe effective strategies were predetermined in thesimulation and were identical for all participants,we could directly examine participants' strategicactions. In the Cellular Industry Business Game,high performance in the interval between year 1and year 8 resulted from a strategy consisting ofentering new markets, buying additional radiowave capacity, and aggressively increasing salesforce and advertising expenditures. After year 8,entering new markets and accumulating radio wavecapacity lost relevance because licenses to operatein new market^ were no longer available and mostparticipants had accumulated spare radio wave ca-pacity. Thus, after year 8, persistence with the paststrategy primarily meant continuing to increase ad-vertising and sales force at rates similar to those ofthe earlier stage.

Since success resulted from the use of an accel-erated strategy, no change or a small change in sales

force and advertising represented nonuse or de-creased use of the old strategy and, therefore, lowpersistence. Thus, we measured persistence withthe past strategy by summing the standardized vari-ance of advertising expenditure and sales force permillion people served between years 8 and 13. Ahigh score indicated high persistence with the oldstrategy, and a low score indicated low persistence.

A large variance score over the interval of timeconsidered might also indicate a pattern of reduc-tion in advertising investment and sales personnel.However, this was not the case in this simulationbecause participants displayed a tendency to in-crease or to hold steady their investments in salesforce and advertising. In fact, the variance of thetwo strategic indicators used was positively corre-lated with their respective difference score (salesforce, r = .90, p < .01; advertising, r = .88, p < .01].We preferred the sum of the standardized varianceof the two strategies over the sum of the differencescores because it provided a finer-grained way todiscriminate among participants.

Change in performance after the environmen-tal change. Our measure was change in operatingprofits between decision period 8 and decision pe-riod 13. In our exploratory analyses, we also usedchange in return on sales between the same twoperiods. Since these two measures produced thesame pattern of results, we decided to use changein operating profit for our main analyses because itwas more directly linked to the use of previouslyeffective strategies.

Mediating variables. We measured satisfactionby computing the means of responses (1 = not atall, 7 = totally) to these two questions (a = .94]: "Iam very satisfied about last year's market share"and "I am very satisfied about last year's overallperformance." Belief in current strategies was mea-sured as the mean of two items expressing confi-dence (0-100; a = .78] that increasing the salesforce and increasing advertising expenditureswould have a positive impact on market share inthe future. Self-efficacy was the sum of the stan-dardized responses to the following two items (a =.97]: "Assess your confidence in achieving the fol-lowing thirteen levels of market share or higher inthe next year by rating Y [yes], for each level, if youthink you can achieve that level of market share orhigher" (range of total number of yes answers,0-13] and "Assess your confidence in achievingthe following thirteen levels of market share orhigher in the next year by using, for each level, anumber between 0 and 100 to indicate how confi-dent you are in achieving that market share level orhigher." The variable goals was measured by thesum of responses to these two items (a = .96]:


"Indicate the level of market share that you aretrying to attain next year" and "What is the mini-mum acceptable level of market share for the nextyear?" These items were filled out at the end of year9 because that was when the deregulation started toaffect the rules of competition. The total amount ofinformation was obtained by summing the informa-tion requests made between years 9 and 13. Favor-able information was the amount of informationsought in those years from executives who hadbeen supportive in the past. Unfavorable informa-tion was the amount of information sought in thoseyears from executives who had been critical in thepast.

Control variables. We used three participantcharacteristics as control variables: gender (male =0, female = 1], class grade (C = 1, B = 2, A = 3],and grade point average.

Analyses

We verified the assumption that the measures ofthe concepts were distinct using confirmatory fac-tor analysis. This procedure permits testing the sig-nificance of the number of factors in a data set aswell as the structure of those factors. Confirmatoryfactor analysis, which utilizes covariance structuremodeling, provides commonly accepted fit statis-tics: the comparative fit index (CFI], the incremen-tal fit index (IFI], and the root-mean-square resid-ual (RMSR] (Bentler, 1990; BoUen, 1988]. Weestimated covariance structure models usingLISREL VII (Joreskog & Sorbom, 1989]. We testedHypotheses 2 and 3 using hierarchical regressionanalysis.

