organizational learning in target setting · proxy for organizational learning compared to changes...

r Academy of Management Journal2017, Vol. 60, No. 3, 1189–1211.https://doi.org/10.5465/amj.2014.0897

ORGANIZATIONAL LEARNING IN TARGET SETTING

CARMEN ARANDAJAVIER ARELLANOUniversity of Navarra

ANTONIO DAVILAIESE Business School

This paper examines organizational learning in a target setting. Organizations com-monly set targets—explicit and quantitative reference points—for their operationalunits that reflect top management aspirations for these units. Targets are commonlythe outcome of a subjective process where supervisors combine their explicit and tacitknowledge to set performance expectations for their units. Using a proprietary da-tabase from a large European travel company during a period of rapid expansion, wedocument the effect of organizational learning by studying how targets change asunits mature. In particular, we examine managers’ experiential learning frombranches’ past performance and their vicarious learning from branches in the sameregion in determining performance expectations over the life cycle of branches. Ourresults indicate that, in setting performance targets, managers increase the weight ofa branch’s past performance and decrease the weight of comparable branches’ per-formance as the branch matures. Vicarious learning, where managers extrapolate theperformance of comparable branches to a new branch, dominates in the early years.Over time, this type of learning is replaced by experiential learning as experienceaccumulates. We document how early on in the life of branches, these two types oflearning interact; this interaction disappears as branches mature. Furthermore, wefind that managers learn differently from successes and failures early in the lives ofthe new units, and this learning is affected by the magnitude of the successes andfailures.

Aspiration levels are crucial reference points fororganizations to gauge observed performance assuccess or failure, which in turn influences perfor-mance evaluation, compensation, strategic behav-ior, risk-taking, and search activities (Audia &Greve, 2006; Baum, Rowley, & Shipilov, 2005;Cyert & March, 1963; Greve, 2003). Hence, an un-derstanding of how companies use available in-formation to set aspiration levels is important formanagement research (Blettner, He, Hu, & Bettis,2015; Bromiley&Harris, 2014; Cyert &March, 1963;Greve, 2002; Kim, Finkelstein, & Haleblian, 2015;Shinkle, 2012).

Literature examining multi-unit firms has fo-cused on two main reference points that organi-zations use in the process of forming aspirations:the past performance of the operational unit andthe performance of comparable units, such as sis-ter units under the same parent organization(Blettner et al., 2015). However, to the best of ourknowledge, no study has analyzed the effect oflearning on the formation of organizational aspi-rations over time. Learning affects the relevance ofa unit’s own performance and its comparability tosister units in the formation of aspirations as theunit matures. Furthermore, we examine howlearning from both one’s own performance and theperformance of others depends on the magnitudeof the deviation from aspirations as well aswhether aspirations are construed as success orfailure. We also consider how these two types oflearning influence each other in the formation ofaspirations.

Our objective is to examine organizational learn-ing within the context of the formation of targets that

We thank participants at the European AccountingAssociation Meetings, and at Rotterdam School ofManagement, Manchester University, Bocconi Univer-sity, and University of Lancaster seminars for helpfulcomments. We gratefully acknowledge the financialsupport of the Catedra de Empresa Volkswagen Navarra-Universidad de Navarra and the Alcatel-Lucent Chair atIESE.

1189

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download, or email articles for individual use only.

https://doi.org/10.5465/amj.2014.0897

are defined as explicit and quantitative aspirations.1

Aspirations interpreted as managers’ anticipatedperformance level (Shinkle, 2012) are not necessar-ily reflected in an explicit reference point. Even so,organizations commonly set targets for their opera-tional units.We posit that experiential learning fromobservation of the performance of each new opera-tional unit (Argote, 2013) and vicarious learningfrom observation of the performance of comparableunits (Haunschild & Miner, 1997) both affect aspi-ration formation. Vicarious learning is enhancedwhen the units are more comparable (Albuquerque,2009; Kim & Miner, 2007).

Organizational learning literature2 has often usedoutcomes to capture changes in knowledge, suchas: learning curve effects (Argote, 2013; Argote &Epple, 1990; Balasubramanian & Lieberman, 2010);investment behavior (Audia & Greve, 2006; Greve,2007); innovation launches (Greve, 2003); growth(Greve, 2008; Sorenson, 2003); patents (Schildt,Keil, & Maula, 2012; Wagner, Hoisl, & Thoma, 2014);foreign subsidiary location decisions (Belderbos,Van Olffen, & Zou, 2011); acquisition patterns(Muehlfeld, Sahib, & Witteloostuijn, 2012); partnercontrol (Dekker & Van den Abbeele, 2010); and theadoption of new management practices (Schwab,2007). Yet, organizational outcomes can be a noisyproxy for organizational learning compared tochanges in organizational practices and can lead toincorrect cause-effect inferences (March & Sutton,1997) and superstitious learning (Zollo & Reuer,2010). Rather than studying the performance con-sequences of learning, we examine how learning

shapes the internal organizational process of set-ting targets.

Only a handful of studies investigate the effect oflearning on internal processes. These studies haveadopted a qualitative, case-based research design.Garud,Dunbar, andBartel (2011) studyhownarrativesincorporate learning from unusual experiences, andRerup and Feldman (2011) provide case-based evi-denceof theco-evolutionof routinesand interpretativeschemata. This paper focuses on aspiration levels ascrystallized inperformance targets usingaquantitativecase study research design. The target-setting processis an information-based routine (Simons, 1995) thatmaterializes the knowledge available tomanagers intotargets (Fisher,Maines, Peffer, &Sprinkle, 2002;Raju&Srinivasan, 1996).Theprocess is commonly subjective(Bol, Keune, Matsumura, & Shin, 2010) to accommo-date tacit knowledge beyond the explicit knowledgeavailable in the organization (Davila & Foster, 2007;Merchant & Van der Stede, 2011). The subjectivitymakes this process particularly suitable to study howorganizational learning affects managers’ use of avail-able information to set aspiration levels.

The paper uses targets from 421 branches ofa travel retail company over a four-year growthperiod. During this period, the company replicatedthe retail model through the opening of newbranches thatwere similar to existing ones inmanyaspects, such as the product portfolio, structure,go-to-market strategies, and organizational routines.For each new branch that was opened, the organiza-tion went through a learning period to understand itscommon and specific characteristics. The evidenceindicates that experiential and vicarious learning af-fect the way that aspirations are formed as units ma-ture.Managers relyon theperformanceofcomparableunits to set targets early in the life of a unit, and aslearning occurs, they shift their attention to past per-formance. A divisional manager’s description of thephenomenon illustrates this process: “Branches ma-ture over a period of three to four years; early on, tar-gets are based on estimates gathered during thedecision about opening the particular office and theperformance of branches in the same neighborhood;as they mature, targets rely to a larger extent on theirpast performance.”

The study further contributes to the learning asso-ciated with the magnitude of deviation from perfor-mance and how this learning differs across young andmature units. Our evidence indicates that the rele-vance of past performance in setting targets is signifi-cantly different between younger andmature units forunfavorable performance deviations (failure). A third

1 How targets and aspirations are set is a topic studiedfrom different perspectives. One line of study exploresaspirations in budget-based incentive systems and theirinfluence on managers’ commitment and effort; anotherline of research examines the effect of aspirations onmanagers’ risk taking and strategic behavior. The variety ofperspectives on the phenomenon has led to a proliferationof terms. For instance, the extent to which targets rely ona unit’s ownpast performance is referred to as “ratcheting”(Leone & Rock, 2002) and “reference point updatingspeed” (Greve, 2002). Similarly, “goals” and “aspirations”are different in that aspirations have been defined as ex-pectations for a specific goal (Cyert & March 1963).

2 Organizational learning captures the dynamic natureof knowledge (Argote & Miron-Spektor, 2011) as partici-pants draw from their experiences and those of othermanagers to alter the knowledge of the organization, cod-ified in routines, rules, and procedures (explicit knowl-edge) and embodied in mental models, schemata, culture,and individual cognition (tacit knowledge).

1190 JuneAcademy of Management Journal

contribution relates to the literature on cross-sectionallearning effects (Brass, Galaskiewicz, Greve, & Tsai,2004; Schwab, 2007). Past performance and perfor-mance from comparable units are likely to provideinformation content relative to each other that couldinform the learning process. Thus, the effect of pastperformance onmanagers’use of informationdependson the performance of comparable units. We find thatvicarious learning is differentially affected by experi-ential learning in young units.

Finally, the paper provides new evidence on thedynamics of target setting. Research on target settinghas studied mature organizations (Bouwens & Kroos,2011; Indjejikian, Matejka, Merchant, & Van derStede, 2014; Leone & Rock, 2002) where operationalunitshaveexisted for longperiods andmanagershavea stable implicit model. These settings assume thatlearning is already incorporated into thesemodels. Incontrast, this study is based on the analysis of ob-served targets for new branches since the date of theirinception. The study explicitly examines how learn-ing affects target setting as managers change their useof information when balancing past performance andpeer performance over time.3

THEORETICAL FRAMEWORK

Performance Feedback and Aspirations

Target setting is a core process of the planning andcontrol functions of management (Fayol, 1916). Tar-gets make aspirations explicit and are frequently usedfor employee evaluation and compensation (Simons,2000).4 At the start of each period, divisions, depart-ments, and often individuals are given a target thatdefines the organization’s aspirations for them, andperformance is routinely evaluated relative to this

target (Matsumura & Shin, 2006; Murphy, 2000). Tar-get setting is commonly a subjective process (Bol et al.,2010) through which the manager synthesizes the dif-ferent sources of information available to him/her(Fisher et al., 2002; Raju & Srinivasan, 1996). Thischaracteristic differs from those of compensation con-tracts for customer-facing units, which are typicallyformula-based. A formula associates the size of thebonus with the difference between the actual and ex-pected performance (deviation). The shape of this re-lationship is often linear (commission) or piecewiselinearwith a floor andaceiling. Learning is reflected inthese contracts only when the formula is modified.

Two questions of interest regarding aspirationsare: (1) how does performance feedback (defined asthe deviation between actual performance and as-pirations) affect organizations? (2) How do man-agers use available information to set aspirationsand targets as their quantitative realization? Thissecond question is the focus of this study.

Performance feedback theory (Cyert & March, 1963)has made significant progress in our understanding ofhow aspiration levels and performance feedback in-fluence a variety of organizational decisions, such asstrategic behavior, risk taking, and the propensity tomake changes (Audia & Brion, 2007; Baum et al., 2005;Greve, 2003, 2008; Sorenson, 2003), which, in turn, in-fluences subsequent performance (Shinkle, 2012). Forinstance, failing to achieve aspirations has been arguedto trigger problemistic search (Argote & Greve, 2007:339; Baum&Dahlin, 2007: 371). The stock of resourcesinfluences how negative performance is perceived—as a repairable gap or alternatively as a threat tosurvival—and hence affects risk tolerance (Audia& Greve, 2006). Additionally, performance above as-pirations has been argued to lead to slack-drivensearch. Decision makers have access to additionalresources and pursue initiatives outside the currentstrategy of the organization (Baum & Dahlin, 2007;Lant, Milliken, & Batra, 1992; Levinthal & March,1981).

