SIGNALING IN SECRET: PAY FOR PERFORMANCE AND THE INCENTIVE AND … · 2015-03-12 · SIGNALING IN SECRET: PAY FOR PERFORMANCE AND THE INCENTIVE AND SORTING EFFECTS OF PAY SECRECY

SIGNALING IN SECRET: PAY FOR PERFORMANCE AND THEINCENTIVE AND SORTING EFFECTS OF PAY SECRECY

ELENA BELOGOLOVSKYCornell University

PETER A. BAMBERGERTel Aviv University

Although the vast majority of U.S. firms follow a policy of pay secrecy, researchprovides a limited understanding of its overall utility to organizations. Building onsignaling theory, we develop and test a model of the incentive and sorting effects of paysecrecy—a pay communication policy that limits employees’ access to pay-relatedinformation and discourages the discussion of pay issues—under varying pay-for-performance (PFP) system characteristics. Results of a multiround laboratory simula-tion largely support the proposed moderated-mediation model. They indicate that paysecrecy has an adverse impact on individual task performance that is mediated by PFPperceptions, amplified when pay determination criteria are relative (as opposed toabsolute), and attenuated when performance assessment is objective (as opposed tosubjective). Results also indicate that pay secrecy has a similar adverse effect onparticipant continuation intentions (mediated through PFP perceptions, amplifiedwhen pay determination criteria are relative, and attenuated when performance as-sessment is objective), particularly among high performers. These findings suggest thatweak signals associated with a particular managerial practice may become salientwhen interpreted in the context of other practice-based signals and that, under suchconditions, even weak signals may drive negative-oriented inferences, having impor-tant behavioral implications.

Although the National Labor Relations Board(NLRB) and courts have struck down organization-al pay secrecy as a violation of employees’ rightsunder the National Labor Relations Act of 1935(Gely & Bierman, 2003), pay secrecy—a pay com-munication policy that limits employees’ access topay-related information and discourages discus-sion among employees about pay issues—contin-ues to be widespread (Colella, Paetzold, Zardkoohi,& Wesson, 2007; Day, 2007). Indeed, most privatesector employees are discouraged or prohibitedfrom discussing compensation-related issues with

others (IWPR & Rockefeller Survey of EconomicSecurity, 2011; Lawler, 2003).

For scholars as well, pay secrecy continues to bea controversial topic, with little agreement regard-ing its overall impact (Colella et al., 2007; Tremblay& Chenevert, 2008). Those favoring pay secrecyover transparent pay communication speculatethat—particularly in the context of pay for perfor-mance (PFP)—transparency has the undesired ef-fect of causing managers to compress performanceratings in order to avoid the pecuniary costs ofdifferential pay. Such centralization of ratings, byequalizing pay, can generate weak incentive andnegative sorting effects (Bartol & Martin, 1989; Lev-enthal, Karuza, & Fry, 1980; Major & Adams, 1983),both of which would undermine performance. Oth-ers argue that transparency can exacerbate em-ployee jealousies and conflicts (Balkin & Gomez-Mejia, 1990). In contrast, drawing from equity(Adams, 1965) and justice (Greenberg, 1990) theo-ries, those favoring pay transparency argue that, bypromoting internal consistency, pay transparencyreinforces employee trust in management, and thus

We are grateful to Michal Biron, Matt Bloom, DvoraGeller, Barry Gerhart, Nina Gupta, Kevin Hallock, DavidWagner, and Michele Williams for their helpful com-ments and suggestions on earlier versions of this manu-script. We also thank Jason Shaw and three anonymousreviewers for their constructive critiques and insightfulrecommendations. This research was supported by thegenerous financial help of an Israel Foundations Trustees(IFT) doctoral grant awarded to the first author, as well asby a grant from the Henry Crown Institute for BusinessResearch at Tel Aviv University.

1706

� Academy of Management Journal2014, Vol. 57, No. 6, 1706–1733.http://dx.doi.org/10.5465/amj.2012.0937

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.

has positive incentive and sorting effects, boostingtask commitment and facilitating the retention ofstrong performers (Cloutier & Vilhuber, 2008;Colquitt, Conlon, Wesson, Porter, & Ng, 2001; Day,2007; Lawler, 1966a, 1966b, 1967; Milkovich, New-man, & Gerhart, 2013).

Unfortunately, the empirical evidence does littleto resolve this conundrum in that, as Colella et al.(2007: 56) concluded, research on pay secrecy isboth “scanty” and “dated.” Most of our knowledgeregarding the consequences of pay secrecy is basedon empirical research conducted 30 or more yearsago (Lawler, 1966a, 1966b, 1967; Leventhal et al.,1980; Milkovich & Anderson, 1972), with the vastmajority of this research examining the effects ofpay secrecy on employee pay perceptions and man-agerial reward allocation. Indeed, only three stud-ies that we know of (i.e., Bamberger & Bel-ogolovsky, 2010; Futrell & Jenkins, 1978; Tremblay& Chenevert, 2008) have empirically examined theperformance-related consequences of pay secrecy.Notably, two of these studies suggest that pay se-crecy has detrimental performance consequences,while the third suggests beneficial effects. More-over, only the most recent of these studies began toexplore the mechanisms potentially underlyingsuch an association. Testing a variety of possiblemediators based on justice and expectancy theo-ries, Bamberger and Belogolovsky (2010) foundonly PFP perceptions (i.e., the perceptions that em-ployees hold regarding the nature of performance–pay contingencies) to partially explain the effects ofpay secrecy on performance, but only for moreinequity-sensitive individuals. Accordingly, schol-ars have yet to develop a comprehensive theoreticalframework able to explain when and how pay se-crecy may affect task performance regardless of in-dividual differences.

Research on the sorting effects of pay secrecy iseven more limited, with only one study examiningits impact on employee turnover (Card, Mas,Moretti, & Saez, 2012). This study found pay com-munication policy to have no significant effect onjob search among those paid at or above the unit oroccupational median, but did find employees inthe lowest pay quartile and working under condi-tions of pay transparency to be 20% more likelythan their quartile peers working under conditionsof pay secrecy to report searching for a new job(Card et al., 2012). Finally, although Colella et al.(2007) suggested that the broader pay contextwithin which pay secrecy is adopted has signifi-cant implications for the policy’s consequences,

scholars have yet to generate (let alone to test) atheory regarding how other pay system character-istics may moderate the effects of pay secrecy. Ac-cordingly, the purpose of the current research is todevelop and test an integrative model of the incen-tive and sorting effects of pay secrecy under vary-ing pay system characteristics.

In developing and testing our model, we focusattention on one particular dimension of thebroader pay system—namely, the nature of PFP, orthe way in which the pay system accounts for dif-ferential employee contribution. We do so for tworeasons. First, in the private sector, PFP is perva-sive (Gerhart, Rynes, & Fulmer, 2009), with 90%ofall U.S. companies reporting that they connect atleast a portion of their employees’ pay to measuresof individual or firm performance, or some combi-nation of the two (Cohen, 2006), most typically inthe form of a bonus (Bloom & Milkovich, 1998;Milkovich et al., 2013). Second, although most PFPsystems are implemented in the context of paysecrecy, research on PFP under varying degrees ofpay transparency has generated inconsistent results(Perry, Engbers, & Jun, 2009). For example, Pfefferand Langton (1993) demonstrated that whilegreater pay dispersion (a typical outcome of PFP)in higher education is associated with decreasedfaculty productivity, this effect is less severe inprivate universities and colleges in which pay isless transparent. In contrast, Shaw and Gupta(2007) found PFP to enhance the commitment ofhigher performers to the extent that the pay systemwas more transparent. Such inconsistent findingsregarding the consequences of the PFP–pay-secrecyinteraction suggest that, in seeking to understandhow the broader pay system may moderate theincentive and sorting effects of pay secrecy, varyingPFP system characteristics may provide a usefulstarting point.

We develop our model of how PFP system char-acteristics moderate the incentive and sorting ef-fects of pay secrecy by integrating research onpositive–negative asymmetry and paranoid cog-nition (Baumeister, Bratslavsky, Finkenauer, &Vohs, 2001; Kramer, 1998) with signaling theory(Spence, 2002). As pointed out by Connelly, Certo,Ireland, and Reutzel (2011), signaling theory fo-cuses on information asymmetries in an effort toexplain whether and how parties attempt to com-municate information and how recipients interpretsuch signals. Signaling theory may be useful forunderstanding the interactive effects of pay secrecyand PFP for two reasons: First, pay secrecy has a

2014 1707Belogolovsky and Bamberger

direct and obvious impact on information asymme-try in employment relations; second, as noted byGomez-Mejia, Berrone, and Franco-Santos (2010:104), agency theory—a central theory upon whichPFP is based (Cadsby, Song, & Tapon, 2007)—“ar-gues that pay is an important signaling device.”Indeed, agency theory suggests that, particularly inan asymmetrical information context, PFP mayhave beneficial incentive effects (by allowing man-agement to emphasize those behaviors most valuedor to signal those objectives of highest priority), aswell as beneficial sorting effects (by signaling thetype of employees whom it seeks to attract andretain).

However, as we will argue, signaling theory alsosuggests that recipients may draw inferences frompay system attributes beyond or different fromthose intended by management (Connelly et al.,2011; Suazo, Martínez, & Sandoval, 2009, 2011). Inparticular, in contrast to the trust-reinforcing ef-fects of pay transparency noted earlier, employeesmay interpret pay secrecy as a signal of deception(Day, 2007; Milkovich et al., 2013). We base thisassertion on social-psychological research, whichsuggests that situational attributes signaling uncer-tainty or potential risk can foster modes of socialinformation processing that elevate distrust andsuspicion (Baumeister et al., 2001; Kramer, 1998).

Moreover, while signaling theory focuses on de-terministic effects of one signal at a time, we sug-gest that signals tend to be interpreted in light ofother signals, with weak signals potentially becom-

ing salient when interpreted in the context of othersignals. As such, pay secrecy may create a contextin which the inferences that employees draw fromtwo main and interactive PFP attributes—namely,the nature of pay determination criteria (i.e.,absolute vs. relative), and the subjectivity of perfor-mance assessment—ultimately combine to ad-versely affect individual performance and continu-ation intentions. More specifically, we argue thatinferred signals of heightened uncertainty regard-ing how performance links to pay (on the basis ofrelative—as opposed to absolute—pay determina-tion criteria) and inferred signals of heightened riskof bias (on the basis of assessment subjectivity)amplify the adverse effects of pay secrecy on per-formance and continuation intentions. Accord-ingly, we develop a moderated-mediation model ofthe incentive and sorting effects of pay secrecy inthe context of PFP (see Figure 1), which we test onthe basis of a laboratory simulation.

By proposing and testing this model, we offerseveral theoretical contributions. First, we extendsignaling theory by proposing and demonstratingthat the inferences drawn from one particular sig-nal may vary as a function of signals inferred fromother related management policies or practices,suggesting that even weak signals may become sa-lient when interpreted in the context of other sig-nals. We also advance signaling theory by inte-grating notions of positive–negative asymmetry(Baumeister et al., 2001) and sinister attribution(Kramer, 1998), and by demonstrating how contex-

FIGURE 1Theoretical Model

Pay Determination Criteria: Absolute vs.

Relative

Pay Communication Policy: Transparent

vs. Secret

Individual Task Performance at T3

Pay-for-PerformancePerceptions at T2

Subjectivity of Performance Assessment: Objective vs. Subjective

Continuation Intention at T3

1708 DecemberAcademy of Management Journal

tual practices signaling uncertainty can result inthe overweighting of malevolent inferences wheninterpreting weak signals. Additionally, while sig-naling theory has been applied in human resourcesresearch to explain employer and candidate behav-ior in the staffing process (Rynes, 1991; Suazo et al.,2011), we extend signaling theory into the realm ofcompensation, using it as a means to better under-stand how pay administration may influence twokey compensation outcomes: performance, andcontinuation intentions. This is important in that,as Gomez-Mejia et al. (2010: 104) point out withrespect to pay, “there is practically no research onhow the signaling process works (or doesn’twork).” In doing so—following the recommenda-tions of Colella et al. (2007), who suggest that thebroader human resources context may influencethe ultimate impact of pay secrecy on employeeoutcomes—we attempt to systematically incorpo-rate several of these elements into a single, integra-tive theoretical model. Accordingly, we provide amore nuanced and context-sensitive (Johns, 2006)framework within which to understand when andhow pay secrecy may interact with other pay sys-tem attributes to affect employee outcomes. Fi-nally, we extend past research on the performance-related consequences of pay secrecy to the issue ofsorting, theorizing, and demonstrating how, in thecontext of alternative PFP system attributes, paysecrecy may also affect continuation intentions,particularly among high performers.

THEORETICAL BACKGROUND ANDHYPOTHESES

Information Asymmetry and Signaling Theory

In economics, information asymmetry occurswhen one party to an exchange has more or higherquality information than the other—or, as Stiglitz(2002: 469) explained, when “different peopleknow different things.” Information asymmetriescan result in less-than-efficient transactions or,when aggregated to the market level, dysfunctionalmarkets (Akerlof, 1970). Accordingly, as originallyproposed by Spence (1973), informed parties at-tempt to communicate or “signal” certain informa-tion to their underinformed exchange partners,hoping that, by reducing uncertainty on the part ofthe underinformed, they can elicit behavior morefavorable to themselves (the more informed). Con-nelly et al. (2011: 45) refer to this as the “strategicaspect of signaling,” noting that, “for signaling to

take place, the signaler should benefit by someaction from the receiver that the receiver would nototherwise have done.”

However, underinformed parties need not waitfor the informed party to signal; rather, they canscan their environment for signals that might re-duce uncertainty and hence lower risk in choicebehavior. Indeed, Connelly et al. (2011: 54) positthat, for the underinformed, “monitoring the envi-ronment for weak signals can be particularly im-portant.” Given that informed parties typicallysignal strategically (Connelly et al., 2011), suchmonitoring allows underinformed recipients to as-sess the signaler’s credibility and intent (Davila,Foster, & Gupta, 2003).