RESULTS, STUDY 2

Treatment Checks

Manipulation checks for degree of success con-sisted of the two following questions: "How do youevaluate your current market share compared toyour market share at year 1?" and "How do youevaluate your current overall performance com-pared to your overall performance at year 1?" (1 =very unsuccessful, 7 = very successful]. Becauseresponses to these questions were highly correlated(a = .94], we used the mean to form a single mea-sure. The means and standard deviations of thismeasure, taken at year 9, were 6.08 (1.16], 5,25(1.90], and 2.63 (1.81] for the high success, moder-ate success, and poor performance conditions, re-spectively. An analysis of variance (ANOVA] re-vealed that these means were significantly different(^2, 165 = 64.94, p < .01].

To verify that the simulation provided sufficientinformation to reveal a radical environmentalchange over time, after participants had receivedthe last performance feedback we asked themwhether, in the last few decision periods, the FCChad made decisions that affected the entire indus-try (1 = strongly disagree, 7 = strongly agree]. Thedistribution of the answers to this question waspositively skewed (x = 5.31, s.d. = 1.98] and soindicated that, on the average, the environmentalchange was noticeable. This item was not corre-lated to persistence with the current strategy. None-theless, we used awareness of the environmentalchange as an additional control variable in ourmediation analyses.

Confirmatory Factor Analysis

We tested the following alternative models toexplore the possibility that some of the conceptswere not distinct: a model with all variables loadedon one factor (model 1]; a hypothesized model(model 2]; a model in which goals and self-efficacyitems were loaded on one factor (model 3]; a modelwhere satisfaction and self-efficacy items wereloaded on one factor (model 4]; a model with sat-isfaction and goals on one factor (model 5]; a modelwhere all information items were loaded on onefactor (model 6]; and, finally, a model in whichfavorable information sought and unfavorable in-formation sought were loaded on one factor (model7). Confirmatory factor analysis revealed thatmodel 2, the hypothesized model with nine dis-tinct factors, had better fit indexes (CFI = .97, IFI =.97, RMSR = .03] than all the other models.

Test of Hypotheses

Table 3 shows means, standard deviations, andcorrelations. Table 4 reports results of the regres-sions conducted to test Hypotheses 2a and 2b andHypothesis 3. Model 2 in Table 4 shows that, afterthe control variables had been included, past suc-cess had a significant effect on strategic persistence(j3 = .40, p < .05], thus confirming Hypothesis 2a,Furthermore, model 5 in Table 4 shows that persis-tence with old strategies decreased performanceafter the environmental change (/3 = -.74, p < .05],thus supporting Hypothesis 2b. Recall that it wasnecessary for us to replicate the conditions of study1 to be able to investigate the psychological factorsthat mediate the success-persistence relationship.

To test Hypothesis 3, the mediation hypothesis,we used a procedure suggested by James and Brett(1984], In this procedure, variable b mediates theeffect of variable a on variable c if all of the follow-


TABLE 3Means, Standard Deviations, and Correlations, Study 2"

Variable

1. Level of pastperformance

2. Satisfaction3. Belief in current

strategies4. Self-efficacy5. Goals6. Total information7. Favorable information8. Unfavorable

information9. Grade point average

10. Glass grade11. Gender12. Persistence13. Change in

performance

Mean

2.01

4.7051.63

0.0022.065.831.021.26

3.203.450.460.00

-33.14

s.d.

