Wage Rigidity in a Competitive Incomplete Contract Market

Author(s): Ernst Fehr and Armin Falk

Source: Journal of Political Economy , Vol. 107, No. 1 (February 1999), pp. 106-134
Published by: The University of Chicago Press
Stable URL: https://www.jstor.org/stable/10.1086/250052

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access
to Journal of Political Economy

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
Wage Rigidity in a Competitive Incomplete
Contract Market

Ernst Fehr and Armin Falk

University of Zurich

Do employers and workers underbid prevailing wages if there is

unemployment? Do employers take advantage of workers’ under-
bidding by lowering wages? We hypothesize that under conditions
of incomplete labor contracts, wage levels may positively affect
workers’ propensity to cooperate. This, in turn, may prevent firms
from underbidding or accepting the underbidding of workers. Ex-
perimental double auctions conducted for the purpose of examin-
ing these hypotheses yield the following results: (i) Workers’ un-
derbidding is very frequent, but employers refuse to accept
workers’ low wage offers in markets with incomplete labor con-
tracts. However, in the presence of complete labor contracts, em-
ployers accept and actively enforce wages close to the competitive
level. (ii) Workers’ effort is positively related to the wage level.
Therefore, wage cutting is costly for the employer if workers have
discretion over their effort level. This holds true even in the pres-
ence of explicit performance incentives.

I. Introduction
Do workers underbid prevailing wages if there is unemployment?
Do employers take advantage of underbidding by lowering wages?
Financial support by the Swiss National Science Foundation (project 12-43590.95)
is gratefully acknowledged. We would like to thank two anonymous referees and
Robert Topel for very helpful comments. The paper also has benefited from discus-
sions at meetings of the Econometric Society, the Economic Science Association, the
European Economic Association, and the Eastern Sociological Association. Valuable
research assistance was provided by Isabelle Busenhart, Martin Brown, Urs Fisch-
bacher, Philip Molitor, Jürg de Spindler, Jean-Robert Tyran, Paolo Vanini, and Edith
Wolfram. We are especially grateful to Simon Gächter for many discussions.

[Journal of Political Economy, 1999, vol. 107, no. 1]

 1999 by The University of Chicago. All rights reserved. 0022-3808/99/0701-0005$02.50

106

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 107
These are the main questions of this paper and of many theories of
wage rigidity.1 Recently performed questionnaire studies with own-
ers and managers of firms suggest that employers are unwilling to
cut wages in the presence of unemployment (Blinder and Choi 1990;
Agell and Lundborg 1995; Bewley 1995, 1997; Campbell and Kam-
lani 1997). Some of these studies even indicate that employers do
not take advantage of workers’ underbidding (Agell and Lundborg
1995; Bewley 1995). Although unemployed workers frequently seem
to be willing to work for less than the going wage, firms report that
they refuse to employ underbidders. According to these studies, a
major reason for firms’ refusal to cut wages is the fear that pay cuts
will adversely affect work morale. In Bewley’s study (1997, chap. 13,
p. 12), for example, there are many indications of owners’ and man-
agers’ reluctance to cut wages. They fear that pay cuts ‘‘express hos-
tility to the work force’’ and will be ‘‘interpreted as an insult.’’ As
a consequence, they were thought to ‘‘provoke hostility’’ against the
firm, with negative consequences for performance. Similar findings
are presented in Agell and Lundborg (1995) and Campbell and
Kamlani (1997).
The findings in the studies mentioned above suggest that, con-
trary to the assumptions of the standard competitive model, the wage
is affected not only by supply and demand conditions but also by its
impact on workers’ effort behavior. In this paper we examine this
possibility in the context of experimental double auctions. Since the
pioneering paper by Smith (1962), a large body of subsequent re-
search has proved the striking competitive properties of experimen-
tal double auctions. It has been shown that under a wide variety of
conditions, double auctions converge quickly and reliably to the
competitive equilibrium (Smith 1982; Plott 1989). This has been
confirmed and reported in hundreds of laboratory double-auction
markets and can be considered as one of the best-established facts
of experimental economics. Even if there are as few as only four
sellers or if ‘‘zero-intelligence’’ traders (traders with simple prepro-
grammed trading rules) are used, the competitive outcome is
reached (Smith 1982; Gode and Sunder 1993). The double auction
generates higher competitive pressure and market efficiency than
any other known institution used by experimental economists

1
Among these theories are the implicit contract theory as developed by, e.g., Aza-
riadis (1975), the insider-outsider theory of Lindbeck and Snower (1988), job search
models as reviewed in Mortensen (1986), and the different versions of the efficiency
wage hypothesis as reviewed in Katz (1986). For important criticisms of those ver-
sions of the efficiency wage approach that rely on standard assumptions about work-
ers’ preferences, see Carmichael (1985, 1990) and Murphy and Topel (1989, 1990).

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
108 journal of political economy
(Friedman and Rust 1993).2 In view of these facts, it seems fair to
say that double-auction experiments provide one of the most rigor-
ous and persuasive pieces of evidence in favor of the standard com-
petitive model. Yet this also means that if one can isolate factors
that prevent convergence to the competitive equilibrium in a double
auction, one has probably identified an important obstacle for con-
vergence.
What are the factors that could lead to nonconvergence? The
above-mentioned studies suggest that many owners and managers
of firms seem to anticipate that workers’ motivation is affected by
considerations of reciprocity and fairness. Managers believe that em-
ployees interpret pay cuts as a hostile (i.e., unfair) action, which in
turn triggers hostile responses from the employees.3 That reciprocity
and perceptions of fairness are potentially important determinants
of motivation and work behavior is also suggested by a large litera-
ture in organization theory (see, e.g., Cropanzano and Folger 1991;
Mowday 1991; Steers and Porter 1991; Lawler 1994) and psychology
(e.g., Adams 1963, 1965; Argyle 1989; Cialdini 1993). Reciprocity
motives seem to affect behavior even among strangers. This is indi-
cated by a large number of experimental studies on ultimatum game
bargaining (Güth, Schmittberger, and Schwarze 1982; Thaler 1988;
Roth 1995; Camerer and Thaler 1995)4 and trust games (Fehr, Kirch-
steiger, and Riedl 1993; Fehr et al. 1994; Berg, Dickhaut, and
McCabe 1995; Miller 1997; Falk, Gächter, and Kovács 1998).5
2
‘‘The striking competitive tendency of the double auction institution . . . has
been confirmed by at least a thousand market sessions in a variety of designs’’ (Holt
1995, p. 370). Most double auctions have been conducted under conditions of pri-
vate information about payoff functions. For the impact of public information, see
Sec. III below.
3
Bewley (1997, chap. 9) interprets his findings in terms of reciprocity: ‘‘Such
reciprocity seems to be real and important . . . . [It] seems to be one of the main
factors making morale important, and perception of the violation of reciprocity is
probably a component of the harm done by pay cuts’’ (p. 11).
4
In an ultimatum game, a proposer has to propose a division of a fixed amount
of money. The responder can accept or reject the proposed division. If the re-
sponder rejects it, both players get nothing; if the responder accepts it, each player
receives the proposed shares. If both players are rational and selfish money maximiz-
ers, the prediction is that proposers demand the whole amount of money (less 1
cent), whereas responders are willing to accept any positive share. Yet responders
reject offers that are well above zero (but below 25 percent) with a very high proba-
bility. Low offers are considered as an insult or a hostile act, which in turn triggers
hostility in the form of a rejection. Proposers anticipate this and offer, therefore,
on average, 40 percent of the available amount of money.
5
In the trust game of Berg et al. and Miller, e.g., first movers are endowed with
an amount of money x, which they can keep or transfer to the second mover. Any
amount y transferred will be tripled. The second mover then has to decide on a
countertransfer z, 0 ⱕ z ⱕ 3y. If both players are rational and selfish money maximiz-
ers, the predicted outcome is y ⫽ z ⫽ 0. However, if second movers are motivated
by positive reciprocity, even selfish first movers may have an incentive to transfer