Aspiration theory predicts how new information,mostly derived from a unit’s prior performance andpeer units’ performance (social comparison), changesaspiration levels (Blettner et al., 2015; Lant, 1992).5

3 Target setting can support learning through the in-teractions between organizational members during the pro-cess; we do not capture this type of learning in our research.

4 Target setting plays other important roles in organi-zations such as motivational, resource allocation, co-ordination, control (management by exception), andthe learning associated with performance evaluations(Merchant & Van der Stede, 2011). Targets are used toquantify aspirations in financial as well as non-financialmeasures. For instance, financial targets are set formeasures such as revenue, revenue growth, and profits.Non-financial measures include measures such as cus-tomer satisfaction, quality levels, productivity, or time tomarket. Together, the main financial and non-financialmeasures are often referred to as key performance in-dicators (KPIs) (Kaplan & Norton, 2001).

5 Theoretical and empirical work on adaptive aspirationformation models has mainly focused on the focal unit’sown past performance and has modeled aspiration as anexponentially weighted average of past performance. So-cial comparison is defined as a direct comparison of theperformance of the focal unit with the performance ofcomparable units (Mezias, Chen, & Murphy, 2002).

2017 1191Aranda, Arellano, and Davila

Theory concerning the formation of aspirations (andtargets) starts froma rational expectations perspectiveand studies deviations from expectations. Targets areset equal to futureperformance expectations—what ismost likely to happen in the future—and managersoptimally use all the available information to setthese targets.6 From a rational expectations per-spective, deviations arise from unexpected perfor-mance changes. These changes are permanent if theunderlying production function varies and managersoptimally include them in next year’s target to main-tain the same level of difficulty and effort. Transitorychanges are shocks not expected to persist going for-ward. Managers will not include them in next year’starget (Webb, Williamson, & Zhang, 2013).

Empirical studies on target setting examine ratio-nal expectations predictions and deviations fromthese predictions associated with organizationalstructures. In particular, the studies investigate theweights of past performance and the performance ofcomparable units on future targets, as well as howthese weights vary with past performance relative toexpectations, seniority, and organizational structure(Aranda, Arellano, & Davila, 2014; Bol et al., 2010;Leone & Rock, 2002). Ratcheting describes the posi-tive relationship between the change in targets forthe current year relative to the previous year’s targetsand last year’s performance deviation, which is de-fined as the difference between last year’s actualperformance and targeted performance (Weitzman,1980). Favorable performance deviations (perfor-mance above targets) are followed by upward targetrevisions (and vice versa, unfavorable deviations areassociated with downward target revisions). Ratch-eting levels depend on managers’ assessment of themix between permanent and transitory componentsof the deviation. Furthermore, aspirations have beenargued to adapt at a slower rate than performancechanges (Cyert & March, 1963). This speed of adap-tation reflects different time perspectives in decisionmaking (Greve, 2002) and affects performance; in thecontext of competitive markets, organizations witha slower speed of adaptation show better timing ofstrategic changes and subsequently better perfor-mance (Greve, 2002). Ratcheting can also affect theshape of the compensation, with lower ratchetingincreasing the level of compensation in future pe-riods (Leone & Rock, 2002). Attenuated ratcheting(Choi, Kim, & Merchant, 2012) refers to managerslowering ratcheting/speed of adoption to avoid

subordinates withholding effort when they reach thetarget (a phenomenon known as the ratchet effect)(Bouwens & Kroos, 2011). This finding is consistentwith adaptive aspiration predictions (Lant, 1992) inwhich managers adapt aspiration levels at a slowerrate thanperformance changes (Cyert &March, 1963).

Empirical evidence shows that ratcheting oftenexplains more than 60% of the change in targets fromone year to the next (Aranda et al., 2014; Leone &Rock, 2002). This finding is consistentwithmodels ofadaptive aspirations (Shinkle, 2012), which assumethat managers “adjust aspiration levels in the di-rection of attainment discrepancy, consciously orunconsciously, with the aim of reducing the dis-crepancy between the actual performance level andthe previous aspiration level” (Mezias, Chen, &Murphy, 2002: 1287).7 Furthermore, evidence in thisliterature suggests an asymmetric response to favor-able andunfavorable deviations (Leone&Rock, 2002)that is consistent with the differences in learningfrom successes and failures (Kim &Miner, 2007).

Prior empirical research has also documented therelevance of peer units’ performance in setting targets(Aranda et al., 2014), which is referred to as relativetarget setting (RTS). These units share a commonproduction function (Albuquerque, 2009) that allowsmanagers to infer performance expectations of thefocal unit from the observation of its peer group.When peers are sister business units under the sameparent company, RTS is similar to the concept ofstriving discrepancy, defined as “the gap between anorganization’s current performance and the perfor-manceof thoseorganizations itdesires tobe like in thefuture” (Shinkle, 2012: 433). Moreover, managersexpect to observe similar performance levels and re-gression to the mean among entities within the samepeer group (Mezias et al., 2002). Thus, greater RTS—focal units performing above their peer group—hasbeen found to be associated with smaller increases intargets (Aranda et al., 2014). The relevance of peers’experiences in setting the aspiration levels of the focalunit depends on the peers’ comparability; the morecomparable these units are, the more relevant peersbecome to target setting (Baum&Dahlin, 2007; Baum,Li, &Usher, 2000;Baumet al., 2005;Greve, 1998). Themost comparable units are often units performinga similar task within the same company.

6 Managers’ expectations are not necessarily the statis-tical expected value of the particular variable.

7 In this quote, the difference between the actual andexpected performance is referred to as a discrepancy. Inthis paper, we refer to this concept as a performance de-viation. The concept is also identified as a variance inmanagerial accounting textbooks.


Organizational Learning

Experiential learning translates the inferences froman organization’s experience into knowledge androutines that systematically alter subsequent behav-iors (Argote, 2013; Cyert & March, 1963; Huber, 1991;Levitt & March, 1988).8 Organizations improve theirroutines through selective repetition that reproducesthe behaviors that are believed to drive performancethrough cause–effect relationships (Argote, 2013).Likewise, organizations improve their ability to sepa-rate relevant data from noise, thus enhancing signal-to-noise ratios in information-based routines (Banker& Datar, 1989). Experiential learning has been sup-ported inavarietyof settings, including learningcurvemodels, innovation adoption (Kraatz, 1998), network-partner selection (Dekker&Van denAbbeele, 2010; Li& Rowley, 2002), and strategic renewal (Audia, Locke,& Smith, 2000; Crossan & Berdrow, 2003).

Vicarious learning refers to changes in the stockof knowledge associated with interorganizationallearning (Beckman & Haunschild, 2002; Ingram &Baum, 1997), where learning accrues through obser-vations of other organizations’ experiences (Baum &Dahlin,2007;Madsen, 2009).Observing theoutcomesof comparable organizations helps infer their actionsand cause–effect relationships, which can be extrap-olated and adapted by the organization observing(Strang & Macy, 2001). Early research examinedlearning through the replication of successful routines(Burns & Wholey, 1993), which led to normative ad-vice on the importance of “best practice” studies and“benchmarking” programs (Collins & Porras, 1994).Vicarious learning reflects the use of available externalinformation. At the unit level, vicarious learning cantrigger changes in routines that seek to imitate or avoid

these external observations. At the supervisory level,vicarious learning can support inferences about thebehavior and performance expectations from observ-ing comparable units. Vicarious learning comple-ments experiential learning as long as it improves theinformativeness beyond individual past experiences(Holmstrom, 1979). The degree of vicarious learningdepends on the comparability of organizations. Evi-dence indicates that vicarious learning is enhancedwhen organizations are similar, in terms of both in-dustry and geography (Kim &Miner, 2007).

Organizational learning is reflected in improvingorganizational practices and outcomes. The dynamicsassociatedwith the formation of aspirations, quantifiedthrough targets, capturehowexperiential andvicariouslearning change this particular organizational practice.

Organizational Learning in Target Setting

Learning enhances managers’ ability to distinguishbetween the permanent and transitory components ofdeviations. Experiential learning relies on past perfor-mance of the focal unit, whereas vicarious learningrelies on the performance of peers. Both types oflearning have been claimed to be important; in someinstances, experiential learning has been found to bemore effective thanvicarious learning (Audia&Brion,2007), whereas in other studies, vicarious learningdominates (Mishina, Dykes, Block, & Pollock, 2010).The relevance of vicarious learning is likely to becontingent on variables such as industry, financialhealth, and organizational size and age (Baum &Dahlin, 2007; Shinkle, 2012; Short & Palmer, 2003).For instance, in stable environments, managers givemore weight to past performance in the formation ofaspirations, whereas in dynamic environments, man-agers rely more on peer units (Lant, 1992; Lant &Shapira, 2008). Age in particular offers managers theopportunity to learn about an operational unit’s busi-ness model and to better interpret and combine dif-ferent sources of information over time. The dynamicnature of learning suggests that the attention allocatedtodifferent sourcesof informationchangesover the lifecycleof the focalunit.Early in the focalunit’s life cycle,managers have fewer data points from which to learnhow to discern between permanent and transitorychanges in performance; hence,managers rely on peerinformation to a larger extent. Relying on others’ ex-periences can be a way for young organizations toovercome theuncertainty associatedwith early phasesof the life cycle. However, for young organizations,there are strong limitations in terms of benefitting fromthe experience of other organizations regarding the

8 Organizational knowledge is the set of concepts and as-sumptions about the cause–effect relationships that the orga-nization uses to form expectations about its activities (Huber,1991) and todefine the representationof its environment (Daft& Weick, 1984). This knowledge shapes the actions that itsmembers take (Madsen & Desai, 2010). Organizational learn-ing is the process of acquiring, translating, and enacting newknowledge through organizational routines (Nonaka, 1994;Polanyi, 1958; Walsh & Ungson, 1991; Weick, 1979) thatsystematically alter subsequent behavior (Argote & Miron-Spektor, 2011). This type of learning emerges from in-dividuals and their intuition (Crossan, Lane,&White,1999)but necessarily involves organizational routines as learn-ing becomes institutionalized (Easterby-Smith, Crossan, &Nicolini, 2000). We interpret organizational learning aschanges in managers’ behavior rather than changes to or-ganizational outputs and outcomes.


core aspects andprinciples of the business, the reasonbeing limited comparability in terms of size, reputa-tion, repertoire of strategic actions, and restricted ca-pacity for assimilating others’ experiences (Blettneret al., 2015). These limitations are less pronouncedwhen peers are sister organizations, as long as theparent organization has established mechanisms forsharing information. As the focal unit matures, expe-riential learning crowds out vicarious learning andmanagers increasingly rely on their own performanceto refine their mental model for each unit: “Decisionmakers may learn about a system given enough time,enough stability in the system, and unambiguous in-formation” (Lant, 1992).