The effectiveness of signaling in resolving infor-mation asymmetries is contingent on the accuracywith which recipients interpret the signals re-ceived. But, for several reasons, recipients maytend to interpret signals in a manner inconsistentwith the intentions of the signals’ senders. First, asConnelly et al. (2011: 55) write, “receivers mayapply weights to signals in accordance with pre-conceived notions about importance or cognitivelydistort signals so that their meanings diverge fromthe original intent of the signaler.” Second, theymay use weak, intent-oriented signals (Stiglitz,2002), distilled from the environment, as a frame-work with which to interpret stronger signals trans-mitted by their exchange partner, such that theinterpretation may be quite different from the in-tended message (Suazo et al., 2011). For example,although the provision of work–life benefits may beused by management to signal to workers that theyare valued in the hope of eliciting enhanced em-ployee commitment (Lambert, 2000), employeesmay also pick up on other, inconsistent signals(e.g., rumors of an intended layoff or union-bust-ing campaign) that may potentially weaken thestrength of the signal or result in its complete mis-interpretation. Third, consistent with the positive–negative asymmetry effect, in drawing inferencesfrom signals, individuals may pay more attentionto, and overweight, negative information (Baumeis-ter et al., 2001; Peeters & Czapinski, 1990). Thisoverweighting of the negative when drawing infer-ences from signals appears to be particularly robustunder conditions of uncertainty (Kramer, 1998,2001). More specifically, research on paranoid cog-nition indicates that individuals uncertain abouttheir social standing in groups and organizationstend to pay more attention to information support-ive of a stance of distrust than to that supportive of


trust, thus leading them to be suspicious regardingothers’ motives and intentions. Such suspicion canlead to a sinister attribution error (i.e., a tendencyto overattribute hostile intentions and malevolentmotives to others), thus motivating behaviorsaimed at loss prevention or reduction (Kramer,1998, 2001).

Finally, economists typically assume informa-tion asymmetry as a given attribute of a market that,while perhaps exploited by a more informed ex-change partner, is not the result of any intentionalact by either party. However, information asymme-try can also be intentional on the part of one ofthe exchange partners, creating what Williamson(1975) terms “information impactedness.” Such isthe case with pay secrecy, in that, in the vast ma-jority of workplaces, pay-related information is in-tentionally made asymmetrical by employers. Andalthough there is little research regarding how dif-ferent signals may interact with one another toshape choice behavior (Connelly et al., 2011), it isconceivable that the underinformed party may in-terpret such intentional information asymmetry asa signal in and of itself, and thus use that under-standing to facilitate the interpretation and assessthe credibility of other signals (Stiglitz, 2002).

Taken together, signaling theory therefore sug-gests the following four main principles relevant tounderstanding the incentive and sorting conse-quences of pay secrecy in the context of PFP.

Principle 1. Informed exchange partners trans-mit signals to their underinformed partners inorder to influence their behavior.

Principle 2. Underinformed partners scan theirenvironments for signals that may reduceuncertainty.

Principle 3. Signals may be misinterpreted byrecipients.

Principle 4. Intentional asymmetries directlyattributable to the informed party may be in-terpreted by an underinformed party as a sig-nal of intent, which may in turn influence theunderinformed party’s interpretation of othersignals transmitted by the informed party.

Signaling, and the Incentive and Sorting Effectsof Pay Secrecy

As noted earlier, empirical evidence regardingthe consequences of pay secrecy is both limited andequivocal. Moreover, while justice and expectancy

theories have been proposed to explain the possibleperformance and sorting effects of pay secrecy(Colella et al., 2007; Colella, Zardkoohi, Paetzold, &Wesson, 2003), in the one study empirically testinghypotheses drawn from both of these theories(Bamberger & Belogolovsky, 2010) no support wasfound for a justice-based explanation—and onlylimited, person-contingent support was found forthe notion that PFP perceptions underlie the effectof pay communication on performance. Given thatpay secrecy, by its very nature, creates a situation ofinformation asymmetry, we build on the four sig-naling-related principles derived above (see “Infor-mation Assymetry and Signaling Theory”) to ex-tend the findings of Bamberger and Belogolovsky(2010), and to develop a more comprehensive, con-text-sensitive theoretical model of the incentiveand sorting effects of pay secrecy.

An understanding of the performance–pay re-lationship is vital to employees whose pay is atleast partially contingent upon performance,since it is difficult to estimate the utility of mar-ginal effort (and hence difficult to make choicesabout how much additional effort to exert on thejob) without such an understanding (Heneman,Greenberger, & Strasser, 1988). Understandingthe nature of performance–pay contingenciesstrictly on the basis of one’s own experience isproblematic in that it demands the considerationof multiple performance–pay events, which, inthe case of a single individual, can often be in-ferred only over time. In contrast, information onthe pay associated with varying levels of co-worker performance facilitates employees’ timelyestimation of performance–pay relations (Naylor,Pritchard, & Ilgen, 1980).

However, when PFP is applied in the context ofpay secrecy, it is more difficult for employees toestimate the nature of performance–pay relations.As suggested in our review of information asymme-try research, under such conditions underinformedagents (i.e., employees) tend to monitor their envi-ronment for signals providing insights into the na-ture of pay-related contingencies. Because some ofthese signals may be weak and hence rather equiv-ocal, they may be open to misinterpretation (Prin-ciples 2 and 3). Frank (1985) offers an excellentexample of how weak signals may affect employ-ees’ pay-related impressions. He suggests that, inattempting to distill others’ pay status in the con-text of information asymmetry, individuals tend toground their inferences on the observation of oth-ers’ acquisition of “positional goods”—that is,


“things whose value depends relatively strongly onhow they compare with things owned by others,”such as cars and clothing (Frank, 1985: 101). Be-cause positional goods are precisely those highlyobservable goods on which individuals tend tooverspend, individuals may tend to overestimatethe pay of those whom they would otherwise ex-pect to be earning the same or less than themselves.

Aside from generating a situation of informa-tion asymmetry demanding employee inferenceof performance–pay relations from weak signals,pay secrecy and the information asymmetry thatit generates may signal a heightened risk of man-agerial opportunism or deceptive intent. Such asignal may both set the basis for the negativemisinterpretation of the weak signals used to es-timate performance–pay contingencies, as well asdirectly result in downwardly biased perceptionsof these contingencies.

Pay secrecy may signal to employees a height-ened risk of managerial opportunism or deceptiveintent because, as suggested by Principle 4, it gen-erates a situation of intentional information asym-metry that, by its very nature, may raise employeeconcerns of managerial ill intent (Wanasika &Adler, 2011; Williamson, 1975). Furthermore,while there may be objectively valid and positivefactors underlying the adoption of pay secrecy (e.g.,to avoid employee jealousies), consistent with re-search on positive–negative asymmetry the infer-ences that employees draw from pay secrecy arelikely to be more strongly influenced by negative(i.e., opportunistic) factors (Baumeister et al., 2001;Peeters & Czapinski, 1990). Finally, because theinformational asymmetry generated by pay secrecyrelates directly to individuals’ social standingwithin the organization, pay secrecy creates a situ-ation conducive to the sinister attribution error(Kramer, 1998, 2001).

To the extent that pay secrecy signals manage-ment opportunism and deception, it may create amalevolent frame for interpreting the weak signalsused to estimate performance–pay contingencies inthat, as noted by Connelly et al. (2011: 55), weaksignals tend to be interpreted “in accordance withpreconceived notions.” For example, in the ab-sence of other information, employees may baseinferences regarding performance–pay contingen-cies on their perceptions of procedural or informa-tional fairness (Greenberg, 2003). But as Colella etal. (2007: 59) suggest, if these are viewed in a neg-ative light (likely because information is beingwithheld), “then distributive judgments are likely

to be negative as well.” The net result, as suggestedby Lawler (1966a, 1966b) and others (Milkovich &Anderson, 1972), is that when employees draw in-ferences from the weak pay-related signals aroundthem, they tend to overestimate the pay of theircoworkers and underestimate the pay of those hi-erarchically superior to them. By perceiving theupper boundary of pay dispersion as lower and thelower boundary as higher, individuals effectivelycompress the perceived range of pay associatedwith varying levels of contribution. Given this re-duction in the perceived range of pay, while re-wards may still be perceived to be monotonicallyincreasing as a function of performance, the per-ceived strength of this contingency is likely to bediminished. To the degree that pay secrecy elicitsthe implicit compression of the perceived range ofpay, it reduces the perception that any incrementalincrease in performance level will be accompaniedby a larger incremental return, and as such is likelyto adversely affect PFP perceptions.

Additionally, this signal of managerial opportun-ism and deception may directly result in employ-ees’ underestimation of performance–pay contin-gencies. In other words, consistent with the notionof Spence (2002) that the costs of signaling positivequalities (in this case, signaling fairness and trust-worthiness as an employer through a policy of paytransparency) are higher for those with less to offeror “what to hide” (i.e., those employers whose paysystems are inherently unfair to begin with), paysecrecy may signal to employees that performance–pay contingencies fail to reflect fairly the addedvalue generated by greater employee contributions.From the employees’ perspective, were these con-tingencies to reflect fairly the added value gener-ated by greater employee contributions, therewould be no real reason to hide this fact. Accord-ingly, the malevolent signals that employees maydirectly infer from pay secrecy may further weakenthe perception that any increase in performancelevel will be accompanied by a larger incrementalreturn. And because the link between PFP percep-tions and individual task performance is well es-tablished in the motivation literature (e.g., Nayloret al., 1980; Vroom, 1964), and has been widelydemonstrated (Kanfer, 1990), this discussion sug-gests that the adverse effect of pay secrecy on em-ployees’ PFP perceptions is likely to have negativeimplications on task performance. Accordingly,we posit:


Hypothesis 1. PFP perceptions mediate the(adverse) impact of pay secrecy on individualtask performance.

Beyond these adverse incentive effects, pay se-crecy’s negative effect on PFP perceptions may alsoelicit detrimental sorting effects (Gerhart et al.,2009; Trevor, Reilly, & Gerhart, 2012), manifestingitself in diminished intentions to remain (i.e., con-tinuation intentions), particularly among high per-formers (Lazear, 1986; Milkovich et al., 2013). Re-search indicates that high performers find PFP to beappealing (Cadsby et al., 2007), tending to remainin firms perceived to reward high performancemore aggressively and to migrate away from thosefirms perceived to offer smaller performance-basedpay differentials (Leventhal et al., 1980; Trevor,Gerhart, & Boudreu, 1997). For example, Lazear(2000) observed that the adoption of piece-rate payat Safelite resulted in reduced turnover among itsmore productive workers and the tendency of itsless productive employees to self-select out of thefirm. Similarly, when organizations communicatedthe nature of performance-based pay contingencies,Shaw and Gupta (2007) found stronger PFP contin-gencies (manifested in higher pay dispersion) to bepositively associated with the retention of high per-formers. Consistent with these findings, researchon the effects of pay dispersion in top managementteams shows that when pay dispersion is high, it isexecutives with a lower relative pay positionwithin the top management team who are morelikely to leave (Messersmith, Guthrie, Ji, & Lee,2011). Moreover, the continuation intentions oflow performers appear to be less related to PFPperceptions than are those of high performers (Har-rison, Virick, & William, 1996). Taken together,these findings suggest that:

Hypothesis 2. Task performance moderates thesecond stage of the mediated relationship be-tween pay secrecy and continuation intentionssuch that the indirect effects of pay secrecy oncontinuation intentions via PFP perceptionsare stronger when task performance is high.

Moderating Effects of PFP Characteristics

The fact that the vast majority of organizationsdo not disclose pay information and that secrecy ismore the rule than the exception suggests, how-ever, that (all else being equal) the level of malev-olence that employees infer from this practice isunlikely to be deterministic. Rather, the degree to

which pay secrecy adversely affects employee cog-nition and behavior is likely to be contextuallycontingent, conditioned by inferences drawn fromother related practices. Indeed, research suggeststhat the nature of the PFP system—and in particu-lar two main PFP system attributes—is likely tomoderate the consequences of pay secrecy.

The first of these attributes has to do with howperformance is measured and, in particular, thesubjectivity of performance assessment. Highlight-ing the potential centrality of this issue, Futrell andJenkins (1978: 214) posited that “one critical vari-able affecting whether to have an open pay systemis the ability of management to measure task per-formance, [and that] organizations not capable ofobjectively measuring performance are likely tohave difficulty with the open system.” Similarly,Colella et al. (2007: 65) noted that “the nature of(performance) measurement helps to determinewhether costs or benefits will be experienced dur-ing pay secrecy.”

The second attribute has to do with how mea-sured performance is translated into an actual in-centive amount, or what may be referred to as paydetermination criteria (Lazear & Oyer, 2013; Mur-phy & Cleveland, 1995). As Gerhart et al. (2009)argue, while some well-known organizations (e.g.,General Electric) base their PFP on employee per-formance determined relative to that of the employ-ee’s peers, others concerned about the competition-related implications of an employee–employeecomparisons ground their PFP on more absolutecriteria under which rates are evaluated againstsome predetermined standard.

Signaling theory offers a useful, overarching the-oretical framework for understanding how variabil-ity in these two aspects of the broader pay contextmoderate the incentive and sorting effects of paysecrecy. As noted by Gomez-Mejia et al. (2010),signaling theory suggests that employers may usePFP system attributes to signal to employees theobjectives that they deem to be of high priority. Forexample, choices made by the employer with re-gard to pay determination criteria may be used tosignal to employees that situational constraints aretaken into account (with subpar performance stillrewarded as long as it exceeds some dynamicnorm). Similarly, choices made with regard to thesubjectivity of performance assessment may signalto employees an emphasis on recognizing and re-warding the employee’s full range of performance,even those not readily measured objectively.