4.70

2.1028.66

0.9915.205.631.541.74

1.701.630.501.41

71.05

1

.56*

.31*

.64*

.71*

.04-.09-.18*

-.09-.08-.08

.34*-.19*

2

.51*

.66*

.61*

.08

.02-.11

-.06-.08-.12

.24*-.19*

3

.48*

.44*

.05-.04-.08

.07-.02

.01

.29*-.22*

4

.82*

.01-.08-.15

.03-.10-.10

.40*-.26*

5

.06

.01-.12

.00-.12-.12

.50*-.32*

6

.80*

.75*

.07

.00-.13

.01-.07

7

.79*

.03-.06-.07

.05- .11

8

.08

.03-.13-.11

.07

9

.41*

.12

.08- .11

10

.24*

.01-.08

11

.01

.06

12

- .73*

n = 168.* p < .05

TABLE 4Results of Regression Analyses for Persistence with Old Strategies and Change in Performance after a

Radical Environmental Change, Study 2

Glass gradeGenderGrade point averageAccuracy of perception of external eventsLevel of past performanceSatisfactionBeliefs in current strategiesSelf-efficacyGoalsTotal information soughtInformation from favorable sourcesInformation from unfavorable sourcesPersistence

Model 1

-.08-.06

.07-.05

.01

Persistence

Model 2

-.07-.03

.11-.10

.40*

.17*

.16*

Model 3

.03-.09

.04-.07

.12-.26*

.15^

.17-̂

.30*-.01

.44*-.40*

.35*

.18*

Change

Model 4

-.07.09

-.10-.02

.02

in Performance

Model 5

-.13.05

-.05-.05

-.74*

.56*

.54*

p < .10p < .05

ing conditions are satisfied: (1) a has an effect on b,(2) b has an effect on c, and (3) the effect of a on cvanishes when b is held constant. The first-ordercorrelations in Tahle 3 reveal that success has asignificant effect, in the expected direction, on themediating variahles, except for total informationand favorable information. Tahle 3 also reveals thatthe hypothesized mediating variahles, again withthe exception of total information and favorahle

information, have significant effects on persistencein the expected direction. Finally, model 3 in Tahle4 shows that, when mediators are entered heforepast success, the effect of past success on persis-tence is vitiated (/3 = .12, n.s.).

Note that the heta coefficients of the mediators inmodel 3 must he interpreted as direct effects. Amediator can also affect success indirectly, throughits effect on another mediator. This is, for example.


the case of self-efficacy, which, as Table 3 reveals,is positively associated with satisfaction, belief incurrent strategies, and goals. A simple way to de-tect total effects (the sum of direct and indirecteffects) is to run regression models including con-trol variahles, past success, and one mediator at atime. These supplemental analyses, not reportedhere but available upon request, confirm the medi-ating effect of goals and information from unfavor-able sources, show that the total effects of self-efficacy and belief in current strategies are muchstronger than their direct effects [p < .05) and, moreimportantly, reveal that the total effect of satisfac-tion is positive (/3 = .23, p < .05), not negative likesatisfaction's direct effect. In view of the links be-tween satisfaction and goals, self-efficacy, and be-liefs (see Table 3), this finding means that, overall,greater satisfaction leads to greater strategic persis-tence. Taken together, these findings show that,with the exception of the amounts of total andfavorable information acquired, the interveningpsychological processes identified in the proposedmodel mediate the effects of past success on indi-vidual strategic decision makers' strategic persis-tence. Thus, except for parts e and f. Hypothesis 3was supported.

DISCUSSION

This research identifies an intriguing pattern thatwe call the paradox of success. The paradox lies inthe fact that the very success that organizationsstrive to achieve plants the seeds of their possiblefuture decline. Once organizations achieve success,their natural tendency is to continue to exploit thestrategies that worked in the past. Indeed, skillfulrefinement and maximum exploitation of pre-viously effective strategies are often at the heartof lasting success (March, 1991). Such success-per-sistence-success cycles, however, become self-destructive when radical external changes imposethe need to use new strategies. After a period ofsuccess, organizations may lose the ability to rec-ognize when it is time to abandon previously effec-tive strategies. Consequently, they may experiencelarger drops in performance than organizationswith lesser histories of success.

