D

Sa

b

a

ARRA

JJJJ

KULL

(it

plcriitstatewp

oct

(

1d

The Journal of Socio-Economics 41 (2012) 445– 454

Contents lists available at ScienceDirect

The Journal of Socio-Economics

j our nal ho me p age: www.elsev ier .com/ loc ate /soceco

oes labor contract completeness drive unionization? Experimental evidence

ean Masaki Flynna,∗, Michael Donnellyb

Department of Economics, Scripps College, United StatesOffice of Research and Quantitative Studies, Division of Domestic Finance, U.S. Department of the Treasury, United States

r t i c l e i n f o

rticle history:eceived 29 January 2010eceived in revised form 8 June 2011ccepted 14 July 2011

EL classification:41

a b s t r a c t

Flynn (2005) proposes that the degree to which labor contracts are complete may be a major driving forcebehind the propensity of employees to unionize. We find behavior consistent with this hypothesis in anexperimental production game in which subjects are assigned to playing either employers or employees.The rate at which employees opt for a proxy for unionization more than triples when the labor-contractingregime under which they are working shifts from incomplete to complete labor contracts. Complete laborcontracts drive out positive reciprocity, anger workers, and increase their desire to unionize.

5153

eywords:nionsabor contracts

© 2011 Elsevier Inc. All rights reserved.

abor contract completeness

This paper reports on two experiments designed to test Flynn’s2005) theory that the degree to which labor contracts are completen a given industry will be a key factor in determining how likelyhat industry’s workers are to unionize.

Flynn’s theory is based on the idea that employees are hap-ier and less likely to unionize when working under incomplete

abor contracts than under complete labor contracts. This is thease because only incomplete labor contracts encourage positiveeciprocity between employers and employees. The essential points that under incomplete labor contracts, employees have the abil-ty to shirk without being caught. This encourages managemento treat employees generously and to rely on carrots rather thanticks. By contrast, employees working under complete labor con-racts cannot shirk because they face heavy monitoring. Under such

system, positive reciprocity is suffocated because managementends to rely on sticks rather than carrots. The result is that work-rs laboring under complete labor contracts are much less satisfiedith how they are being treated and, consequently, much morerone to joining a union in order to seek a redress of grievances.

The results presented here are noteworthy because, to the best

f our knowledge, Flynn’s (2005) hypothesis that labor contractompleteness can explain the distribution of unions across indus-ries is the only theory of its type. That is, while other authors

∗ Corresponding author. Tel.: +1 909 607 3398.E-mail addresses: sflynn@scrippscollege.edu, sean.flynn@scrippscollege.edu

S.M. Flynn).

053-5357/$ – see front matter © 2011 Elsevier Inc. All rights reserved.oi:10.1016/j.socec.2011.07.002

discuss why particular employees in particular industries at partic-ular times may be more or less likely to unionize under particularcircumstances, there exists to our knowledge no other general the-ory of unionization that makes any claim to explaining the differentrates of unionization found across various industries in terms of acore set of characteristics or variables.

The remainder of this paper proceeds as follows. Section 1describes the production environment and explains the Bonus Con-tract, Monitoring Contract, and Minimum Wage Contract that weuse to model, respectively, incomplete labor contracts, completelabor contracts, and the option to unionize. Section 2 reports theresults of the primary experiment in which incomplete contractsprecede complete contracts. Section 3 reports the results of thesecondary experiment in which complete contracts precede incom-plete contracts. Section 4 concludes.

1. Modeling labor contract completeness and unionizationwith experimental games

Experiments reviewed by Fehr and Gächter (2000a) demon-strate that labor contract completeness has significant effects onthe major outcome variables of employer-employee productiongames, including employee effort levels, the size of the total sur-plus generated by production, and the division of the total surplusbetween employees and employers. These effects had not been

noted in previous work (see Prendergast, 1999; MacLeod andParent, 1999).

The completeness of the labor contract appears to have pro-nounced effects on these outcome variables by either facilitating

446 S.M. Flynn, M. Donnelly / The Journal of Socio-Economics 41 (2012) 445– 454

Table 1Effort cost to employees.

Actual effort level, e

1 2 3 4 5 6 7 8 9 10

oes(pbeeettt

rfttstflt

piswbcrCrAittoC

vTkWthi

pjewp

tmits

proposal consisting of three elements to the employee with whomthe employer is anonymously paired:

Cost to employee, c(e) 0 1 2 4

r impeding reciprocal gift exchanges between employers andmployees. Indeed, reciprocal gift exchanges of the sort envi-ioned by Akerlof (1982), Leibenstein (1987), and Akerlof and Yellen1990) can only take place in labor markets operating under incom-lete labor contracts. The key intuition is that generous reciprocalehavior is the only way for employers to effectively motivatemployees in situations where employees are free to shirk. Andmployees only have this option in incomplete labor contractingnvironments in which they possess significant freedom to varyheir output levels without any substantial likelihood of causinghemselves harm (either by getting themselves fired or by reducinghe level of compensation that they receive).

By contrast, employees operating under a complete contractingegime are either guaranteed to suffer harm or are very likely to suf-er harm if they reduce their effort levels. This is true either becausehe nature of the labor contract directly ties their compensation toheir effort levels (as is the case with piece-work compensationchemes) or because the nature of the production process allowshe employer to fully monitor effort levels and punish any deviationrom the contracted labor agreement (as is the case with assemblyine work where an inability or an unwillingness to keep up withhe line is immediately visible to supervisors).

Since the point of this paper is to examine whether theropensity of employees to unionize varies under complete and

ncomplete labor contracting regimes, it is necessary to definepecific contract types that will hopefully capture the essential real-orld features of the decision to unionize and how it is affected

y incomplete and complete labor contracts. When we wish toapture the essential features of an incomplete labor contractingegime, we require that employers only offer incomplete “Bonusontracts.” When we wish to impose a complete labor contractingegime, employers can only offer complete “Monitoring Contracts.”n alternative “Minimum Wage Contract” that proxies for union-

zation is available to employees throughout the experiment sohat we can see whether their propensity to opt for unioniza-ion varies depending up on whether employers are constrained toffering only incomplete Bonus Contracts or complete Monitoringontracts.

The Bonus Contract and the Monitoring Contract have been pre-iously studied in great detail, most notably by Fehr et al. (2007).his is very useful when interpreting our results since it is alreadynown how subjects behave when presented with these games.e can thus concentrate on how they behave with regard to the

wo novel features that we introduce—the option to unionize andow the popularity of that option is affected by the switch from an

ncomplete contracting regime to a complete contracting regime.Before we proceed to describe the three contracts in detail,

lease note that in the instructions that were read by our sub-ects, the Bonus Contract and the Monitoring Contract were fullyxplained but never explicitly named. Thus, we do not have toorry about whether those names in any way biased subjects’erceptions as to the attractiveness of those two contracts.