Thus, we expect relative target setting (RTS) to bemore salient early in a unit’s life cycle and to de-crease in importance as managers shift to learningfrom past performance. These arguments lead to thefollowing hypotheses:

Hypothesis 1. The relationship between changein targets and the current year’s performancedeviation is weaker for young units than formature units.

Hypothesis 2. The relationship between changein targets and relative target setting is strongerfor young units than for mature units.

Learning from Successful and Failed Performance

Early studies on experiential learning focusedon learning from successes; researchers have onlyrecently studied learning from failures (Chuang &Baum, 2003; Haunschild & Sullivan, 2002), dif-ferential learning from successes and failures(Baum & Dahlin, 2007), the magnitude thereof(Madsen & Desai, 2010), and concentration (Desai,2015). Targets are often used as reference points toconstruct performance as either a success or fail-ure (March & Simon, 1958). Performance belowaspirations (failures) has been argued to lead tomore changes (Shinkle, 2012), with importantlearning episodes derived from extreme failures(rare events) (Lampel, Shamsie, & Shapira, 2009;Madsen & Desai, 2010): “the desire to overcomea performance failure is stronger than the desire toextend success, so decision makers below the as-piration level accept more risks than decisionsmakers above aspiration levels” (Audia & Greve,2006: 84). Failures question existing assumptionsabout cause–effect relationships, which forcesorganizations into non-local searches (Cyert &March, 1963; Weick & Roberts, 1993). Learning

from failure leads to a focus on outside organiza-tions’ performance. In contrast, success has orga-nizations confirm their existing assumptions (Lant,1992), reinforce their current behaviors and existingstocks of knowledge (Bromiley, Miller, & Rau, 2001),stimulate local searches, simplify decision making,avoid risky actions that can result in unfavorabledeviations (March & Shapira, 1987), and ignoreinformation that potentially conflicts with theshared schemata (Hayward, Rindova, & Pollock,2004). These arguments suggest a greater focus onpast performance after experiencing successfulperformance and on peer performance after expe-riencing failure.

Alternative arguments predict failures to be as-sociatedwith less vicarious learning and successesto be associated with more experiential learning.Failure may result in stagnation (Blettner et al.,2015). In contrast, success leads to non-local ex-plorative searches because of the organization’sconfidence in its own abilities and the resourcesavailable for new experiments (Lant et al., 1992;March & Shapira, 1992). Moreover, an organiza-tion’s resource endowment allows managers toviewminor failures as repairable gaps (Audia&Greve,2006); large firmswith considerable resources actuallyincrease their risk in response to minor failures,whereas smaller firms, usually resource-constrained,reduce their risk taking. Similarly, R&D expendituresincrease in response to sub-par performance (prob-lemistic search) or excess resources (slack search)(Greve, 2003).

The asymmetry of responses to successes andfailures thus becomes an empirical question.Moreover, these arguments can have different ef-fects over the course of an organization’s life cycle.Early on, failures are a threat to the survival of theorganization; this threat is more severe for new op-erational units of an existing company where oftenthe decision to close the unit is made at headquar-ters. In addition, failing tomeet performance targetsis likely to trigger search routines outside the unitand to reduce the relevance of past performance insetting targets. Failures in younger units lead man-agers to use external reference points as sources ofinformation when setting targets. If this is the case,we expect the following:

Hypothesis 3. The relationship between changein targets and the current year’s performancedeviation is weaker for unfavorable perfor-mance deviations of young units than for thoseof mature units.


Learning from Large and Small PerformanceDeviations

Much about learning from failures has beenexamined within the context of extreme nega-tive outcomes, such as companies going out ofbusiness or barely avoiding it (Ingram & Baum,1997; Kim &Miner, 2007) and disasters (Madsen &Desai, 2010). For instance, “problemistic searches”are enhanced after extreme failures (Cyert &March, 1963; Simons, 1995). This evidence sug-gests that organizations’ reactions to performancedeviations—and managers’ learning—dependsnot only on the sign of the deviation but also onits magnitude (Audia & Greve, 2006; Greve, 1998,2003).

Research on the informational content of a unit’sown past performance for target setting extendsthis argument. The magnitude of the deviation isinformative about the likelihood of permanentversus transitory components of the deviation.Permanent performance changes are associatedwith changes in the unit’s productivity; pro-ductivity changes often occur slowly. In contrast,large deviations aremore likely to occur because oftransitory shocks unrelated to changes in the pro-duction function.9 For instance, a drop in trafficwhen there are unexpected public works in front ofthe retail unit leads to a transitory performancedecrease (until the works are finished). This setassumptions based informational content predictslarger deviations to be less informative for experi-ential learning.

The performance feedback literature makes sim-ilar predictions: “organizations emphasize learn-ing from their own experience when performanceis near aspirations, and emphasize learning fromothers’ experience when performance deviates fromaspirations” (Baum & Dahlin, 2007: 369). Smallerlosses lead to more effective experiential learning,triggering search routines to aid understanding ofthe causes of the losses and with less attention de-voted to identifying the people responsible so thatthey can be held accountable. These circumstancesfacilitate information sharing instead of inspiringpeople to create excuses to protect themselves. Forsmaller deviations, the risk of exploring new routines

is perceived as unnecessary (Baum & Dahlin, 2007;Baum et al., 2005). The arguments predict the preva-lence of experiential learning for small performancedeviations and the prevalence of vicarious learningfor large unfavorable deviations.

The foregoing arguments do not consider whetherthe indicated effects change across different infor-mation environments as captured in the age of theunit. Similar to the reasoning leading to the first threehypotheses, the relevance of these arguments is ex-pected to vary over the life cycle of a unit. Man-agers’ models for young units are in flux, whereasthey are much more established for mature units.Thus, learning opportunities from historical andpeers’ performance information are likely to bemore important for young units. In particular, largeunfavorable performance deviations in young unitswill reinforce managers’ use of the performance ofpeer units as amore important referencepointwhilelowering the weight that they give to performancedeviations. The corresponding hypothesis is asfollows:

Hypothesis 4. The relationship between changein targets and the current year’s performancedeviation is weaker for large unfavorable per-formance deviations of young units than forthose of mature units.

Cross-Learning Interaction

The focus on different reference points affects howmanagers interpret performance deviations (March &Shapira, 1992): “rather than shifting attention moretoward historical or social aspirations, decisionmakerscombine information on social and historical aspira-tionswhen interpreting performance feedback” (Baumet al., 2005: 543). Basedon this observation, researchershave argued for a better understanding of how experi-ential and vicarious learning interact (Denrell, 2003;Levinthal & March, 1993). However, evidence of suchcross-learning within organizations is scant (Baum &Dahlin, 2007; Crossan, Lane, & White, 1999; March,1991; Schwab, 2007). In particular, the relevance ofperformancecomparedwith thatofpeersmayaffect theinterpretation of a unit’s performance deviation, in-stead of being two independent sources of informationas prior hypotheses assume. For instance, a favorabledeviation in a unit that is performing below its peers islikely tohavea largerpermanent component as theunitregresses back to its mean performance. Conversely,a favorable deviation together with a performanceabove that of the peer group ismore likely to be caused

9 Interviews conducted in our research site indicate thatmanagers believe this argument to bemore characteristic ofour particular research setting. For instance, the divisionalmanager indicated that public works can be so disruptivethat a branch may have to temporarily close.


by a transitory shock and thus will not be incorporatedinto the next year’s target. In such cases, managerscan use experiential learning to evaluate the quality oftheir vicarious learning, and vice-versa.

The interaction is expected to be more useful earlyin an operational unit’s life cycle as the company ex-trapolates from comparable units to interpret pastperformance. Later, managers have enough experi-ence from the unit, and the reference point frompeers’performance is thus less informative. Learning affectshow managers use information, and thus we expectthe following:

Hypothesis 5. The relationship between changein targets and relative target setting depends toa larger extent on the current year’s performancedeviation for young units than for mature units.

METHODS

Research Setting

We collect yearly data from 421 branches of the in-dividual vacation division of a large European travelcompanyover a four-year period.All of thebranches arelocated in the same country. The company has fivelines of business: hotel chains, receiving agencies,airlines, tour operators, and travel agencies. The travelagencies are further organized into four divisions: in-dividual vacations, business travel, convention travel-ing, and conference organizing. During the period ofstudy, the individual vacation division experiencedsignificant growth, reaching revenues of V444 million(approximately 60% of the total of the travel agencies’business revenues) in 2006. The division is organizedinto branches that are divided over 13 regions.

The branches are responsible for commercializingtravel products at the retail level. All branches sell thesame product portfolio, use similar operational prac-tices, have comparable marketing programs, and are ofsimilar size and complexity. The combination of a pe-riod of expansion within a homogeneous network andheterogeneous environments provides a unique settingfor examining organizational learning in information-based routines, such as target setting. Each branchoperates as a profit center and receives a yearly targetfor the forthcoming year. The target is set for “guidedsales,”which are sales from other lines of business ofthe company (hotel rooms, plane tickets, etc.) andservices from some suppliers. Once the target forguided sales is set, the rest of the income statement—total sales and the various components of expenses,such as selling expenses over sales—are set using amathematical relationship determined at the division

level. For instance, in 2006, guided sales were set to76.8% of total sales; thus, fixing guided sales mathe-matically determined total sales and the rest of the in-come statement. The target setting process works asfollows. The branchmanager and the regionalmanagerinformally discuss targets for the coming year; the re-gionalmanager proposes a target for eachbranch inhis/her region to the divisional manager. Based on the di-vision’s overall growth objectives, the divisional andregionalmanagers revieweachbranch’s targets tomakesure that the overall growth objective is met. Branchmanagers are then informed of their final targets for thecoming year. This process is subjective rather thanformula-based and integrates the different sources ofinformation available to managers. These sources in-clude their ongoing reports, visits, and conversationswith branchmanagers throughout the year, which pro-vides a detailed understanding of each branch’s uniqueaspects. Regional and divisional managers also obtaininformation by comparing activities across branchesand regions, industry trends, competitors’moves, localconditions, and discussions across divisions in thecompany. Thus, the target combines knowledge that isexplicit to the various actors in the process as well astacit knowledge that managers are not necessarily ableto articulate but can reflect in the final target.