Study 1 demonstrates such detrimental effects ofpast success in the airline and trucking industries.After controlling for market diversity, size, andCEO change—three variables often associated withinertia—we found that success increased strategicpersistence in the face of dramatic environmentalchanges and that this persistence had negative per-formance consequences. Study 2, in addition toreplicating the paradox of success at the individual

level, shed some light on the underlying mecha-nisms, pointing to the psychological consequencesof success for decision makers as one cause of thisparadox. Our findings show that the detrimentaleffect of success is due not to any one particularprocess, such as restricted information seeking ordevelopment of rigid beliefs, but rather, to the com-bined effects of several processes.

As predicted, past success increased strategic de-cision makers' satisfaction, and satisfaction led de-cision makers to increase their use of .past strate-gies. However, the effect of satisfaction is morecomplex than is generally thought. Typically, sat-isfaction is expected to lead to complacency—thatis, drifting with no attempt at improvement (e.g..Miller & Chen, 1994). However, that did not occurhere with respect to motivation to perform. First-order correlations revealed that higher satisfactionwas associated with higher self-efficacy and highergoals. Satisfaction did lead to complacency, how-ever, in another way. It was associated with stron-ger belief in the validity of current strategies, whichincreased strategic decision makers' tendency tostick to past strategies. As consecutive successesaccumulated, past achievements led to a strongerconviction that the current course of action was

. correct.Moreover, the higher level of self-efficacy and

higher goals that accompany past success inducedfurther dysfunctional persistence. This result ap-parently contradicts previous micro research thathas consistently shown that high goals and highself-efficacy lead to higher-quality planning andbetter selection of effective strategies (Smith,Locke, & Barry, 1990; Wood & Bandura, 1989). Theresolution of this enigma is that, if effective strate-gies are already known, then high goals and highself-efficacy increase the likelihood that such strat-egies will be used. But high goals and high self-efficacy will also increase the likelihood that inef-fective strategies will be used if individualsbelieve, mistakenly, that they will work (cf. Earley& Perry, 1987). This is consistent with argumentsmade by Lindsley, Brass, and Thomas (1995), whosuggested that when higher performance andhigher efficacy build upon each other, creating anupward spiral, individuals expect easy results andso are less likely to adapt to a changing task envi-ronment.

Not all the psychological mechanisms leading tostrategic persistence here were motivational. Infor-mation seeking also mediated the success-persis-tence relationship. However, such an effect did notinvolve the amount of information sought as such.Rather, it involved the type of information solic-ited. Following success, the amount of information


sought from unfavorable sources decreased. Such apattern would tend to reinforce assumptions aboutthe effectiveness of past strategies and underminethe ability to foresee the need to develop new ones.Our findings are consistent with those of Ashfordand Tsui (1991), who found that failing to seeknegative feedback fostered managerial ineffective-

ness.

Implications for Future Research

Together, these findings are relevant to the liter-ature on organizational inertia. Typically, organiza-tions' lack of responsiveness to radical environ-mental changes has been seen as stemming fromconstraints that emanate from organizational struc-tures, institutional pressures, organizational ideol-ogies, investments in specialized assets, and soforth (e.g., Hannan & Freeman, 1989). Althoughauthors have assumed that these constraints workthrough decision making (Child, 1972; Hambrick &Finkelstein, 1987), they have not directly examinedthe mediating effects of individual strategic deci-sion makers' psychological processes. Our researchaddresses this gap by pointing to a path that is analternative or, at least, a complementary path toinertia, a path that links organizational success tostrategic rigidity through a well-identified set ofpsychological processes present in strategic deci-sion makers.

In the context of micro explanations of macroorganizational phenomena (Staw & Sutton, 1992),our research provides an interesting complement tothe series of experimental and field studies con-ducted by Staw and his associates on the escalationof commitment (e.g., Staw & Ross, 1987). Althoughour studies focus on strategic persistence caused bypast success, those of Staw and his associates fo-cused on persistence in the face of failure. Staw'sstudies examined situations in which strategic de-cision makers have taken actions that lead to lossesand, subsequently, rather than changing their be-havior, stick to the current, losing course of action.His research shows that decision makers' motives,such as the desire not to admit failure, in combina-tion with other factors, are responsible for theseinstances of pathological organizational persis-tence.