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 109
The stylized facts of ultimatum bargaining games indicate that
many people are willing to punish behavior that they view as hostile
or unfair even if this punishment is costly for them. This behavior
can be called negative reciprocity because hostile actions trigger hos-
tile responses. Trust games, in contrast, indicate the presence of a
behavior that can be termed positive reciprocity. Positively reciprocal
behavior is based on a willingness to pay in order to reward actions
that are perceived as generous, kind, or fair: Generous actions trig-
ger cooperative responses even if they are costly for the responder.
It is worthwhile to point out that negative as well as positive reciproc-
ity has been observed in games with rather high stake levels. Cam-
eron (1995) reports results that show that negative reciprocity affects
the results of ultimatum games even if the available money amounts
to three months’ income in a single one-shot game. Fehr and Toug-
areva (1995) confirmed the presence of positive reciprocity in an
environment in which subjects earned three times their monthly in-
comes in an experimental session that lasted for two hours.
What are the conditions that allow reciprocal motivations to play
a role in the work process? In our view an essential condition is the
existence of incomplete labor contracts; that is, the obligations of
the employer and the employee are not specified in each possible
state of the world. This incompleteness has been stressed, for exam-
ple, by Williamson (1975, 1985) and more recently by Milgrom and
Roberts (1992) and Hart (1995).6 In practice, incomplete labor con-
tracts often take the form of a fixed-wage contract 7 without explicit
performance incentives and a considerable degree of worker discre-
tion over the work effort. The absence of explicit performance in-
centives can be viewed as a rational response to the difficulties of
measuring and verifying a worker’s performance in a multitask envi-
ronment (Holmström and Milgrom 1991; Baker 1992). If a worker
should perform n tasks and in some of these tasks explicit incentives
are not possible because performance is difficult to measure and

some money. This is also what is observed in the experiments. Most first movers
choose a positive transfer y in anticipation of a positive countertransfer z, and most
second movers in fact choose a positive countertransfer z.
6
Milgrom and Roberts neatly describe the incompleteness of the labor contract
as follows: ‘‘The employment contract is typically quite imprecise. The employees
agree that—within limits that are rarely completely described and only partly under-
stood—they will use their minds and muscles to undertake the tasks that the em-
ployer directs them to do. The employer agrees to pay the employees. The range
of actions that might be requested or required is unclear’’ (p. 329). In the absence
of incomplete contracting, it is, in fact, not clear why there should be firms at all.
This viewpoint is forcefully put forward by Grossman and Hart (1986) and Hart
(1995).
7
By a fixed wage we mean a monthly or weekly or hourly wage rate that does not
directly depend on the performance level.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
110 journal of political economy
verify, it is more efficient to provide no explicit incentives for the
other tasks, too. Under conditions of incompletely specified obliga-
tions and only weak or no explicit performance incentives, a work-
er’s general job attitude, or what Williamson (1985, p. 262) called
‘‘consummate cooperation,’’ becomes important. ‘‘Consummate
cooperation is an affirmative job attitude whereby gaps are filled,
initiative is taken, and judgment is exercised in an instrumental
way.’’ If it were possible to specify, verify, and enforce a course of
action for each potentially arising gap, to specify, verify, and enforce
‘‘initiative’’ and ‘‘good judgment,’’ a generally cooperative job atti-
tude would be superfluous because all relevant actions could be en-
forced by a complete contract. However, managers are aware of the
fact that in the absence of complete contracts ‘‘workers have so many
opportunities to take advantage of employers that it is not wise to
depend on coercion and financial incentives alone as motivators’’
(Bewley 1995, p. 252).
In our view it is the requirement of a generally cooperative job
attitude that renders reciprocal motivations important. That reci-
procity motives may have a powerful impact on effort decisions and
the efficiency of trades has been shown in a recent paper by Fehr,
Gächter, and Kirchsteiger (1997).8 The results of this paper as well
as the work from psychology and organization theory cited above
suggest that a firm can build up or sustain a cooperative attitude if
its actions are considered as fair. On the other hand, it seems quite
likely that a cooperative attitude is destroyed if a firm’s actions (e.g.,
wage cuts) are perceived as hostile. Therefore, firms may not want
to cut wages because of the implications for workers’ performance.
Notice that if reciprocity motives provide a channel for firms to af-
fect workers’ cooperative propensities, firms’ unwillingness to cut
wages may be an efficient response that increases the total gains
from trade.
Whether reciprocity motives are sufficiently strong to affect firms’
wage setting in a competitive market is an empirical question. To
examine this question we have conducted competitive double-
auction experiments. The essential feature of an experimental dou-
ble auction is that both workers and firms can propose wages and
can accept wages proposed by the other side of the market. Workers
can, thus, freely accept low wage offers made by the firms or they
can make low wage offers themselves. Since both sides can initiate
a process of underbidding, this trading institution is particularly well
suited for studying the existence and the impact of underbidding
on wage formation.
8
Similar results are reported in Fehr, Kirchsteiger, and Riedl (1993).

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 111
Almost all double-auction experiments conducted so far imple-
mented complete contracts.9 Since we argued above that many labor
relations are characterized by incomplete contracts, which renders
workers’ reciprocal motivations and the associated cooperative atti-
tudes potentially important, we conducted double auctions with
complete and incomplete contracts. In our main treatment condi-
tion, contracts were incomplete and there were two stages. At the
first stage, wages were determined under double-auction trading
rules. If a pair of traders agreed on a wage level, a labor contract
was concluded. Contractual incompleteness was captured by the fea-
ture that the effort level could not be stipulated in the contract. At
the second stage, workers had to make a costly effort decision. The
effort cost function implied that a worker who wanted to maximize
his or her monetary payoff had to choose the minimal effort level.
However, in principle, workers could also exhibit a cooperative be-
havior by choosing nonminimal (i.e., more costly) effort levels.
In the main treatment, workers could thus vary their effort levels.
This provided the opportunity to vary the effort choice in response
to the wage level, that is, to make reciprocal choices. In the control
treatment, the effort decision at the second stage was removed. In
this treatment, reciprocal effort choices were, therefore, ruled out.
By comparing wages in the two treatment conditions, we are able to
test for the impact of contractual incompleteness and the associated
reciprocation opportunities on bidding behavior and wage forma-
tion. Our most important results are the following: (i) Workers mas-
sively underbid but firms refuse to accept low offers. Firms’ wage bids
are, on average, higher than workers’ wage offers, although firms
know that workers are willing to work for lower wages. (ii) Workers
choose low effort levels in response to low wages and high effort
levels in response to high wages; that is, wage levels positively affect
workers’ propensity to cooperate. Therefore, firms’ high-wage strat-
egy increases total revenue and renders both workers and firms bet-
ter off. (iii) In the control treatment, where effort is exogenously
fixed, firms take advantage of workers’ underbidding. In addition,
firms are themselves actively involved in enforcing low wages. As a
result, in the control treatment, wages are much lower than in the
main treatment. Results i, ii, and iii indicate that firms did not want
to enforce lower wages when effort was an endogenous variable be-
cause they anticipated that wages had a significant impact on work-
ers’ willingness to cooperate.
These results were obtained in an environment in which workers