By contrast, the Minimum Wage Contract was explicitly referredo in the instructions as the “Minimum Wage Option.” We therefore

ust take into consideration the possibility of that name caus-ng framing effects. In particular, we might be concerned that theerm “minimum wage” carries with it connotations of low socialtatus.

6 8 10 13 16 20

We believe that our experimental design should be robust tothese potential framing effects and that our results should not bebiased by them. We hold this opinion because the “Minimum WageOption” was presented to subjects both when the alternative wasthe Bonus Contract and when the alternative was the MonitoringContract. Thus, any framing effects or biases related to the use of thephrase “minimum wage” should have been held constant over thecourse of the experiment. With the potential biases held constantin this way, the experiment tests whether—relative to the Mini-mum Wage Option and any frames that come with it—the MonitoringContract or the Bonus Contract is better received by employees.

1.1. The production environment

All three labor contracts share the same general environment. Tobegin with, both employers and employees play for tokens.1 Theyare informed in the instructions that each token will be convertedinto $0.10 at the end of the experiment.2 Furthermore, each player(whether assigned the role of employer or employee) is given aninitial endowment of 125 tokens ($12.50) at the start of the exper-iment and informed that if their total stock of tokens falls below10 tokens at any time during the game (due to losses incurredas they play the game), they will be dismissed. As it turned out,however, no players were dismissed and average earnings at theend of the experiment were $33.37 per person. The highest earnerwas an employee who finished with $46.20 (including the initialendowment). The lowest earner was an employer who finishedwith $22.10 (including the initial endowment).

The earnings of employers are referred to as profits, P, whilethe earnings of employees are referred to as income, I. The pro-duction function under all possible labor contracts is linear in theactual amount of effort, e, that employees exert, so that employ-ers receive 10e units of revenue for any given level of actual effort,e, that employees put forth. The actual effort level is selected byemployees after seeing the labor contract offers made by employersand can range over e = {1, 2, 3, . . ., 10}.

Actual effort, however, is costly to employees, so that employeesmust be incentivized by employers to provide effort. An employee’seffort cost, c, is measured in tokens and varies with actual effort lev-els as described in Table 1. As is clear from Table 1, the marginalcost of providing effort increases with higher effort levels. It is only1 token when moving from the lowest effort level e = 1 to e = 2, butgradually increases until it is 4 tokens when moving from effortlevel e = 9 up to the maximal effort level, e = 10. Thus, employersmust not only incentivize employees to provide effort, they mustalso provide successively greater incentives if they wish to incen-tivize successively higher effort levels.

1.2. The Bonus Contract

Under the Bonus Contract, an employer makes a labor contract

1 These were small glass discs that were referred to as “pebbles” in the instructionsgiven to participants.

2 The instructions are available upon request from the corresponding author.

l of So

•

•

•

imtrPI

peCaeve

tnpevagtceh

Btwlstfieai

1

ieeei

e

te

the propensity to unionize. If it does, then, presumably, there wouldbe an even larger increase in the propensity to unionize if a moredraconian switch were imposed.

and punishments—rather than the positive incentives and bonuses available underthe Bonus Contract—we think that referring to it as an “Incentive Contract” wouldonly cause confusion in terms of this paper and the distinction between completeand incomplete contracts. Consequently, we have opted instead to refer to it as a“Monitoring Contract.”

5 As a historical example, supervisory costs at Ford Motors were much higherafter the firm switched from craft production (and incomplete labor contracts) toassembly line production (and complete labor contracts) in 1914 (Raff, 1988; Flynn,2005). Indeed, the ratio of supervisory personnel to production employees rose from

S.M. Flynn, M. Donnelly / The Journa

w, a fixed wage which will be paid by the employer to theemployee no matter how much (or how little) actual effort, e,the employee chooses to put forward.e*, a demanded effort level that the employer would like theemployee to exert.b*, an announced bonus payment that the employer claims shewill pay if the employee’s actual effort is equal to or greater thanthe demanded effort level (if, that is, e ≥ e*). Once the employerhas seen the employee’s actual effort level, she is free to set anactual bonus, b, that can be different from the announced bonuspayment, b*. In particular, it can be zero, even if e ≥ e*.

Also note that for both Bonus Contracts as well as Monitor-ng Contracts, employer wage offers and demanded effort levels

ust satisfy w > c(e*) in order to rule out employee losses.3 Ifhe employee agrees to accept the employer’s offer, then theirespective profits and incomes are calculated as, respectively,

= 10 × actual effort level (e) − fixed wage (w) − actual bonus (b); and = fixed wage (w) + actual bonus (b) − cost of actual effort level (c).

The key point about the Bonus Contract that makes it an incom-lete contract is that, with the exception of the fixed wage, neithermployees nor employers have any way of enforcing the Bonusontract. Employees have no way of holding employers to theirnnounced bonus, b*, even if they do provide an actual effort level

≥ e*. And employers have no guarantee that employees will pro-ide an actual effort level higher than the minimal effort level of

= 1 in exchange for the fixed wage and the promised bonus.The interesting thing about the Bonus Contract is that despite

he incomplete nature of the contract and the fact that it haso enforcement mechanisms whatsoever, employees do tend torovide positive (and in some cases very high) effort levels andmployers do often pay their promised bonuses if employees pro-ide the demanded level of effort. Because this is true even innonymous one-shot games in which there is no scope for repeatedame punishment strategies, Fehr and Gächter (2000a,b) argue thathis behavior must be the result of reciprocity. Employees recipro-ate high fixed wages and high promised bonuses with high actualffort levels; employers reciprocate high actual effort levels withigh actual bonuses.

For the purposes of this paper, it is important to note that theonus Contract replicates several key features of incomplete con-racts found in the real world. First, the fact that the fixed wageill be paid and received irrespective of how large the actual effort

evel turns out to be implies that employees have the chance tohirk. Second, effort is costly, so that employees have an incentiveo shirk. Third, employers are free to vary the values of both thexed wage and the promised bonus in order to signal generosity tomployees. Fourth, because employees are free to modulate theirctual effort levels, they have the ability to reciprocate generouslyf they feel that they are being treated generously.

.3. The Monitoring Contract

The Monitoring Contract is our proxy for complete contract-ng regimes. It modifies the Bonus Contract, removing the bonus

ntirely while simultaneously imposing a costly but imperfectlyffective monitoring regime that can lead to costly punishment ifmployees are discovered to have given an actual effort level thats less than the demanded effort level.4

3 This constraint is imposed to ensure that loss aversion does not affect themployee’s behavior. See Fehr et al. (2007).4 What we are referring to as the “Monitoring Contract” is referred to as an “Incen-

ive Contract” by other authors, including Fehr and Gächter (2000a,b) and Fehrt al. (2007). However, because this type of contract only features disincentives

cio-Economics 41 (2012) 445– 454 447

Under the Monitoring Contract, an employer makes a labor con-tract proposal consisting of three elements to the employee withwhom the employer is anonymously paired:

• w, a fixed wage that will be paid by the employer to the employeeno matter how much actual effort, e, the employee puts forth.