Targets areused for compensationpurposes. Branchmanagers and other branch employees receive a bonusdepending on the branch’s performance relative totargets.Bonusesaccount for10%to20%ofemployee’ssalaries. Bonuses are distributed among branch em-ployees according to their job titles. In other words,bonuses are determined at the branch level and arethen allocated to employees. The branch manager re-ceives the largest bonus. The bonus percentage is de-terminedaccording to the following formula (Figure1):

Bonus5�50%1 50%3

�Ip 3

profit performancebudgeted profit

��3 guided sales

Ip takes a value of zero if the ratio of profit perfor-mance to budgeted profit is lower than 70%, one if theratio is between 70% and 130%, and 1.3 times the in-verse of the ratio to cap this part of the bonus at 130%.The first half of the bonus is distributed bimonthly andworks much like commission does, as a percentage ofguided sales that kicks in with the first sale. The otherhalf is distributed at the end of the year. This secondhalf has a floor at 70% and a cap at 130%. In between,the bonus also serves as a commission for guided sales,but the slope depends on the profit performance


relative to the profit target. If the profit does notreach 70%of the target, the second half of the bonus islost, and the branch only receives a bonus of 50% ofguided sales (the bimonthly part). If a branch profitexceeds 130% of its target, the second half of thebonus is 65% of guided sales (50%*130%) for a totalbonus of 115% (50% 1 50%*130%) of guided sales.For instance, a branch that exceeds its profit target by5% receives a total bonus of 102.5% of guided sales.The commission-like component is 50%, and the ad-justment based on the profit target is 52.5% for a totalof 102.5% (50% 1 50%*105%) of guided sales. Inaddition to the bonus, branches receive prizes and/orpenalties based on non-financial objectives, such asinternal audits, the quality of customer information inthe database, discounts offered to customers, and baddebt expenses.

The division experienced a period of rapid growth,which provides two natural sub-samples: branchesthat have been operating for a few years and newbranches. Thus, there is an opportunity to observehow managers use information as they learn aboutthese new branches. All the branches belong to thesame division, thus providing a homogeneous orga-nizational context. The branches are similar to eachother, with a comparable personnel structure, thecommercialization of the same product portfolio, andaccess to the same resources from the company. Thedivision groups the branches by geographical region.This organizational structure creates a natural set ofbranches that experience a comparable environmentand that can be used as benchmarks for learning.

This particular research setting offers a unique op-portunity for probing learning in target setting. Thegrowth in the number of branches, their maturationprocess, and the availability of comparable branchesoffer fertile ground for examining how learning affectsthe use of different sources of information. Comparinghow managers use information over the branches’ lifecycles allows this study to capture learning. A priorpaper (Aranda et al., 2014) using this research settingshowed how both past performance and relative per-formance evaluationwere correlatedwith target setting.This evidence was required to then explore how learn-ing affects the relevance of these two sources of in-formation over time.

Data Description

The number of branches in the sample grew from244 at the beginning of 2003 to 390 branches bythe end of the fourth year.10 Table 1 provides

FIGURE 1Incentive System for Branches

Adjusted accordingto profit performancerelative to profit target

0% 50% 100% 115%

Profit performance lowerthan 70% of profit target

Profit performance higherthan 130% of profit target

Profit performanceequal to profit target

Given duringthe year

SALESPERFORMANCE

10 The total number of branches for which we have in-formation at some point in time is 421. These are thebranches for which we have data in 2003 (237) plus thenumber of branches opened (177) plus the number ofbranches closed in 2003 (not included in the 237 becausethey do not have actual end-of-year sales). Because ourresearch specification requires at least one full year oftargets and operations, we start tracking learning in thesecond year of each branch’s life cycle. The specificationdoes not capture the learning accrued during the first year,when learning is likely to be steeper.


descriptive background information. The signifi-cant growth rate is the result of a combination of thehigh rate of branch openings and the low rate ofbranch closings. For instance, 34 of the 238 branchesin 2003 opened in 2002; in 2005, 78 branches openedand 11 closed. In total, 177 branches opened duringthe four-year period and 31 closed.11 Of the branchesthat closed, 26 opened before and five opened after2003. Thus, only five of the 177 branches that opened

during the observation period subsequently closed.One branch with two years of budgeted and actualsales in 2004 (first line Table 1, Panel B) was closed.Similarly, twooutof the60brancheswith twoyearsofdata in 2005 were closed. Finally, two of the 78branches with two years of data in 2006 were alsoclosed.

Thenumber of observations per year is the numberof branches for which we have budgeted and actualguided sales for the current and theprevious year. Thisset excludes branches that openedduring the previousyear and closed during the current year because theydonot have a full calendar year for either the current orprevious year’s sales (Table 1, Panel A). For instance,

TABLE 1Descriptive Statistics for All Branches

Panel A: Construction of Full Sample

Years

2003 2004 2005 2006 Total

Number of branches with end-of-yearactual and budgeted guided sales

238 294 361 390

Number of branches closed 7 4 11 9 31Number of branches without budget for

the previous year1 60 78 38 177

Missing observation* 1Total number of observations per year** 237 234 283 351 1,105

Panel B: Age of Branches per Year

Years

2003 2004 2005 2006 Total

Branches with two years of budgetedand actual sales (one observationfor the budgeted sales revision)

34 1 60 78 169

Minus branches closed during the year 22 22Branches with three years of budgeted

and actual sales9 34 1 58 101

Minus branches closed during the year 21Total number of young branches

(branches with one or two targetrevisions)

43 35 59 133 270

Branches with more than four yearsof budgeted and actual sales

194 203 233 225

Minus branches closed during the year 24 29 26Total number of mature branches 194 199 224 219 836Missing observation 1

194 199 224 218 835Total number of observations per year 237 234 283 351 1,105Percentage of young branches 18.1% 15% 20.8% 37.8% 24.4%Percentage of mature branches 81.8% 84.1% 79.1% 62.2% 75.6%

*The number of budgeted employees was missing for one mature branch in 2006.**A branch-year is an observation if the branch has information on its budgeted and actual sales for the previous and current years.

11 The division manager mentioned two main reasonsfor closing a recently opened branch: a mistake inchoosing the location or a mistake in the expectedmix ofcustomers.


at the end of 2004, there were 294 branches, 60 ofwhich opened in 2003 and did not have a full calendaryear for the previous year’s sales. The total number ofbranch-years is 1,105 observations.12

A branch’s age is the number of full years of op-eration. For instance, a branch that opened in 2003is one year old by the end of 2004 (2004 is its first fullyear of operation). Panel B of Table 1 describes theage distribution of the branch network from 2003through 2006. Branches that have been in operationfor two to three years are described as youngbranches, and branches that have been in operationfor more than three years are mature branches.13

One-year-old branches are not included becausetheir previous year (the opening year) does notcount as a full year. Of the young branches, 63%(169 out of 270) are two-year-old branches. Thedistribution of branches between the two groupsremains relatively stable over time, showing ex-pansion over that four-year period. We have 270young branch observations and 836 mature branchobservations.

The branches included in the study are groupedinto 13 regions that are situated around large met-ropolitan areas.14 This territorial organization iscommon to the majority of companies in the sector,according to the managers interviewed. Headquar-ters are located in one of these regions. The fourmain regions have 70% of the mature branches and67% of the young branches. The company is there-fore expanding its business into regions where italready has a large presence. In addition, the level ofcompetition in these four regions is high. If we rankthe 13 regions using the number of travel agencybranches permillion people in the region, these fourregions rank second to fifth, only after the region

where the headquarters is located, which occupiesthe first place.15

Variable Definition

Our dependent variable is the change in budg-eted sales (DBi,t) for each period and each branch.Sales have also been used in previous literatureto study how targets—aspirations—are set (Lant,1992; Lant & Montgomery, 1987; Mezias et al.,2002).Wedefine the variable asDBi,t5 (Bi,t –Bi,t–1)/Bi, t–1 to account for the differences in branch size.Bi,t is the budgeted guided sales for branch i in yeart. This definition is also consistent with previouswork on target setting (Bouwens & Kroos, 2011;Leone & Rock, 2002). The variable measures theincrease or decrease in a branch’s target relativeto its previous year’s target.

We use last year’s actual guided sales minus thetargeted guided sales as our measure of past per-formance deviation. We define the measure as PDi,

t–1 5 (Ai,t–1 – Bi,t–1)/Bi,t–1, where Ai,t refers to theactual sales for branch i in year t. In addition toa branch’s past performance, managers also usethe performance of comparable branches to inferthe expected performance of a particular branch.We measure relative target setting (RTSi,t–1) asfollows:

RTSii,t21 5

"Bi,t21/employeesi,t21

2

+n

j51Aj,t21/employeesj,t21

!/n

#,"

+n

j51Aj,t21/employeesj,t21

!/n

#

n is the number of branches in branch i’s region.RTSi,t–1 compares last year’s expected performancefor a particular branch with the actual average per-formance of the branches in its region. We normal-ize by employees to compare across branches withdifferent numbers of employees. We use the ex-pected performance (Bi,t–1) to eliminate idiosyn-cratic risks associated with the actual realization ofperformance. The average performance of branches

12 We lose one observation because the budgeted num-ber of employees was missing.

13 These two categories are based on discussions withcompany managers regarding the evolution and life cycleof branches. The findings are robust to defining youngbranches as only two years old. Becausewe need to have atleast one target review, branches are at least two years old.This fact also reduces the potential effect of higher or lowerthan normal transitory shocks associated with the noveltyeffect. In certain industries, such as fast food, the initialdemand can be above the normal demand as people try theconcept, or lower than normal as they become accustomedto the concept. The divisionalmanager described branchesas maturing over a three- to four-year period.

14 The languageused in all regions and inheadquarters isthe same.

15 When we asked the managers why they were operat-ing in the same regions as their competitors, we receivedthe following answer: “Everybody is there, because busi-ness is there, and . . . we had to be there, too.”


in a region balances the individual branch risks.16 Apositive RTS indicates a target (per employee) forthe branch above its peers, whereas a negative valueindicates a target (per employee) below peers.

