EFFICIENCY WAGES AND SUBJECTIVE PERFORMANCE PAY

HUANXING YANG*

This paper studies optimal relational contracts in motivating workers in a market

setting. We find that labor markets with higher turnover costs will use more subjective
performance pay and less efficiency wages and that in those markets, the total wage
payment is lower and the equilibrium employment level is higher. Surprisingly, under
certain conditions, an increase in turnover costs leads to higher social welfare.
Incorporating workers’ search costs, we show that wages are procyclical in booms
and are either rigid or countercyclical during recessions. The predictions of the
model are consistent with some empirical evidence. (JEL D82, J33, J41, J63)

I. INTRODUCTION this paper provides a theory of contract selec-

How to design incentive schemes to moti- tion in a market setting.
vate workers is an important topic in econom- Inaseminalpaper,MacLeodandMalcomson
ics. The shirking models of efficiency wages, (1998, MM hereafter) provided a model of
such as Shapiro and Stiglitz (1984), establish contract selection between efficiency wages
that firms need to pay a wage premium (effi- and subjective performance pay. The driving
ciency wages) to motivate workers, with un- force in their model is the market condition.
employment serving as a punishment device. In a market with more workers than jobs,
However, one shortcoming of these models a firm can always immediately and costlessly
is that performance pay plays no role. One jus- fill its vacancy after reneging on the promised
tification for their omission of performance bonus. Therefore, no subjective performance
pay is that individual performance may not pay is credible, and firms have to use solely
be verifiable. Nevertheless, if workers’ perfor- efficiency wages to motivate workers. On
mance is observable and employment relation- the other hand, in a market with more jobs
ships are repeated, firms can use implicit than workers, efficiency wages are useless
bonuses or relational contracts based on in providing incentives because a worker
workers’ subjectively assessed performance can find another job immediately after being
to motivate workers. Since subjective perfor- fired. As a result, firms use solely subjective
mance pay cannot be legally enforced, it has performance pay to motivate workers.
to be self-enforcing. In MM, there are no exogenous turnover
Given that both efficiency wages and sub- costs. Complementary to MM, this paper pro-
jective performance pay motivate workers, vides a model of contract selection driven by
what is the optimal wage contract from the exogenous turnover costs in labor markets.
firm’s perspective? Will different labor mar- We focus on the situation in which unemploy-
kets (occupations) use different forms of wage ment always exists in labor markets, thus rul-
contracts? What are the impacts of different ing out market condition as a determinant of
forms of wage contracts on unemployment
and social welfare? To answer these questions,
ABBREVIATIONS
*I wish to thank two anonymous referees, Ken ALC: Average Labor Cost
Burdett, Bill Dupor, Jan Eeckhout, Paul Evans, Rafael BO: Blanchflower and Oswald
Rob, and the participants at the Fall 2003 Midwest The- GSS: General Social Survey
ory Conference for helpful comments and discussions. MM: MacLeod and Malcomson
I am truly indebted to Steven Matthews for his invaluable
advice and guidance. The usual disclaimers apply. NLSY: National Longitudinal Surveys of
Youth
Yang: Assistant Professor of Economics, Ohio State
University, 405 Arps Hall, 1945 N. High Street, WIRS: Workplace Industrial Relations
Columbus, OH 43210. Phone 614-292-6523, Fax Survey
614-292-3906, Email yang.1041@osu.edu

179
Economic Inquiry doi:10.1111/j.1465-7295.2007.00069.x
(ISSN 0095-2583) Online Early publication July 28, 2007
Vol. 46, No. 2, April 2008, 179–196 Ó 2007 Western Economic Association International
180 ECONOMIC INQUIRY