• e*, a demanded effort level.• f, a zero-sum fine that will be imposed on the employee and paid

to the employer if (a) a die that is rolled turns up 1 or 2, and (b)the employee’s actual effort level is less than the demanded effortlevel.

If an employee accepts the Monitoring Contract, then employerprofits and employee incomes are calculated, respectively,as Employer profit P = 10 × actual effort level (e) − fixed wage(w) − monitoring fee (10) + fine (f) if e < e* AND if die roll is 1 or 2; andEmployee income I = fixed wage (w) − cost of actual effort (c) − fine (f)if e < e* AND if die roll is 1 or 2. Notice that monitoring is costly, sothat employers must bear a 10-token fee for monitoring under theMonitoring Contract.5

In our setup, monitoring only takes place with probability 1/3,so that the fine for shirking, f, is only imposed if the employee givesless effort than demanded AND if the die roll comes up 1 or 2. The1/3 probability of being able to catch an employee shirking is meantto replicate the fact that even in many real-world labor contractingregimes that we would tend to classify as featuring complete laborcontracts, there is still some ability on the part of employees toshirk.6

We used the relatively low monitoring probability of 1/3 for sev-eral reasons. First, 1/3 is the probability used by Fehr et al. (2007).Thus, the behavior of their subjects can be used as a point of com-parison when interpreting the behavior of our subjects. Second,a moderate value for the monitoring probability implies that ourproxy for a complete labor contracting regime is conservative as faras the objective of this paper is concerned. That is, the MonitoringContract used here features a much lower chance of an employeegetting caught if he shirks than is the case on a highly monitoredassembly line where an employee has an extremely high chanceof getting caught if he shirks.7 The result is that we can determinewhether changing from a very incomplete labor contracting regime(the Bonus Contract) to a moderately complete labor contactingregime (the Monitoring Contract) causes a substantial increase in

1:53 in 1914 to 1:15 in 1917 even as the total number of Ford employees more thantripled over this time period (Raff, 1988; Slichter, 1919, footnote 21, pp. 243–244).

6 For instance, seamstresses operating under a piece-work scheme can presum-ably do a marginally bad job sewing shirts without guaranteeing that they will becaught shirking. Similarly, press operators can most likely do a slightly shoddy jobon an assembly line without necessarily being discovered.

7 For <fn0035>instance, after Ford Motor’s switch to assembly line production(and therefore complete labor contracts) in 1914 (see footnote 4), supervisorsstood vigil on platforms above the assembly line. They would immediately fire anyemployee who could not keep up with the line and replace him with one of thesubstitute employees who “were constantly kept on hand, at the factory’s expense,to meet all emergencies” (Martin, 1915).

4 l of So

1

eBett

cnwaizrittWa

Wtrwe

taeohtWteettc

cwIcetCeta

1

eTte

pir

48 S.M. Flynn, M. Donnelly / The Journa

.4. The Minimum Wage Contract

The Minimum Wage Contract is our proxy for unionization. Inach period, employees have two alternatives when faced with theonus Contract or Monitoring Contract offered by their respectivemployers. They can either reject the contract outright and refuseo work at all, or they can opt for a Minimum Wage Contract thathe employer is obliged to accept.

The first alternative of outright rejection allows employees thehance to opt for voluntary unemployment, presumably becauseeither the contract offered by the employer nor the minimumage contract is attractive enough to continue onward and expend

ny effort. It is also a way for fair-minded employees to pun-sh unfair offers because rejection also forces unemployment andero earnings for the period onto any employer whose offer isejected. As things turned out, however, we only observed onenstance of an employee opting for outright rejection and volun-ary unemployment.8 In all other cases, employees either acceptedhe contract offered by the employer or opted for the Minimum

age Contract, which sets a fixed wage of w = 20 in exchange forn actual effort level that cannot be less than e = 3.

Neither bonuses nor fines are possible under the Minimumage Contract. If an employee opts for the Minimum Wage Con-

ract, then employer profits and employee incomes are calculated,espectively, as, Employer profit P = 10 × actual effort level (e) − fixedage (20); and Employee income I = fixed wage (20) − cost of actual

ffort level (c).It should be noted that while the Minimum Wage Contract fixes

he wage at w = 20, it still allows employees the option of choosingn effort level greater than the minimum required effort level of

= 3. This allows for the possibility that a given employee’s sensef fairness might dictate that he put forth an actual effort leveligher than the minimum required effort level in exchange forhe fixed wage of w = 20 that he will receive under the Minimum

age Contract. In the course of the experiments reported here,his only happened three times—twice out of the 51 cases in whichmployees opted for the Minimum Wage Contract in the primaryxperiment in which incomplete contracts preceded complete con-racts, and once out of the 20 cases in which employees opted forhe Minimum Wage Contract in the second experiment in whichomplete contracts preceded incomplete contracts.

As a result, the actual profits and incomes in all otherases where the Minimum Wage Contract was put into effectere, respectively, P = 3 × 10 − 20 = 10 for the employers and

= 20 − 2 = 18 for the employees. Please note that with such an out-ome, employees receive the lion’s share of the total surplus, so thatmployees concerned about relative winnings could always ensurehemselves a superior position by opting for the Minimum Wageontract. It also implies that employers hoping to entice employ-es into providing higher actual effort levels must entice them withhe possibility of earning an income higher than the I = 18 that isvailable under the Minimum Wage Contract.

.5. The Minimum Wage Contract as a union proxy

We considered several different mechanisms to model a prefer-nce for unions before settling on the Minimum Wage Contract.

he major challenge was to choose a mechanism that captureshe essential features of unionization. We take these to be: strongmployee control of the contract that is agreed to; very spe-

8 This one instance happened during the reverse treatment order, in which com-lete contracts preceded incomplete contracts. In the normal treatment order where

ncomplete contracts preceded complete contracts, workers never chose outrightejection and voluntary unemployment.

cio-Economics 41 (2012) 445– 454

cific delimitations of the compensation that will be paid by theemployer; and very specific delimitations of the work that will beperformed by the employee—indeed, union contracts are typicallyvery specific and have the flavor of complete contracts.