Whereas the escalation of commitment phenom-enon is caused by the belief that a previously losingcourse of action will succeed in the future, theparadox of success is caused by the belief that apreviously winning course of action will succeed inthe future. In the first case, decision makers assumethat conditions will be somehow different in thefuture, whereas in the second case, they assume

that the conditions will be the same. The sad ironyof Staw's findings combined with ours is that bothfailure and success may be dysfunctional, thoughfor different reasons. Although the escalation ofcommitment phenomenon is, at least in part, due tothe fact that decision makers who are responsiblefor having chosen the current, failed strategies seethe decision to abandon them as ego-threatening,the paradox of success phenomenon is due to thefact that decision makers become overconfident be-cause of their past success.

There may also be other psychological processesmediating the effect of success on strategic decisionmakers' dysfunctional persistence, but we believethat more can be learned by widening the analysisto other organizational processes. Enduring successis likely to affect the strategy-making processthrough other causal paths. Past success may alterinternal dynamics within top management teamsby favoring the development and sharing of strongbeliefs based on the past and by limiting dissent(e.g., Dutton & Duncan, 1987). Past success mayalso affect the political context of an organizationby strengthening the power of its dominant coali-tion and creating a context in which demands forchange are suffocated before they are voiced (Pfef-fer, 1981).

The paradox of success, however, is not inevita-ble. Some successful organizations are able to gen-erate streams of incremental innovation as well asinnovation that redefines their own industry (Tush-man & O'Reilly, 1997). The investigation of thefactors that prevent such organizations from fallinginto a pattern of dysfunctional persistence strikesus as another important area for future research.For example, their long-lasting success might bedue to the fact that their executives are endowedwith skills that make them immune to the paradoxof success. Future research could try to identify theindividual skills that act as antidotes. Alterna-tively, their invulnerability might stem from theeffective use of organizational processes such asexecutive turnover and board monitoring.

Managerial Implications

From a practical standpoint, our research sug-gests that successful strategic decision makersshould develop routines to counteract the naturaltendency to rely excessively on past achievementsin their strategic assessments. More specifically,our findings point to information-seeking patternsas a major area of concern. Being open to informa-tion and advice from critics seems to be an effectiveway to increase one's adaptive capacity. Seekingand being open to critics' opinions, however.


should not necessarily imply the need to followtheir advice. As noted, great husiness leaders typi-cally succeed hy ignoring critics and defying thestatus quo. Thus, it is critical that invalid criticismhe distinguished from valid criticism. One way todo this is to check information coming from onesource against views coming from different per-spectives (Nystrom & Starhuck, 1984).

Another way to weaken the pernicious effect ofsuccess might he to redefine the concept of perfor-mance hy adding new kinds of goals. We haveshown that when a firm is performing extremelywell on a salient performance dimension, the nat-ural tendency is to use the same strategies. In thesesituations, executives can facilitate change hy di-recting attention to other performance dimensions,for which current performance is inadequate. Thisshift in the allocation of attention can he achievedhy setting new goals. At Boeing, for example, exec-utives sought to reduce the perception of successderived from having the largest market share in theaerospace industry hy focusing on a different typeof market share—the proportion of new aircraft op-erated hy airline companies [Fortune, 1994). Thelogic hehind this new goal was that old airplanes,even if they were made hy Boeing, could he themost dangerous competitors if airline companiesput off huying new aircraft and decided to keep oldplanes in use. Since high prices were the mainreason why airline companies postponed purchasedecisions, this new goal reduced strategic inertia hyfavoring change initiatives aimed at ohtaining dras-tic cost reductions.