Our control variables include the increase in re-sources available to branch managers. Brancheshave a commercial orientation such that the mostimportant resources are people. An increase in thenumber of employees implies more available re-sources. Additional resourcesmay also lead tomoredemanding targets. We measure these resources asthe change in the number of budgeted employeesfrom the previous year to the current year (DΒEmpi,t).In our setting, decisions about the slots available atthe branch level are made during the planning pro-cess. Our measure of the number of employees is esti-mated as the total number of hours divided by thework-load of a full-time employee. Because travel isa cyclical business, branches use part-time contractsduring busy periods. To control for the changes in thelevel of competition, we measure the change in thenumber of competitors. An increase in competitionmay have a negative effect on targets because newentrants increase supply. We use regional-level datafrom the government statistics office. We measurethe competition change as the change in the numberof travel agency branches per million people in theregion from the previous year to the current year(DCompi,t–1). We also include two dummies to

control for changes in branch and regional man-agers (DDirectori,t–1 and DManageri,t–1, respectively).Regional managers accumulate experience andknowledge over time, and replacing them can af-fect target setting in the short run. In addition,different regional managers have different man-agement styles and may be more stringent or le-nient when compared with each another. Branchmanagers also have different styles and skills formotivating their employees.17 Finally, we includeyearly dummies to control for time-related, company-wide effects.18

Table 2 provides sample statistics on the abso-lute values of variables to better visualize the re-search setting. Target “guided sales” increasedover the study period from V748,818 to V944,547;actual “guided sales” also grew from V739,477to V913,017. The percentage of “guided sales” tototal sales increased over the period reachinga target level of 77% and an actual level of 78%. Intotal, 41% of the branches exceeded their target; in2005, only 25% of the branches surpassed theirobjective.

In Table 3, Panel A reports the descriptive sta-tistics for the full sample. The average increase intargets is 0.14 (DB). However,more than 25%of thebranches have reduced their targets, whereas theupper quartile has increased its targets by 23%.Branches fail to meet their targets by 0.04 on av-erage (PD). Only one-year-old branches have fa-vorable performance deviations (not reported).After the third year, the percentage of brancheswith favorable performance deviations remainsstable at approximately 40%. The RTS is posi-tive, which again reflects targets that are higherthan actual performances in a large number ofbranches. The average number of employees perbranch was 2.74. Panels B and C provide the samedescriptive statistics but group branches by age.Young branches show significantly larger targetincreases and larger standard deviations thanmature branches. Hence, the distribution shifts tothe left, with lower values for the standard deviationas branches age. The shape of the performance

16 This measure of relative target setting is referred to asrelative target difficulty. An alternative way to measurethis vicarious learning is to use a relative outcome thatreplaces expected branch performance with actual branchperformance.Relative target difficulty is less noisybecauseit operates with expectations and excludes the noise termof the actual outcome of the individual branch. This noiseterm in the average performance of the branches in theregion is much smaller because independent error termscancel each other out. Both of these measures of relativetarget setting have a comparable correlation with Bi,t–1 at0.35. Relative target difficulty has a correlation of 20.10with (Ai,t–1 – Bi,t–1), and relative outcome and (Ai,t–1 – Bi,t–1)have a correlation of 0.50. Another advantage of relativetarget difficulty is that it captures easy versus difficulttargets. In contrast, relative outcome captures good versusbad performers. We ran our models with relative out-comes. This formulation has a lower R-squared value,probably because of the noise associated with the actualperformance, and the coefficients have higher p values. Athird alternative for operationalizing relative target set-tings is to compare the branch’s budget to the averagebudget of the region’s branches. This alternative is referredto as relative expected performance. The results are con-sistent when using these alternative specifications.

17 There was only one change in regional managers in2005. The company informed us that the manager hadreached retirement age and was replaced by one of thepeople that had beenworking in his team for several years.During the sample period, there were eight changes inbranch managers of mature branches.

18 Regional dummies are constant over time and aretherefore included in the branches’ fixed effects.


deviations (PD) is more compressed in maturebranches. The standard deviation for young branches(0.25) is significantly higher than the standard de-viation for mature branches (0.18). If deviations aresplit between positive and negative values, youngbranches also exhibit a significantly higher standarddeviation.

Consistent with prior results on theweight of pastperformance on target setting, performance de-viation (PD) is positively correlated with targetchanges (DB). This correlation is higher for maturebranches. RTS is negatively correlated with targetchanges. This relationship reflects branches withtargets that are tougher than those of their referencegroup, which means having a smaller increase innext year’s target.

RESULTS

We test our hypotheses using a within-groupfixed effect estimation19 with panel robust

standard errors to correct for serial correlation andheteroskedasticity. We assume the structure of theerror term in our model to be «i,t 5mi 1 yit; the twoorthogonal components include the fixed effect,20

or unobserved time-invariant heterogeneity acrossbranches, mi, and idiosyncratic shocks, yit. The ex-planatory variables are uncorrelated with the idi-osyncratic component but can be correlated withthe unobserved heterogeneity.

Our hypotheses predict a change in the managers’use of available information as they learn to interpretthe branch’s past performance and peers’ perfor-mance. In particular, hypothesis 1 predicts an in-crease in the relevance of past branch performance(PD) over time as the managers learn the character-istics of each branch. Hypothesis 2 predicts a de-crease in the relevance of the performance ofcomparable branches (RTS) over time as vicariouslearning is replaced by experiential learning. Weexamine these predictions using the followingbaseline model, which includes controls for years(Model 2 of Table 4):21

DBi,t 5a0 1 lPDi,t21 1 lYYi,t21PDi,t21 1b RTSi,t21

1bYYi,t21RTSi,t21 1a1DDirectori,t1a2DManagern,t21 1a3DBEmpi,t

1a4DCompi,t21 1 «i,t

TABLE 2Sample Statistics

(in V)

Average “Guidedsales” per branch

Average percentage“Guided sales” tototal sales per

branch

Averagenumber ofemployees

perbranch

Averageperformancedeviation of“guidedsales”

Percentage ofbranches thatexceeded“guided

sales” budgetYearsNumber ofobservations

Budgeted(B)

Actual(A)

Budgeted(B/BR)

Actual(A/AR) (A-B)

2003 237 748,818 739,477 0.72 0.74 2.6 29,341 0.462004 234 871,423 874,940 0.75 0.78 2.8 3,517 0.512005 283 959,451 867,491 0.77 0.78 2.9 291,960 0.252006 351 944,547 913,017 0.77 0.78 2.9 229,192 0.43Average 1,105 890,899 856,125 0.75 0.77 2.8 234,079 0.41

B is budgeted guided sales, BR is budgeted sales, A is actual guided sales, and AR is actual sales.

19 The within-group estimator is the pooled ordinaryleast square (OLS) of the demeaned variables. We alsoadjust our measure of goodness of fit R2 for fixed effects.Difference Generalized Method of Moments (GMM) in-cludes lag variables as instruments to address potentialadditional endogeneity issues (Bascle, 2008; Roodman,2009; Semadeni, Withers, & Certo, 2014); however, work-ing with young branches with little past experience limitsthe option of using a difference GMM specification;branches with one of two years of operations lack thenecessary lags in the explanatory variables. Our specifi-cation controls for time-invariant differences across branchesas well as yearly division-wide variation. Furthermore, theresults of the baseline model are robust to difference GMMspecification.

20 The fixed effect in our setting includes factors such asthe distance to headquarters, the region where the branchis located, the branch’s culture, and the level of wealth ofthe surrounding neighborhood,whichwe assume does notchange during the time span of our study.

21 We include the direct effect of the relevant dummiesin the estimation of all models.


TABLE 3Descriptive Statistics

Panel A: Full Sample

Variable Mean SD Q1 Median Q3 1 2 3 4 5 6

1. Target revision (DB) 0.14 0.31 20.02 0.10 0.232. Performancedeviation

(PD)20.04 0.20 20.16 20.04 0.08 0.51 **

3. Favorable PD 0.14 0.15 0.05 0.10 0.19 0.55 **4. Unfavorable PD 20.17 0.14 20.24 20.13 20.06 0.11 **5. Relative target setting

(RTS)0.06 0.26 20.09 0.05 0.20 20.44 ** 20.13 ** 20.25 ** 0.11 **

6. Change in budgetedemployees (DEmp)

0.06 0.23 0.00 0.00 0.13 0.32 ** 0.23 ** 0.20 ** 0.15 ** 0.27 **

7. Change in competition(DComp)

0.05 0.04 0.01 0.04 0.08 20.08 ** 20.08 ** 20.08 0.04 0.06 ** 20.06

N5 1,105 except for Favorable PD and Unfavorable PD; for these two variables, the numbers of observations are 462 and 643, respectively.* p, 0.05**p , 0.01; two-tailed test.

Panel B: Young Branches


1. Target revision (DB) 0.27 0.50 0 0.17 0.332. Performance

deviation (PD)20.03 0.25 20.18 20.03 0.09 0.40 **

3. Favorable PD 0.17 0.21 0.05 0.11 0.19 0.55 **4. Unfavorable PD 20.20 0.14 20.27 20.17 20.10 20.135. Relative target

setting (RTS)20.06 0.27 20.20 20.03 0.10 20.64 ** 20.20 ** 20.37 ** 0.18 **


2.65 1.07 2.00 2.50 3.00 20.37 ** 0.06 20.07 0.22 ** 0.43 **


0.05 0.04 0.02 0.07 0.08 20.11 20.06 20.16 0.15 0.09 20.07

N 5 270 except for Favorable PD and Unfavorable PD; for these two variables, the numbers of observations are 120 and 150, respectively.* p, .05**p , 0.01; two-tailed test.

Panel C: Mature Branches


1. Target revision (DB) 0.09 0.19 20.03 0.09 0.212. Performance

deviation (PD)20.03 0.18 –0.14 20.04 0.07 0.72 **

3. Favorable PD 0.13 0.13 0.04 0.10 0.19 0.60 **4. Unfavorable PD 20.14 0.11 20.21 20.12 20.06 0.51 **5. Relative target

setting (RTS)0.12 0.24 20.04 0.10 0.25 20.23 ** 20.11 ** 20.11 * 0.03


0.07 0.23 0.00 0.00 0.13 0.43 ** 0.24 ** 0.19 ** 0.18 ** 0.35 **


0.04 0.04 0.01 0.04 0.08 20.10 ** 20.10 ** 20.06 0.01 0.08 * 20.05

N 5 835 except for Favorable PD and Unfavorable PD; for these two variables, the numbers of observations are 342 and 493, respectively.*p , 0.05

**p , 0.01; two-tailed test.


Yi,t–1 is a dummy variable that takes a value of one foryoung branches leaving mature branches as the refer-ence group. Thus, lY and bY capture the incrementaleffect for young branches. A value of zero implies thatthe effect of the independent variables is the same foryoung and mature branches. Conversely, lY , 0 andbY , 0 will support hypotheses 1 and 2, respectively.Given thedefinitionof ourYdummy, the sumofl andlY measures the totalweight ofperformancedeviationfor young branches. Similarly,b plusbY measures theweight of RTS for young branches.