We believe that the Minimum Wage Contract captures thesefeatures of union contracts. To begin with, our employees havefull control over whether the Minimum Wage Contract is imple-mented. This is analogous to unions having a strong say in thedetails of the labor contracts that they negotiate with employers.Next, the Minimum Wage Contract is extremely detailed about thecompensation that must be paid by the employer. This is similarto the very detailed compensation agreements that are negotiatedinto real-world union contracts. Finally, the Minimum Wage Con-tract is specific about the work that must be performed by theemployee. This is again similar to real-world union contracts, whichare very specific about both the quality and quantity of labor thatmust be supplied. Because of these similarities, we feel that theMinimum Wage Contract used in this experiment is a reasonableproxy for union contacts in the real world—both because it is verycomplete and because it can be imposed by employees onto theiremployers.

The Minimum Wage Contract is also helpful because it avoidshaving to model any sort of collective action process on the part ofemployees in terms of whether or not to form a union. Since weare trying to determine whether an individual’s propensity to join aunion is modulated by the type of labor contracting regime in whichhe or she works, there is no need—as far as the purpose of this paperis concerned—to utilize a proxy for unionization that would simu-late the collective action nature of real-world union organizationor collective bargaining. The Minimum Wage Contract is very use-ful because it can elicit individual preferences about unionizationwhile avoiding voting paradoxes and other problems that mightarise if we implemented a collective action mechanism for deter-mining preferences about unionization. We would like to thankSimon Gächter for suggesting it to us after we told him of our desireto abstract from the collective action aspects of real-world unioncontracting.

As for the specifics of the Minimum Wage Contract that weimplemented, a month before running the experiments presentedin this paper we did several rounds of pre-testing with differentsubjects to make sure that a Minimum Wage Contract with e = 3 andw = 20 would be appealing to employees as an alternative to bothBonus Contracts and Monitoring Contracts (depending, of course,upon the generosity employer offers). We chose these values basedupon the results of the Bonus Games and Monitoring Games runby Fehr et al. (2007), but a more theoretical underpinning can begiven by reference to the theoretical work in Fehr et al. (2007) thatdeals with how both fair and self-interested agents would act underBonus Contracts and Monitoring Contracts—assuming that theirbehavior is correctly captured by the model of inequity aversionproposed by Fehr and Schmidt (1999).

We refer the reader to those papers, noting here only that theattractiveness of our Minimum Wage Contract under the Fehr-Schmidt model of inequity aversion depends crucially upon theassumptions that subjects make about what fraction of their fellowsubjects are fair-minded or self-interested. Thus, there are theo-retical reasons to believe that our results about the propensity tounionize are specific not only to the details of the Minimum WageContract that we actually implemented but also depend upon theassumptions made by our subject pool about their fellow subjects.Since there is no way to tell whether a similar distribution of beliefswould hold for any given set of real-world employees, we make no

claim to universal generality regarding our results. We only claimthat the Minimum Wage Contract that we implemented is a plau-sible proxy for unionization and that our results can be interpretedaccordingly.

l of Socio-Economics 41 (2012) 445– 454 449

2

dToddorancsobak

CwMetpuepto

wpCajtaMth

erupwapTslWsdgs

impboC

1 2 3 4 5 6 7 8 9 10 11 120

0.2

0.4

0.6

0.8

1

Fraction of Employees Accepting Employer Offer

Fraction of Employees Taking Minimum Wage Option

Fraction of Employees Rejecting Employer Offer

S.M. Flynn, M. Donnelly / The Journa

. Incomplete contracts followed by complete contracts

Our main experiment runs 12 periods. It begins with six periodsuring which employers can only offer incomplete labor contracts.hese are followed by six periods during which employers can onlyffer complete labor contracts. All players are made aware in theirections that one employer is matched with one employee for theuration of each period, with new random parings made at the startf each period. Employers and employees were kept in separateooms to ensure anonymity between employers and employees,nd within each room partitions were used so that subjects couldot see the actions taken by their fellow players. As a further pre-aution to ensure anonymity between employers and employees,ubjects were randomized to playing the role of either an employeer an employer before being given directions about where to assem-le for the experiment. This allowed us to have the two groupsssemble in separate areas so that they were at no time able tonow the identity of anyone in the other group.

During the first six periods, each employer can only offer a Bonusontract to the employee with whom she is paired in each round,hile during the final six periods, each employer can only offer aonitoring Contract to the employee with whom she is paired in

ach round. Throughout all 12 periods, however, employees havehe choice of opting for the Minimum Wage Contract that acts as ourroxy for unionization. The constant availability of this proxy fornionization throughout all 12 periods allows us to see if employ-es are more or less likely to opt for unionization during the first sixeriods in which employers can only offer incomplete labor con-racts or during the final six periods during which employers cannly offer complete labor contracts.

It is important to understand that the subjects playing the gameere at no time told that it had anything to do with unionization. Inarticular, the directions make no mention of the Minimum Wageontract being utilized as a proxy for unionization. This was done tovoid triggering any biases (either positive or negative) that sub-ects might have had about unions. It should also be noted thathe Minimum Wage Contract was the only contract that was givenn explicit name in the directions. Both the Bonus Contract and theonitoring Contract were explained—but not named—in the direc-

ions. This was also done to avoid any framing effects that mightave affected game play had we assigned names to these contracts.9

We ran two trials of the main experiment, 2 weeks apart. Thexperiments were done with pen and paper in order to make theesults directly comparable to those of Fehr et al. (2007), who alsosed a pen-and-paper implementation rather than linking partici-ants through a computer network. All participants were providedith an electronic hand calculator. There were 44 total participants,

ll of them Vassar College undergraduates. Each session used 22articipants, 11 acting as employees and 11 acting as employers.he participants were restricted to being non-economics majorsince previous research has found that economics students behaveess reciprocally than people who have not studied economics.10

e also excluded any students who had previously played anyimilar games. A post-experiment analysis revealed no substantial

ifferences in play during the course of the experiment either byender or by major. Since the results of the two trials were veryimilar, we will present combined results for the two trials.

9 By contrast, explicitly referring to the Minimum Wage Contract as the “Min-mum Wage Contract” in the directions may have triggered associations with

inimum wage laws. But because the Minimum Wage Contract is available in everyeriod, any feelings or associations triggered by that particular name should haveeen present in every period. As such, they should not have disturbed the likelihoodf noticing a change in behavior when switching from Bonus Contracts to Monitoringontracts at the start of period 7.10 See, for instance, Frank et al. (1993) as well as Yezer et al. (1996).

Period

Fig. 1. Employee contract choices.

Fig. 1 shows the share of the 22 employees in each of the 12periods who took each of the three possible actions available tothem: accepting the employer’s offer, opting instead for the Mini-mum Wage Contract, or choosing to be unemployed for the round.The first thing to notice is that employees in this primary experi-ment never chose to be unemployed. In all cases, employees eitheraccepted the offer made to them by their employers or optedinstead for the Minimum Wage Contract.