Limitations

Like all research, our studies have limitations.One limitation pertains to the generalizahility ofthe findings of study 1. Because we focused on twoindustries that underwent discrete and radical en-vironmental changes, it is unclear whether ourfindings apply to industries undergoing differentkinds of shifts, like high-technology industries thatexperience continuous and unanticipated changes.Nonetheless, the fact that we were ahle to ohtainsimilar findings in hoth industries studied (airlinesand trucking) and to replicate these findings in alahoratory setting gives us confidence that the par-adox of success is a phenomenon worthy of addi-tional research.

One should also he cautious in generalizing thefindings of study 2 to executives. Some of the dif-ferences hetween undergraduates and executives,such as differences in age and joh experience, mayalter the success-persistence relationship. Clearly,future research should try to replicate the paradox

of success with other kinds of participants. How-ever, until then, given the difficulties in gainingaccess to executives for in-depth examinations ofthe psychological dynamics hehind strategic deci-sion making, we think that our research is a usefulfirst step that can guide and spur future research.

REFERENCES

Ashford, S. J. 1989. Self-assessment in organizations: Aliterature review and an integrative model. In L. L.Cummings & B. M. Staw (Eds.), Research in orga-nizational behavior, vol. 11: 133-174. Greenwich,CT: JAI Press.

Ashford, S. J., & Cummings, L. L. 1983. Feedback as anindividual resource: Personal strategies of creatinginformation. Organizational Behavior and HumanPerformance, 32: 370-398.

Ashford, S. J., & Tsui, A. S. 1991. Self-regulation formanagerial effectiveness: The role of active feed-back-seeking. Academy of Management Journal,34: 251-280.

Bandura, A. 1986. Social foundations of thought andaction. Englevkfood Cliffs, NJ: Prentice-Hall.

Bandura, A. 1997. Self-efficacy: The exercise of control.New York: W. H. Freeman.

Bentler, P. M. 1990. Comparative fit indexes in structuralmodels. Psychological Bulletin, 107: 238-246.

Boeker, W. 1997. Strategic change: The influence of rrian-agerial characteristics and organizational growth.Academy of Management Journal, 40: 152-170.

Bollen, K. A. 1988. A new incremental fit index forgeneral equation models. Sociological Methods andResearch, 17: 303-316.

Child, J. 1972. Organizational structure, environmentand performance: The role of strategic choice. Soci-ology, 6: 1-22.

Corsi, T. M., & Stowers, J. R. 1991. Effects of a deregu-lated environment on motor carriers: A systematic,multi-segment analysis. Transportation Journal, 30(spring): 4-28.

Cyert, R. M., & March, J. G. 1963. A behavioral theory ofthe firm. Englewood Gliffs, NJ: Prentice-Hall.

Daft, R. L., & Weick, K. E. 1984. Toward a model oforganizations as interpretation systems. Academy ofManagement Review, 9: 284-295.

Dutton, J. E., & Duncan, R. B. 1987. The creation ofmomentum for change through the process of strate-gic issue diagnosis. Strategic Management Journal,8: 279-295.

Earley, P. G., & Perry, B. G. 1987. Work plan availabilityand performance: An assessment of task strategypriming on subsequent task completion. Organiza-tional Behavior and Human Decision Processes,39: 279-302.


Finkelstein, S., & Hambrick, D. C. 1990. Top managementteam tenure and organizational outcomes: The mod-erating role of managerial discretion. Administra-tive Science Quarterly, 35: 484-503,

Finkelstein, S,, & Hambrick, D. C, 1996, Strategic lead-ership: Top executives and their effects on organi-zations. New York: West.

Fortune. 1994. Why to go for stretch targets. November14: 145-155,

Greve, H, R, 1998. Performance, aspirations, and riskyorganizational change. Administrative ScienceQuarterly, 43: 58-86,

Hambrick, D. C, & Finkelstein, S. 1987. Managerial dis-cretion: A bridge between polar views of organiza-tions. In L. L, Cummings & B, M. Staw (Eds.], fle-search in organizational behavior, vol. 9: 369-406.Greenwich, GT: JAI Press.