Table 4 presents a first model with only controlvariables. Consistent with the relevance of perfor-mance deviation, we find (Model 2) a positive andsignificant coefficient for last year’s deviation of 0.61.Thus, for each percentage point that performance de-viates from last year’s target, next year’s target in-creases by 0.61 relative to last year’s target. Themagnitude of the coefficient is comparable to that re-ported in previous studies (Leone & Rock, 2002). Thecoefficient on performance deviation for youngbranches (Model 2: lY5 –0.35,p, 0.05) is significant,

indicating that theweight of past performance in thesebranches is only 0.26 (0.61–0.35 5 0.26). Thus, man-agers assign less weight to past performance in youngbranches compared with their mature peers. As ex-pected, relative target setting isnegative (–0.45).Youngbranches have a significant negative coefficient forrelative target setting (bY5–1.20, p , 0.01), as pre-dicted (hypothesis 2). RTS is significantly larger inyoung branches (–0.45–1.20 5 21.65), a result con-sistent with the presence of vicarious learning.

The learning content of successes and failures hasbeen argued to be different. Hypothesis 3 predictsthis effect to affect the weight on past performanceand peer performance over a branch’s life cycle.Model 3 extends the baseline model to include twoadditional terms:

Baseline Model1 lUNi,t21 3PDi,t21

1 lYUYi,t21 3Ni,t21 3PDi,t21 1 «i,t

Ni,t–1 is a dummy variable that takes a value of one iflast year’s deviation was unfavorable (PDi,t–1 , 0) and

TABLE 4Experiential and Vicarious Learning in Target Setting

Variable Coef. Model 1 Model 2 Model 3 Model 4

Past performance deviation (PD) l 0.61 *** 0.55 *** 0.34 **Young x PD lY (H1)20.35 ** 0.08 20.94Unfavorable x PD lU 0.07 0.31Young x unfavorable x PD lYU (H3)21.00 ** 1.09Relative target setting (RTS) b 20.45 *** 20.44 *** 20.43 ***Young x RTS bY (H2)21.20 *** 21.10 *** 20.97 ***Large deviation (L) x PD lL 0.48 **Young x L x PD lYL (H4) 0.79L x unfavorable x PD lLU 20.82 ***Young x L x unfavorable x PD lYLU (H4)23.09 ***Constant a0 0.08 *** 0.18 *** 0.19 *** 0.21 ***Change in branch’s director a1 20.30 *** 20.29 *** 20.30 *** 20.29 ***Change in regional manager a2 20.01 0.01 0.02 0.02Change in budgeted employees a3 0.30 *** 0.44 *** 0.43 *** 0.42 ***Change in competition a4 0.41 20.23 20.20 20.24Dummy for young 0.46 *** 0.05 0.05 0.11Dummy for unfavorable PD 20.01 20.03Dummy for young x dummy for unfavorable PD 20.12 20.05Dummy for L 20.12 ***Dummy for young x dummy for L 20.04Dummy for L x dummy for unfavorable PD 0.003Dummy for young x dummy for L x dummy for

unfavorable PD20.55 *

Branch fixed effects and year dummies Yes Yes Yes YesWithin R2 32.4% 74.6% 75.8% 77.39%Adjusted R2 28.8% 73.1% 74.3% 76.59%Number of observations 1,105 1,105 1,105 1,105

*p , 0.10**p , 0.05

***p , 0.01; two-tailed test.


separates failure from success. The coefficient lU re-flects the incremental weight on past performance forfailures. The coefficient lYU captures the incrementalweight of failures in young branches.

Consistent with hypothesis 3, the coefficient foryoung branches when performance deviations areunfavorable is significant and negative—the weightgiven for young branches (–0.30) is smaller than theweight given formature branches (0.625 0.551 0.07).In contrast, the coefficient for favorable performancedeviations is not significantly different for youngbranches (0.63 5 0.55 1 0.08) and mature branches(0.55) (Figure 2).

The response to successes and failures is not signif-icantly different in mature branches, suggesting thatthe information content of the two events is compara-ble. In contrast, the response in young branches isasymmetrical. The coefficient for young branches thatperform above target is 0.63 (0.551 0.08), whereas thecoefficient for young branches that perform belowtarget is20.30 (0.551 0.081 0.07–1.00).22 Thus, in theearly phases of the life cycle, managers appear to rely

on successes more than they do on failures when set-ting targets.

Hypothesis 4 examines learning for extreme per-formance deviations. It predicts a difference in theweight of information across branches’ life cycles forlarger unfavorable performance deviations. Model 4incorporates the following terms into the baselinemodel:

Baseline Model1 lLLi,t21 3PDi,t21 1 lYLLi,t213Yi,t21 3PDi,t21 1 lLULi,t21 3Ni,t21 3PDi,t21

1 lYLULi,t21 3Yi,t21 3Ni,t21 3PDi,t21 1 «i,t

Li,t–1 is a dummy variable for large performancedeviations, both favorable and unfavorable. Thevariable takes a value of one for the branch-yearobservations with a performance deviation valuelarger than the 87.5 percentile or smaller than the12.5 percentile of the performance deviationdistribution.23

For large performance deviations, young andmature branches show different behaviors forunfavorable performance deviations. The coeffi-cient for large unfavorable deviations is 21.84

FIGURE 2Response to Performance Deviation for Young and Mature Branches (Hypothesis 3)

0.50

0.45

0.40

0.35

0.30

0.25

0.20

0.15

0.10

0.05

0.00–0.3 –0.2 –0.1 0 0.1 0.2

Unfavorable PerformanceDeviation

λ= 0.62

λ= 0.55

λ= 0.63

λ= −0.30

Targ

et R

evis

ion

Solid line represents the response for mature branches. The dotted line representsthe response for young branches.

Favorable PerformanceDeviation

0.3

22 l1 lY (the coefficient for successful young branches)is significantly different from l 1 lY 1 l2 1 l 2 Y (thecoefficient for unsuccessful young branches) at the 5%level (p 5 0.04).

23 The results are similar for percentiles 10 and 90 andpercentiles 15 and 85.


(0.801 0.48–0.821 0.79–3.09) for young branchesand 0.31 (0.34 1 0.31 1 0.48–0.82) for maturebranches. However, the difference across thesecoefficients for large favorable deviations is in-significant. Managers behave as if they perceivedifferent learning opportunities for unfavorableversus favorable large deviations. Figure 3 plotsthe slopes of the response coefficients to large de-viations for young and mature branches. Further-more, young branches show significantly differentbehaviors for small and large unfavorable perfor-mance deviations (p , 0.01). Finally, large favor-able performance deviations increase the nextyear’s target by 0.48 (p , 0.05) more than smallerfavorable performance deviations, regardless ofthe branch’s life cycle.

Our last hypothesis (hypothesis 5) examines cross-learning between experiential and vicarious learn-ing. We test this hypothesis by estimating thefollowing model (Model 5):

Baseline Model1 d RTSi,t21 3PDi,t21 1 dYRYi,t21

3RTSi,t21 3PDi,t21 1 «i,t

In addition, given our prior results, we also studywhether this relationship differs across favorableand unfavorable deviations using Model 6:

Baseline Model1 d RTSi,t21 3PDi,t21

1 dYYi,t21 3RTSi,t21 3PDi,t21

1 lUNi,t21 3PDi,t21 1 lYYi,t21 3PDi,t21

1 dU 3Ni,t21 3RTSi,t21 3PDi,t21

1 dYUYi,t21 3Ni,t21 3RTSi,t21 3PDi,t21 1 «i,t

The results are reported inTable 5. The interactionbetween PD and RTS is not significant when de-viations are not split into success and failure (Model5). However, when the distinction between favor-able and unfavorable performance deviations is in-cluded, the interaction is not significant in maturebranches but is highly significant in young branches.The interaction effect of RTS in young brancheswithfavorable performance deviations is PD * (–0.26 –

1.09)521.35*PDand (–0.2621.0910.1912.62) *PD 5 1.46 * PD for unfavorable deviations, bothsignificantly different from the behavior of maturebranches (–0.26*PD and 20.07*PD). Furthermore,the slopes are not significantly different fromzero formature branches; thus, past performance deviationsdo not affect the way in whichmanagers incorporateinformation about the reference group in revisingtargets. Figure 4 graphs this interaction effect for

FIGURE 3Response to Performance Deviation for Young and Mature Branches for Large Deviations (Hypothesis 4)

0.60

0.50

0.40

0.30

0.20

0.10

0.00

–0.10

–0.1 0.1 0.1 0.1–0.1–0.1

–0.20

λ= −1.84

λ= 0.31

λ= 0.82

λ= 0.67

Large Unfavorable PerformanceDeviation

Large Favorable PerformanceDeviation


Targ

et R

evis

ion


young andmature branches. Young branches exhibita steeper slope in both panels.

DISCUSSION AND CONCLUSIONS

To affect organizational outcomes, organizationallearning must change organizational practices, suchas how aspirations are formed. Understanding howmanagers change their processing of information canhelp better model learning activities and comple-ment prior research.

This study takes advantage of a unique researchsetting to examine how managers use different

sources of information as they learn about the char-acteristics of new operational units. The case studydesign controls for potential confounding factorsthat appear when studying multiple organizations.We examine this effect of organizational learning bystudying how targets that describe the aspirations oforganizational units change over their life cycles.The study takes advantage of a period of rapidgrowth for an organization with a large number ofnew openings.

Consistent with prior research, the weight of pastperformance is positive and less than one, reflectingthe fact that only aportion of theunexpected changesare permanent and, in turn, affect future perfor-mance; the weight of relative target setting is alsonegative, reflecting branches’ regression back to theaverage performance of their reference groups. Theresults indicate the important learning effects oninformation use as operational branches mature.Early in the lives of these branches, managers relymore on peers’ performance to set targets; as man-agers accumulate experience, they shift their use ofinformation, relying to a larger extent on the unit’spast performance. These findings suggest a change inthe role of experiential and vicarious learning asbranches mature.

Furthermore, we find that managers learn differ-ently from successes and failures early in the lives ofthe new units, and this learning is affected by themagnitude of successes and failures. Thus, in oursetting, learning from failures depends not only onthemagnitude of the failure but also on the age of theunits. This evidence is consistentwith prior researchcomparing successes and failures (Madsen & Desai,2010), but it extends it beyond extreme failures toroutine failures and conditions it on the stage of theunit’s life cycle. This paper also shows thatmanagersbehave as though they interpret past performancebased on how the branch performed relative to itspeers. The interaction between both sources of in-formation in younger branches disappears inmaturebranches. This finding informs existing evidence onhow different types of experience affect organiza-tional learning (Baum & Dahlin, 2007), extending itto different types of information.