This paper’s most important result is that the fraction of employ-ees opting for the Minimum Wage Contract that serves as theproxy for unionization more than triples from the first six peri-ods to the last six periods. For each group of six periods, therewere 132 labor contracts completed (22 contracts in each periodfor each employer-employee pair times six periods). During thefirst six periods when employers had to offer Bonus Contracts, onlyin 16 of the 132 possible cases did employees opt for the Mini-mum Wage Contract instead of the Bonus Contract. But during thefinal six periods during which employers were restricted to offer-ing only Monitoring Contracts, employees opted for the MinimumWage Contract in 51 out of the 132 possible cases. The percentage ofcontracts in which employees opted for the Minimum Wage Con-tract therefore jumped from 12.1% when incomplete contracts werebeing offered to 38.6% when complete contracts were being offered.Since this difference is highly statistically significant (Wilcoxonsigned-rank test: z-stat = −4.18; p < 0.001), we take this as evidencestrongly consistent with Flynn’s (2005) hypothesis that labor con-tract completeness strongly influences the propensity of employeesto unionize.

The employees’ much greater propensity to choose the Mini-mum Wage Contract during the final six periods is likely due topositive reciprocity being extinguished by the Monitoring Contract,an effect noted by Fehr and Gächter (2000b). By contrast, the major-ity of subjects in experimental games in which positive reciprocityis possible do in fact demonstrate positive reciprocity—meaningthat if they feel that they are being treated generously, they recip-rocate generously. Crucially, part of being treated well is the feelingthat one is being trusted to reciprocate kind actions with kindactions.

That feeling can be conveyed under the Bonus Contract butnot under the Monitoring Contract. This is true because underthe Bonus Contract, employers offer a fixed wage combined with

the possibility of a bonus, whereas under the Monitoring Contractemployers offer a fixed wage combined with the threat of a fine.In the former, employers offer employees the opportunity to harmthem (by taking the fixed wage and then shirking) while at the

450 S.M. Flynn, M. Donnelly / The Journal of So

1 2 3 4 5 6 7 8 9 10 11 125

10

15

20

25

30

35

Period

Peb

bles

Average Offered Income to the Employee

Average Offered Profit for the Employer

Fe

stosgt

tlbiIdiMr

itbmwtttdsd

fopaiaaswep

ba

contract that was offered to them or chose to opt instead for the

ig. 2. Average incomes and profits that would have resulted each period if bothmployees and employers had exactly fulfilled the contracts offered by employers.

ame time holding out the possibility that if employees cooperate,hey will be rewarded with a bonus. Implicitly, employers are notnly showing trust in the employees not to harm them, but are alsoaying that the worst punishment they will meet out if employeesive less than the demanded effort level is to reduce or withholdhe bonus.

By contrast, an employee is threatened with a large fine underhe Monitoring Contract if he does not provide the demanded effortevel. This means that he is at no time shown the courtesy of trusty the employer, nor is he allowed an opportunity to voluntar-

ly reciprocate to the demanded effort level without compunction.nstead, he is threatened with a 1/3 chance of punishment if heoes not comply. The result—as shown by Fehr and Gächter (2000b)

n a series of experimental games involving Bonus Contracts andonitoring Contracts—is that voluntary cooperation and positive

eciprocity are eliminated when Monitoring Contracts are imposed.We believe that employee resentment of Monitoring Contracts

s responsible for the higher rate at which the Minimum Wage Con-ract is chosen during the final six periods. This belief is bolsteredy the responses that we received on exit surveys that our experi-ental subjects completed immediately after period 12 and whileaiting to be paid. Only seven out of the 22 employees considered

he situation they faced during the final six periods to be “more fair”han the situation they faced during the first six periods. In addi-ion, when asked whether they treated their employers differentlyuring the final six periods, we received comments from employeesuch as, “I treated them more as adversaries,” and, “I had a secretesire to screw them over.”

The huge change in reciprocity that takes place when switchingrom the incomplete Bonus Contracts used during the first six peri-ds to the complete Monitoring Contracts used during the final sixeriods can be seen most directly in Fig. 2, which shows the aver-ge incomes and profits that would have resulted in each periodf all employer contract offers had been accepted by employeesnd if both parties had then acted in accordance with the requestsnd promises made in each contract. For instance, during the firstix periods during which Bonus Contracts were being offered, thisould have involved the employee providing exactly the requested

ffort level, e*, and the employer following through by actuallyaying the promised bonus, b*.

Fig. 2 shows that the generosity of employer offers plummetsoth in absolute and relative terms when Monitoring Contractsre introduced in period seven. During the first six periods when

cio-Economics 41 (2012) 445– 454

Bonus Contracts were being offered, employees would have earnedan average income of 31.8 tokens per round had both parties ful-filled the offered contracts. But as soon as Monitoring Contracts areintroduced in period 7, employer generosity plummets so muchthat employees would have earned an average of just 12.9 tokensper round. This huge decline in employer generosity and fair-ness almost undoubtedly caused at least some employees to growresentful.

Employee ire during the final six rounds was also probablystoked by the size of the fines with which they were threatened.Recall that employers had the option of setting the potential finefor shirking at any value from zero to 13 tokens. Low values couldhave potentially been used by employers as a way to signal non-adversarial generosity to their anonymous employees. It was thecase, however, that employers went almost entirely for large fines,with the average potential fine being 12.1 tokens and the modalfine being 13 tokens. Low potential fines were extremely rare,with there only being five observations of a fine of five or fewertokens (out of 132 contract proposals made during the final sixperiods).

These facts indicate that employers under the Monitoring Con-tract behaved in a non-reciprocal fashion in so far as they didnot appear to attempt to utilize a low or zero potential fine as away of signaling trust or generosity. What is more, this behaviorbegan as soon as Monitoring Contracts were introduced: the aver-age fine proposed during period 7 was 11.4 tokens. This means thateven after participating in 6 rounds of largely reciprocal behav-ior with employees, employers immediately switched to a morenon-trusting, non-reciprocal stance when the nature of the laborcontracting regime switched from incomplete to complete con-tracting. In addition, since employers move first each period, theirbehavior in period 7 implies that they acted in this way as soon asthe new game was introduced—setting huge fines in period 7 pre-emptively before even seeing how employees might react to low orzero fines. It is our assumption that their doing so was interpretednegatively by employees, who most likely interpreted it as non-trusting, non-reciprocal behavior on the part of employers. If so,their dissatisfaction presumably grew worse in subsequent peri-ods, as the average potential fine trended upward from 11.4 tokensin period 7 to 12.7 tokens in period 12.