Hannan, M. T., & Freeman, J. 1984. Structural inertia andorganizational change. American Sociological Re-view, 49: 149-164,

Hannan, M, T., & Freeman, J. 1989. Organizational ecol-ogy. Gambridge, MA: Harvard University Press,

Haveman, H. A. 1992. Between a rock and a hard place:Organizational change and performance under con-ditions of fundamental environmental transforma-tion. Administrative Science Quarterly, 37: 48-75,

Hedberg, B. L, T, 1981. How organizations learn andunlearn. In W. H, Starbuck & P. G, Nystrom (Eds,),Handbook of organizational design, vol, 1: 3-27.New York: Oxford University Press.

Hollander, E, P., & Julian, J, W, 1969. Gontemporarytrends in the analysis of leadership processes. Psy-chological RuUetin, 71: 387-397.

Isen, A. M., & Baron, R, A, 1991. Positive affect as a factorin organizational behavior. In B. M, Staw & L, L,Gummings [Eds.], Research in organizational be-havior, vol, 13: 1-53, Greenwich, GT: JAI Press.

James, L, R., & Brett, J, M. 1984. Mediators, moderatorsand tests for mediation. Journal of Applied Psychol-ogy, 69: 307-321.

Joreskog, K, G,, & Sorbom, D, 1989, LISREL VII: A guideto the program and applications. Ghicago: Scien-tific Software,

Kelly, D., & Amburgey, T, L, 1991. Organizational inertiaand momentum: A dynamic model of strategicchange. Academy of Management Journal, 34: 591-612,

Kiesler, S,, & Sproull, L. 1982, Managerial response to achanging environment: Perspectives on problemsensing from social cognition. Administrative Sci-ence Quarterly, 27: 548-570,

Lant, T, K,, Milliken, F, J,, & Batra, B. 1992. The role ofmanagerial learning and interpretation in strategicpersistence and reorientation: An empirical investi-

gation. Strategic Management Journal, 13:585-608.

Lant, T. K., & Montgomery, G, 1987, Learning from stra-tegic success and failure. Journal of Business Re-search, 15: 503-518,

Levine, M. 1971. Hypothesis theory and nonlearning de-spite ideal S-R reinforcement contingencies. Psycho-logical Review, 78:, 130-140.

Levitt, B., & March, J, G. 1988. Organizational learning. InW. R. Scott & J, F, Short, Jr. (Eds.], Annual review ofsociology, vol. 14: 319-340, Palo Alto, GA: AnnualReviews.

Lindsley, D. H., Brass, D, J., & Thomas, J. B. 1995. Effi-cacy-performance spirals: A multilevel perspective.Academy of Management Review, 20: 645-678,

Locke, E, A., & Latham, G, P. 1990, A theory of goalsetting and task performance. Englewood Gliffs, NJ:Prentice-Hall.

March, J. G, 1991, Exploration and exploitation in organi-zational learning. Organization Science, 2: 71-87.

March, J. G., & Olsen, J, P, 1976, Ambiguity and choice inorganizations. Bergen, Norway: Universitetsfor-laget,

Meyer, A. D., Goes, J. B., & Brooks, G. R. 1993. Organi-zations reacting to hyperturbulence. In G, P. Huber &W. H. Glick (Eds.], Organizational change and re-design: 66-111. New York: Oxford University Press,

Miller, D, 1993, The architecture of simplicity. Academyof Management Review, 18: 116-138.

Miller, D., & Ghen, M-J. 1994, Sources and consequencesof competitive inertia. Administrative ScienceQuarterly, 39: 1-23,

Milliken, F. J,, & Lant, T. K. 1991, The effect of an organi-zation's recent performance history on strategic persis-tence and change. In P, Shrivastava, A. Huff, & J. E,Dutton (Eds,], Advances in strategic management,vol. 7: 129-156, Greenwich, GT: JAI Press,

Nystrom, P, G,, & Starbuck, W. H. 1984. To avoid orga-nizational crises, unlearn. Organizational Dynam-ics, 12(4): 53-65.