Ultimately, however, the moderating effect ofboth of these PFP system attributes depends on theinferences that employees—the signal recipients—draw from them. And, as suggested earlier, theseemployee inferences may not necessarily be thoseintended by the employer. As such, inferencesdrawn from both PFP system attributes may haveunintended consequences with respect to individ-ual beliefs that a given reward is contingent uponassessed performance and/or the individual’s be-liefs regarding the degree to which measured per-formance captures actual performance. Based onsuch a notion, and as we detail next, we posit thatpay determination criteria (i.e., relative vs. abso-lute) are likely to interact with pay secrecy in shap-ing PFP perceptions, thus moderating the first stageof the mediation model developed above. Simi-larly, we posit that the subjectivity of performanceassessment is likely to influence inferences regard-ing the degree to which variance in actual perfor-mance is recognized, hence moderating the secondstage of the mediation model developed above (i.e.,the effects of PFP perceptions on performance andcontinuation intentions).

Moderating effect of pay determination crite-ria. Research differentiates between PFP systemsbased on relative vs. absolute performance criteria(Baron & Kreps, 1999; Lazear & Oyer, 2013). Signal-ing theory suggests that employees are likely todraw different inferences from these alternativepay determination criteria and that these inferencesare likely to interact with those drawn from thetransparency of the overall pay system to affect PFPperceptions.

Relative performance criteria signal to employeesthat incentive pay is based not only on their ownperformance, but also takes that of others into ac-count as well, with incentive magnitude being afunction of the individual’s assessed performancerelative to that of others. Employees may infer cer-tain benefits from such a signal—namely, that thePFP system has been designed to “filter out shockscommon to the whole peer group” (Lazear & Oyer,2013: 484). However, such an attribute may alsosend the signal that reward-eligible performancestandards are dynamic and that meeting them is amatter of luck (Lazear, 1995), depending not onlyon one’s own measured performance, but also onthat of some comparative set of peers (Murphy &Cleveland, 1995). To the degree that employeesoverweight this negative aspect of relative pay de-termination (as would be expected on the basis ofthe positive–negative attribution effect), relative

pay determination is likely to heighten employees’uncertainty regarding their status in the organiza-tion. And because such a condition is conducive tosinister attribution error (Kramer, 1998), it maystimulate or intensify employee concerns regardingthe motives underlying the pay system overall. Inthe context of such concerns, the (negative) infer-ences that employees draw from pay secrecy re-garding the nature of performance–pay relationsare likely to be more salient, thus intensifying anytendency of pay secrecy to be associated with aweakening of subjective expectations that shifts inmeasured performance will be associated withcommensurate shifts in pay (Lazear, 1995; Sloof &van Praag, 2008). Accordingly, we posit:

Hypothesis 3a. Pay determination criteriamoderate the first stage of the mediated rela-tionship between pay secrecy and task perfor-mance such that the indirect effects of paysecrecy on task performance via PFP percep-tions are stronger when pay determination cri-teria are relative.

Hypothesis 3b. Pay determination criteriamoderate the first stage of the mediated rela-tionship between pay secrecy and continuationintentions such that the indirect effects of paysecrecy on continuation intentions via PFPperceptions are stronger when pay determina-tion criteria are relative.

Moderating effect of the subjectivity of perfor-mance assessment. In the same way that PFP sys-tem attributes transmit signals about how assessedperformance will be rewarded (i.e., performance–reward contingencies), they also transmit signalsabout how performance will be assessed and, morespecifically, the ability of the system to recognizevariance in employee contribution accurately. Thisis important in that, beginning with the seminalwork of Vroom (1964), organizational psychologistsand economists have posited, and found, that em-ployee perceptions regarding performance–rewardcontingencies interact with perceptions regardingthe validity of performance assessments to affectemployee performance and continuation intentions(Kanfer, 1987; Lazear & Oyer, 2013). More specifi-cally, this body of research suggests that the impactof PFP perceptions on task performance and thecontinuation intentions of high-performing em-ployees are likely to be conditional upon inferencesdrawn from the PFP system regarding the degree towhich any increase in effort will be reflected by a


commensurate increase in assessed performance.Such effort–performance perceptions are affectedby the degree of imprecision, or “noise,” withwhich the performance measure reflects actual ef-fort. To the degree that employees associate the PFPsystem with noisier performance measures, “therelationship between effort and measured perfor-mance is weakened” (Sloof & van Praag, 2010: 752),and any positive association between employees’PFP perceptions and performance is likely to beattenuated (Kanfer, 1987). Similar attenuation ef-fects can be expected with regard to the impact ofPFP perceptions on the continuation intentions ofhigh performers. To the degree that the approachtaken to measure performance signals that “em-ployees’ (measured) output is imperfectly related toactual output (that is, measures of the worker’sproductivity),” even those higher-performing em-ployees with strong PFP perceptions are likely to bemotivated to seek alternative employment, signal-ing an enhanced ability to better recognize theirdifferential contribution (Lazear & Oyer, 2013:481).

Such noise problems may be avoided when ob-jective criteria providing a comprehensive pictureof employee contribution are available (Bamberger& Meshoulam, 2000). The performance appraisalliterature (DeNisi, 1996; Murphy & Cleveland,1995) suggests that when performance appraisal isbased on such objective (as opposed to subjective)performance criteria, pay is perceived as less dis-posed to effort–performance weakening rater biasessuch as leniency and centrality (Moers, 2005;Rynes, Gerhart, & Parks, 2005). However, whensuch objective data are either less readily available,or do not provide a real-time and comprehensiveindication of both output and quality, organiza-tions are likely to have little choice but to rely onnoisier, more subjective performance assessmentmethods (Gerhart et al., 2009). Because subjectiveperformance assessments are based on personaljudgments, they are difficult to predict ex ante andnon-verifiable ex post, thus signaling to employeesa greater degree of uncertainty regarding which ac-tions prompt which performance ratings, and aheightened likelihood that bias and favoritism in-fluenced appraised performance (Ittner, Larcker, &Meyer, 2003). Accordingly, we posit:

Hypothesis 4a. Subjectivity of performanceassessment moderates the second stage of themediated relationship between pay secrecyand task performance such that the indirect

effects of pay secrecy on task performance viaPFP perceptions are stronger when perfor-mance assessment is subjective.

Similarly, we posit that the subjectivity of per-formance appraisal attenuates the otherwise posi-tive association between PFP perceptions and con-tinuation intentions—particularly among higherperformers, who tend to be more sensitive to inac-curacies in performance assessment, such as thoseassociated with more subjective performance ap-praisal (Bol, 2011). Overall, more precise forms ofperformance measurement may signal a higherlevel of procedural fairness to such employees (Bol,2011). Research suggests that employees may recip-rocate such fair treatment with a heightened senseof loyalty to the employer (Jones & Skarlicki, 2003).Indeed, meta-analytic results indicate a correctedcorrelation of �.46 between procedural justice andturnover intention (Colquitt et al., 2001). Such aheightened sense of loyalty would serve only toreinforce the positive association between PFPperceptions and continuation intentions, partic-ularly among high performers who, by definition,have more to gain from stronger performance–pay contingencies. Moreover, PFP systems char-acterized by a combination of strong PFP percep-tions and more objective performance appraisalsignal to employees that even if, in the short run,owing to performance deficiencies, their rewardswere to be lower than hoped for, they may stillgain a favorable return on their effort in the longrun (Tyler, 1994). To the degree that (particularlyhigh-performing) employees infer from more pre-cise approaches to performance measurementmore favorable organizational treatment overall(Colquitt, Greenberg, & Zapata-Phelan, 2005;Greenberg, 2003), they may be more willing toaccept unfavorable PFP outcomes in the short run(Brockner & Wiesenfeld, 1996) and to delay anyconsideration of alternative employment oppor-tunities. In contrast, even high-performing em-ployees with strong PFP perceptions may not beso patient in the context of a PFP system charac-terized by the potential for imprecise perfor-mance measurement. This is because PFP sys-tems characterized by less precise performancemeasurement (such as more subjective assess-ment) can signal a heightened potential for unfairtreatment, leading particularly high-performingemployees to infer a lack of utility in waiting formore favorable pay-related outcomes in the longrun. Hence:


Hypothesis 4b. Subjectivity of performanceassessment and task performance jointly mod-erate the second stage of the mediated relation-ship between pay secrecy and continuation in-tentions such that the indirect effects of paysecrecy on continuation intentions via PFPperceptions are stronger when performance as-sessment is subjective and task performanceis high.

But what if employees perceive that the subjec-tivity of performance assessment is likely to workin their favor? The notion that positive (i.e., leni-ency) bias may actually amplify the link betweenPFP perceptions and task performance (Bol, 2011)suggests that the relationships posited in Hypothe-ses 4a and 4b may themselves be contingent on paysystem attributes’ signaling as to whether the sub-jectivity of performance appraisal is likely to workin or against the employee’s interest. Pay secrecymay play a key role in signaling to employees whatto expect from subjective performance assessment.More specifically, Principle 4 of signaling theorynoted earlier (see “Information Assymetry and Sig-naling Theory”) suggests that the adverse effects ofthe subjectivity of performance assessment on thepositive association between PFP perceptions andtask performance may themselves be amplified un-der conditions of pay secrecy (vs. transparency). Tothe extent that the intentional nature of informa-tional asymmetry created by pay secrecy may beinferred as signaling managerial opportunism ordeceptive intent, it may generate suspicion on thepart of employees as to the underlying purpose ofassessment subjectivity. For example, in the con-text of PFP, employees may interpret pay secrecy assignaling that subjective assessment is being used“to give management more freedom in distributingcompensation dollars to employees” (Gomez-Mejiaet al., 2010: 98). To the degree that employees drawsuch inferences from pay secrecy, perceptions ofdownward bias under conditions of subjective per-formance assessment may be accentuated, thus am-plifying the adverse effects of subjective assess-ment on the link between PFP perceptions and bothperformance and continuation intentions. Thuswe posit:

Hypothesis 5a. Pay secrecy and performanceassessment subjectivity jointly moderate thesecond stage of the mediated relationship be-tween pay secrecy and individual task perfor-mance such that the indirect effects of paysecrecy on individual task performance via

PFP perceptions are stronger when pay is se-cret and performance assessment is subjective.

Hypothesis 5b. Pay secrecy, performance as-sessment subjectivity, and individual task per-formance jointly moderate the second stage ofthe mediated relationship between pay secrecyand continuation intentions such that the in-direct effects of pay secrecy on continuationintentions via PFP perceptions are stronger forhigher performers when pay is secret and per-formance assessment is subjective.

METHOD

Participants

A total of 320 undergraduate students in an Is-raeli university participated in the experiment.Owing to a bug in the simulation software, no datawere recorded for five participants. An additional35 observations were excluded from our analysesbecause of either suspect or excessive missing data(e.g., 33% or more incomplete items) resulting fromthe same bug. This left us with a final sample of 280(138 and 142 in the transparent and secret condi-tions, 137 and 143 in the objective and subjectiveconditions, 141 and 139 in the absolute and relativeconditions, respectively). To assess any biasing ef-fect of participants excluded from the analysis be-cause of this bug, we followed the approach ofGoodman and Blum (1996), applying logistic re-gression to test a model in which the dependentvariable was a dichotomous variable (1 � Obser-vations used in the analyses, 0 � Excluded obser-vations) and the independent variables were all ofthe variables specified in our model. With all of thecoefficients statistically insignificant, the resultssuggest that all attrition was random and henceunlikely to bias our findings (Little & Rubin, 1987).

Design

Although performing their tasks autonomously,participants were randomly assigned to four-per-son groups, with each group randomly assigned toone of the conditions specified by the 2 (pay com-munication conditions: secret vs. transparent) � 2(pay determination criteria conditions: relative vs.absolute) � 2 (subjectivity of assessment condi-tions: objective vs. subjective) � 3 (performancephases) repeated measures, mixed-factorial design.The between-subject factors were pay communica-


tion policy, pay determination criteria, and subjec-tivity of assessment conditions.

Given that the majority of contemporary, privatesector employees are at least partially rewarded forindividual performance, with a substantial portionof PFP coming in the form of a bonus (Milkovich etal., 2013), to maximize the study’s external validityeach participant received NIS20 (US$5.70) basepay, as well as a bonus of up to NIS8 ($2.28) foreach task round.

Procedure

The experiment included three task rounds thateach lasted 5 minutes and three breaks betweenthese rounds, each lasting 10 minutes (see Appen-dix A). The task (i.e., a computer-based simulation)was adopted from that employed by Bamberger andBelogolovsky (2010). The object of the task was toplace “magic stones” in each of several squaresappearing in rows on the screen, such that eachsquare turned into “gold.” Stones were accepted forplacement if they matched adjacent stones alreadyon the screen in either color, or shape, or both.Participants scored points by turning as many rowsas possible into “gold” before the end of the perfor-mance phase. The experimental simulation wasdesigned to emulate a virtual work group, the mem-bers of which, working autonomously, may occa-sionally communicate with each other. Thereforewhile participants were unable to communicatewith each other during task rounds, they were ableto do so between task rounds using an intragroupe-mail system.

Upon arrival to the laboratory, participants wereasked to fill out a questionnaire including demo-graphic items. Next, the participants were informedthat they would participate in a study of perfor-mance in a computer task. The experimenter thenexplained the pay system noted above (see “De-sign”)—i.e., base pay of NIS20 ($5.70) and an op-portunity to earn an additional bonus—and the in-tragroup e-mail system, which participants coulduse during scheduled breaks between task roundsto communicate with each other. Participants werethen assigned to a computer and guided through anonline tutorial. The first of three task rounds fol-lowed. At the end of each task round, the computerscreen of each participant displayed either (a) hisor her score on the task for that round and a bargraph of the bonus pay to be received for that round(pay secrecy condition), or (b) his or her total scoreand bar graph of the additional pay to be received

for that task round and a graph of his or her payrelative to that of the others (by code number) inhis or her group (pay transparency condition).After the third task round, participants com-pleted the questionnaire accessing their studycontinuation intentions and the manipulationcheck questionnaire.

Measures

Exogenous variables. Pay communication pol-icy variables comprised two conditions: pay se-crecy vs. pay transparency. Consistent with Bam-berger and Belogolovsky (2010), participants in thepay secrecy condition received information ontheir own level of performance and bonus pay. Ad-ditionally, prior to the first task round, participantswere asked not to discuss any pay-related issuewith others at any time. All communication be-tween participants (even during breaks) was mon-itored, allowing us to confirm that participants inthis condition did not attempt to disclose pay orperformance information to other participants. Incontrast, participants in the pay transparency con-dition received information regarding (a) their ownperformance and bonus pay, and (b) the pay (butnot the performance levels) of their fellow groupmembers (listed by code numbers to ensure pri-vacy). Participants in this condition were told thate-mail communications with their fellow groupmembers would be unrestricted. In fact, 90% of theparticipants in this condition did disclose pay- andperformance-related information (typically, theirassigned study code number) to at least one othergroup member in at least one of the task rounds.