The higher unionization rates during the final six periods areindicative of increased dissatisfaction and reduced cooperationon the part of employees when offered Monitoring Contracts. Animportant consequence of this change is that employer profits aregreatly reduced relative to what they had been during the first sixperiods, even in cases where employees accepted the employer’soffer. In Fig. 3, this shows up clearly if one examines the trianglesthat plot out the average employer earnings in each round for thoseemployers in each period whose contract proposals were acceptedby their respective employees. For the first six periods during whichBonus Contracts are offered, average employer profits never fallbelow 22.9 tokens in any period and range up to a maximum of30.9 tokens in period 3. By contrast, average profits for acceptedcontracts are at or below zero for all of the six periods during whichMonitoring Contracts are in effect.

The fall in employer profits is too large to be ascribed just tothe 10-token monitoring fee that must be incurred during the finalsix periods during which employers can only offer Monitoring Con-tracts. The real problem for employers is that effort levels fall offprecipitously under Monitoring Contracts. This can be seen clearlyin Fig. 4, which gives average demanded and actual effort lev-els each period broken down by whether employers accepted the

Minimum Wage Contract.A quick scan of the solid circles that give the average actual effort

levels put forth in each period by the employees who accepted the

S.M. Flynn, M. Donnelly / The Journal of Socio-Economics 41 (2012) 445– 454 451

1 2 3 4 5 6 7 8 9 10 11 12-20

-10

0

10

20

30

40

50

60

Period

Peb

bles

Avg. Employee Income if Min Wage Chosen

Avg. Employee Income if Offer Accepted

Avg. Employer Profit if Min Wage Chosen

Avg. Employer Profit if Offer Accepted

Fig. 3. Employee incomes and employer profits, separated into cases where theemployer’s contract offer was accepted and those in which employees opted insteadfor the minimum wage contract.

1 2 3 4 5 6 7 8 9 10 11 121

2

3

4

5

6

7

8

9

10

Avg. Demanded Effort Across All Offers

Avg. Demanded Effort for Accepted Offers

Avg. Demanded Effort for Offers Rejected in favor of Min WageAvg. Actual Effort for Accepted Offer

Avg. Actual Effort for Offers Rejected in favor of Min Wage

csiudifet

spat

p

1 2 3 4 5 6 7 8 9 10 11 120

0.2

0.4

0.6

0.8

1

Period

Peb

bles

Share of Employee s Accep ting Employer Off er

Share of Employee s Taking Minimum Wage Option

Share of Employee s Rejecting Employer Off er

Period

Fig. 4. Actual and demanded effort levels.

ontracts that were offered to them shows that effort levels wereubstantially higher during the Bonus Contract periods than dur-ng the Monitoring Contract periods. In fact, the average effort levelnder contracts that were accepted was 6.4 units of actual efforturing the first six periods but only 3.0 units of actual effort dur-

ng the final six periods. It is this difference that causes the sharpall-off in employer profits even when employers are able to offermployees a contract that the employees find more enticing thanhe (always available) Minimum Wage Contract.

Even worse for employers is the fact that the amount of effortupplied by employees who accepted contracts during the final six

eriods was, in the majority of cases, less than the three units ofctual effort that they would have been required to provide underhe Minimum Wage Contract.11 Interestingly, some employers

11 Employees accepted 81 of the 132 contracts that were offered to them duringeriods 7–12. In 46 of the 81 cases, employees provided an actual effort level of less

Fig. 5. Employee contract choices when treatment order is reversed.

appear to have noticed this behavior; they reacted by makingoffers that were so bad that employees felt compelled to opt for theMinimum Wage Contract. Indeed, one employee wrote on her exitsurvey that she was annoyed that the employers developed thisstrategy to “force me” to opt for the Minimum Wage. That beingsaid, this strategy allowed employers to guarantee themselves aprofit of 10 tokens rather than the negative profits typical underaccepted Monitoring Contracts.

Along these lines, it should be pointed out that Berg and Hoffrage(2008) present a theoretical explanation for why employers mightprefer a contracting regime that ignores information about effortlevels. Their work provides an intuition for why the payoffs to adecision maker are sometimes maximized by ignoring informa-tion even if that information is costlessly available (as it is herewhen employers choose to offer Monitoring Contracts rather thanBonus Contracts). Thus, the low profits that result from monitor-ing workers in this experiment are not the only theoretical channelfor explaining why we observe so many employers opting againstMonitoring Contracts (despite the fact that they provide moreinformation about what workers are doing). The extension of thisintuition to the real world implies an additional theoretical channel(beyond reduced profits) for explaining why incomplete contractsare so commonly observed in labor markets.

3. Reversing the treatment order: ill will is permanent

In this section, we report the results of a secondary experimentin which we reverse the treatment order so that complete Moni-toring Contracts precede incomplete Bonus Contracts. Our resultssuggest that once reciprocity between employers and employees ishobbled by six periods of complete contracts, it cannot be reestab-lished simply by switching to incomplete contracts. The ill willgenerated under complete Monitoring Contracts lingers and poi-sons labor relations even after the switch is made to incompleteBonus Contracts.

This is seen most directly in Fig. 5, which reports the fraction of

employees in each round who either accept the contract offered tothem, opt for the Minimum Wage contract, or reject totally. Duringthe first six periods, employers can only offer Monitoring Contracts,

than 3; in 4 cases, they chose an actual effort level equal to 3; and in 31 cases theychose an actual effort level greater than 3.

4 l of So

wC

spiiMfido

eiwociMwopabioi3ot

tahtraocimcmbtltc

ieiNebiaa

paaet

52 S.M. Flynn, M. Donnelly / The Journa

hile during the final six periods, employers can offer only Bonusontracts.12

When encountering the contract types in this order, there is noignificant difference between the first six periods and the final sixeriods in terms of the propensity of employees to opt for the Min-

mum Wage Contract. This result differs considerably from whats seen in Fig. 1, in which the rate at which employees opt for the

inimum Wage contract more than triples from 12.1% during therst six periods when Bonus Contracts are being offered to 38.6%uring the final six periods when Monitoring Contracts are beingffered.

Here, with the treatment order reversed, the rate at whichmployees opt for the union proxy rises from 24.1% when Monitor-ng Contracts are being offered during the first six periods to 31.5%

hen Bonus Contracts are being offered during the final six peri-ds. This difference is not, however, statistically significant, and,rucially, the percentage opting for the Minimum Wage contracts the same in periods 6, 7, and 8, indicating that the switch from

onitoring Contracts to Bonus Contracts that takes place startingith period 7 does not cause a sudden change in the propensity to

pt for the union proxy. Furthermore, learning seems to be takinglace during the first six periods, with employees during periods 1nd 2 apparently reluctant to utilize the Minimum Wage Contractefore taking greater advantage of it in periods 3–6. Consequently,

f one focuses on periods 3–6 and then compares them with peri-ds 7–12, one finds that the rate at which the union proxy is useds virtually identical, with 30.6% of the 36 contract offers in periods–6 being rejected in favor of the union proxy compared with 31.8%f the 54 contract offers in periods 7–12 being rejected in favor ofhe union proxy.