Pfeffer, J. 1981, Power in organizations. New York: Pitman,

Prahalad, G. K., & Bettis, R, A, 1986, The dominant logic:A new linkage between diversity and performance.Strategic Management Journal, 7: 485-501.

Schwartz, B, 1982. Reinforcement-induced behavioralstereotype: How not to teach people to discoverrules. Journal of Experimental Psychology: Gen-eral, 1: 23-59.

Shaw, M. E. 1976, Group dynamics: The psychology ofsmall group behavior. New York: McGraw-Hill.

Smith, K. G., & Grimm, G. M. 1987, Environmental vari-ation, strategic change and firm performance: Astudy of railroad deregulation. Strategic Manage-ment Journal, 8: 363-376,


Smith, K. G., Grimm, G. M., & Gannon, M. J. 1992. Dy-namics of competitive strategy. London: Sage.

Smith, K. G., Locke, E. A., & Barry, D. 1990. Goal setting,planning, and organizational performance; An ex-perimental simulation. Organizational Behaviorand Human Performance. 46: 118-134.

Staw, B. M., & Ross, J. 1987. Behavior in escalation situ-ations: Antecedents, prototypes, and solutions. InB. M. Staw & L. L. Gummings (Eds.), Research inorganizational behavior, vol. 9: 39-78. Greenwich,GT: JAI Press.

Staw, B. M., & Sutton, R. I. 1992. Macro organizationalpsychology. In J. K. Murnighan (Ed.), Social psy-chology in organizations: Advances in theory andresearch: 350-384. Englewood Gliffs, NJ: Prentice-Hall.

Tushman, M. L., & O'Reilly, G. 1997. Winning throughinnovation. Boston: Harvard Business School Press.

Tushman, M. L., & Romanelli, E. 1985. Organizationalevolution: A metamorphosis model of convergenceand reorientation. In B. M. Staw & L. L. Gummings(Eds.), Research in organizational behavior, vol. 7:171-222. Greenwich, GT: JAI Press.

Vietor, R. H. K. 1990. Gontrived competition: Airlineregulation and deregulation, 1925-1988. BusinessHistory Review. 64(spring): 61-108.

Vroom, V. 1964. Work and motivation. New York:Wiley.

Wood, R., & Bandura, A. 1989. Impact of conceptions ofability on self-regulatory mechanisms and complexdecision making. Journal of Personality and SocialPsychology. 56: 407-415.

Zajac, E. J., & Kraatz, M. S. 1993. A diametric model ofstrategic change: Assessing the antecedents and con-sequences of restructuring in the higher educationindustry. Strategic Management Journal. 14: 83-102.

Pino G. Audia is an assistant professor of organizationalbehavior at the London Business School. He received hisPh.D. from the University of Maryland. His current re-search focuses on micro explanations of macro phenom-ena such as organizational inertia and the emergence ofindustrial clusters.

Edwin A. Locke is the Dean's Professor of Leadershipand Motivation at the Robert H. Smith School of Businessat the University of Maryland. He received his Ph.D. fromGornell University. He is now doing research on therelationship of motivation and cognition, job satisfactionand the traits of great wealth creators.

Ken G. Smith is a professor of management- and organi-zation at the Robert H. Smith School of Business at theUniversity of Maryland. He received his Ph.D. from theUniversity of Washington. His current research interestsinclude competitive dynamics, strategy, and adaptive ef-ficiency.

THE PARADOX OF SUCCESS: AN ARCHIVAL AND A LABORATORY STUDY OF STRATEGIC … 12 The paradox of...

Documents

Transcript of THE PARADOX OF SUCCESS: AN ARCHIVAL AND A LABORATORY STUDY OF STRATEGIC … 12 The paradox of...