Pay determination criteria variables comprisedtwo conditions: absolute vs. relative. Participantsin the absolute condition were informed that bonuspay would be allocated on an absolute (i.e., set andpredetermined) basis according to their level ofperformance. In contrast, participants assigned tothe relative condition were informed that bonuspay would be allocated according to the points-based ranking of each participant within his orher group.

Subjectivity of performance assessment variablescomprised two conditions: objective vs. subjective.Participants in the objective condition were toldthat the magnitude of each rounds’ performancescore would be calculated objectively, entirely con-tingent upon the individual’s point count in thatphase. In contrast, given that, in many organiza-tions, subjective assessments are often used in com-


bination with objective assessments (Gibbs, Mer-chant, Van der Stede, & Vargus, 2004), participantsassigned to the subjective condition were told thatthe magnitude of each rounds’ performance scorewould be calculated according to the combinationof the individual’s objective point count in thatphase with the investigator’s subjective assess-ment, based on his or her observation of the indi-vidual’s performance. In reality, however, for thosein the subjective condition, in all rounds the per-formance score was calculated according to the fol-lowing formula:

66% � Objective score � Random number(range: 0.50 –1.50)

Endogenous variable. Pay-for-performance per-ceptions were assessed after the second round oftask performance using a modified version of ameasure developed by Erez and Isen (2002). Par-ticipants were asked to estimate the probabilitythat a given level of performance (i.e., score)would result in specific levels of bonus pay.The performance levels given ranged from thelowest performance level (–2SD, i.e., 100) to thehighest performance level (�2SD, i.e., 900) thatwe identified in pretest samples. The bonus paylevels ranged from no pay to NIS8 ($2.28). Anexpected bonus pay score was calculated for eachperformance level by multiplying the probabili-ties that participants assigned to each level ofbonus pay by the performance levels. These prod-ucts were then summed and divided by 100, suchthat each performance level (i.e., 100, 300, 500,700, 900) had an expected bonus pay level asso-ciated with it. Accordingly, each participant inthe data set had a five-point vector representingan expected level of bonus pay for each level ofperformance. The PFP perceptions were calcu-lated by correlating the vector of expected bonuspay levels with the specified performance levelson the basis of regression analysis.

Dependent variables. Individual task perfor-mance was assessed in terms of participants’ levelof performance (i.e., score) in the third and finaltask round (T3). Scores ranged between 90 and 1261(M � 730.42, SD � 186.52).

Continuation intention was assessed after thethird and final task round (T3) using a single-itemmeasure. Participants were asked to indicate, usinga seven-point Likert scale (1 � Not at all willingand 7 � Very willing), the degree to which theywould be willing to participate in additionalrounds of this experiment.

Manipulation Checks

The effectiveness of the pay communication pol-icy manipulation was assessed using a three-itemmeasure developed by Bamberger and Bel-ogolovsky (2010). Participants were asked to in-dicate, using a Likert scale (1 � Low and7 � High), the degree to which they received in-formation about (for example) other participants’level of pay. Cronbach’s alpha was .79.

The effectiveness of the pay determination crite-ria manipulation was measured by asking partici-pants to indicate the degree (1 � Completely dis-agree and 7 � Completely agree) to which theyagreed or disagreed with each of the followingstatements: (a) “My pay was influenced only by mylevel of performance (i.e., number of pointsearned),” (b) “The amount I was paid was deter-mined strictly on the basis of my level of perfor-mance,” (c) “The amount I was paid was influencedby how well others performed” (reversed item), and(d) “My pay was influenced by my performancerelative to my coworkers” (reversed item). Cron-bach’s alpha was .70.

The effectiveness of the subjectivity of perfor-mance assessment manipulation was measuredby asking participants to indicate the degree(1 � Completely disagree and 7 � Completelyagree) to which they agreed or disagreed with eachof the following statements: (a) “My pay was calcu-lated entirely on the basis of my objective perfor-mance of the assigned task” (reversed item), and (b)“In this experiment, pay is entirely determined byone’s objective performance” (reversed item). Cron-bach’s alpha was .70.

Analytical Procedure

Level of analysis. Given that all participantswere nested in groups of four, our analysis beganwith a test for the random effects of group assign-ment (Singer, 1998). The results indicated that therandom effect of group assignment was significantin all of the models tested. We therefore tested ourhypotheses using a multilevel moderated-media-tion model (Bauer, Preacher, & Gil, 2006). However,in our study, all variables were measured as fixedeffects at the individual level of analysis. Addition-ally, our model implies that the mediation effect isconsistent across the upper level of analysis (i.e.,groups), such that there is no variation in the me-diation effect across the groups. Nevertheless, sinceparticipants were nested within groups, we took


into account this nested structure by allowing theintercepts to vary across the groups. Accordingly,we used a modified version of the multilevel mod-erated-mediation SAS macro (Preacher & Hayes,2004), in which the indirect effect does not varyacross the upper level (i.e., the groups), yet thevariation in the intercepts is taken into account.

Data were analyzed using maximum likelihood(ML) and restricted maximum likelihood (REML)methods, which, while yielding similar results interms of fixed effects (regression parameters), pro-duce different estimates of variance components(Snijders & Bosker, 1999): REML takes into accountthe degrees of freedom from the fixed effects andtherefore creates variance components estimatesthat are less biased. While in small samples withbalanced data REML is preferable to ML for esti-mating variance components because it is unbi-ased, ML is preferable to REML for testing model fit(McCoach, 2010). Therefore we present the resultsof the regression analyses using REML and the re-sults of model fit tests using ML.

Procedure. We tested our hypotheses followingthe approach recommended by Edwards and Lam-bert (2007). In testing our hypotheses, we con-trolled for pay determination criteria and subjec-tivity of performance assessment in order tounderstand the impact of pay secrecy on individ-ual task performance and continuation intentionsabove and beyond these two PFP characteristics.Furthermore, in modeling the base-free measureof the change in task performance, we used therepressor variable method, taking participants’starting levels of performance (i.e., performanceat T1) into account because they had a true causaleffect on performance at T3 (r � .47, p � .01)(Allison, 1990).

The significance of the indirect, direct, and totaleffects was calculated by means of parametric boot-strap tests (Preacher & Hayes, 2004). The relativepredictive utility of each model was assessed onthe basis of the significance of the likelihood ratio-based R2 (R2

LR), estimated as:

1 � exp(�2n(logLM � logLo))

where:logLM is the maximum log-likelihood of the

model of interest,logLo is the maximum log-likelihood of the inter-

cept-only model, andn is the number of observations (Cox & Snell,

1989; Magee, 1990).

The calculation of the R2LR is based on ML, rather

than REML, because the REML log likelihood can-not be used to compare models with different fixed-effects specifications.

RESULTS

Manipulation Checks

The effectiveness of the experimental manipula-tions was confirmed by t-tests that were performedon the three manipulation check measures. First, at-test confirmed that participants in the transparentpay condition reported significantly higher per-ceived pay transparency perceptions (M � 5.11,SD � 1.43) than those assigned to the pay secrecycondition (M � 2.33, SD � 1.40; t(277) � 16.49,p � .01). Second, the participants in the absolutepay determination criteria condition rated absolutepay determination criteria higher (M � 3.43,SD � 1.52) than those in the relative condition(M � 3.01, SD � 1.58; t(277) � 2.23, p � .05).Finally, the participants assigned to the objectiveperformance assessment condition reported thesubjectivity of performance assessment to be signif-icantly lower (M � 4.16, SD � 1.46) than thoseassigned to the subjective condition (M � 4.46,SD � 1.39; t(277) � 1.70, p � .05).

These results indicate that all three manipula-tions were effective.

Descriptive Statistics

Table 1 shows the means, standard deviations,and intercorrelations among the variables. Thefindings indicate significant negative correlationsbetween pay communication policy (pay se-crecy � 1) and performance at T3 (r � –.13,p � .05), and between pay communication policyand continuation intentions at T3 (r � –.13,p � .05). In addition, the results showed a signif-icant positive correlation between performance atT3 and continuation intentions at T3 (r � .16,p � .01). Moreover, the results indicate a signifi-cant positive correlation between performance atT1 and performance at T3 (r � .47, p � .01). Sep-arate calculations show that there is no significantdifference in mean performance at T1 betweenthose in the transparent (M � 537.1, SD � 206.9)versus secret (M � 560.5, SD � 181.2; t(278) �–1.01, n.s.) conditions. Nevertheless, mean perfor-mance in the transparency condition rose to 756.93(SD � 188.78) at T3, while the mean performance


in the pay secrecy condition rose to only 702.03(SD � 175.96) at T3. In other words, mean perfor-mance in the transparency condition increasedfrom T1 to T3 by 28.93% versus 20.16% in paysecret condition. Notable also is a significant posi-tive correlation between PFP perceptions at T2 andperformance at T3 (r � .12, p � .05).

Hypotheses Testing

The results of an analysis of the baseline controlmodels (Model 1 in Tables 2 and 3) showed thatPFP characteristics were not significantly associ-ated with either T3 task performance or continua-tion intentions. They also indicate that task perfor-mance at T1 (estimate � .46, p � .01) explained23% of the variance in the task performance at T3.Following the recommendation of Becker (2005),we tested all of the models both with and withoutcontrolling for T1 performance. The analyses with-out controlling for T1 performance yielded similareffects with similar relative magnitude of the esti-mates. However, because “failing to include rele-vant control variables could inflate the amount ofexplainable variance in the dependent variable,thereby increasing the chances of a Type I error”(Becker, 2005: 288)—that is, concluding that thereis an effect when, in fact, there is not—we presentthe results of the analyses when controlling for T1

performance.Incentive and sorting effects of pay secrecy:

The mediating effect of PFP perceptions. The re-gression results for the tests of mediation indicatethat, after controlling for performance at T1, paydetermination criteria, and subjectivity of perfor-mance assessment, pay secrecy is inversely relatedto performance at T3 (estimate � �60.14,p � .01) (see Model 2 in Table 2). They also indi-

cate that, after controlling for pay determinationcriteria and subjectivity of assessment, pay secrecyis inversely related to continuation intentions (es-timate � �.38, p � .05) (Model 2 in Table 3).However, the results of Model 3a in Tables 2 and 3show that pay secrecy is inversely, but not signifi-cantly (estimate � �.02, n.s.), related to PFP per-ceptions at T2, thus suggesting that, in contrast toHypotheses 1 and 2, PFP perceptions fail to medi-ate the effects of pay secrecy on task performance.This is not surprising given that, as suggested byHypotheses 3–5, such a mediated effect may becontingent upon the nature of pay determinationcriteria and performance assessment subjectivity.

Moderated mediation analysis. The results oftests of the moderated mediation models specifiedin Hypotheses 3–5 are presented in Model 4 ofTables 2 and 3.

Moderating effect of PFP system attributes onthe performance consequences of pay secrecy.Model 4a in Table 2 shows a negative and signifi-cant interaction of pay communication policy (i.e.,pay secrecy) and pay determination criteria (esti-mate � �.21, p � .05). As recommended by Ai-ken and West (1991), we used interaction plots andsimple slopes analyses (i.e., comparing each of thesimple slopes to 0) to assess the degree to whichthese effects were consistent with those hypothe-sized. The slope of pay secrecy on PFP perceptionsis negative and significantly different from 0 whenpay determination criterion is relative (esti-mate � �.12, p � .05), but positive and not sig-nificantly different from 0 when pay determinationcriterion is absolute (see Figure 2). Accordingly,our data suggest support for Hypotheses 3a and3b, demonstrating that the impact of pay secrecy(relative to pay transparency) on PFP perceptionsis contingent upon the nature of pay determi-

TABLE 1Means, Standard Deviations, and Intercorrelations of the Variablesa

Variable M SD (1) (2) (3) (4) (5) (6)

1. Pay communication policy (Transparency � 0, Secrecy � 1) .49 .502. Performance at T1 549.22 197.56 .073. Performance at T3 730.42 186.52 �.13** .47*

4. Continuation intentions at T3 5.74 1.45 �.13** .06 .16*

5. PFP perceptions at T2 .91 .22 �.02 .002 .12** .016. Pay determination criteria (Absolute � 0, Relative � 1) .53 .50 .008 .08 �.05 �.01 �.017. Subjectivity of performance (Objective � 0, Subjective � 1) .50 .50 �.01 .11 .05 �.03 �.09 �.03

a n � 280.* p � .01

** p � .05


TA

BL

E2

Eff

ects

ofP

ayC

omm

un

icat

ion

Pol

icy

onIn

div

idu

alT

ask

Per

form

ance

:R

esu

lts

ofM

oder

ated

-Med

iati

onA

nal

ysis

Dep

end

ent

Var

iabl

e

Con

trol

Mod

elT

otal

effe

ctm

odel

Med

iati

onm

odel

Mod

erat

ed-m

edia

tion

mod

el

Mod

el1

Mod

el2

Mod

el3a

Mod

el3b

Mod

el4a

Mod

el4b

Mod

el4c

Tas

kp

erfo

rman

ceat

T3

Tas

kp

erfo

rman

ceat

T3

PF

Pp

erce

pti

ons

atT

2

Tas

kp

erfo

rman

ceat

T3

PF

Pp

erce

pti

ons

atT

2

Tas

kp

erfo

rman

ceat

T3

Tas

kp

erfo

rman

ceat

T3

Ind

epen

den

tV

aria

ble

Est

.S

EE

st.

SE

Est

.S

EE

st.

SE

Est

.S

EE

st.

SE

Est

.S

E

Inte

rcep

t49

5.94

*32

.42

522.

60*

32.9

5.9

3*.0

443

2.38

*52

.81

.88*

.05

299.