9
An exception is Lynch et al. (1991). In a previous version of this paper we discuss
the differences between Lynch et al. and our experimental design in detail.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
112 journal of political economy
had the opportunity to exhibit positive reciprocity by responding to
high wages with high (i.e., more costly) effort levels. Therefore, the
results do not show the relevance of negative reciprocity for wage
formation. To study the impact of negative reciprocity, we con-
ducted additional experiments. As before there was a main treat-
ment condition in which effort levels were determined by the work-
ers and a control treatment in which effort was exogenously fixed.
The distinguishing feature of the ‘‘new’’ main treatment was a modi-
fied cost function for the workers. Under the modified cost function
a money-maximizing worker would always choose the maximum ef-
fort level because lower levels are more costly. This cost function cap-
tures a situation in which a firm’s enforcement technology ensures
that workers will in general put forward a relatively high effort level.
In this situation there is, therefore, no need for firms to elicit positive
reciprocity. Yet there is still the danger that wage cuts upset workers
and trigger negative reciprocity, that is, submaximal effort levels.
The modified main treatment allows us to study the impact of reci-
procity under conditions of relatively strong monetary incentives for
high performance. If it turns out that workers in fact choose low
effort levels in response to low wages, although this is costly for them,
and if this gives rise to downward wage rigidity, the case for reciproc-
ity as a determinant of wages is strengthened.
The results of the modified main treatment indeed show that
workers respond to low wages with submaximal effort levels. Their
negatively reciprocal behavior induces firms to pay significantly
higher wages in the main treatment than in the control treatment.
Thus, even if firms are capable of providing performance incentives
that will in general ensure high effort levels, they are reluctant to
force wages to competitive levels if workers have some opportunity
to ‘‘retaliate’’ via their effort choices.
Our paper is organized as follows: In Section II we describe our
experimental design in more detail. Section III derives the predic-
tions. Section IV presents and interprets our results and discusses
links to the above-mentioned survey studies. Concluding remarks
are presented in Section V.

II. Experimental Design

In total we conducted four double-auction sessions with incomplete
labor contracts (henceforth called ‘‘auction with effort’’). As a con-
trol treatment we conducted two double-auction sessions with com-
plete contracts in which effort was exogenously fixed and only the

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 113
first stage of the two-stage design was implemented.10 In each session
we had 18 experimental subjects. Before each session started they
were randomly divided into two groups, seven firms and 11 workers,
and were then assigned to two different rooms. They were given their
instructions and had to answer several control questions to check
their understanding of the experimental procedures.11
At the first stage of an auction with effort session, seven firms and
11 workers traded for jobs with each other. Trading time was 4 min-
utes per period. Both firms and workers were free to submit and
accept bids and offers at any time during the trading period. Bids
and offers had to obey the improvement rule. In each room all wages
were written on two blackboards (one for workers’ offers and one
for firms’ bids). Every worker and every firm always knew all offers
and bids that had been submitted so far. We used two telephone
lines to transmit offers/bids into the other room. Whenever a bid
or an offer was accepted, a contract was concluded. In each period
a firm could employ only one worker and each worker could accept
only one job. We thus had an excess supply of four workers in each
period. No firm was ever informed about the identity of the worker with
whom it had traded, and vice versa. When all seven firms had concluded
a contract or when trading time was over, the market was closed and
the second stage began.
In our auction with effort treatment, contractual incompleteness
was captured by the feature that effort was not contractible. At the
second stage, after the closing of the market, all the workers who
had concluded a contract had to choose an effort level e. The choice
of the effort level was completely private and was revealed only to the firm
with which the worker was matched. To ensure the privacy of the effort
decision, subjects sat remotely from each other. This information
condition ruled out that group pressure could affect the effort
choices of workers.
The first stage of our auction with effort sessions represents a dou-
ble auction in the sense that both workers and firms could propose
wages. In double auctions in which a homogeneous good is traded,
there is usually the requirement that only the best price offers of
the other side can be accepted. In the context of our auction with
effort sessions, this would have implied that a worker could accept
only the highest current wage bid by firms whereas a firm could

10
The following description concerns only the experiments that aimed at isolating
the impact of positive reciprocity on wage formation. We provide a description of
the experiments that isolate negative reciprocity in Sec. IVE.
11
Instructions and control questions are available on request.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
114 journal of political economy
TABLE 1
Effort Levels and Costs of Efforts

Effort Level e
.1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
Effort cost c(e) 0 1 2 4 6 8 10 12 15 18

accept only the lowest going wage offer by workers. However, since
effort is a noncontractible variable in our setting, the wage offered
is no unambiguous indicator of the overall profitability of a contract.
If, for example, low (high) wage offers are systematically associated
with low (high) effort levels, accepting a higher wage may be more
profitable for firms. Moreover, since we were interested in studying
whether workers would underbid the going wage and whether firms
would accept such low wages, it would have been useless to force
firms to accept the lowest wage. Therefore, we allowed firms to ac-
cept any current wage offer. To maintain symmetry we also allowed
workers to accept any going wage bid. Firms’ payoff in experimental
money units (guilders) was given by
π ⫽ (v ⫺ w)e. (1)
The redemption value v of an employed worker was 120; w denotes
the wage paid to the employed worker and e the effort level chosen
by the worker. Notice that wage costs varied with the chosen effort
level. Since wage bids had to obey w ⱕ v, this feature of the payoff
function ruled out losses.12 Yet we would like to stress that the com-
petitive prediction remains the same compared to the more com-
mon profit equation π ⫽ ve ⫺ w.
A worker who accepted a wage offer had to bear fixed costs c 0 of
20. The payoff function in guilders of an employed worker who
chose an effort level of e was therefore given by
U ⫽ w ⫺ c(e) ⫺ c 0 , (2)
where c(e) denotes the effort costs. The range of feasible effort levels
and the associated effort costs are shown in table 1.
For both workers and firms, one guilder represented 0.10 Swiss
francs (⬇ U.S. $0.08). At the end of the experiment the total income
in guilders was converted into Swiss francs and paid out in cash. Both
firms and workers earned a profit of zero if they did not conclude
12
It is a well-known fact that loss aversion can affect behavior (Tversky and Kahne-
man 1991). The payoff function (1) was implemented to prevent the interaction
between loss aversion and fairness effects.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 115
a contract. All payoff functions were common knowledge. Both sides
were thus able to determine the implications of their choices for
the payoffs of their trading partners. This information condition was
essential to allow for fairness considerations. If cost and profit func-
tions had been private information, there would have been no refer-
ence standard against which (un-) fair behavior could have been
evaluated.13
In order to study the impact of opportunities for reciprocity on
wage formation, we conducted two control sessions. These sessions
consisted only of the first stage, the double-auction labor market.
Because no effort level could be chosen, the labor contract was com-
plete. The payoffs in the control treatment were given by
π ⫽ 120 ⫺ w (3)
for the firms and
U ⫽ w ⫺ 20 (4)
for the workers. If someone did not trade, payoffs were zero. Again,
payoff functions were common knowledge.14 By comparing the wage
outcome in the control sessions with the outcome in the auction
with effort sessions, one can isolate the impact of endogenous effort
choices on wage formation. In order to study convergence proper-
ties of wages and effort levels, we implemented the common method
of stationary replication; that is, we replicated each constituent game
10 times per session. To make subjects familiar with the bidding
rules, we started each session with one training period (only the first
stage), which did not get remunerated.