We take this behavior to indicate that complete labor con-racts cause lasting damage to the labor–management relationshipnd believe that it is further evidence consistent with Flynn’sypothesis that labor contract completeness is an important fac-or behind unionization and the differences seen in unionizationates across different industries. This interpretation is, however,

refinement of Flynn’s original hypothesis, which asserts that anbservation of a given industry’s current state of labor contractompleteness has a strong influence on how likely it is that thendustry is heavily unionized. The results of our secondary experi-

ent suggest that observing only the current state of labor contractompleteness is not sufficient. History matters here, so that weust observe not only the completeness of current labor contracts

ut also the completeness of past labor contracts as well. In par-icular, employees who have previously worked under completeabor contracts may still have a strong preference for unioniza-ion even if they are currently working under incomplete laborontracts.

The fact that high rates of unionization persist after the switchs made to incomplete contracts appears to be a type of “afterglowffect” similar to those noted in other studies in which behav-ors persist despite changes in contracting regimes. For instance,albantian and Schotter (1997) report on subjects who, after havingxperienced free-riding by other subjects in a voluntary contri-ution game, continue to exert low effort levels even if group

ncentive schemes like team competition or group monitoringre subsequently introduced. Gneezy and Rustichini (2000) find

similar effect in a field study at day care centers when small

12 There was only one session of this treatment order. It involved 18articipants—nine randomized to roles as employees and nine randomized to roless employers. All instructions were the same as in the primary experiment reportedbove and none of the participants in this game had participated in the earlierxperiment or had seen any similar labor market games. The only innovation inhis experiment was the reversal of the treatment order.

cio-Economics 41 (2012) 445– 454

monetary fines for picking up children late were introduced andthen removed. The change in behavior precipitated by the intro-duction of fines persisted even after the fines were subsequentlyremoved. We suggest that an afterglow effect may also be takingplace here.

4. Summary and discussion

In an experimental game that mimics a firm shifting fromincomplete labor contracts to complete contract labor contractsand in which a proxy for unionization is always available, the rate atwhich employees opt for the unionization proxy more than tripleswhen the shift is made from incomplete to complete labor con-tracts. This behavior is consistent with Flynn’s (2005) hypothesisthat employees will be more likely to unionize if they work undercomplete labor contracts because such contracts eliminate positivereciprocity between employers and employees. A second experi-ment, in which the treatment order is reversed so that completecontracts precede incomplete contracts, demonstrates that the illwill generated by complete labor contracts appears to linger. Ifcomplete contracts come first, they generate a high propensity tounionize that does not dissipate even if incomplete labor contractsare subsequently introduced.

To the extent that the manufacturing sector is more prone toutilizing complete labor contracts than the service sector, theseresults provide an explanation for why the manufacturing sec-tor tends to be more heavily unionized than the service sector.And under the same assumption, they also provide a deeperunderstanding of the decline of unionism in recent decades. Thisdecline has often been taken as a natural consequence of theongoing shift in employment from the heavily unionized manu-facturing sector to the sparsely unionized service sector (Farberand Western, 2001). What our results demonstrate is why themanufacturing sector is more heavily unionized than the servicesector.

Along these lines, it is interesting to note that after doing exten-sive interviews with many real-world managers, Bewley (1995)summarized their beliefs about the importance of maintainingpositive reciprocity and good feelings with employees as follows:“Managers claim that employees have so many opportunities totake advantage of employers that it is not wise to depend on coer-cion and financial incentives alone as motivators.” If this is true,then the results of the experiments presented here suggest thatmanagers who are free to choose their labor contracting regimesshould, other things equal, opt for incomplete labor contracts pre-cisely because they do not depend on “coercion and financialincentive alone.” But, given that the ill effects of complete con-tracting appear to be largely irreversible, managers should also notbe naïve. If they have already utilized complete contracts with agiven set of workers, simply switching to incomplete contracts willnot by itself restore positive reciprocity or reduce the propensity ofworkers to unionize.

We should not, however, oversell our results or our interpre-tation. While we believe that workers dislike monitoring in mostcases because it can be experienced as a degrading power inequalitythat also signals distrust, there could in fact be instances in whichworkers might prefer more monitoring to less. As pointed out by ananonymous referee, workers doing piece work might prefer occa-sional random quality checks over no monitoring because such aregime would allow workers substantial autonomy while also pro-viding a mechanism to observe and reward the quality of individual

worker’s outputs.

It should also be acknowledged that we would be going too farif we were to argue that real-world decisions about monitoringwere driven solely by the effect that monitoring has on profits. As

l of So

aoc

bNerttcipferrw

tavclet

Rmasfobctdaabe

dpFnfoiiuBw

aTtfarsecu

w

S.M. Flynn, M. Donnelly / The Journa

n anonymous referee put it, “Monitoring may be a componentf contracts for historical reasons and not necessarily because anyompetitive pressure forces an industry there.”

Along those lines, it is also important to state clearly ourelief that monitoring does not directly cause disutility to workers.either do we believe that it directly provides utility to employ-rs. As a result, the presence or absence of monitoring in anyeal-world situation is likely to depend on a wide variety of fac-ors including historical traditions, the availability of monitoringechnologies, and whether monitoring in any particular situationomes to be viewed positively or negatively by those being mon-tored. And, in the same vein, it seems sensible to imagine theossibility of multiple equilibria over a contracting set rangingrom low-monitoring to high-monitoring. Thus, while the employ-es in our experiment reacted decisively against the monitoringegime that we utilized, a much wider variety of monitoringegimes (and worker reactions) could be expected in the realorld.

Our experimental design could be improved upon in four impor-ant ways. First, our design abstracts entirely from the collectivection aspects of unionism. It is not clear if the distaste that indi-idual workers show in our experiment toward complete laborontracts would translate into a greater propensity to join col-ectively with other disgruntled workers to form a union. Anxperimental design that allows for collective action would resolvehis question.

Second, authors including Andreoni et al. (2003), Fehr andockenbach (2003), and Fehr and Schmidt (2007) have found asym-etries in the incentive effects of rewards and punishments. As

result, we must be cautious in interpreting the results pre-ented here because the incomplete Bonus Contract only allowsor rewards (bonuses) while the complete Monitoring Contractnly allows for punishments (fines). A way around this woulde to design an experiment under which both the incompleteontract and the complete contract would have the same incen-ive/disincentive types (i.e., bonuses and/or fines). One possibleesign would be to pair the (incomplete) Bonus Contract studiedbove with a complete contract that allows for a bonus—perhaps

complete contract in which the bonus is awarded automaticallyy the experimenter when the actual effort level is greater than orqual to the desired effort level.