36*

76.6

327

8.05

*78

.20

Per

form

ance

atT

1.4

6*.0

5.4

7*.0

5—

—.4

7*.0

5—

—.4

6*.0

5.4

7*.0

5P

ayd

eter

min

atio

ncr

iter

ia(A

bsol

ute

�0,

Rel

ativ

e�

1)�

33.3

322

.61

�33

.44

21.5

8�

.000

4.0

4�

32.8

621

.44

.10

.06

�23

.24

30.6

8�

22.6

431

.10

Su

bjec

tivi

tyof

asse

ssm

ent

(Obj

ecti

ve�

0,S

ubj

ecti

ve�

1)�

3.91

22.6

6�

4.03

21.6

4�

.05

.04

.04

21.5

7�

.05

.04

205.

95**

92.3

938

8.86

*14

8.98

Pay

com

mu

nic

atio

np

olic

y(T

ran

spar

ency

�0,

Sec

recy

�1)

�60

.14*

21.5

5�

.02

.04

�59

.05*

21.4

1.0

9.0

6�

48.8

331

.51

344.

8638

3.74

PF

Pp

erce

pti

ons

atT

298

.47*

*45

.23

245.

96*

80.6

825

5.61

*82

.73

Pay

pol

icy

�R

elat

ive

pay

det

erm

inat

ion

crit

eria

�.2

1**

.08

�33

.48

43.6

7�

33.3

644

.23

PF

Pp

erce

pti

ons

atT

2�

Su

bjec

tivi

tyof

asse

ssm

ent

�22

6.49

**98

.52

�41

0.90

*15

9.47

PF

Pp

erce

pti

ons

atT

2�

Pay

pol

icy

�39

9.99

402.

13P

ayp

olic

y�

Su

bjec

tivi

tyof

asse

ssm

ent

�59

4.53

408.

81P

FP

per

cep

tion

sat

T2

�S

ubj

ecti

vity

ofas

sess

men

t�

Pay

pol

icy

604.

8743

0.19

�2

log

like

lih

ood

3646

.136

38.3

�83

.236

33.4

�89

.636

27.4

3624

.20

�–2

log

like

lih

ood

—7.

8*—

4.9*

6.4*

*10

.9*

9.20

(rel

ativ

eto

Mod

el1)

(rel

ativ

eto

Mod

el2)

(rel

ativ

eto

Mod

el3a

)(r

elat

ive

toM

odel

3b)

(rel

ativ

eto

Mod

el3b

)L

ikel

ihoo

dra

tio-

base

dR

2.2

3.2

5.0

06.2

7.0

3.2

9.2

9�

Lik

elih

ood

rati

o-ba

sed

R2

—.0

2*—

.02*

*.0

2*.0

2*.0

2(r

elat

ive

toM

odel

1)(r

elat

ive

toM

odel

2)(r

elat

ive

toM

odel

3a)

(rel

ativ

eto

Mod

el3b

)(r

elat

ive

toM

odel

3b)

*p

�.0

1**

p�

.05

TA

BL

E3

Eff

ects

ofP

ayC

omm

un

icat

ion

Pol

icy

onC

onti

nu

atio

nIn

ten

tion

:R

esu

lts

ofM

oder

ated

Med

iati

onA

nal

ysis

Con

trol

Mod

elT

otal

Eff

ect

Mod

elM

edia

tion

Mod

elM

oder

ated

-med

iati

onM

odel

Mod

el1

Mod

el2

Mod

el3a

Mod

el3b

Mod

el3c

Mod

el4a

Mod

el4b

Mod

el4c

Dep

end

ent

Var

iabl

eIn

dep

end

ent

Var

iabl

eC

onti

nu

atio

nIn

ten

tion

atT

3

Con

tin

uat

ion

Inte

nti

onat

T3

PF

PP

erce

pti

ons

atT

2

Con

tin

uat

ion

Inte

nti

onat

T3

Con

tin

uat

ion

Inte

nti

onat

T3

PF

PP

erce

pti

ons

atT

2

Con

tin

uat

ion

Inte

nti

onat

T3

Con

tin

uat

ion

Inte

nti

onat

T3

Est

.S

EE

st.

SE

Est

.S

EE

st.

SE

Est

.S

EE

st.

SE

Est

.S

EE

st.

SE

Inte

rcep

t5.

79*

.16

5.98

*.1

8.9

3*.0

45.

94*

.41

5.98

*.4

7.8

8*.0

54.

04*

1.02

4.74

*1.

07P

ayd

eter

min

atio

ncr

iter

ia(A

bsol

ute

�0,

Rel

ativ

e�

1)�

.03

.18

�.0

3.1

7�

.00

.04

�.0

3.1

7�

.005

.18

.10

.06

�.1

7.2

5�

.19

.26

Su

bjec

tivi

tyof

asse

ssm

ent

(Obj

ecti

ve�

0,S

ubj

ecti

ve�

1)�

.09

.18

�.0

9.1

7�

.05

.04

�.0

9.1

8�

.11

.18

�.0

5.0

42.

30a

1.14

1.95

1.74

Pay

pol

icy

(Tra

nsp

aren

cy�

0,S

ecre

cy�

1)�

.38*

*.1

7�

.02

.04

�.3

8**

.17

�.3

2.1

8.0

9.0

6�

.50

.26

�8.

14**

3.55

PF

Pp

erce

pti

ons

atT

2.0

5.4

0�

.03

.45

2.13

**1.

101.

491.

16P

ayp

olic

y�

Pay

det

erm

inat

ion

crit

eria

�.2

1**

.08

�.2

8.3

6.2

9.3

7P

erfo

rman

ceat

T3

.00

.002

�.0

03.0

02�

.002

.002

PF

Pp

erce

pti

ons

atT

2�

Su

bjec

tivi

tyof

asse

ssm

ent

�2.

58**

1.21

�2.

391.

86S

ubj

ecti

vity

ofas

sess

men

t�

Per

form

ance

atT

3.0

09**

.004

.001

.008

PF

Pp

erce

pti

ons

atT

2�

Per

form

ance

atT

3�

.000

.002

.005

**.0

02.0

04.0

03P

FP

per

cep

tion

sat

T2

�S

ubj

ecti

vity

ofas

sess

men

t�

Per

form

ance

atT

3

�.0

1**

.004

.000

.009

PF

Pp

erce

pti

ons

atT

2�

Pay

pol

icy

7.83

**3.

73P

ayp

olic

y�

Su

bjec

tivi

tyof

per

form

ance

7.66

**3.

84P

FP

per

cep

tion

sat

T2

�P

ayp

olic

y�

Su

bjec

tivi

tyof

asse

ssm

ent

�7.

654.

05P

ayp

olic

y�

Per

form

ance

atT

3�

.008

.03

PF

Pp

erce

pti

ons

atT

2�

Pay

pol

icy

�P

erfo

rman

ceat

T3

.008

.03

Pay

pol

icy

�P

erfo

rman

ceat

T3

�S

ubj

ecti

vity

ofas

sess

men

t.0

2.0

3P

FP

per

cep

tion

sat

T2

�P

ayp

olic

y�

Su

bjec

tivi

tyof

asse

ssm

ent

�P

erfo

rman

ceat

T3

�.0

2.0

3

�2

log

like

lih

ood

1002

.17

997.

41�

83.2

997.

3999

0.78

�89

.698

3.73

976.

10�

–2lo

gli

keli

hoo

d—

4.76

**—

.02

6.63

6.4*

*13

.66*

*14

.68

(rel

ativ

eto

Mod

el1)

(rel

ativ

eto

Mod

el2)

(rel

ativ

eto

Mod

el2)

(rel

ativ

eto

Mod

el3a

)(r

elat

ive

toM

odel

3b)

(rel

ativ

eto

Mod

el3b

)L

ikel

ihoo

dra

tio-

base

dR

2.0

01.0

2.0

06.0

2.0

4.0

3.0

6.0

9�

Lik

elih

ood

rati

o-ba

sed

R2

—.0

19**

—0

.02

.02*

.04*

*.0

5(r

elat

ive

toM

odel

1)(r

elat

ive

toM

odel

2)(r

elat

ive

toM

odel

2)(r

elat

ive

toM

odel

3a)

(rel

ativ

eto

Mod

el3b

)(r

elat

ive

toM

odel

3c)

*p

�.0

1**

p�

.05

nation criteria and, more specifically, more adversewhen these criteria are relative (as opposed toabsolute).

As can be seen from Model 4b, the interactionterm of the PFP perceptions with subjectivity ofperformance assessment is significant and in-versely associated with the dependent variable(i.e., task performance at T3) (estimate � �226.49,p � .05). Simple slopes analyses (Aiken & West,1991; see Figure 3) show that, when performanceassessment is subjective, the slope of PFP percep-tions on task performance, while positive, is notsignificantly different from 0 (estimate � 19.46,n.s.). In contrast, when performance assessment is

objective, the positive slope is significantly differ-ent from 0 (estimate � 245.96, p � .01). Theseresults are consistent with Hypothesis 4a, positingthat the subjectivity of performance assessmentmoderates the second stage of mediation, with thepositive impact of PFP perceptions on performancebeing amplified when performance assessment isobjective (as opposed to subjective).

Results of parametric bootstrap analysis—that is,drawing 1,000 random samples with replacementfrom the full sample (Bauer et al., 2006)—indicatethat, under conditions of relative pay determina-tion criteria and objective assessment, the expectedindirect effect of pay secrecy on individual task

FIGURE 2Moderation of the Effect of Pay Communication Policy on PFP Perceptions (T2)

0.85 0.86 0.87 0.88 0.89 0.90 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Pay Determination Criteria

Pay Communication Policy

Transparent Secret

Absolute Relative

−.12, p < .05

.09, ns

Pay

-for

-Per

form

ance

Per

cep

tion

s

FIGURE 3Moderation of the Effect of (Centered) PFP Perceptions at T2 on Individual Task Performance at T3

720

730

740

750

760

770

780

790

800

810

820

830

0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15

Subjectivity of Performance Assessment

Objective Subjective

Pay-for-Performance Perceptions (centered)

In

div

idu

al T

ask

Per

form

ance

245.96, p < .01

19.46, ns


performance is �30.02 (90% CI � �54.45, �8.88)and the expected total effect is �112.61 (90%CI � �158.34, �67.29). With mediation presentwhen the indirect effect differs significantly from 0(Edwards & Lambert, 2007), our results, indicatingthat the 90% confidence interval (CI) of this indi-rect effect excluded 0, are consistent with Hypoth-esis 4a. Moreover, this moderated indirect effect(via PFP perceptions) accounts for 26.66% of thetotal effect of pay secrecy on task performance. Italso explains 7.41% more of the variance in taskperformance than the unmoderated mediationmodel (�R2

relative to Model 3b � .02, p � .01).1

However, we found no support for Hypothe-sis 5a, which posited that this attenuation of anypositive impact of PFP perceptions on task perfor-mance by subjective performance assessmentwould itself be amplified by pay secrecy. More

specifically, as can be seen in Model 4c in Table 2,the three-way interaction of the PFP perceptionswith the subjectivity of performance assessmentand the pay communication policy is not statisti-cally significant.

Moderating effects of PFP system attributes onthe sorting consequences of pay secrecy. The re-sults of Model 4b in Table 3 indicate that, consis-tent with Hypothesis 4b, the three-way interactionof PFP perceptions, subjectivity of assessment, andtask performance is significantly associated withcontinuation intentions at T3 (estimate � �.01,p � .05). Moreover, as illustrated in Figure 4, sim-ple slopes analyses (Aiken & West, 1991) showedthat, among high performers (i.e., �1SD of perfor-mance), when performance assessment is objective,the slope of PFP perceptions on continuation inten-tions is positive and significantly different from 0(estimate � 3.09, p � .05). In contrast, for averageperformers, these same objective performance as-sessment conditions yield a flatter (yet still signif-icantly different from 0) positive slope (esti-mate � 2.13, p � .05). Figure 4 also shows that,consistent with Hypothesis 4b, this amplificationof the impact of PFP perceptions on continuationintentions is absent under conditions of subjectiveperformance assessment. More specifically, underconditions of subjective performance assessment,

1 Although the effect sizes reported in Table 2 appearto be rather small, it must be emphasized that effect sizesestimated in the context of hierarchical models reflect theimpact of explanatory variables on multiple, level-specificcomponents of variance (with some variables reducing thevariance at one level, but potentially increasing it at an-other). Accordingly, it is likely that the absolute size of theeffects estimated in the current study are systematicallysmaller than those typically obtained in single-level linearregression (Recchia, 2010; Snijders & Bosker, 1999).

FIGURE 4Moderation of the Effect of (Centered) PFP Perceptions at T2 on Continuation Intentions at T3

5.45.55.65.75.85.96.06.16.26.36.46.56.66.76.86.9

0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15

Pay-for-Performance Perceptions (centered)

Objective performance assessment and high (centered) performance level (+1SD) = 3.09, p < .05Objective performance assessment and mean (centered) performance level (Mean) = 2.13, p < .05

Objective performance assessment and low (centered) performance level (−1SD) = 1.18, nsSubjective performance assessment and high (centered) performance level (+1SD) = −1.32, nsSubjective performance assessment and mean (centered) performance level (Mean)= −.45, nsSubjective performance assessment and low (centered) performance level (−1SD) = .43, ns

Con

tin

uat

ion

In

ten

tion

s


the PFP perceptions–continuation intentions slopeis not significantly different from 0 for high, aver-age, and low performers. In sum, as posited, thesubjectivity of performance assessment and indi-vidual task performance jointly moderate the sec-ond stage of the mediated relationship, with thepositive impact of PFP perceptions on continuationintentions being amplified when performance as-sessment is objective and this amplification beingparticularly pronounced among high performers.