III. Predictions
With purely selfish money maximizers the prediction of the market
outcome is straightforward. In the market without effort the compet-
itive equilibrium wage is given by w ⫽ 20. In the market with effort it
is also given by w ⫽ 20 because rational firms anticipate that money-
maximizing workers will always choose e ⫽ 0.1. Therefore, the com-
petitive wage prediction in both sessions is identical.

13
The fact that payoff functions were common knowledge does not imply that sub-
jects’ payoffs were common knowledge because effort choices were known only by
the parties involved in a trade. This means that workers (firms) knew only their own
income and the income of the trading partner. They did not know the payoffs of
other firms or other workers.
14
Since in the control sessions effort was set equal to one, each guilder repre-
sented 0.06 Swiss francs (⬇ U.S. $0.05). This ensured that total average earnings
were similar across treatments.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
116 journal of political economy
In view of the questionnaire evidence and the experimental evi-
dence cited in the Introduction, it is, however, likely that many peo-
ple are not motivated solely by selfish objectives. Note that in the
competitive equilibrium described above, firms reap the whole gains
from trade whereas workers earn nothing. Moreover, since payoff
functions are common knowledge, all market participants know that
at w ⫽ 20 an extreme earnings inequality prevails. In our view the
combination of an extreme earnings inequality with public informa-
tion about payoff functions opens the way ‘‘for ‘equity consider-
ations’ to modify self-interest choices’’ (Smith 1976, p. 278). Smith
already speculated that such equity considerations might slow down
the convergence to the competitive equilibrium.15
In the meantime there is more evidence available that shows that
the combination of an extreme earnings inequality with public infor-
mation about payoff functions may prevent full convergence to the
competitive equilibrium in markets with complete contracts. For ex-
ample, in the experiments conducted by Kachelmeier and Shehata
(1992), the long side of the market is capable of seizing between 13
and 17 percent of the surplus after 10 periods, although according
to the competitive prediction, they should receive nothing. How-
ever, the results of Kachelmeier and Shehata also indicate the strong
drawing power of the competitive equilibrium in markets with com-
plete contracts. If the authors switch from an excess demand to an
excess supply regime, the buyers’ share of the surplus quickly in-
creases from roughly 16 percent to 90 percent.16
In our view the big difference between markets with complete and
markets with incomplete contracts is that in the latter equity and
fairness considerations can play a much bigger role. In these mar-
kets, reciprocally motivated workers can ‘‘punish’’ firms that pay low
wages with low effort choices. Contractual incompleteness ensures
a kind of private property right in punishment. In contrast, in mar-
kets with complete contracts, a worker can try to punish a firm only
by refusing to trade with the firm. Yet if other workers are willing
to trade with low-wage firms, the refusal to trade is a futile attempt
to punish them. Under complete labor contracts, fairness and equity
considerations are thus likely to have a much smaller impact on
wages. This argument has been made precise recently by Fehr and
Schmidt (1997). These authors provide a rigorous proof of the lim-

15
Smith’s conjecture is based on an experiment similar to our control design. In
the final period (i.e., period 4) of his experiment, workers (sellers) were still able
to earn 27 percent of the available surplus, although in equilibrium they should
earn only 4 percent.
16
Kachelmeier and Shehata conducted experiments in China and in Western
countries. We refer here to the auction results generated by Western subjects.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 117
ited impact of fairness and equity considerations in markets with
complete contracts. They show that although fairness preferences
may give rise to Nash equilibria above the competitive equilibrium,
the least fair player is decisive for the Nash equilibrium outcome. In
contrast, under incomplete contracts, the least fair player is not
more influential than the other players. If this argument is correct,
we should observe systematically higher wages in auction with effort
sessions relative to the control sessions.

IV. Results and Discussion

All participants of the experiment were students of the University
of Zurich or the Swiss Federal Institute of Technology. All of them
received $12.50 as a show-up fee plus their earnings in the experi-
ment. The two-stage sessions lasted about 2.5 hours, and a partici-
pant earned, on average, between $34 (firms) and $40 (workers),
including the show-up fee. In the control sessions, which lasted
about 1.5 hours, participants earned, on average, between $18
(workers) and $49 (firms), including the show-up fee. In compari-
son to the usual earnings in economic experiments, the average
earnings of our subjects were quite high. As a result they were well
motivated and took the experiments very seriously. In the four auc-
tion with effort sessions, there were 280 potential trades; in the two
control sessions, there were 140 potential trades. The actual number
of trades in the auction with effort sessions was 280. In the control
sessions, workers refused to accept offered wages in 11 cases; that
is, there were 129 actual trades.

A. The Drawing Power of the Competitive Equilibrium

Remember that if all agents are rational and selfish, we should not
observe any significant difference between average wages in the auc-
tion with effort sessions and the two control sessions. In both treat-
ments wages should converge rather quickly toward the competitive
equilibrium wage of w ⫽ 20 and should be indistinguishable from
each other over time. Yet if the extreme earnings inequality in this
equilibrium evokes fairness and equity considerations, wages may
not fully adjust to w ⫽ 20 in the control sessions. In addition, and
more important, wages in the auction with effort sessions may be
considerably above wages in the control sessions because recipro-
cally motivated workers can punish low-wage firms with low effort
levels in the auction with effort sessions.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
118 journal of political economy
In table 2 and figure 1 we present the behavior of wages in both
treatment conditions. As one can see, wages in the auction with ef-
fort (AE) sessions are far above wages in the control (A) sessions.
There is not a single period in which wages in the control treatment
exceed wages in the auction with effort treatment. Moreover, over
time the difference in the average wage across treatments even in-
creases. While the average wage decreases from 39 to 34 in the con-
trol sessions, it increases from 54 to 65 in the auction with effort
sessions. The fact that in period 1 there is already a large wage gap,
which even increases over time, is consistent with our view that con-
tractual incompleteness allows for a much bigger impact of fairness
and equity considerations on wages. The fact that wages in the con-
trol sessions are much lower and decrease over time indicates that
the competitive equilibrium still exerts a considerable drawing
power under complete contracts.17 The upward trend of wages in
the auction with effort sessions, however, casts serious doubts on
the drawing power of the competitive equilibrium under incomplete
contracts.

B. Do Workers Underbid?
To what extent is the apparent wage rigidity in the auction with ef-
fort sessions due to a lack of underbidding on the workers’ side?
According to Solow (1990), the existence of a fairness rule or a social
convention may deter the unemployed from underbidding. How-
ever, the results of our experiments show that Solow’s conjecture
does not hold in a double-auction environment. In our experiments
it is not a lack of underbidding that hinders wages from falling. We
observe massive underbidding by the workers. In figure 2a we show work-
ers’ underbidding behavior in session AE4. All other auction with
effort sessions show the same qualitative pattern.18 Except for the
first few periods, there is always a large number of wage offers below
contract wages in these sessions. This reveals severe competition for
scarce jobs among workers. Underbidding is thus a strong and persis-
tent phenomenon in our auction with effort treatment. The absence
of underbidding can therefore be ruled out as a reason for the ap-
parent rigidity of wages.