contract selection. It turns out that turnover problem and gives them commitment power,
costs affect the amount of efficiency wages which in turn grants firms more flexibility to
and performance pay in optimal contracts. alleviate the workers’ moral hazard problem.
While in MM efficiency wages and subjective A general interpretation of this result is that
performance pay cannot be used together to exogenous friction might be more efficient
motivate workers, in our model firms are than endogenously created friction (efficiency
able to use combinations of both methods wages) in overcoming double moral hazard
of payments. By affecting optimal contracts, problems in markets.
turnover costs also have impacts on the equi- In an extended model, we incorporate
librium employment level and social welfare. workers’ search costs. Now, wage contracts
Finally, our model generates rich empirical should not only motivate workers to exert
implications about the relationships among effort (IC condition) but also induce unem-
turnover costs, forms of employment con- ployed workers to search (IR condition). It
tracts, and levels of employment. turns out that inducing workers to search
Our basic model studies how turnover costs becomes more difficult as the unemployment
borne by firms affect contract selection. From rate increases. The main result in the extended
the firm’s point of view, subjective perfor- model is that wages are increasing in the
mance pay is ‘‘cheaper’’ since efficiency wages employment level only when employment is
entail a wage premium. Thus, the optimal high and are completely rigid when employment
wage contract uses the maximum amount of is low. This implies that wage-unemployment
bonus to motivate workers. However, subjec- relationship changes over the course of busi-
tive performance pay may not be credible due ness cycles: wages are procyclical in booms
to the firm’s moral hazard problem: in labor and rigid during recessions.
markets with positive unemployment, a firm Our model generates rich empirical impli-
can immediately hire a new worker after cations. First, different labor markets (occu-
reneging on the implicit bonus. The presence pations) will adopt different forms of wage
of turnover costs can alleviate this moral haz- contracts. In particular, the efficiency wage
ard problem. This is because a worker can component (wage premium) is negatively
punish a reneging firm by quitting and the firm related to and the amount of bonuses is pos-
has to incur turnover costs in hiring new work- itively related to the turnover costs borne by
ers. Therefore, in the optimal contract, the firms in labor markets. Second, workers paid
amount of bonus is increasing in turnover by bonuses on average earn less than workers
costs. Since subjective performance pay is paid fixed wages (efficiency wages). Third,
cheaper, as the turnover costs increase, more occupations paid bonuses should have lower
subjective performance pay leads to a lower unemployment rates than occupations in
total wage payment. which bonuses are seldom used. Fourth, wages
After deriving the optimal contracts, we are procyclical during booms and are either
turn to study market equilibrium, which is rigid or countercyclical in recessions. Finally,
determined by firms’ free entry condition. In the wage-unemployment elasticity is decreas-
market equilibrium, the revenue product of ing in turnover costs in labor markets. All
labor equals the average labor cost (ALC), these predictions are consistent with some
which consists of the wage payment each empirical evidence.1
period and the average turnover costs incurred Relational contracts have been studied by
per period on the equilibrium path. Interest- Bull (1987), MM (1989, 1998), Baker, Gibbons,
ingly, up to some threshold (when the wage and Murphy (1994), and Levin (2003). How-
premium is positive), an increase in turnover ever, except for MM (1989, 1998), all the other
costs reduces the ALC and leads to an increase papers study the one-firm-one-worker setting;
in the equilibrium employment level. More- hence, both the firm’s and the worker’s outside
over, when the revenue product of labor is options are exogenously given. Moreover,
elastic enough, an increase in turnover costs these papers do not study contract selection
leads to higher social welfare. This is a surpris- between efficiency wages and subjective per-
ing result: a little bit of (exogenous) friction in formance pay. MM (1989) offered a complete
markets is beneficial for social welfare. The
main reason behind this result is that friction
in markets alleviates the firms’ moral hazard 1. See Section VI for details.
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 181