The results of parametric bootstrap analysis(Bauer et al., 2006) provide further support for Hy-pothesis 4b. More specifically, among average per-formers, when the PFP system is characterized bythe relative pay determination criterion and objec-tive assessment, the expected indirect effect of paysecrecy on continuation intentions is �.26 (90%CI � �.52, �.04) and the expected total effect is�.47 (90% CI � �.89, �.07). In other words, un-der such PFP conditions, the indirect effect of paysecrecy via PFP perceptions accounts for 55.32% ofpay secrecy’s total effect on continuation inten-tions. In contrast, this same moderated indirect ef-fect is, as predicted, more robust among high per-formers. For them, the expected indirect effect is�.38 (90% CI � �.74, �.07) and the expected to-tal effect �.59 (90% CI � �1.08, �.13), with thisindirect effect explaining 64.41% of the total effect.This moderated-mediation model (Model 4b in Ta-ble 3) also explains a significantly greater share of thevariance in continuation intentions than the unmod-erated mediation model (�R²relative to Model 3b � .05,p � .05).

Finally, the results of Model 4c in Table 3 showthat the four-way interaction term of PFP percep-tions, pay secrecy, subjectivity of performance as-sessment, and task performance is inversely (esti-mate � �.02, n.s.), but not significantly associatedwith continuation intentions. Accordingly, Hy-pothesis 5b—positing that the attenuation of thebeneficial impact of PFP perceptions on continua-tion intentions by subjective performance assess-ment would itself be amplified by pay secrecy—was not supported.

DISCUSSION

Pay secrecy continues to be a contentious issueamong scholars and practitioners alike, with boththeory and empirical evidence regarding the impli-cations of pay secrecy remaining equivocal. Draw-ing from signaling theory, we aimed to enhanceunderstanding of the psychological mechanisms

underlying pay secrecy’s performance and sortingeffects, and to explore how aspects of the broaderpay context within which pay communication pol-icies are embedded may condition these effects.Based on the notion that pay secrecy generatesinformation asymmetries that heighten the uncer-tainty surrounding employees’ effort allocation de-cisions, we argued that employees seek to gleanwhatever information they can from the broaderpay context. More specifically, we proposed thatemployees may infer signals from pay secrecy it-self, as well as from two attributes of the PFP sys-tem—a dominant aspect of the broader pay contextin many organizations enforcing pay secrecy—inorder to reduce such uncertainty. Accordingly, wedeveloped a moderated-mediation model specify-ing that these two PFP system attributes—namely,pay determination criteria and the subjectivity ofperformance assessment—condition the PFP-re-lated inferences that employees draw from pay se-crecy, as well as the impact of these inferences onboth performance and continuation intentions.

Our findings largely support this moderated-me-diation model. First, consistent with earlier find-ings (Bamberger & Belogolovsky, 2010; Futrell &Jenkins, 1978), our results indicate that pay secrecyhas an adverse effect on individual task perfor-mance. Second, although we found no support forHypotheses 1 and 2, suggesting that PFP percep-tions unconditionally mediate the performance andsorting (respectively) effects of pay secrecy, we didfind support for the moderated-mediation effectsspecified by Hypotheses 3, 4a, and 4b. More spe-cifically, we found that pay determination criteriamoderate the effects of pay secrecy on PFP percep-tions (consistent with Hypothesis 3). As posited,while PFP perceptions are relatively insensitive topay transparency or secrecy when pay determina-tion criteria are absolute, they are highly sensitivewhen these criteria are relative. Moreover, while, asexpected, PFP perceptions were the lowest underconditions of pay secrecy and relative pay determi-nation criteria, they were highest under conditionsof pay transparency and relative (vs. absolute) cri-teria. This is interesting in that while pay–perfor-mance contingencies unbounded by some absolutestandard may signal potentially higher rewards forsuperior performance, the uncertainties associatedwith relative criteria should theoretically heightenrisk, thus potentially limiting PFP perceptions. Ourresults suggest that, in the context of relative paydetermination criteria, transparency may actually


signal reduced risk, thus facilitating higher PFPperceptions.

Additionally, we found that the positive effectsof PFP perceptions on task performance are ampli-fied under conditions of objective (vs. subjective)performance assessment (consistent with Hypothe-sis 4a). Accordingly, our findings suggest that, re-gardless of the interactive effect of pay transpar-ency and pay determination criteria on PFPperceptions, any positive impact that heightenedPFP perceptions may have on individual task per-formance is likely to be limited to the extent thatperformance assessment methods signal to employ-ees (as they do when assessment is less objective)that “noise” factors may limit the organization’sability to accurately differentiate strong from weakperformers.

Similar support was found for Hypothesis 4b re-garding the sorting effects of pay secrecy. As pos-ited, we found pay secrecy’s indirect and inverseeffect (through PFP perceptions) on continuationintentions to be attenuated under conditions ofobjective (vs.subjective) performance assessment,particularly among higher-performing study partic-ipants. These results suggest that high performers(relative to average and low performers), when con-sidering whether to stay or leave, tend to be themore sensitive to signals stemming from the natureof pay communication and the broader attributes ofthe PFP system. In sum, these findings suggest that,when implemented under conditions of relativepay determination criteria and subjective (vs. ob-jective) assessment, the adverse impact of pay se-crecy (vs. pay transparency) is unlikely to be lim-ited only to task performance. Pay secrecy, whenimplemented under such conditions, may also takea toll on the ability of the firm to retain its bestperformers.

Interestingly, we found no evidence that pay se-crecy amplifies the adverse impact of performanceassessment subjectivity on the association betweenPFP perceptions and individual task performance(Hypothesis 5a), and between PFP perceptions andcontinuation intentions among high performers(Hypothesis 5b). One explanation for this may bethat the size of our sample did not provide thestatistical power necessary to detect the indirectrelationships embedded within the three- and four-way interactions that we posited (Murphy & Myors,2004; Preacher & Kelley, 2011). Alternatively, itmay be that while subjectivity in assessment cansometimes result in positive (i.e., leniency) bias,consistent with the notion of Baumeister et al.

(2001: 323) that “bad impressions are . . . moreresistant to disconfirmation than good ones,” evenin the absence of an intent signal such as pay se-crecy individuals implicitly associate assessmentsubjectivity with the potential for negative (i.e.,harshness) bias. Accordingly, any marginal in-crease in signal strength generated by pay secrecy isneither psychologically meaningful nor statisti-cally significant.

Theoretical Implications

The current study offers several important con-tributions to theory and research on both signalingand pay communication policy.

Signaling. Our study offers three main implica-tions for signaling theory. First, while signalingtheory and research tend to focus on the determin-istic effects of individual signals, we posited, andfound, that the impact of pay secrecy as a discretesignal may be more dynamic—contingent upon theimplicit signals suggested by relevant aspects of thesurrounding context. The fact that pay secrecy’seffect on participants’ PFP perceptions was signif-icant only under conditions of relative (but notabsolute) pay determination criteria suggests thatinferences drawn from related policies or practiceslikely frame the interpretation of the target signaland influence its salience. Accordingly, our find-ings offer a more nuanced, context-sensitive (Johns,2006) approach to understanding signaling in man-agement, taking into account the potentially inter-active effects of multiple signals inferred from avariety of related policies, practices, and artifacts.They also reinforce the argument of Bowen andOstroff (2004) regarding the beneficial effect ofadopting human resources policies and practicesthat transmit unequivocal, consistent, and consen-sual signals to employees. As Bowen and Ostroff(2004: 207) note, “the more [human resources man-agement] practices send strong signals about whatstrategic goals are most important and what em-ployee behaviors are expected, supported and re-warded relative to those goals, the more likely it isthose goals will be achieved.”

Second, our findings extend signaling theory bysuggesting that, consistent with the literature onpositive–negative asymmetry (Baumeister et al.,2001; Peeters & Czapinski, 1990), even weak signalsmay have a tendency to drive negative impressionsand suspicions of opportunism and deceptionwhen interpreted in the context of heightened un-certainty. Moreover, they suggest that these nega-


tive impressions and suspicions can elicit precau-tionary or loss-reduction behaviors, such as effortreduction or withdrawal. They do so in that whilepay secrecy in isolation did not generate reducedPFP perceptions (and, through them, loss-reduc-tion behaviors), it did so when framed within thecontext of other pay-related policies and practicessignaling heightened uncertainty. These findingsare important because they raise questions aboutthe efficacy of management policies, practices, andsymbols intended to signal or “give sense” to em-ployees (Gioia & Chittipeddi, 1991). Particularlywhen applied in the context of other policies andpractices heightening uncertainty or in other wayspotentially raising employee concerns or suspi-cions, the results of this study suggest that the“sense” employees infer from such policies andpractices may be more negative than intended.

Third, we extend signaling theory to the realm ofpay administration, shifting away from the focus ofSpence (1973; 2002) on employees as signaling dif-ferences in ability, toward employers and how theirpay systems may implicitly signal differences intheir qualities as employers. Applied in such away, we use the principles of signaling to enhanceour understanding of how pay communication pol-icy may influence key compensation outcomes—namely, performance, and continuation intentions.Our finding that the effect of pay communicationpolicy on PFP perceptions is conditioned by thenature of pay determination criteria suggests thatthe inferences employees glean from one pay pol-icy are contingent upon signals inferred from oth-ers. Our finding—that the indirect effects of paysecrecy (via PFP perceptions) on incentive andsorting effects are contingent on the broader attri-butes of the PFP system—provides further evidencethat signals drawn from multiple elements of thepay system interact to affect key compensation out-comes. This has several key theoretical implica-tions. First, it suggests that the performance andsorting effects of pay secrecy may be different de-pending upon the nature of the PFP context withinwhich it is implemented. Second, while in the cur-rent study we limited our analysis to the contin-gency effects of two attributes of the PFP systems, itsuggests that other pay system attributes (e.g., paymix, dispersion) may also affect the inferences thatemployees draw from pay secrecy and thus moder-ate the impact of pay communication policy onthese same pay outcomes. Finally, it suggests that,in modeling the impact of pay secrecy, scholars

may need to adopt the more nuanced and context-sensitive frameworks noted here.

Pay communication policy. Our findings alsoprovide important insights into the manner inwhich pay communication policy may affect keycompensation outcomes. Although scholars pro-posed as early as 30 years ago that weakened PFPperceptions likely underlie any adverse impact ofpay secrecy on performance (Futrell & Jenkins,1978), the empirical evidence in support of thisexplanation has been limited at best. Indeed, al-though Bamberger and Belogolovsky (2010) foundsupport for this explanation, they did so onlyamong those highly sensitive to inequity. The cur-rent research extends these earlier findings in sev-eral important ways. First, our findings demon-strate that PFP perceptions play a key role inlinking pay transparency to individual task perfor-mance, as well as continuation intentions, irrespec-tive of individual differences. We demonstrate that,in the context of PFP, the adverse effects of paysecrecy are likely to be manifested in, as well asmediated by, reduced PFP perceptions. In particu-lar, the findings indicate that such effects are par-ticularly pronounced when pay secrecy is com-bined with PFP systems characterized by relative(as opposed to absolute) pay determination criteria.The findings also indicate that the indirect (via PFPperceptions) and adverse effects of pay secrecy arefurther amplified only under conditions of subjec-tive (relative to objective) assessment. Both of thesemoderation effects are important in that in the vastmajority of firms applying pay secrecy, relative paydetermination and subjective performance assess-ment are also typically in place (Gibbs et al., 2004;Kuhn & Yockey, 2003).

Additionally, we extend past research on the per-formance-related consequences of pay secrecy tothe issue of sorting, theorizing, and demonstratinghow—particularly in the context of varying PFPsystem attributes—pay secrecy may also affect con-tinuation intentions. More specifically, our find-ings indicate that, particularly when pay secrecy isapplied in the context of a PFP system, character-ized by relative pay determination criteria and sub-jective performance assessment (in contrast to asystem characterized by absolute pay determina-tion and more objective assessment), such a policymay have an adverse impact on the continuationintentions of better-performing staff. This is impor-tant not only because these are precisely the con-ditions under which pay secrecy is often adopted(Gibbs et al., 2004), but also because agency theory


suggests that PFP generally signals managerial in-tentions to better reward higher-performing em-ployees, and thus serves as a useful means bywhich to increase the likelihood of their attractionand retention (Gomez-Mejia et al., 2010). Our find-ings are also important in that, based on the prin-ciples of signaling theory, they shed light on how,under conditions of information asymmetry, thebeneficial signals often attributed to PFP may “getcrossed,” with the net result being that rather thanstrengthening the continuation intentions of better-performing employees as suggested by agency the-ory, PFP may actually weaken them.

Practical Implications

The results of the current investigation also offera number of significant practical implications.First, they suggest that while pay secrecy may in-deed provide managers with greater flexibility inallocating rewards (Colella et al., 2007), when com-bined with a PFP system in which pay determina-tion criteria are more relative (than absolute) andperformance assessment that is largely subjective(as opposed to objective), such a pay communica-tion policy may ultimately do more harm to indi-vidual task performance and continuation inten-tions than good. In this regard, managers may eitherconsider relaxing pay communication policies orchanging the nature of their organization’s PFP sys-tem. Indeed, although we contrasted a policy ofcomplete pay secrecy with complete transparency,there are more moderate options, often referred toas “partial openness” (Lawler & Jenkins, 1992). Forexample, organizations might provide employeeswith information on pay ranges and/or the sched-ule detailing how, at the aggregate level, merit in-creases and/or bonus payments link to performancerating and position in the pay range. Althoughscholars have yet to investigate the incentive andsorting effects of such moderate pay communica-tion practices, they may offer managers a partialsolution to the challenge of reducing informationasymmetries (and the problems associated withthem) while retaining pay-related flexibility andprivacy.

Second, they suggest that, to the extent that PFPperceptions explain a substantial portion of thedocumented indirect effects of pay secrecy on per-formance and continuation intentions, managersunable to adjust their organization’s pay communi-cation policy or PFP system attributes might con-sider taking steps to otherwise bolster employee

PFP perceptions. For example, organizations mighttake visible steps to heighten rater accountability,such as incorporating assessment quality as a cri-terion to be considered when assessing the perfor-mance of the rater. Such steps might signal toemployees that, despite the combination of infor-mation asymmetry and subjective assessment, rat-ers have a vested interest to do their utmost toensure a tighter link between actual effort and ap-praised performance (Bamberger, 2007; Baron &Kreps, 1999).