17
Note that our results in the control sessions are in line with the results in Kachel-
meier and Shehata (1992). In the final period of their experiments, the long side
of the market was capable of seizing between 13 and 17 percent of the available
surplus. In our control sessions, workers receive 14 percent of the available surplus.
18 The pattern of all auction with effort sessions is presented in a previous version

of this paper, which is available on request.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
 TABLE 2
Mean Wages

Auction without Effort

Auction with Effort Sessions Sessions
1 2 3 4 1–4 1 2 1–2
Mean wage over all periods 59.21 61.64 61.15 62.51 61.13 32.05 37.76 34.96
(10.96) (13.1) (15.14) (15.11) (13.68) (6.32) (4.32) (6.09)
Mean wage of the last period 61.57 67.71 56.85 72.51 64.67 29.28 38.71 34.00
(10.69) (8.18) (24.20) (8.01) (14.88) (1.11) (1.60) (5.06)
Note.—Standard deviations are in parentheses.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
120 journal of political economy

Fig. 1.—Evolution of average wages in the auction with effort (AE1–4), the auc-
tion without effort (A1–2), and the bilateral case.

Workers’ underbidding in the auction with effort sessions indi-

cates that firms could have enforced lower wages than they actually
paid. To achieve a reduction in wages, they would have had to accept
only the available low wage offers. It seems that they did not want
to enforce lower wages; that is, they voluntarily paid workers above
their reservation wages. To show this in more detail, we have com-
puted an upper bound for each worker’s reservation wage. On the
basis of our data we measure the upper bound for worker i in each
period by min(w a, w o ), where w a denotes a bid accepted by worker
i and w o denotes i ’s lowest wage offer. Notice that min(w a, w o ) gives
us only an upper bound for the reservation wage because a worker
may in fact accept less. The average of the seven lowest upper
bounds in each period of an auction with effort session indicates
the average wage that (the seven) firms could have enforced in this
period with certainty.19 The gap between actual average wages and

19
Each of the seven firms has traded or could have traded (if it wanted) at one
of these upper bounds.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
 Fig. 2.—Workers’ offers and mean contract wages: a, in the auction with effort
session AE4; b, in the auction without effort session A1.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
122 journal of political economy

Fig. 3.—Mean contract wages and upper bounds for workers’ reservation wages

this average upper bound shows, therefore, that firms voluntarily

paid workers above their reservation wages. As figure 3 reveals, there
is always a positive gap in these sessions.
This is, however, not yet the full story because our control treat-
ment provides further information about workers’ reservation
wages. The relation between workers’ wage offers and actual wage
payments differs radically across treatments. While the majority (71
percent) of workers’ offers in the auction with effort treatment lie
below the average contract wage, almost all offers in the control treat-
ment lie above the average contract wage. This difference is illus-
trated by the comparison between figures 2a and 2b, which are both
representative of the data pattern in the respective treatments. More-
over, as figure 3 shows, the upper bound of workers’ reservation
wages is much lower in the control sessions. This is a further indica-
tion that firms in the auction with effort sessions voluntarily paid
higher wages because there is no reason why workers’ reservation
wages should be different across treatment conditions. Figure 3 fi-
nally shows that there is almost no difference between actual wages
and our measure of the upper bound on workers’ reservation wages

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 123
in the control sessions.20 Thus, in these sessions, firms took advan-
tage of workers’ low wage offers.
Taken together, figures 2 and 3 provide strong evidence that firms
voluntarily paid workers above their reservation wages in the auction
with effort sessions. In contrast, in the control sessions, they took
advantage of workers’ underbidding and were actively pushing
wages toward workers’ reservation levels. This is further evidence
that the competitive equilibrium exerts a strong drawing power in
the control sessions, whereas in the auction with effort sessions its
drawing power is severely weakened or, perhaps, even absent.
Both results—frequent underbidding and the refusal of firms to
employ underbidders—confirm what different authors have found
in their survey studies. Bewley (1995, 1997) finds no support for So-
low’s conjecture of an absence of underbidding. Agell and Lund-
borg (1995) report that underbidding is not ‘‘all that uncommon’’
(p. 298). Even though Swedish unemployment rates used to be very
low, 43 percent of firms had at least once encountered underbidding
blue-collar workers and 53 percent of firms had experienced under-
bidding white-collar workers. However, this is only part of the story.
While both studies report underbidding behavior of unemployed
workers, they also mention that firms refuse to employ underbidders.
In Agell and Lundborg, for example, firms that had encountered
underbidding blue-collar workers refused to accept their offers in
95 percent of the cases. Underbidding white-collar workers, on the
other hand, were rejected by 82 percent of the firms.

C. Does Underbidding Cause a Wage Decrease?

Even though workers’ underbidding did not seem to have an impact
on wage formation in the auction with effort sessions, we cannot
completely rule out that it had some wage-decreasing effect. To test
for the wage-decreasing effect of underbidding, we can, however,
compare our results with those of Fehr et al. (1994). Fehr et al. have
reported the data of experiments that differ in one respect from our
auction with effort treatment. Instead of a double auction, the wage
is determined by a bilateral interaction between an exogenously
matched firm-worker pair: The firm proposes a wage w. Then the

20
In periods 3, 5, 8, and 9 the upper bound is slightly larger than mean contract
wages because fewer than seven workers traded in these periods. Our measure of
the upper bound is the average of the seven lowest values of min(w a, w o ), whereas
the mean contract wage is the average over actually concluded contracts. Therefore,
when there are fewer than seven trades, the upper bound may exceed the mean
contract wage.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
124 journal of political economy
worker accepts or rejects it. If he accepts it, he has to choose an
effort level e ; if he rejects it, both earn nothing. Each subject is
matched with 10 different subjects from the other side. Payoff func-
tions are common knowledge and are given by equations (1) and
(2) above. All parameters are set as in our auction with effort treat-
ment. Thus, while in Fehr et al. there is no competition at all, in
our double auction, both firms and workers can underbid.
By comparing the wages in Fehr et al. with wages in our auction
with effort treatment, we are therefore able to check whether under-
bidding has any wage-decreasing effect. Figure 1 shows that wages
in the bilateral treatment are, on average, close to wages in the auc-
tion with effort treatment. While they are higher during the first few
periods of the sessions, they are even somewhat lower from period
5 onward. Thus, after a few periods, underbidding has no wage-
decreasing effect. This suggests that in the auction with effort sessions
the competitive equilibrium has no drawing power.

D. Why Do Firms Refuse to Hire Underbidders?

In our view, it is workers’ effort behavior that explains firms’ unwill-
ingness to employ low-wage workers. As we shall see, a majority of
workers made reciprocal effort choices. Yet there also was a signifi-
cant minority who behaved in a completely selfish way. At the aggre-
gate level this mixture of reciprocal and selfish effort choices gave
rise to a significantly positive aggregate wage-effort relation.
To examine reciprocity at the individual level, we have computed
the (nonparametric) Spearman rank correlation ρ(w, e) between w
and e for each worker. It turns out that 10 of 44 workers always chose
the minimum effort irrespective of the wage they received. One
worker always chose the maximum effort. Hence, 11 workers (25
percent) have ρ(w, e) ⫽ 0. On the other hand, 29 workers (66 per-
cent) exhibit a positive correlation; for all but one of these workers
ρ(w, e) ⬎ .2 holds. Four workers have a slightly negative correlation.
Because of the small number of observations, for many workers the
correlation is not significantly positive for all those with a positive
ρ. Despite this, the correlation is significantly positive at the 10 per-
cent level for 18 workers (43 percent).21 Taken together, these data
suggest that reciprocal effort behavior is the dominant pattern at
the individual level.