characterization of the set of relational con- If employed, a worker gets utility Wt
vet,
tracts that can be implemented in a market where Wt is the total wage payment, v . 0 is
setting, but it does not study the problem of the disutility of work, and et is effort in period t.
contract selection. For simplicity, we assume that a worker can
Akerlof and Katz (1989) incorporated either work (et 5 1) or shirk (et 5 0). The profit
a performance bond into a shirking model of a filled job vacancy in period t is pet
Wt.
of efficiency wages. They, however, assumed Workers without a job receive an unemploy-
that firms are able to commit: firms never for- ment benefit u . 0 per period, which is exog-
feit a worker’s bond if he does not shirk. In enously given. Consistent with employment at
contrast to their assumption, our model, fol- will, we assume that only spot contracts are
lowing the literature of relational contracts, legally enforceable. Following the incomplete
assumes that firms are not able to commit. contract literature, we assume that et is observ-
The labor turnover models of efficiency wages, able but not verifiable.2 Therefore, spot con-
such as Salop (1979) and Stiglitz (1974), treat tracts that are contingent on et cannot be
the turnover rate as endogenous. They derive enforced by the court.
the result that firms with higher turnover costs Nevertheless, since employment relation-
may pay higher wages in order to reduce total ships have the potential to be long term, firms
turnover costs. This result is in direct contrast may use implicit bonuses. In particular, Wt
to the prediction of our model. consists of a fixed wage wt  0 that the firm
The structure of the paper is as follows. Sec- pays regardless of et and a bonus bt  0 that
tion II sets up the basic model. Section III the firm agrees to pay only if et 5 1. While wt is
studies the optimal wage contracts. In Section legally enforceable, bt cannot be enforced by
IV, first the stationary market equilibrium is the court; hence, it has to be self-enforcing.
derived and then comparative statics and wel- Note that if an employed worker shirks,
fare analysis are conducted. In Section V, we then his employer’s period profit is negative.
extend the basic model to incorporate search Thus, firms have to motivate workers to exert
costs. Section VI presents some empirical evi- effort. To make sure there is positive employ-
dence. Section VII concludes the paper. ment, we assume that p(0) . u + v/(dq). In
addition, we assume that p(L) u + v/(dq).
II. THE BASIC MODEL
This assumption ensures that J  L; that is,
there is always unemployment and workers
Consider a labor market over time, with are always on the long side of the market.
time t being discrete. There are L workers There are turnover costs in the labor mar-
and many firms which create J job vacancies ket. Firms incur recruiting costs in finding job
in total. While L is exogenously given, there candidates.3 Workers incur search costs in
is free entry on the firms’ side, thus J is endog- finding jobs. Moreover, each firm may require
enous. Both workers and firms are risk neutral a skill which is firm specific.4 For simplicity,
and share the same discount factor d. Each job we assume that it takes one period for
has the same revenue product of labor p. Each a new employee to acquire the firm-specific
firm takes p as given, but in the aggregate, p is skill. During that period, the output of
a decreasing function of the total employment a new worker is less than that of an old worker
E. At the beginning of each period, unem- by cF  0.5 The parameter cF captures the
ployed workers and unfilled vacancies are ran-
domly matched. Note that agents on the long
side of the market may not get a match. At the 2. An alternative setting would be that, while et is not
end of each period, each existing match breaks observable, the performance is observable but not verifi-
able, and et determines the worker’s performance with
up with probability 1
q for exogenous rea- some noise. The only difference this alternative setting
sons. In any existing match that survives this makes is that shirking cannot be detected for sure. Since
shock, the worker and the firm simultaneously workers are risk neutral, the qualitative results of the
paper still hold under this alternative setting.
decide whether to continue the relationship 3. For some jobs, Lang (1991) found that firms are
next period. If either party decides to leave, willing to pay employment agencies large sums for finding
then the match is broken up endogenously. workers.
All the agents without a match enter into 4. See Hashimoto and Yu (1980) for the issue of firm-
specific human capital.
the unmatched pool at the beginning of the 5. cF can be interpreted as the training cost to acquire
next period. firm-specific skills.
182 ECONOMIC INQUIRY