Limitations and Avenues for Future Research

Despite these theoretical and practical contribu-tions, a number of limitations of this study shouldbe mentioned, many having to do with the fact thatwe tested our model by means of a laboratory-basedsimulation. As noted by others (Falk & Fehr, 2003;Falk & Heckman, 2009), lab experiments offer sev-eral significant advantages over naturally occurringfield data, such as allowing us to systematicallyvary our exogenous variables (e.g., transparency ofpay communications) while keeping everythingelse fixed, as well providing the ability to ascertaincausality. Nevertheless, several questions may beraised regarding the study’s mundane realism andhence the external validity of our findings.

First, the setting and procedure were not veryconducive for the development of negative atti-tudes regarding pay, such as jealousy, among thestudy participants—something that, in the realworld, could potentially weaken the positive in-centive and sorting effects of pay transparency.Similarly, because—regardless of pay communica-tion condition—assessments were based on objec-tive performance (modified randomly in the sub-jective performance assessment condition), ourfindings fail to take into consideration that moretransparent pay systems may motivate appraisers tobias their ratings centrally so as to avoid the pecu-niary costs associated with more extreme scores(Bartol & Martin, 1989; Leventhal, 1976; Leventhalet al., 1980). To the extent that this may occur inactual organizations, it suggests that transparencymay be associated with an actual (rather than per-ceived) attenuation of PFP link, which, like jeal-ousy, could potentially counterbalance the adverseincentive and sorting effects of pay secrecy withsimilarly adverse pay transparency effects. Indeed,to the extent that transparency might drive central-ized ratings and thus result in reduced pay differ-entials, more talented individuals may not only be


less motivated to perform,but also less motivated tojoin or to remain (Harrison et al., 1996; Lazear,2000). In order to address such limitations, re-searchers may consider replicating the currentstudy in the context of a simulation in which theoutcomes are more meaningful to the participants(thus having the potential to create jealousies) andin which subjective ratings are centrally (ratherthan randomly) biased.

Second, mundane realism may have been limitedby assessing participants’ performance on a com-puter game as opposed to a regular work task(Aronson, Wilson, & Brewer, 1998). To the degreethat study participants may have been less seriousabout and engaged in the task than employees in aregular job, one may question the degree to whichtheir responses to varying pay communication andPFP conditions truly reflect those of actual work-ers. Still, participants indicated to us that the ex-perimental situation was involving and meaningfulto them, that they took it seriously, and that theycared about the scores they generated and the in-come they would receive, suggesting a high degreeof experimental and psychological realism (Aron-son et al., 1998). Moreover, Brewer (2000: 12)points out that “an experimental setting may havelittle mundane realism but still capture processesthat are highly representative of those that underlieevents in the real world.” Nevertheless, future re-search might attempt to address this limitation byreplicating our study in a more job-like experimen-tal context.

Mundane realism may also have been limitedby the somewhat minimalist nature of the subjec-tive vs. objective appraisal manipulation, withthe result being empirical effects that, while sta-tistically significant, were of a small magnitude.However, as noted by Prentice and Miller (1992:161), the strength of such a manipulation “de-rives not from the proportion of variance [theindependent variable] can account for, but in-stead from the fact that such a slight manipula-tion . . . can account for any variance at all.” Inthis context, the minimalist nature of our manip-ulation might best be viewed as heightening therisk of Type II error and generating effects that, ifanything, err on the conservative. Moreover, par-ticularly when it comes to task performance andthe continuation intentions of employees, evensmall effects can be practically meaningful (Cor-tina & Landis, 2009; Prentice & Miller, 1992).Finally, it is important to reinforce that, as indi-cated earlier (see fn. 1), effect size may be under-

estimated when testing a hierarchical model(Recchia, 2010; Snijders & Bosker, 1999). Never-theless, scholars seeking to replicate and extendthis study should consider the adoption of stron-ger manipulations in order to capture more accu-rately the true magnitude of the hypothesizedeffects. For example, researchers might moreclearly differentiate between subjective and ob-jective performance assessment by highlightinghow the former will take into account impres-sions of participants’ contextual behaviors, suchas helping and sportsmanship.

Third, although this experiment took about90 minutes, it may still have been too brief toallow for the development of the interpersonalcoworker dynamics found in actual workplaces.On the one hand, participants in the transparentcondition, lacking close relationships, may havebeen more cautious about comparing pay- or per-formance-related information than employeesworking in “real” transparent pay environments.To the extent that this may have occurred, itwould have only weakened the effects notedabove, thus suggesting that our findings may erron the conservative side. On the other hand, thelimited time frame over which our design al-lowed pay secrecy to affect perceptions and be-haviors may have allowed us to capture only themore immediate effects of pay communicationpolicies. To the degree that the accuracy withwhich employees estimate performance–pay con-tingencies increases with their own personal ex-perience over time, or that they find ways overtime to compensate cognitively for the uncer-tainty created by pay secrecy, our findingsmay not necessarily be indicative of the longer-term consequences of pay secrecy on perfor-mance and continuation intentions. In futurestudies, researchers might consider examiningempirically the extent to which seniority attenu-ates the negative effects of pay secrecy on PFPperceptions, task performance, and continuationintentions.

These context- and time-based threats to mun-dane realism cannot be discounted, and should beaddressed in future replication research in both thelab and the field. Nevertheless, the existing litera-ture on the generalizability of lab-based organiza-tional research in general, and lab-based compen-sation research in particular, suggest that ourresults and effect sizes would be more robust in thefield, rather than null or in a different direction(Anderson, Lindsay, & Bushman, 1999; Cohen-Cha-


rash & Spector, 2001; Dipboye, 1990; Locke, 1986).Anderson et al. (1999) observed that the correla-tions between effect sizes obtained in laboratoryand field settings generally exceed .70, indicatingthe similarity between results obtained from fieldand laboratory studies. Interestingly, in the case ofresearch on incentive pay, results indicate thatfield studies on average yield larger effects thanlaboratory studies (Jenkins, Mitra, Gupta, & Shaw,1998). Indeed, Jenkins et al. (1998), in their meta-analysis examining the impact of incentives on per-formance, found that incentives had a more pow-erful effect on performance in field settings than inlaboratory settings. Such findings suggest that, ifanything, the results presented in our study mayactually be conservative, with actual, field-basedestimates likely to be of substantially greatermagnitude.

Finally, it may be important to test the cross-cultural generalizability of our findings. For exam-ple, it may be that in cultures characterized bynorms and values promoting transparency aboutpay-related issues, pay secrecy may be more likelyto signal deceptive intent. In contrast, in cultures inwhich the disclosure of any information related toincome or wealth is considered taboo, employeesmay be less likely to infer from pay secrecy malev-olent managerial intent.

REFERENCES

Adams, J. S. 1965. Inequity in social exchange. In L.Berkowitz (Ed.), Advances in experimental socialpsychology, vol. 2: 267–299. New York: AcademicPress.

Aiken, L. S., & West, S. G. 1991. Multiple regression:Testing and interpreting interactions. NewburyPark, CA: Sage Publications.

Akerlof, G. A. 1970. The market for “Lemons”: Qualityuncertainty and the market mechanism. QuarterlyJournal of Economics, 84: 488–500.

Allison, P. D. 1990. Change scores as dependent variablesin regression analysis. Sociological Methodology,20: 93–114.

Anderson, C. A., Lindsay, J. J., & Bushman, B. J. 1999.Research in the psychological laboratory: Truth ortriviality? Current Directions in Psychological Sci-ence, 8: 3–9.

Aronson, E., Wilson, T. D., & Brewer, M. B. 1998. Exper-imentation in social psychology. In D. T. Gilbert,S. T. Fiske, & G. Lindzey (Eds.), The handbook ofsocial psychology: 99–142. Boston, MA: McGraw-Hill.

Balkin, D. B., & Gomez-Mejia, L. R. 1990. Matching com-pensation and organizational strategies. StrategicManagement Journal, 4: 153–169.

Bamberger, P. A. 2007. Competitive appraising: A socialdilemma perspective on the conditions in whichmulti-round peer evaluation may result in counter-productive team dynamics. Human Resource Man-agement Review, 17: 1–18.

Bamberger, P. A., & Belogolovsky, E. 2010. The impact ofpay secrecy on individual task performance. Person-nel Psychology, 63: 965–996.

Bamberger, P. A., & Meshoulam, I. 2000. Human re-source strategy: Formulation, implementation andimpact. Thousand Oaks, CA: Sage.

Baron, J. N., & Kreps, D. M. 1999. Strategic humanresources: Frameworks for general managers.New York: Wiley.

Bartol, K. M., & Martin, D. C. 1989. Effects of depen-dence, dependency threats, and pay secrecy on man-agerial pay allocations. Journal of Applied Psychol-ogy, 74: 105–113.

Bauer, D. J., Preacher, K. J., & Gil, K. M. 2006. Conceptu-alizing and testing random indirect effects and mod-erated mediation in multilevel models: New proce-dures and recommendations. PsychologicalMethods, 11: 142–163.

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., &Vohs, K. D. 2001. Bad is stronger than good. Reviewof General Psychology, 5: 323–370.

Becker, T. E. 2005. Potential problems in the statisticalcontrol of variables in organizational research: Aqualitative analysis with recommendations. Organ-izational Research Methods, 8: 274–289.

Bloom, M., & Milkovich, G. T. 1998. Relationships amongrisk, incentive pay, and organizational performance.Academy of Management Journal, 41: 283–297.

Bol, J. C. 2011. The determinants and performance effectsof managers’ performance evaluation biases. Ac-counting Review, 86: 1549–1575.

Bowen, D., & Ostroff, C. 2004. Understanding the HRM-firm performance linkages: The role of the “strength”of the HRM system. Academy of Management Re-view, 29: 203–221.

Brewer, M. B. 2000. Research design and issues of valid-ity. In H. T. Reis & C. M. Judd (Eds.), Handbook ofresearch methods in social and personality psy-chology: 3–16. New York: Cambridge UniversityPress.

Brockner, J., & Wiesenfeld, B. M. 1996. An integrativeframework for explaining reactions to decisions: In-teractive effects of outcomes and procedures. Psy-chological Bulletin, 120: 189–208.


Cadsby, C. B., Song, F., & Tapon, F. 2007. Sorting andincentive effects of pay-for-performance: An experi-mental investigation. Academy of ManagementJournal, 50: 387–405.

Card, D., Mas, A., Moretti, E., & Saez, E. 2012. Inequal-ity at work: The effect of peer salaries on job sat-isfaction. American Economic Review, 102: 2981–3003.

Cloutier, J., & Vilhuber, L. 2008. Procedural justice crite-ria in salary determination. Journal of ManagerialPsychology, 23: 713–740.

Cohen, K. 2006. The pulse of the profession: 2006 – 07salary budget survey. Workspan, September: 23–26.

Cohen-Charash, Y., & Spector, P. E. 2001. The role ofjustice in organizations: A meta-analysis. Organiza-tional Behavior and Human Decision Processes,86: 287–321.

Colella, A., Paetzold, R. L., Zardkoohi, A., & Wesson,M. J. 2007. Exposing pay secrecy. Academy of Man-agement Review, 32: 55–71.

Colella, A., Zardkoohi, A., Paetzold, R., & Wesson, M.2003. Day secrecy revisited: An integrative model.Paper presented at Academy of Management Meet-ing, Seattle, WA.

Colquitt, J. A., Conlon, D. E., Wesson, M. J., Porter, C. O.,& Ng, K. Y. 2001. Justice at the millennium: A meta-analytic review of 25 years of organizational justiceresearch. Journal of Applied Psychology, 86: 425–445.

Colquitt, J. A., Greenberg, J., & Zapata-Phelan, C. P. 2005.What is organizational justice? A historical over-view. In J. Greenberg & J. A. Colquitt (Eds.), Hand-book of organizational justice: 3–56. Mahwah, NJ:Lawrence Erlbaum.

Connelly, B. L., Certo, S. T., Ireland, R. D., & Reutzel,C. R. 2011. Signaling theory: A review and assess-ment. Journal of Management, 37: 39–67.

Cortina, J. M., & Landis, R. S. 2009. When small effectsizes tell a big story, and when large effect sizesdon’t. In C. E. Lance & R. J. Vandenberg (Eds.), Sta-tistical and methodological myths and urban leg-ends: Doctrine, verity, and fable in the organiza-tional and social sciences: 287–308. New York:Routledge.

Cox, D. R., & Snell, E. J. 1989. Analysis of binary data(2nd ed.). London: Chapman & Hall.

Davila, A., Foster, G., & Gupta, M. 2003. Venture capitalfinancing and the growth of startup firms. Journal ofBusiness Venturing, 18: 689–708.

Day, N. E. 2007. An investigation into pay communica-tion: Is ignorance bliss? Personnel Review, 36: 739–762.

DeNisi, A. S. 1996. Cognitive approach to performanceappraisal: A program of research. London: Rout-ledge.

Dipboye, R. L. 1990. Laboratory vs. field research inindustrial and organizational psychology. Interna-tional Review of Industrial and OrganizationalPsychology, 5: 1–34.

Edwards, J. R., & Lambert, L. S. 2007. Methods for inte-grating moderation and mediation: A general analyt-ical framework using moderated path analysis. Psy-chological Methods, 12: 1–22.

Erez, A., & Isen, A. M. 2002. The influence of positiveaffect on the components of expectancy motiva-tion. Journal of Applied Psychology, 87: 1055–1067.

Falk, A., & Fehr, E. 2003. Why labour market experi-ments? Labour Economics, 10: 399–406.

Falk, A., & Heckman, J. J. 2009. Lab experiments are amajor source of knowledge in the social sciences.Science, 326: 535–538.

Frank, R. H. 1985. The demand for unobservable andother nonpositional goods. American Economic Re-view, 75: 101–116.

Futrell, C. M., & Jenkins, O. C. 1978. Pay secrecy versuspay disclosure for salesmen: A longitudinal study.Journal of Marketing Research, 15: 214–219.

Gely, R., & Bierman, L. 2003. Pay secrecy/confidenti-ality rules and the National Labor Relations Act.Journal of Labor and Employment Law, 6: 120–156.