Third, a future experiment should also be designed to clearlyistinguish between the effects of wage rents and contract com-leteness on the propensity to unionize. This is necessary becauselynn (2005) only addresses the possibility that contract complete-ess influences the propensity to unionize. A plausible alternative

actor that could affect the propensity to unionize is the paymentf wage rents. It is possible that the propensity to choose the Min-mum Wage Contract that we have reported on above for workersn our experiments was driven not by resentment over workingnder a complete labor contract but by differences between theonus Contract and the Monitoring Contract in terms of expectedage rents.

Fourth, a future experiment could attempt to identify the rel-tive importance of contractual completeness versus wage rents.his could be accomplished by comparing the results of onereatment in which incomplete contracts yield no wage rentsor workers (for example, they are taxed away) with those ofnother treatment in which complete contracts do yield wageents (for instance, the government or the employer pays a sub-idy to all workers irrespective of actual effort provided). Such anxperiment would be able to distinguish between wage rents and

ontract completeness as drivers of the propensity of workers tonionize.

While these four modifications of our experimental designould help to clarify the generalizability of our results, it should

cio-Economics 41 (2012) 445– 454 453

be noted that these proposed experiments—as well as the exper-iment reported upon here—are all based upon an earlier body ofwork relating to fairness and contract designed that is reviewedand championed by Fehr et al. (2007). That body of work, how-ever, has been criticized by Berg and Gigerenzer (2010) andBerg (2010) as failing to base itself upon the actual psycho-logical and mental processes that determine choice behavior.Under this critique, the experiment reported upon here, as wellas the previous work upon which it builds, are fatally depen-dent on Friedman’s (1953) contention that a model only needmake reliable predictions—that it is OK if the model fails to accu-rately describe the process choice behavior just as long as its “asif” assumptions lead to accurate predictions about the choicesmade.

In response to such criticisms (which were mentioned by ananonymous referee), we would like to point out that the experimentreported upon here is in fact a contribution to our attempts to bet-ter understand the actual motivations of employers and employees.Berg (2006) points out that there has been very little work doneon unions in the subfield of behavioral labor economics. We viewour results as an important first step in developing more psycho-logically realistic models of labor–management interactions. Bycreating experiments in which subjects can express a wide range ofbehaviors and then seeing what behaviors actually result, we willdevelop the body of knowledge necessary to—hopefully—constructmodels that not only make good predictions but also make thembased upon accurate representations of the thought processes fol-lowed by both sides of labor contracts.

References

Akerlof, G., 1982. Labor contracts as a partial gift exchange. Quarterly Journal ofEconomics 97 ((November) 4), 543–569.

Akerlof, G., Yellen, J., 1990. The fair wage–effort hypothesis and unemployment.Quarterly Journal of Economics 105 ((May) 2), 255–283.

Andreoni, J., Harbaugh, W., Vesterlund, L., 2003. The carrot or the stick: rewards,punishments, and cooperation. American Economic Review 93 ((June) 3),893–902.

Berg, N., 2006. Behavioral labor economics. In: Altman, M. (Ed.), Handbook of Con-temporary Behavioral Economics. M.E. Sharpe, New York, pp. 457–478.

Berg, N., 2010. Behavioral economics. In: Free, R. (Ed.), 21st Century Economics: AReference Handbook, vol. 2. Sage, Los Angeles, pp. 861–872.

Berg, N., Gigerenzer, G., 2010. As-if behavioral economics: neoclassical economicsin disguise? History of Economic Ideas 18 (1), 133–166.

Berg, N., Hoffrage, U., 2008. Rational ignoring with unbounded cognitive capacity.Journal of Economic Psychology 29, 792–809.

Bewley, T., 1995. A depressed labor market as explained by participants. AmericanEconomic Review 85 ((May) 2), 250–254.

Farber, H.S., Western, B., 2001. Accounting for the decline of unions in the privatesector, 1973–1998. Journal of Labor Research 22 ((July) 3), 459–485.

Fehr, E., Gächter, S., 2000a. Fairness and retaliation: the economics of reciprocity.Journal of Economic Perspectives 14 ((Summer) 3), 159–181.

Fehr, E., Gächter, S., 2000b. Do incentive contracts crowd out voluntary cooperation?Institute for Empirical Research in Economics, University of Zurich, WorkingPaper No. 34.

Fehr, E., Alexander, K., Schmidt, K., 2007. Fairness and contract design. Econometrica75 ((January) 1), 121–154.

Fehr, E., Rockenbach, B., 2003. Detrimental effects of sanctions on human altruism.Nature 422 ((March) 6928), 137–140.

Fehr, E., Schmidt, K., 1999. A theory of fairness, competition, and cooperation. TheQuarterly Journal of Economics 114 ((August) 3), 817–868.

Fehr, E., Schmidt, K., 2007. Adding a stick to the carrot? The interaction of bonusesand fines. American Economic Review 97 ((May) 2), 177–181.

Flynn, S.M., 2005. Why only some industries unionize: insights from reciprocitytheory. Journal of Institutional Economics 1 ((May) 1), 88–120.

Frank, R., Gilovich, T., Regan, D., 1993. Does studying economics inhibit cooperation?Journal of Economic Perspectives 7 ((January) 1), 159–171.

Friedman, M., 1953. Essays in Positive Economics. University of Chicago Press,Chicago.

Gneezy, U., Rustichini, A., 2000. A fine is a price. Journal of Legal Studies 29 ((January)1), 1–17.

Leibenstein, H., 1987. Inside the Firm: The Inefficiencies of Hierarchy. Harvard Uni-versity Press, Cambridge.

MacLeod, W.B., Parent, D., 1999. Job characteristics and the form of compensation.In: Polachek, S.W. (Ed.), Research in Labor Economics, vol. 18. JAI Press, London,pp. 177–242.

4 l of So

M

N

P

54 S.M. Flynn, M. Donnelly / The Journa

artin, I.T., 1915. The melting pot at Ford: conditions under the new profit shar-ing system. Official Journal of the Carriage, Wagon, and Automobile Employees

Union 4 ((August) 8).

albantian, H.R., Schotter, A., 1997. Productivity under group incentives: an exper-imental study. American Economic Review 87 ((June) 3), 314–341.

rendergast, C., 1999. The provision of incentives in firms. Journal of EconomicLiterature 37 ((March) 1), 7–63.

cio-Economics 41 (2012) 445– 454

Raff, D., 1988. Wage determination theory and the five dollar day at Ford. Journal ofEconomic History 48 ((June) 2), 387–399.

Slichter, S., 1919. The Turnover of Factory Labor. D. Appleton and Company, NewYork.

Yezer, A.M., Goldfarb, R.S., Poppen, P.J., 1996. Does studying economics discouragecooperation? Watch what we do, not what we say, or how we play. Journal ofEconomic Perspectives 10 ((Winter) 1), 177–186.