FIGURE 1
Timing

degree of firm specificity of jobs: the more firm The model is similar to MM but with two
specific are jobs in the market, the larger the departures. First, we introduce turnover costs
cF. We denote c 5 cF + cR (cR represents in labor markets, which are not studied in their
recruiting costs) as the turnover costs. For model. Second, we only consider the case J  L,
simplicity, we assume that c is borne by firms while in their model both cases J  L and J .
alone.6 Later on, we will discuss how relaxing L are studied.8 Note that there are double
this assumption affects the results of the moral hazard problems: workers may want
model. In the basic model, we ignore workers’ to shirk and collect the fixed wage wt, and
search costs, which will be studied in Section firms may want to renege on the bonus bt after
V. For simplicity, we assume that if a worker is workers exert effort. The concern for external
separated from a firm, his skill which is specific reputation may mitigate the moral hazard prob-
to that firm degenerates immediately. There- lems. We rule out such a reputation effect.9
fore, all workers in the unmatched pool are We are interested in equilibria in which
homogeneous regardless of their employment employed workers exert effort and firms do
history. not renege in each period. There are many
This is essentially an infinitely repeated relational contracts that can be supported as
game with some employment relationships perfect equilibria, some of which may involve
reshuffled in each period. The timing of a typ- complicated strategies. For simplicity, we
ical period is shown in Figure 1. At the begin- restrict our attention to trigger strategies. In
ning of period t, unemployed workers and particular, if a worker shirks, the firm will fire
unfilled job vacancies are randomly matched. him immediately. Similarly, if a firm reneges
In each match, the firm offers the worker on a bonus (pays less than bt), the worker quits
a contract. The worker can either accept or immediately.10
reject the offer. If he rejects the offer, the
worker is unemployed for that period. Then,
each employed worker chooses effort and gets III. THE OPTIMAL CONTRACTS
paid. After the payments have been made, a
1
q fraction of existing matches exogenously We are interested in stationary equilibrium,
break up. Then, in each surviving match, the in which the employment level, the fixed wage,
worker and the firm simultaneously decide and implicit bonus are constant over time, so
whether to continue the relationship into the we can drop all the time subscripts. Denote UN
next period.7 All the unmatched agents enter (US) as the expected discounted lifetime utility
into the unmatched pool in the beginning of of an employed nonshirker (shirker) and U as
the next period.
8. Another minor difference is that in MM, there is
a job creation cost. Incorporating a job creation cost
6. This assumption is not unrealistic. An empirical would not change the qualitative results of our paper.
study by Green et al. (2000) shows that even expenses 9. External reputation may not work for two reasons.
on most transferable training are paid by employers. First, outsiders do not know exactly whether a separation
7. Alternatively, it is also possible that endogenous occurs due to exogenous reasons or cheating. Second, even
separation decisions are made before exogenous separa- if a separation is known to be due to cheating, outsiders do
tion occurs. But whether exogenous separation occurs not know exactly which party is at fault.
before or after endogenous separation decisions are made 10. The trigger strategies are appealing in practice: if
does not matter, since a match breaks up if either separa- at least one party violates the relational contract, the infor-
tion occurs. mal relationship sours and is unlikely to continue.
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 183