We model the target-setting process using fixed-effects specifications analyzing within-unit ratherthan between-unit variation. Consequently, we cap-ture learning as reflected in managers’ evolving useof different sources of information as a unit matures.The research design proxies learning as the differ-ential use of information across the unit’s life cycle,instead of using organizational outcomes as a proxy

TABLE 5Cross-Learning Interaction

Variable Coef. Model 5 Model 6

Past performancedeviation (PD)

l 0.61 *** 0.54 ***

Young x PD lY 20.41 ** 20.57 **Relative target setting

(RTS)b 20.47 *** 20.44 ***

Young x RTS bY 21.19 *** 20.75 ***RTS x PD d 20.14 20.26Young x RTS x PD dY (H5) 0.002 (H5) 21.09 *Unfavorable x PD lU 0.08Young x unfavorable

x PDlYU 20.02

Unfavorable x RTS xPD

dU 0.19

Young x unfavorablex RTS x PD

dYU (H5) 2.62 **

Constant a0 0.18 *** 0.19 ***Change in branch

directora1 20.30 *** 20.31 ***

Change in regionalmanager

a2 0.02 0.02

Change in budgetedemployees

a3 0.43 *** 0.44 ***

Change incompetition

a4 20.20 20.16

Dummy for young 0.05 0.11Dummy for

unfavorable PD20.02

Dummy for young xdummyfor unfavorable PD

20.12

Branch fixed effectsand year dummies

Yes Yes

Within R2 74.8% 77.0%Adjusted R2 73.1% 75.4%Number of

observations1,105 1,105

*p , 0.10**p , 0.05

***p , 0.01; two-tailed test.


for learning. Future research can use qualitativedesigns to capture the mechanics of this learningprocess. Although the use of information brings uscloser to understanding the learning mechanisms, itlacks a direct relationship with organizational out-comes to assess the quality of learning. Furthermore,a single case study design does not allow for the ef-fect of different organizational characteristics on thelearning dynamics to be examined. This approachalso limits conclusions to the particular character-istics of the research setting, such as its unique target-setting process, the compensation system used, thestrategy of the division, and the replicability of thebusinessmodel across operational units. Alternativeorganizational designs are likely to affect organiza-tional learning. Similarly, we study a period of rapidgrowth, yet the learning dynamicsmight be differentfor organizations that face a period of crisis andmustdecide which units to close.

The study also contributes to the target-settingliterature by highlighting the importance of learningas new units come into operation. Prior research ontarget setting has taken a static view (Bouwens &Kroos, 2011; Leone & Rock, 2002), implicitly as-suming that organizations are in a stable situationand that whenever a new unit opens, it instan-taneously reaches an equilibrium. Our findings

suggest otherwise. Target setting is a learning pro-cess in which managers constantly change their useof different sources of information as they learn.Consistent with the adaptive aspirations literature,we document this phenomenon in new units as theybecome operational. We extend this literature fo-cusing on managers’ use of information as capturingex ante evidence of learning, rather than examiningthe consequences of learning on organizational out-comes. Going forward, the dynamic nature of targetsetting could be examined in other settings, suchas economic periods that experience considerablechanges or across economies with different levels ofpredictability.

Finally, the findings can be interpreted to revealnot only a learning process inherent in the evolutionof an organization’s operations but also managers’shifting attitudes toward learning over the course ofunits’ life cycles. The shift from vicarious to experi-ential learning as operational units mature can re-flect managers’ attitudes rather than a change inlearning opportunities. Over time, managers maydisregard learning from peers in favor of learningfrom experience, even though the information envi-ronment does notmerit such a shift. If this is the case,then managers are ignoring important informationbecause their learning attitude changes.

FIGURE 4Target Revision and Prior Performance Given Relative Target Setting (Hypothesis 5)


0.40

0.30

0.20

0.10

0.000.05–0.05–0.1–0.15–0.2–0.25–0.3

Inte

ract

ion

eff

ect

:RT

SxP

D

Performance Deviation

0.1 0.15 0.2

–0.10

–0.20

–0.30

–0.40

–0.50

–0.60

λ=–1.35∗PD

λ=–0.26∗PD

λ=1.46∗PD

λ=–0.07∗PD


The findings of this paper have several practicalimplications. First, they suggest that early in the lifecycle of an organization or operational unit, vicari-ous learning plays a central role, and having otherorganizations as reference points and benchmarkscan accelerate learning. Over time, vicariouslearning loses much of its prominent role. Disen-tangling these explanations for learning dynamicsis an interesting avenue for future research. A sec-ond practical implication is the complexity of theorganizational learning process and its importancetopresentmanagerswithdiverse sets of informationsources. The interaction between experiential andvicarious learning early in the life of operationalunits illustrates how learning benefits from com-bining different sources of learning. Finally, thepaper reinforces theoretical predictions regardingdifferential learning from failures and successesand the need for organizations to critically assesstheir learning during periods of success.

REFERENCES

Albuquerque, A. 2009. Peer firms in relative performanceevaluation. Journal of Accounting and Economics,48: 69–89.

Aranda, C., Arellano, J., & Davila, A. 2014. Ratcheting andthe role of relative target setting. The Accounting Re-view, 18: 1197–1226.

Argote, L. 2013. Organizational learning: Creating,retaining and transferring knowledge (2nd edn).New York, NY: Springer.

Argote, L., & Epple, D. 1990. Learning curves in manufactur-ing. Science, 247: 920–924.

Argote, L., & Greve, H. R. 2007. A behavioral theory of thefirm—40 years and counting: Introduction and im-pact. Organization Science, 18: 337–349.

Argote, L., & Miron-Spektor, E. 2011. Organizationallearning: From experience to knowledge. Organiza-tion Science, 22: 1123–1137.

Audia, P. G., & Brion, S. 2007. Reluctant to change:Self-enhancing responses to diverging performancemeasures. Organizational Behavior and HumanDecision Processes, 102: 255–269.

Audia, P. G., & Greve, H. R. 2006. Less likely to fail: Lowperformance, firm size, and factory expansion in theshipbuilding industry. Management Science, 52:83–94.

Audia, P. G., Locke, E. A., & Smith, K. G. 2000. Theparadox of success: An archival and a laboratorystudy of strategic persistence following a radical

environmental change. Academy of ManagementJournal, 43: 837–853.

Balasubramanian, N., & Lieberman, M. B. 2010. Industrylearning environments and the heterogeneity of firmperformance. Strategic Management Journal, 31:390–412.

Baum, J.A., &Dahlin,K. 2007.Aspirationperformance andrailroads’ patterns of learning from train wrecks andcrashes. Organization Science, 18: 368–385.

Baum, J. A., Li, S. X., & Usher, J. M. 2000. Making the nextmove: How experiential and vicarious learning shapethe locations of chains’ acquisitions. AdministrativeScience Quarterly, 45: 766–801.

Baum, J. A., Rowley, T. J., & Shipilov, A. V. 2005. Dancingwith strangers:Aspirationperformance and the searchfor underwriting syndicate partners. AdministrativeScience Quarterly, 50: 536–575.

Banker, R. D., & Datar, S. M. 1989. Sensitivity, precision,and linear aggregation of signals for performanceevaluation. Journal of Accounting Research, 27:21–39.

Bascle, G. 2008. Controlling for endogeneity with in-strumental variables in strategic management re-search. Strategic Organization, 6: 285–327.

Beckman, C., & Haunschild, P. 2002. Network learning:The effects of partners’ heterogeneity of experience oncorporate acquisitions. Administrative ScienceQuarterly, 47: 92–124.

Belderbos, R., Van Olffen, W., & Zou, J. 2011. Generic andspecific social learning mechanisms in foreign entrylocation choice. Strategic Management Journal, 32:1309–1330.

Blettner, D. P., He, Z.-L., Hu, S., & Bettis, R. A. 2015.Adaptative aspirations and performance heterogene-ity: Attention allocation among multiple referencepoints. Strategic Management Journal, 36: 987–1005.

Bol, J. C., Keune, T. M., Matsumura, E.M, & Shin, J.Y.2010. Supervisor discretion in target setting: Anempirical investigation. The Accounting Review,85: 1861–1886.

Bouwens, J., & Kroos, P. 2011. Target ratcheting and ef-fort reduction. Journal of Accounting and Economics,51: 171–185.

Brass, D. J., Galaskiewicz, J. H. R., Greve, H. R., & Tsai, W.2004. Taking stock of networks and organizations: Amultilevel perspective. Academy of ManagementJournal, 47: 795–817.

Bromiley, P.,Miller, K.D., &Rau,D. 2001. Risk in strategicmanagement research. InM. A. Hitt, R. E. Freeman, &J. S. Harrison (Eds.), The Blackwell handbook ofstrategic management: 259–288. Malden, MA:Blackwell.


Bromiley, P., & Harris, J. D. 2014. A comparison of alter-native measures of organizational aspirations. Stra-tegic Management Journal, 35: 338–357.

Burns, L., &Wholey, D. 1993. Adoption and abandonmentof matrix management programs: Effects of orga-nizational characteristics and interorganizationalnetworks. Academy of Management Journal, 36:106–138.

Choi, H. C., Kim, J., & Merchant, K. A. 2012. Firms’ com-mitment to mitigate the ratchet effect under annualbonus plans with multiple performance measures.Working paper, University of Southern California.

Collins, J. C., & Porras, J. L. 1994. Built to last: Successfulhabits of visionary companies. New York, NY:Harper Business.

Chuang, Y. T., & Baum, J. A. C. 2003. It’s all in the name:Failure-induced learning by multiunit chains. Ad-ministrative Science Quarterly, 48: 33–61.

Crossan, M. M., Lane, H. W., & White, R. E. 1999. An or-ganizational learning framework: From intuition toinstitution. Academy of Management Review, 24:522–537.

Crossan, M. M., & Berdrow, I. 2003. Organizational learn-ing and strategic renewal. Strategic ManagementJournal, 24: 1087–1105.

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

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

Davila, T., & Foster, G. 2007. Startup firms growth, man-agement control systems adoption, and performance.The Accounting Review, 82: 907–937.

Dekker, H. C., & Van den Abbeele, A. 2010. Organizationallearning and interfirm control: The effects of partnersearch and prior exchange experiences.OrganizationScience, 21: 1233–1250.

Denrell, J. 2003. Vicarious learning, undersampling offailure, and the myths of management. OrganizationScience, 14: 227–243.

Desai, V. 2015. Learning through the distribution of fail-ures within an organization: Evidence from heart by-pass surgery performance.Academy ofManagementReview, 58: 1032–1050.

Easterby-Smith, M., Crossan, M., & Nicolini, D. 2000. Or-ganizational learning: Debates past, present and fu-ture. Journal of Management Studies, 37: 783–796.

Fayol, H. 1916. Administration industrielle et generale.Bulletin de la Societe de l’Industrie Minerale, 10:5–164.

Fisher, J. G., Maines, L. A., Peffer, S. A., & Sprinkle, G. B.2002. Using budgets for performance evaluation:

Effects of resource allocation and horizontal in-formation asymmetry on budget proposals, budgetslack, and performance.TheAccounting Review, 77:847–866.