Gerhart, B., Rynes, S. L., & Fulmer, I. S. 2009. Pay andperformance: Individuals, groups, and executives.Academy of Management Annals, 3: 251–315.

Gibbs, M., Merchant, K. A., Van der Stede, W. A., &Vargus, M. E. 2004. Determinants and effects of sub-jectivity in incentives. Accounting Review, 79: 409–436.

Gioia, D. A., & Chittipeddi, K. 1991. Sensemaking andsensegiving in strategic change initiation. StrategicManagement Journal, 12: 433–448.

Gomez-Mejia, L. R., Berrone, P., & Franco-Santos, M.2010. Compensation and organizational perfor-mance: Theory, research, and practice. New York:M. E. Sharpe.

Goodman, J. S., & Blum, T. C. 1996. Assessing the non-random sampling effects of subject attrition in lon-gitudinal research. Journal of Management, 4: 627–652.

Greenberg, J. 1990. Organizational justice: Yesterday, to-day and tomorrow. Journal of Management, 16:399–432.

Greenberg, J. 2003. Creating unfairness by mandating fair


procedures: The hidden hazards of a pay-for-perfor-mance plan. Human Resource Management Re-view, 13: 41–57.

Harrison, D. A., Virick, M., & William, S. 1996. Workingwithout a net: Time, performance, and turnover un-der maximally contingent rewards. Journal of Ap-plied Psychology, 81: 331–345.

Heneman, R. L., Greenberger, D. B., & Strasser, S. 1988.The relationship between pay-for-performance per-ceptions and pay satisfaction. Personnel Psychol-ogy, 41: 745–759.

Institute for Women’s Policy Research (IWPR) & Rocke-feller Survey of Economic Security. 2011. Pay se-crecy and wage discrimination. Fact Sheet #C382.Washington, DC: Institute for Women’s Policy Re-search.

Ittner, C. D., Larcker, D. F., & Meyer, M. W. 2003. Sub-jectivity and the weighting of performance measures:Evidence from a balanced scorecard. Accounting Re-view, 78: 725–758.

Jenkins, G. D. J., Jr., Mitra, A., Gupta, N., & Shaw, J. D.1998. Are financial incentives related to perfor-mance? A meta-analytic review of empirical re-search. Journal of Applied Psychology, 83: 777–778.

Johns, G. 2006. The essential impact of context on organ-izational behavior. Academy of Management Re-view, 31: 386–408.

Jones, D. A., & Skarlicki, D. P. 2003. The relationshipbetween perceptions of fairness and voluntary turn-over among retail employees. Journal of AppliedSocial Psychology, 33: 1226–1243.

Kanfer, R. 1987. Task-specific motivation: An integrativeapproach to issues of measurement, mechanisms,processes, and determinants. Journal of Social andClinical Psychology, 5: 237–264.

Kanfer, R. 1990. Motivation theory and industrial/orga-nizational psychology. In M. D. Dunnette & L. Hough(Eds.), Handbook of industrial and organizationalpsychology, vol. 1: 75–170. Palo Alto, CA: Consult-ing Psychologists Press.

Kramer, R. M. 1998. Paranoid cognition in social sys-tems: Thinking and acting in the shadow of doubt.Personality and Social Psychology Review, 2: 251–275.

Kramer, R. M. 2001. Organizational paranoia: Originsand dynamics. Research in Organizational Behav-ior, 23: 1–42.

Kuhn, K. M., & Yockey, M. D. 2003. Variable pay as arisky choice: Determinants of the relative attrac-tiveness of incentive plans. OrganizationalBehavior and Human Decision Processes, 90: 323–341.

Lambert, S. J. 2000. Added benefits: The link betweenwork–life benefits and organizational citizenship be-havior. Academy of Management Journal, 43: 801–815.

Lawler, E. E. 1966a. Managers’ attitudes toward howtheir pay is and should be determined. Journal ofApplied Psychology, 50: 273–279.

Lawler, E. E. 1966b. The mythology of management com-pensation. California Management Review, 9: 11–22.

Lawler, E. E. 1967. Secrecy about management com-pensation: Are there hidden costs? Organization-al Behavior and Human Performance, 2: 182–189.

Lawler, E. E. 2003. Reward practices and performancemanagement system effectiveness. OrganizationalDynamics, 32: 396–404.

Lawler, E. E, & Jenkins, G. D. 1992. Strategic rewardsystems. In M. D. Dunnette & L. M. Hough (Eds.),Handbook of industrial and organizational psy-chology (3rd ed.): 1009–1055. Palo Alto, CA: Coun-seling Psychologists Press.

Lazear, E. P. 1986. Salaries and piece rates. Journal ofBusiness, 59: 405–431.

Lazear, E. P. 1995. Personnel economics. Cambridge,MA: The MIT Press.

Lazear, E. P. 2000. Performance pay and productivity.American Economic Review, 90: 1346–1361.

Lazear, E. P., & Oyer, P. 2013. Personnel economics. In R.Gibsons & J. D. Roberts (Eds.), Handbook of organ-izational economics: 479 –519. Princeton, NJ:Princeton University Press.

Leventhal, G. S. 1976. The distribution of rewards andresources in groups and organizations. Advancesin Experimental Social Psychology, 9: 92–131.

Leventhal, G. S., Karuza, J., & Fry, W. R. 1980. Beyondfairness: A theory of allocation preferences. In G.Mikula (Ed.), Justice and social interaction: 167–218. New York: Springer-Verlag.

Little, R. J. A., & Rubin, D. B. 1987. Statistical analysiswith missing data. New York: John Wiley.

Locke, E. A. (Ed.). 1986. Generalizing from labora-tory to field settings. Lexington, MA: LexingtonBooks.

Magee, L. 1990. R2 measures based on Wald and likeli-hood ratio joint significance tests. American Statis-tician, 44: 250–253.

Major, B., & Adams, J. B. 1983. Role of gender, interper-sonal orientation, and self-presentation in distribu-tive-justice behavior. Journal of Personality and So-cial Psychology, 45: 598–608.

McCoach, D. B. 2010. Hierarchical linear modeling. InG. R. Hancock & R. O. Mueller (Eds.), The reviewer’s


guide to quantitative methods in the social sci-ences: 123–140. New York: Routledge.

Messersmith, J. G., Guthrie, J. P., Ji, Y. Y., & Lee, J. Y.2011. Executive turnover: The influence of disper-sion and other pay system characteristics. Journal ofApplied Psychology, 96: 457–469.

Milkovich, G. T., & Anderson, P. H. 1972. Managementcompensation and secrecy policies. Personnel Psy-chology, 25: 293–302.

Milkovich, G. T., Newman, J. M., & Gerhart, B. 2013.Compensation (11th ed.). New York: McGraw-HillIrwin.

Moers, F. 2005. Discretion and bias in performanceevaluation: The impact of diversity and subjectiv-ity. Accounting, Organizations and Society, 30:67–80.

Murphy, K. R., & Cleveland, J. N. 1995. Understandingperformance appraisal: Social, organizational,and goal based perspectives. Thousand Oaks, CA:Sage.

Murphy, K. R., & Myors, B. 2004. Statistical poweranalysis (2nd ed.). Mahwah, NJ: Lawrence Erl-baum.

Naylor, J. C., Pritchard, R. D., & Ilgen, D. R. 1980. Atheory of behavior in organizations. New York: Ac-ademic Press, Inc.

Peeters, G., & Czapinski, J. 1990. Positive–negativeasymmetry in evaluations: The distinction be-tween affective and informational negativity ef-fects. European Review of Social Psychology, 1:33–60.

Perry, J. L., Engbers, T. A., & Jun, S. Y. 2009. Back to thefuture? Performance-related pay, empirical research,and the perils of persistence. Public AdministrationReview, 69: 39–51.

Pfeffer, J., & Langton, N. 1993. The effect of wage disper-sion on satisfaction, productivity, and working col-laboratively: Evidence from college and universityfaculty. Administrative Science Quarterly, 38:382–407.

Preacher, K. J., & Hayes, A. F. 2004. SPSS and SASprocedures for estimating indirect effects in sim-ple mediation models. Behaviour Research Meth-ods, Instruments and Computers, 36: 717–731.

Preacher, K. J., & Kelley, K. 2011. Effect size measures formediation models: Quantitative strategies for com-municating indirect effects. Psychological Methods,16: 93–115.

Prentice, D. A., & Miller, D. T. 1992. When small effectsare impressive. Psychological Bulletin, 112: 160–164.

Recchia, A. 2010. R-squared measures for two-level hier-

archical linear models using SAS. Journal of Statis-tical Software, 32: Code Snippet 2. Retrieved fromhttp://www.jstatsoft.org/v32/c02/paper

Rynes, S. L. 1991. Recruitment, job choice, and post-hireconsequences: A call for new research directions. InM. D. Dunnette & L. Hough (Eds.), Handbook ofindustrial and organizational psychology, vol. 2(2nd ed.): 399–444. Palo Alto, CA: Consulting Psy-chologists.

Rynes, S. L., Gerhart, B., & Parks, L. 2005. Performanceevaluation and pay for performance. Annual Reviewof Psychology, 56: 571–600.

Shaw, J. D., & Gupta, N. 2007. Pay system characteristicsand quit patterns of good, average, and poor perform-ers. Personnel Psychology, 60: 903–923.

Singer, J. D. 1998. Using SAS PROC MIXED to fit multi-level models, hierarchical models, and individualgrowth models. Journal of Educational and Behav-ioral Statistics, 23: 323–355.

Sloof, R., & van Praag, C. M. 2008. Performance measure-ment, expectancy and agency theory: An experimen-tal study. Journal of Economic Behavior and Or-ganization, 67: 794–809.

Sloof, R., & van Praag, C. M. 2010. The effect of noise ina performance measure on work motivation: A realeffort laboratory experiment. Labour Economics, 17:751–765.

Snijders, T. A. B., & Bosker, R. J. 1999. Multilevel anal-ysis. An introduction to basic and advanced multi-level modeling. London: Sage.

Spence, M. 1973. Job market signaling. Quarterly Jour-nal of Economics, 87: 355–374.

Spence, M. 2002. Signaling in retrospect and the infor-mational structure of markets. American EconomicReview, 92: 434–459.

Stiglitz, J. E. 2002. Information and the change in theparadigm in economics. American Economic Re-view, 92: 460–501.

Suazo, M. M., Martínez, P. G., & Sandoval, R. 2009.Creating psychological and legal contracts throughhuman resource practices: A signaling theory per-spective. Human Resource Management Review,19: 154–166.

Suazo, M. M., Martínez, P. G., & Sandoval, R. 2011.Creating psychological and legal contracts throughHRM practices: A strength of signals perspective.Employee Responsibilities and Rights Journal, 23:187–204.

Tremblay, M., & Chenevert, D. 2008. Influence of com-pensation strategies in Canadian technology-inten-sive firms on organizational and human resources


performance. Group and Organization Manage-ment, 33: 269–302.

Trevor, C. O., Gerhart, B., & Boudreu, W. 1997. Voluntaryturnover and job performance: Curvilinearity and themoderating influences of salary growth and promotions.Journal of Applied Psychology, 55: 585–610.

Trevor, C. O., Reilly, G., & Gerhart, B. 2012. Reconsidering paydispersion’s effect on the performance of interdependentwork: Reconciling sorting and pay inequality. Academyof Management Journal, 55: 585–610.

Tyler, T. R. 1994. Psychological models of the justice

motive: Antecedents of distributive and proceduraljustice. Journal of Personality and Social Psychol-ogy, 67: 850–863.

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

Wanasika, I., & Adler, T. 2011. Deception as strategy:Context and dynamics. Journal of Managerial Is-sues, 23: 364–378.

Williamson, O. E. 1975. Markets and hierarchies, analysisand antitrust implications: A study in the economics ofinternal organization. New York: Free Press.

APPENDIX ATimeline of the Procedure

PhaseProcedure/Information

Provided Manipulation

Start Instructions Absolute condition: Participants were informed that bonus pay would be allocated on an absolute (i.e., setand predetermined) basis according to their level of performance.

Relative condition: Participants were informed that bonus pay would be allocated according to the point-based ranking of each participant within his or her group.

Objective condition: Participants were told that the magnitude of each rounds’ performance score would becalculated objectively, entirely contingent upon the individual’s point count in that phase.

Subjective condition: Participants were told that the magnitude of each rounds’ performance score wouldbe calculated according to the combination of the individual’s objective point count in that phase withthe investigators’ subjective assessment, based on his or her observation of the individual’s performance.

TutorialTask 1

Break 1 Graph(s) Pay secrecy condition: Participant’s score on the task for that round and a bar graph of the bonus pay to bereceived for that round.

Pay transparency condition: Participant’s total score and bar graph of the additional pay to be received forthat task round and a graph of his or her pay relative to that of the others (by code number) inthe group.

E-mails Pay secrecy condition: Participants were requested not to discuss any pay-related issue with others.Pay transparency condition: Participants in this condition were told that their e-mail communications

with their fellow group members would be unrestricted.Questionnaire (PFP

perceptions)Task 2

Break 2 Repeat of Break 1Task 3

Break 3 Graph(s)Questionnaire (continuation

intentions)Manipulation check

Elena Belogolovsky ([email protected]) is an assistantprofessor of human resource studies in the ILR School atCornell University. She received her PhD in behavioralsciences and management from Technion–Israel Institutefor Technology, Israel. Her research interests includecompensation administration and strategy, organization-al citizenship behavior (OCB), and occupational stressand employee well-being.

Peter A. Bamberger ([email protected]) (PhD Cornell

University) is a professor of organizational behavior andhuman resource studies at the Recanati Business Schoolof Tel Aviv University, and director of research of theSmithers Institute at Cornell University’s ILR School. Hiscurrent research interests include peer relations and em-ployee helping processes, occupational health psychol-ogy, and compensation strategy.


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.

Documents

SIGNALING IN SECRET: PAY FOR PERFORMANCE AND THE INCENTIVE AND … · 2015-03-12 · SIGNALING IN SECRET: PAY FOR PERFORMANCE AND THE INCENTIVE AND SORTING EFFECTS OF PAY SECRECY