the expected discounted lifetime utility of an The continuation value of VN has two com-
unemployed worker. Let E be the number ponents. With probability 1
q, exogenous
of employed workers and a be the job acqui- separation occurs, so the firm gets a continua-
sition rate. Note that both E and a are endog- tion value V ; with probability q, the existing
enous variables. The value function UN is the match remains in the next period, and the
following (supposing firms do not renege): firm’s continuation value is VN(w, b). VR is dif-
ferent from VN in continuation values: after
ð1Þ U N ðw; bÞ 5 ðw þ b
vÞ þ d½qU N ðw; bÞ the firm reneges, the existing match breaks
þ ð1
qÞU ðw; bÞ: for sure so the firm’s continuation value is
always V . V is the firm’s fallback position if
The value of UN consists of two terms. The it reneges. Since there is unemployment in
first term in Equation (1) is a nonshirker’s cur- the market, the firm can immediately find
rent period payoff. With probability 1
q, the a new worker to fill its vacancy, by incurring
current match breaks up exogenously so the turnover cost c. Therefore, V ðw; bÞ 5 V N
c.
worker gets a continuation value U ðw; bÞ.
With probability q, the current match contin- A. Programming Problem
ues in the next period and the worker’s contin- Each firm offers a relational contract to
uation value is UN(w, b). Similarly, other value maximize its profit, taking the employment E
functions are: (hence a) as given. The relational contract
ð2Þ U S ðw; bÞ 5 w þ dU ðw; bÞ should satisfy the following conditions. Unem-
ployed workers should be willing to accept the
contract, and employed workers should have
ð3Þ U ðw; bÞ 5 a  U N ðw; bÞ incentives to exert effort. Firms should earn
zero profit due to free entry, and firms should
þ ð1
aÞ½u þ dU ðw; bÞ: not have incentives to renege. Note that max-
imizing profit is equivalent to minimizing total
If an employed worker shirks, the current wage payment W. Mathematically, the pro-
relationship breaks up for sure, so the contin- gramming problem is as follows:
uation value is U ðw; bÞ. For an unemployed
worker (at the beginning of a period) with min W 5 w þ b
w0;b0
probability a, he will find a job in that period
and his payoff is UN(w, b); with probability subject to
1
a, he will stay unemployed and get UN  U ðIRWÞ
u þ dU ðw; bÞ.
Similarly, denote VN (VR) as the expected V 5 0 ðIRFÞ
discounted lifetime profit of a filled job UN  US ðICWÞ
vacancy if the firm does not renege (reneges)11
N R
and V as the expected discounted lifetime V V ðICFÞ:
profit of an unfilled job vacancy. The value
functions are the following (supposing work- The IRF condition depends on the revenue
ers do not shirk): product of labor p(E). For the time being, we
ignore this condition and discuss it in the next
ð4Þ V N ðw; bÞ 5 ðp
w
bÞ þ d½qV N ðw; bÞ section when we study market equilibria.
þ ð1
qÞV ðw; bÞ Substituting Equations (1) and (2), the no-
shirking condition (ICW) becomes:

ð5Þ V R ðw; bÞ 5 ðp
wÞ þ dV ðw; bÞ ð7Þ dqðU N
U Þ  ðv
bÞ:

The left-hand side of Equation (7) is the

ð6Þ V ðw; bÞ 5 V N
c: rent of continued employment enjoyed by
an employed worker, and the right-hand side
is the current period gain from shirking. Note
11. Given that workers are playing trigger strategy,
for a reneging firm, paying 0 instead of any positive bonus that efficiency wages (measured by U N
U )
less than b is the most profitable deviation. and performance pay b are substitutes in
184 ECONOMIC INQUIRY

motivating workers: either the bonus b should on the bonus if it is too high. This moral haz-
be big enough to reduce the gain of shirking or ard can be alleviated by the presence of turn-
the worker’s rent from continued employ- over costs c: the firm will incur c in hiring
ment, which is created by paying efficiency a new worker next period if it reneges on
wages, should be big enough. Note that firms the bonus. The upper bound of credible
can also use a combination of both. bonuses thus is increasing in turnover costs
Similarly, the no-reneging condition (ICF) c. When dqc , v, the firm should set the high-
can be simplified as: est credible bonus b* 5 dqc to reduce the nec-
essary wage premium required to motivate
ð8Þ b  dqc: workers. When dqc  v, setting any b 2 [v,
dqc] is enough to motivate workers, and the
The left-hand side of Equation (8) is the firm does not need to pay wage premiums.
firm’s current period gain from reneging; the But now the IRW condition is binding, so
right-hand side is the expected cost of reneg- the total wage payment cannot be reduced
ing. Since it can always have a job vacancy further.
filled immediately, the firm’s rent from contin- Define the wage premium wp as the extra
ued employment comes solely from the fact utility per period enjoyed by an employed
that retaining an old worker saves turnover worker relative to an unemployed worker.
costs c.12 Therefore, any credible b has an By Lemma 1, we can calculate wp in the opti-
upper bound dqc. mal contracts.