Garud, R., Dunbar, L.M., & Bartel, C. A. 2011. Dealingwithunusual experiences: A narrative perspective on or-ganizational learning. Organization Science, 22: 587–601.

Greve, H. R. 1998. Performance, aspirations, and risky or-ganizational change. Administrative Science Quar-terly, 43: 58–86.

Greve, H. R. 2002. Sticky aspirations: Organizational timeperspective and competitiveness.Organization Science,13: 1–17.

Greve, H. R. 2003. A behavioral theory of R&D expendi-tures and innovations: Evidence from shipbuilding.Academy of Management Journal, 46: 685–702.

Greve, H. R. 2007. Exploration and exploitation in productinnovation. Industrial and Corporate Change, 16:945–975.

Greve, H. R. 2008. A behavioral theory firm growth: Se-quential attention to size and performance goals.Academy of Management Journal, 51: 476–494.

Haunschild, P. R., & Miner, A. S. 1997. Modes of in-terorganizational imitation: The effects of outcomesalience and uncertainty. Administrative ScienceQuarterly, 42: 472–500.

Haunschild, P. R., & Sullivan, B. N. 2002. Learning fromcomplexity: Effects of prior accidents and incidents onairlines’ learning.AdministrativeScienceQuarterly,47: 609–643.

Hayward, M. L., Rindova, V. P., & Pollock, T. G. 2004.Believing one’s own peers: The causes and conse-quences of CEO celebrity. Strategic ManagementJournal, 25: 637–653.

Holmstrom, B. 1979. Moral hazard and observability.The Bell Journal of Economics, 10: 74–91.

Huber, G. P. 1991. Organizational learning: The contrib-uting processes and the literatures. OrganizationScience, 2: 88–115.

Indjejikian, R., Matejka, M., Merchant, K. A., & Van derStede, W. A. 2014. Earning targets and annual bonusincentives. The Accounting Review, 89: 1227–1256.

Ingram, P., & Baum, J. A. 1997. Chain affiliation and thefailure of Manhattan hotels, 1898–1980. Administra-tive Science Quarterly, 42: 68–102.

Kaplan, R. S., &Norton,D. 2001. Transforming the balancedscorecard from performance measurement to strategicmanagement: part I.AccountingHorizons, 15: 87–104.

Kim, J. Y., Finkelstein, S., & Haleblian, J. 2015. All aspira-tions are not created equal: The differential effectsof historical and social aspirations on acquisition


behavior. Academy of Management Journal, 58:1351–1388.

Kim, J. Y., &Miner, A. S. 2007. Vicarious learning from thefailures and near-failures of others: Evidence from theU. S. commercial banking industry. Academy ofManagement Journal, 50: 687–714.

Kraatz, M. S. 1998. Learning by association? Interorganiza-tionalnetworks andadaptation to environmental change.Academy of Management Journal, 41: 621–643.

Lampel, J., Shamsie, J., & Shapira, Z. 2009. Experiencingthe improbable: Rare events and organizationallearning. Organization Science, 20: 835–845.

Lant, T. K. 1992. Aspiration level adaptation: An empiricalexploration.Management Science, 38: 623–644.

Lant, T. K., Milliken, F. J., & Batra, B. 1992. The role ofmanagerial learning and interpretation in strategicpersistence and reorientation: An empirical explo-ration. Strategic Management Journal, 13: 585–608.

Lant, T. K., & Montgomery, D. B. 1987. Learning fromstrategic success and failure. Journal of Business Re-search, 15: 503–518.

Lant, T. K., & Shapira, Z. 2008.Managerial reasoning aboutaspirations and expectations. Journal of EconomicBehavior & Organization, 66: 60–73.

Leone, A. J., & Rock, S. 2002. Empirical tests of budgetratcheting and its effect on managers’ discretionaryaccrual choices. Journal of Accounting and Eco-nomics, 33: 43–67.

Levinthal, D., & March, J. G. 1981. A model of adaptiveorganizational search. Journal of Economic Behavior& Organization, 2: 307–333.

Levinthal, D., & March, J. G. 1993. The myopia of learning.Strategic Management Journal, 14: 95–112.

Levitt, B., & March, J. G. 1988. Organizational learning.Annual Review of Sociology, 14: 319–340.

Li, S., & Rowley, T. J. 2002. Inertia and evaluation mecha-nisms in interorganizational alliance selection decisions.Academy of Management Journal, 45: 1104–1119.

Madsen, P. M. 2009. These lives will not be lost in vain:Organizational learning from direct and vicarious ex-perience with disasters and minor accidents inU.S. coal mining.Organization Science, 20: 861–875.

Madsen, P. M., & Desai, D. 2010. Failing to learn? The ef-fects of failure and success on organizational learningin the global orbital launch vehicle industry. Acad-emy of Management Journal, 53: 451–476.

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

March, J. G., & Shapira, Z. 1987. Managerial perspectiveson risk and risk taking. Management Science, 33:1404–1418.

March, J. G., & Shapira, Z. 1992. Variable risk preferencesand the focus of attention. Psychological Review, 99:172–183.

March, J. G., & Simon, H. A. 1958. Organizations. NewYork, NY: Wiley.

March, J. G., & Sutton, R. I. 1997. Organizational perfor-mance as a dependent variable. Organization Sci-ence, 8: 698–706.

Matsumura, E.M., & Shin, J. Y. 2006.An empirical analysisof an incentive plan with relative performance mea-sures: Evidence fromapostal service.TheAccountingReview, 81: 533–566.

Merchant, K. A., & Van der Stede, W. A. 2011. Manage-ment control systems: Performance measurement,evaluation and incentives (3rd edn). Upper SaddleRiver, NJ: Prentice Hall.

Mezias, S. J., Chen,Y.-R., &Murphy, P. R. 2002.Aspiration-level adaptation in an American financial servicesorganization: A field study. Management Science,48: 1285–1300.

Mishina, Y., Dykes, B. J., Block, E. S., & Pollock, T. G. 2010.Why “good” firms do bad things: The effects of highaspirations, high expectations, andprominence on theincidence of corporate illegality. Academy of Man-agement Journal, 53: 701–722.

Muehlfeld,K., Shabib, P. R., &Witteloostuijn,A.V. 2012.Acontextual theory of organizational learning fromfailures and successes: A study of acquisition completionin the global newspaper industry, 1981–2008. StrategicManagement Journal, 33: 938–964.

Murphy, K. J. 2000. Performance standards in incentivecontracts. Journal of Accounting and Economics,30: 245–278.

Nonaka, I. 1994. A dynamic theory of knowledge creation.Organization Science, 5: 14–37.

Polanyi, M. 1958. Personal knowledge: Towards a post-critical philosophy, Chicago, IL: University of Chi-cago Press.

Raju, J. S., & Srinivasan, V. 1996. Quota-based compensa-tion plans for multiterritory heterogeneous sales-forces. Management Science, 42: 1454–1462.

Rerup, C., & Feldman, M. S. 2011. Routines as a source ofchange in organizational schemata: The role of trial-and-error learning. Academy of Management Jour-nal, 54: 577–610.

Roodman, D. 2009. How to do xtabond2: An introductionto difference and system GMM. The Stata Journal, 9:86–136.

Schildt,H.,Keil, T., &Maula,M. 2012.The temporal effectsof relative and firm-level absorptive capacity on in-terorganizational learning. Strategic ManagementJournal, 33: 1154–1173.


Schwab, A. 2007. Incremental organizational learningfrom multilevel information sources: Evidence forcross-level interactions. Organization Science, 18:233–251.

Semadeni, M., Withers, M. C., & Certo, S. T. 2014. Orga-nizational aspirations, reference points, and goals:Building on the past and aiming for the future. Stra-tegic Management Journal, 35: 1070–1079.

Shinkle, G. A. 2012. Organizational aspirations, referencepoints, and goals: Building on the past and aiming forthe future. Journal of Management, 38: 415–455.

Short, J. C., & Palmer, T. B. 2003. Organizational perfor-mance referents: An empirical examination of theircontent and influences. Organizational Behaviorand Human Decision Processes, 90: 209–224.

Simons, R. 1995. Levers of control: How managers useinnovative control systems to drive strategic re-newal. Boston, MA: Harvard Business School Press.

Simons, R. 2000. Performancemeasurement and controlsystems for implementing strategy. Upper SaddleRiver, NJ: Prentice Hall.

Sorenson, O. 2003. Independence and adaptability: Orga-nizational learning and the long-term effect of inte-gration. Management Science, 94: 446–463.

Strang, D., & Macy, M. W. 2001. In search of excellence:Fads, success stories, and adaptive emulation.American Journal of Sociology, 107: 147–182.

Wagner, S., Hoisl, K., & Thoma, G. 2014. Overcoming local-ization of knowledge: The role of professional servicefirms. Strategic Management Journal, 35: 1671–1688.

Walsh, J. P., &Ungson, G. R. 1991.Organizationalmemory.Academy of Management Review, 16: 57–91.

Webb, A., Williamson, M. G., & Zhang, Y. 2013. Pro-ductivity: Target difficulty, performance-based payand outside-the-box-thinking. The Accounting Re-view, 88: 1433–1457.

Weick, K. E. 1979. The social psychology of organizing(2nd edn). New York, NY: McGraw-Hill.

Weick, K. E., & Roberts, K. H. 1993. Collective mindin organizations: Heedful interrelating on flightdecks. Administrative Science Quarterly, 38:357–381.

Weitzman, M. 1980. The “ratchet principle” and per-formance incentives. The Bell Journal of Economics,11: 302–308.

Zollo, M., & Reuer, J. J. 2010. Experience spillovers acrosscorporate development activities. Organization Sci-ence, 21: 1195–1212.

Carmen Aranda ([email protected]) is professor of fi-nance at the School of Economics andBusiness,Universityof Navarra (Spain). She received her PhD in accountingand finance from the University of Navarra. Her currentresearch interests are in the areas of target setting, perfor-mance evaluation, and strategic alliances.

Javier Arellano ([email protected]) is professor of ac-counting at the School of Economics and Business, Uni-versity of Navarra (Spain). He received his PhD inaccounting and finance from theUniversity ofNavarra.Hisresearch centers on how companies set and revise targets,with a current focus on the use of subjective information inthis process.

Antonio Davila ([email protected]) is professor at IESEBusiness School, University of Navarra. He received hisdoctorate from the Harvard Business School. His researchinterests include organizational learning and how com-panies integrate information from the events happening inthe environment.


mailto:[email protected]



Copyright of Academy of Management Journal is the property of Academy of Managementand its content may not be copied or emailed to multiple sites or posted to a listserv withoutthe copyright holder's express written permission. However, users may print, download, oremail articles for individual use.

organizational learning in target setting · proxy for organizational learning compared to changes...

Documents