21
The small number of individual observations is due to the excess supply of work-
ers. Notice that, by definition, ρ(w, e) cannot be significant below the 5 percent
level for fewer than five observations. There are, e.g., several workers with ρ ⬎ .8,
but because n ⬍ 5, the correlation is not significant.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 125

Fig. 4.—Evolution of average effort for given wage intervals

The aggregate wage-effort relation is displayed in figures 4 and

5. Figure 4 depicts the evolution of average effort over time for dif-
ferent wage intervals. It shows that for wages between 20 and 45, the
average effort was between 0.1 and 0.2; for wages in the interval [46,
65], workers chose, on average, 0.3; whereas for wages above 65, the
average effort was slightly below 0.5. In figure 5 we show a scatter
plot of the wage-effort combinations in AE1–4 together with a re-
gression line. Since there are multiple observations, for given wage-
effort combinations the scatter plot exhibits different symbols, which
indicate the number of observations at each combination. Figure 5
indicates that, on average, higher wages are associated with higher
effort levels.
Formal evidence for the existence of a positive wage-effort rela-
tion at the aggregate level is provided by table 3, which shows the
results of several two-sided censored Tobit regressions (censoring
occurs because e ⬍ 0.1, and e ⬎ 1 was ruled out by the design).
Regression model 1 is given by the simple equation
α ⫹ βw ⫹ ⑀ if 0.1 ⬍ RHS ⬍ 1
e⫽
冦 0.1
1
if RHS ⱕ 0.1
if RHS ⱖ 1,
(5)

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
 Fig. 5.—The effort-wage relation: scatter plot and regression line (model 2)

TABLE 3
Tobit Regression Results for Auction with Effort Sessions 1–4

Independent
Variable Model 1 Model 2 Model 3 Model 4 Model 5
Constant ⫺.0522 ⫺.2744 ⫺.2892
(.0653) (.1833) (.1849)
Wage .00640 .0147 .0149 .0051 .0056
(.0010) (.0065) (.0065) (.0005) (.0011)
Wage squared ⫺.000073 ⫺.000076
(.000056) (.000056)
Period .0030 .00061
(.0052) (.0025)
Wald statistic 10E⫹10 8.29
p(W ) .0000 .6000
Note.—Dependent variable is effort. The Wald statistic refers to the null hypothesis that all dummy vari-
ables in model 4 or model 5, respectively, are equal to the constant in a model without dummies. Model 4
allows for individual fixed effects, and model 5 includes period-dependent intercepts. p(W ) is the p-value
for the corresponding Wald statistics. The number of observations is 280. Standard errors are in parentheses.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 127
where ⑀ denotes the normally distributed error term, and RHS
(right-hand side) is defined as RHS ⫽ α ⫹ βw ⫹ ⑀. Model 2 adds
squared wages to equation (5). Model 3 includes, in addition, the
period of observation as an explanatory variable. Since w 2 always
turned out to be highly insignificant, we did not include it in models
4 and 5. Model 4 allows for individual fixed effects, whereas model
5 allows for period-dependent intercepts.
The most important result of table 3 is that the coefficient on
wages is always highly significant and positive. In addition, the regres-
sions suggest that the positive wage-effort relation is not weakened
over time. This follows from the fact that the coefficient on the pe-
riod is always nonnegative and insignificant. Moreover, the Wald
statistic for H 0 that all period-dependent intercepts in model 5 are
equal to the constant in model 1 does not reject H 0 (see col. 5 in
table 3). This also indicates that the position of the wage-effort
schedule does not change over time. Model 4 allows us to test for
individual differences. The p-value of the Wald statistic for H 0 that
all individual intercepts are equal to the constant in an equation
without individual dummies is below .001. Thus H 0 is clearly re-
jected, which indicates individual differences in effort choices.
Firms that believe and experience that there is a positive relation
between wages and effort have a reason to pay high wages. It is there-
fore a natural question whether firms’ wage policy was optimal, given
workers’ effort behavior. Figure 6 provides the relevant information to
answer this question. It shows firms’ average profits and the relative
frequency of trades in different wage intervals. Most important, the
fact that many workers responded reciprocally to the actual wage
payment made it profitable for firms not to enforce wages in the
interval [20, 35]. If firms paid wages in this interval, they received
effort levels close to e ⫽ 0.1. Their profit was, therefore, little higher
than 10. Notice that only 6 percent of all trades took place in this
interval. Yet if they paid wages in the interval [66, 75], they earned,
on average, more than 20, which is more than twice the competitive
equilibrium payoff of 10. Figure 6 shows that firms did pretty well
in maximizing their payoff: 32 percent of the observations are in the
profit-maximizing wage interval [66, 75] and 23 percent are in the
next-best interval [56, 65]. Moreover, the relative frequency of
trades decreases the lower the average profit earned in an interval.22
This evidence suggests that firms’ behavior was quite rational.

22
The Spearman rank correlation between average profits in a given wage interval
and the relative frequency of trades in this interval is .964 (p ⬍ .0001).

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
128 journal of political economy

Fig. 6.—Wage-profit relation (relative frequency of observations above bars)

E. Wage Formation under Negative Reciprocity

So far our examination has focused on the impact of positive reci-
procity on wage formation. Since positive reciprocity affects workers’
cooperative attitudes, it is likely to be important under conditions of
incomplete labor contracts and weak performance incentives. Such
conditions arise if firms face the difficulties of measuring and veri-
fying workers’ performance in a multitask environment (Holmström
and Milgrom 1991; Baker 1992). The impact of reciprocity motives
on wage formation is, however, not necessarily confined to this envi-
ronment. Negative reciprocity, in particular, may also be important
in an environment in which firms are endowed with a relatively pow-
erful enforcement technology. Remember that negative reciprocity
means that somebody is even willing to bear some cost to punish
actions that are viewed as hostile. The language used by the manag-
ers in Bewley’s interview study indicates that they were afraid of work-
ers’ negative reciprocity. They feared that wage cuts ‘‘express hostil-

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 129
TABLE 4
Cost Associated with Different Effort Levels in the Modified Design

Effort Level e
.1 .2 .3 .4 .5 .6 .7 .8 .9 1.0
Effort cost c(e) 18 15 12 10 8 6 4 2 1 0

ity to the work force’’ and are ‘‘interpreted as an insult’’ by the

workers, which in turn triggers their hostility.
To examine the impact of negative reciprocity on wage formation,
we conducted double auctions with a modified cost function. In con-
trast to the previous design, the costs associated with a worker’s effort
are now lowest at the maximum effort level (see table 4). Conse-
quently, a money-maximizing worker will always choose the maxi-
mum effort.
This cost function can be interpreted as follows: It not only repre-
sents a worker’s ‘‘subjective’’ effort cost but incorporates, in addi-
tion, firms’ enforcement technology. This enforcement technology
renders the choice of the maximum effort optimal for a nonrecipro-
cal, money-maximizing subject and imposes increasing pecuniary
costs on shirking workers.23
To isolate the impact of negative reciprocity on wage formation,
we compared a modified main treatment in which effort was endoge-
nous with a control treatment in which effort was no choice variable.
In the modified main treatment, monetary payoffs from a trade were
given by π ⫽ ve ⫺ w and U ⫽ w ⫺ c(e) ⫺ 20, with v ⫽ 120 and c(e)
as given in table 4. In the control treatment, the payoff functions
were the same but e was fixed at the maximum level of e ⫽ 1. As in
the previous design, there was an excess supply of four workers, and
the same information conditions prevailed. In total we conducted
two experimental sessions for each treatment condition.
23
A simple example may illustrate how this cost function can be constructed. Sup-
pose that a worker’s subjective effort cost x(e) is a convex and increasing function
of e and that the probability of being caught performing below the maximum effort
e 0, σ(e 0 ⫺ e), is a rising and strictly convex function of the deviation e 0 ⫺ e, with
σ(0) ⫽ 0 and σ′(0) ⬎ 0. Suppose further that f ⬎ 0 is the pecuniary loss imposed
on the worker if he is caught shirking. The total expected costs c(e) associated with
e are then given by c(e) ⫽ x(e) ⫹ σ(e 0 ⫺ e)f. If c ′(e 0) ⫽ x′(e 0) ⫺ σ′(0)f ⬍ 0, an
expected cost function with the same qualitative properties as the function in table
4 emerges. Instead of having a fixed penalty f, one could also assume that f in-
creases with e 0 ⫺ e. Notice that in the experiments we did not implement a probabi-
listic cost function. This has the advantage that e ⬍ e 0 causes a sure loss. Therefore,
the choice of e ⬍ e 0 cannot be attributed to risk seeking but is a clean expression
of negative reciprocity.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
130 journal of political economy

Fig. 7.—Evolution of average wages in the auction with effort sessions with nega-
tive reciprocity and in the auction without effort sessions.