ð9Þ wp 5 ðb þ w
vÞ
u
B. Optimal Contracts 5 f1=½ð1
aÞdq
1gmaxfv
dqc;0g
With a positive c, firms have some freedom
From Equation (9), it is clear that the wage
to choose b in relational contracts. From the
premium is decreasing in c. Moreover, when
no-reneging condition (Equation 8), the cred-
c  v/(dq), the wage premium equals to 0.
ible bonuses the firm can choose are in the
On the other hand, when c 5 0, b* 5 0 and
range [0, dqc]. The bigger the c, the more free-
firms have to use solely efficiency wages to
dom the firm has to choose b. To facilitate
motivate workers. Note that this case corre-
analysis, we first prove a lemma.
sponds to the situation studied by Shapiro
and Stiglitz (1984). The following proposition
LEMMA 1. If c 2 [0, v/(dq)), in the optimal
summarizes the above results.
relational contract, the firm should set b* 5
dqc. If c  v/(dq), in the optimal relational con-
PROPOSITION 1. (a) Suppose c  v/(dq).
tracts, the firm can set any b* 2 [v, dqc] subject
Then, the optimal contracts have the following
to w* 5 u + v
b*  0.
form: b* 2 [v, dqc] subject to w* 5 u + v
b*  0.
Workers receive no wage premium, and the opti-
C. Proof. See the Appendix. j
mal contract is purely in the form of perfor-
The intuition for Lemma 1 is the following. mance pay. (b) Suppose c 2 (0, v/(dq)).
There are two ways to motivate workers: effi- Then, the optimal contract is the following:
ciency wages and performance pay. Perfor- b* 5 dqc and w* 5 u + (v
dqc)/[(1
a)dq].
mance pay is less costly than efficiency Employed workers receive a positive wage pre-
wages from the firms’ perspective. This is mium, which is decreasing in c. The optimal con-
because performance pay discourages shirking tract is a combination of performance pay and
directly without increasing total compensa- efficiency wages. (c) Suppose c 5 0. Then, the
tion, while efficiency wages require firms to optimal contract has no performance pay compo-
increase total compensation. However, perfor- nent and is in the form of pure efficiency wages.
mance pay is restricted by the moral hazard
problem on the part of the firm: it may renege
D. Robustness
12. In MM, a firm’s rent from continued employment Note that the trigger strategies associated
comes from the fact that the firm may not have its vacancy
filled immediately, which is possible only when J . L. This with the optimal contract do constitute a
rules out the use of performance pay when J  L. subgame-perfect Nash equilibrium even when
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 185

c 2 [0, v/(dq)) (so employed workers enjoy only due to exogenous separation. Denote
a positive wage premium). This is because E* as the employment level(s) in stationary
the worker and the firm make their separation equilibria, which is determined by the firms’
decisions simultaneously. Under this assump- free entry condition:
tion, after one party’s deviation, the strategy
profile (quit, fire) is optimal since unilateral ð10Þ V ðw ;b ;E Þ 5 ½pðE Þ
w
b
deviation in separation decision would not
change the outcome.13
dð1
qÞc=
So far, we have assumed that c is borne by ð1
dÞ
c 5 05pðE Þ
firms solely. Now suppose that the firm and
5 w þb þð1
dqÞc:
the new worker share the turnover cost c
according to some bargaining rule, with the
firm bearing cost hc and the new worker bear- The equilibrium job acquisition rate a is
ing (1
h)c (h 2 (0, 1)). We can go through the
same analysis again. Now, the firm can enforce ð11Þ a 5 ð1
qÞE =ðL
qE Þ:
a credible bonus only up to h(dqc), which is
less than dqc in the basic model. However,
now the worker has less incentive to shirk Using Equation (11) and the results in Prop-
since if he shirks, he will be fired and bear osition 1, Equation (10) can be written more
the additional cost (1
h)c in case he finds explicitly. Specifically, for c 2 [0, v/(dq)),
a new job later. The amount of necessary effi- Equation (10) becomes
ciency wages is still decreasing in c if c 2 [0, v/
(dq)). Overall, the optimal bonus is decreased, ð12Þ pðE Þ 5 u þ m=ðdpÞ þ ½m=ðdqÞ
c
but the efficiency wage is more or less the ð1
qÞE =ðL
E Þ:
same. The qualitative results of the model
remain the same.
Note that any severance pay has no value in For c  v/(dq), Equation (10) can be rewritten
overcoming the moral hazard problems. Sup- as:
pose employment contracts specify that firms
pay workers s whenever separation occurs. ð13Þ pðE Þ 5 u þ m=ðdqÞ
This severance pay s enables firms to credibly
þ ð1
dqÞ½c
m=ðdqÞ:
enforce a bonus s. Under this circumstance,
however, the workers’ moral hazard problem
is not altered: though workers do not receive The right-hand side of Equation (10) speci-
the bonus s in the case of shirking, they receive fies the ALC per period on the equilibrium
a severance pay s instead, thus there is no pun- path, with (1
dq)c being the average turn-
ishment for shirking. over costs incurred per period. Now, define