In the presence of money-maximizing subjects the predictions for

both treatment conditions are again straightforward. Since workers
will always choose the maximum effort in the modified main treat-
ment, the wage outcome in both treatment conditions should be
the same. However, if wage cuts are indeed interpreted as an insult
triggering workers’ hostility, we should observe submaximal effort
levels at low wages in the modified main treatment. As a conse-
quence, firms may constrain wage cutting. Figure 7 indicates that
this in fact happened. If workers had the opportunity to choose sub-
maximal effort levels, average wages were substantially higher in each
period. The difference in the average wage is significant according
to the nonparametric Mann-Whitney and median tests at all conven-
tional significance levels (p ⬍ .0001). Moreover, while in the control
condition there is a downward trend in wages over time, no such
trend can be observed in the main treatment: wages in period 10
are as high as wages in period 1.
Space limitations prevent us from presenting the results of the
modified experiments in the same detail as the results of the previ-
ous design. However, apart from the fact that it is now negative reci-

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 131
procity that drives the results, the qualitative data pattern is very simi-
lar. Tobit and ordinary least squares regressions of effort on wages
show that there is a significantly positive wage-effort relation. This
relation is stable across periods, and as in the previous design, there
are significant individual differences. In the main treatment we
again observe much underbidding by the workers, but in the face
of reciprocal effort choices, firms refuse to accept low wage offers.
In the control treatment, however, firms make wage bids close to
the competitive equilibrium wage and do not hesitate to accept
workers’ low offers. These regularities suggest that workers’ negative
reciprocity generates downwardly rigid wages in the main treatment,
whereas no such rigidity is present in the control treatment.

V. Concluding Remarks
Labor relations are frequently characterized by incomplete labor
contracts and weak performance incentives. The absence of explicit
performance incentives can be viewed as a rational response of firms
to the difficulties of measuring and verifying performance in a
multitask environment. In practice, incomplete labor contracts of-
ten take the form of a fixed-wage contract without explicit perfor-
mance incentives and a considerable degree of workers’ discretion
over their work effort. Under these conditions, workers have many
opportunities to take advantage of employers, and therefore, firms
have an interest in creating and maintaining a generally cooperative
attitude among the workers. It is hypothesized that fairness and reci-
procity considerations are an important determinant of workers’
propensity to cooperate and that wage cuts may cause a significant
decrease in workers’ willingness to cooperate with the firm. As a con-
sequence, firms may be reluctant to cut wages even if they face an
excess supply of workers.
To examine whether people’s willingness to cooperate is indeed
affected by the wage level, we have conducted experimental double
auctions with complete and incomplete contracts. Double auctions
are well known for their striking competitive properties. They consti-
tute, therefore, a particularly hostile environment for the emer-
gence of reciprocity effects. By comparing the wages in markets with
complete labor contracts with wages in markets with incomplete la-
bor contracts, we are able to isolate the impact of reciprocation op-
portunities on wage formation.
It turns out that workers’ effort levels are indeed positively related
to the wages paid. This positive wage-effort relation prevails when
there are weak and when there are strong pecuniary performance
incentives. As a consequence, firms face a cost when they reduce
their wages, which gives rise to downwardly rigid wages in the market

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
132 journal of political economy
with incomplete contracts. Despite the fact that there is an immense
amount of underbidding on the workers’ side, firms refuse to accept
workers’ low wage offers in this market. The positive wage-effort rela-
tion also implies that higher wages increase the total gains from
trade. Therefore, the refusal to cut wages is efficiency enhancing for
those who are involved in the trade. In the market with complete
contracts, however, firms accept workers’ underbidding so that
wages come close to the competitive level. Our analysis thus suggests
that in the absence of complete labor contracts, reciprocity motives
become important in two ways: First, they severely restrict the impact
of underbidding on wage formation, and, second, they are a deter-
minant of the total gains from trade for those who are involved in
the trade.

References
Adams, J. Stacy. ‘‘Wage Inequities, Productivity and Work Quality.’’ Indus.
Relations 3 (October 1963): 9–16.
———. ‘‘Inequity in Social Exchange.’’ In Advances in Experimental Social
Psychology, vol. 2, edited by Leonard Berkowitz. New York: Academic
Press, 1965.
Agell, Jonas, and Lundborg, Per. ‘‘Theories of Pay and Unemployment:
Survey Evidence from Swedish Manufacturing Firms.’’ Scandinavian J.
Econ. 97 ( June 1995): 295–307.
Argyle, Michael. The Social Psychology of Work. 2d ed. Harmondsworth, U.K.:
Penguin, 1989.
Azariadis, Costas. ‘‘Implicit Contracts and Underemployment Equilibria.’’
J.P.E. 83 (December 1975): 1183–1202.
Baker, George P. ‘‘Incentive Contracts and Performance Measurement.’’
J.P.E. 100 ( June 1992): 598–614.
Berg, Joyce; Dickhaut John W.; and McCabe Kevin A. ‘‘Trust, Reciprocity,
and Social History.’’ Games and Econ. Behavior 10 ( July 1995): 122–42.
Bewley, Truman F. ‘‘A Depressed Labor Market as Explained by Partici-
pants.’’ A.E.R. Papers and Proc. 85 (May 1995): 250–54.
———. ‘‘A Depressed Labor Market as Explained by Participants.’’ Book
manuscript. New Haven, Conn.: Yale Univ., Cowles Found., July 31, 1997.
Blinder, Alan S., and Choi, Don H. ‘‘A Shred of Evidence on Theories of
Wage Stickiness.’’ Q .J.E . 105 (November 1990): 1003–15.
Camerer, Colin F., and Thaler, Richard H. ‘‘Ultimatums, Dictators and
Manners.’’ J. Econ. Perspectives 9 (Spring 1995): 209–19.
Cameron, Lisa. ‘‘Raising the Stakes in the Ultimatum Game: Experimental
Evidence from Indonesia.’’ Manuscript. Princeton, N.J.: Princeton Univ.,
Dept. Econ., 1995.
Campbell, Carl M., III, and Kamlani, Kunal S. ‘‘The Reasons for Wage Rigid-
ity: Evidence from a Survey of Firms.’’ Q .J.E . 112 (August 1997): 759–
89.
Carmichael, H. Lorne. ‘‘Can Unemployment Be Involuntary? Comment.’’
A.E .R. 75 (December 1985): 1213–14.
———. ‘‘Efficiency Wage Models of Unemployment—One View.’’ Econ.
Inquiry 28 (April 1990): 269–95.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
wage rigidity 133
Cialdini, Robert B. Influence: Science and Practice. 3d ed. New York: Quill
William Morrow, 1993.
Cropanzano, Russell, and Folger, Robert. ‘‘Procedural Justice and Worker
Motivation.’’ In Motivation and Work Behavior, 5th ed., edited by Richard
M. Steers and Lyman W. Porter. New York: McGraw-Hill, 1991.
Falk, Armin; Gächter, Simon; and Kovács, Judit. ‘‘Intrinsic Motivation and
Extrinsic Incentives in a Repeated Game with Incomplete Contracts.’’ J.
Econ. Psychology (1998), in press.
Fehr, Ernst; Gächter, Simon; and Kirchsteiger, Georg. ‘‘Reciprocity as a
Contract Enforcement Device: Experimental Evidence.’’ Econometrica 65
( July 1997): 833–60.
Fehr, Ernst; Kirchler, Erich; Weichbold, Andreas; and Gächter, Simon.
‘‘When Social Norms Overpower Competition.’’ Manuscript. Zurich:
Univ. Zurich, Dept. Econ., 1994.
Fehr, Ernst; Kirchsteiger, Georg; and Riedl, Arno. ‘‘Does Fairness Prevent
Market Clearing? An Experimental Investigation.’’ Q .J.E . 108 (May
1993): 437–59.
Fehr, Ernst, and Schmidt, Klaus. ‘‘A Theory of Fairness, Competition and
Cooperation.’’ Manuscript. Zurich: Univ. Zurich, Dept. Econ., 1997.
Fehr, Ernst, and Tougareva, Elena. ‘‘Do High Stakes Remove Reciprocal
Fairness—Evidence from Russia.’’ Manuscript. Zurich: Univ. Zurich,
Dept. Econ., 1995.
Friedman, Daniel, and Rust, John, eds. The Double Auction Market: Institu-
tions, Theories and Evidence. Santa Fe Inst. Series in the Science of the
Complexity, Proceedings vol. 14. Reading, Mass.: Addison-Wesley, 1993.
Gode, Dhananjan K., and Sunder, Shyam. ‘‘Allocative Efficiency of Markets
with Zero-Intelligence Traders: Market as a Partial Substitute for Individ-
ual Rationality.’’ J.P.E . 101 (February 1993): 119–37.
Grossman, Sanford J., and Hart, Oliver D. ‘‘The Costs and Benefits of Own-
ership: A Theory of Vertical and Lateral Integration.’’ J.P.E . 94 (August
1986): 691–719.
Güth, Werner; Schmittberger, Rolf; and Schwarze, Bernd. ‘‘An Experimen-
tal Analysis of Ultimatum Bargaining.’’ J. Econ. Behavior and Organization
3 (December 1982): 367–88.
Hart, Oliver D. Firms, Contracts, and Financial Structure. Oxford: Clarendon,
1995.
Holmström, Bengt, and Milgrom, Paul. ‘‘Multitask Principal-Agent Analy-
ses: Incentive Contracts, Asset Ownership and Job Design.’’ J. Law, Econ.
and Organization 7 (special issue; 1991): 24–52.
Holt, Charles A. ‘‘Industrial Organization: A Survey of Laboratory Re-
search.’’ In The Handbook of Experimental Economics, edited by John H. Ka-
gel and Alvin E. Roth. Princeton, N.J.: Princeton Univ. Press, 1995.
Kachelmeier, Steven J., and Shehata, Mohamed. ‘‘Culture and Competi-
tion: A Laboratory Market Comparison between China and the West.’’
J. Econ. Behavior and Organization 19 (October 1992): 145–68.
Katz, Lawrence F. ‘‘Efficiency Wage Theories: A Partial Evaluation.’’ In
NBER Macroeconomics Annual, edited by Stanley Fischer. Cambridge,
Mass.: MIT Press, 1986.
Lawler, Edward E. Motivation in Work Organizations. San Francisco: Jossey-
Bass, 1994.
Lindbeck, Assar, and Snower, Dennis J. The Insider-Outsider Theory of Employ-
ment and Unemployment. Cambridge, Mass.: MIT Press, 1988.
Lynch, Michael; Miller, Ross M.; Plott, Charles R.; and Porter, Russell.
‘‘Product Quality, Informational Efficiency, and Regulations in Experi-

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms
134 journal of political economy
mental Markets.’’ In Research in Experimental Economics, vol. 4, edited by
R. Mark Isaac. Greenwich, Conn.: JAI, 1991.
Milgrom, Paul, and Roberts, John. Economics, Organization, and Management.
Englewood Cliffs, N.J.: Prentice-Hall, 1992.
Miller, Sven. ‘‘Strategienuntersuchung zum Investitionsspiel von Berg,
Dickhaut, McCabe.’’ Diploma thesis, Univ. Bonn, 1997.
Mortensen, Dale T. ‘‘Job Search and Labor Market Analysis.’’ In Handbook
of Labor Economics, vol. 2, edited by Orley Ashenfelter and Richard Layard.
Amsterdam: North-Holland, 1986.
Mowday, Richard T. ‘‘Equity Theory Predictions of Behavior in Organiza-
tions.’’ In Motivation and Work Behavior, 5th ed., edited by Richard M.
Steers and Lyman W. Porter. New York: McGraw-Hill, 1991.
Murphy, Kevin M., and Topel, Robert H. ‘‘Ability Bias in Models of Earn-
ings: New Methods and Evidence.’’ Manuscript. Chicago: Univ. Chicago,
Dept. Econ., 1989.
———. ‘‘Efficiency Wages Reconsidered: Theory and Evidence.’’ In Ad-
vances in the Theory and Measurement of Unemployment, edited by Yoram
Weiss and Gideon Fishelson. London: Macmillan, 1990.
Plott, Charles R. ‘‘An Updated Review of Industrial Organization: Applica-
tions of Experimental Methods.’’ In Handbook of Industrial Organization,
vol. 2, edited by Richard Schmalensee and Robert D. Willig. Amsterdam:
North-Holland, 1989.
Roth, Alvin E. ‘‘Bargaining Experiments.’’ In The Handbook of Experimental
Economics, edited by John H. Kagel and Alvin E. Roth. Princeton, N.J.:
Princeton Univ. Press, 1995.
Smith, Vernon L. ‘‘An Experimental Study of Competitive Market Behav-
ior.’’ J.P.E . 70 (April 1962): 111–37.
———. ‘‘Experimental Economics: Induced Value Theory.’’ A.E .R. Papers
and Proc. 66 (May 1976): 274–79.
———. ‘‘Microeconomic Systems as an Experimental Science.’’ A.E .R . 72
(December 1982): 923–55.
Solow, Robert M. The Labour Market as a Social Institution. Oxford: Blackwell,
1990.
Steers, Richard M., and Porter, Lyman W., eds. Motivation and Work Behavior.
5th ed. New York: McGraw-Hill, 1991.
Thaler, Richard H. ‘‘Anomalies: The Ultimatum Game.’’ J. Econ. Perspectives
2 (Fall 1988): 195–206.
Tversky, Amos, and Kahneman, Daniel. ‘‘Loss Aversion in Riskless Choice:
A Reference-Dependent Model.’’ Q .J.E . 106 (November 1991): 1039–
61.
Williamson, Oliver E. Markets and Hierarchies: Analysis and Antitrust Implica-
tions. New York: Free Press, 1975.
———. The Economic Institutions of Capitalism: Firms, Markets, Relational Con-
tracting. New York, Free Press, 1985.

This content downloaded from

150.217.1.66 on Wed, 13 Mar 2019 01:33:53 UTC
All use subject to https://about.jstor.org/terms