ð14Þ ALCðEÞ[w ðEÞþb ðEÞþð1
dqÞc

IV. MARKET EQUILIBRIUM 8
> uþv=ðdqÞþ½v=ðdqÞ
c½ð1
qÞE 
>
>
In stationary market equilibria, all the firms < =ðL
E Þ if c2½0;v=ðdqÞÞ
in the market will offer the same optimal con- 5 :
tract derived in the previous section since they >
> uþv=ðdqÞþð1
dqÞ½c
v=ðdqÞ
>
:
face the same programming problem. More- if cv=ðdqÞ
over, in equilibria, all employed workers exert
effort, all firms pay the implicit bonus (if there
Thus, ALC(E) specifies an ALC curve.
is any), and employment relationships end According to Equations (12) and (13), market
equilibria are determined by the intersection(s)
13. Note that since any endogenous separation results
in welfare loss (turnover costs c have to be incurred), the
of the revenue product curve P(E) and the
equilibria derived under trigger strategies are not renego- ALC curve ALC(E). It can be easily seen from
tiation proof. In the previous version, we show that the (14) that ALC is strictly increasing in E if c 2
optimal contract derived under trigger strategies can be [0, v/(dq)), and ALC is independent of E if c 
supported as the optimal contract in a strongly renegoti-
ation proof equilibrium in the sense of Farrell and Maskin v/(dq). On the other hand, the revenue product
(1989). The proof is available upon request. curve p(E) is downward sloping. Therefore,
186 ECONOMIC INQUIRY

the market equilibrium must be unique.14 FIGURE 2

Figure 2 illustrates the determination of mar- Market Equilibrium
ket equilibrium for the case c 2 [0, v/(dq)),
where point A denotes the market equilibrium.

A. Comparative Statics
We are interested in how changes in c affect
the market equilibrium.

PROPOSITION 2. For c 2 [0, v/(dq)), the

equilibrium employment level E* is increasing
in c. For c  v/(dq), E* is decreasing in c.

Proof. First consider the case c 2 [0, v/(dq)).

Differentiating Equation (12) with respect to
c, we get
ð15Þ @E =@c 5 ð1
qÞE ðL
E Þ
=f½v=ðdqÞ
cð1
qÞL

p9ðE ÞðL
E Þ2 g . 0; age turnover costs per period. Hence, the equi-
librium employment level E* is decreasing in
c. By Equation (14), it can be easily seen that
since E* 2 (0, L) and p9(E*) , 0. If c  v/(dq),
an increase in q or d shifts the ALC curve
then by Equation (13)
down. Thus, as the exogenous separation rate
(1
q) decreases or the discount factor d
@E =@c 5 ð1
dqÞ=p9ðE Þ , 0: j
increases, the equilibrium employment level
E* increases.
It is surprising that the equilibrium employ- Changesincalsoaffectthewage-employment
ment level E* is increasing in c when c is small. relationship. For c 2 [0, v/(dq)),
The intuition for this result is the following. An
increase in turnover costs by Dc can decrease @ðw þ b Þ=@E 5 ½v=ðdqÞ
cð1
qÞL
the total wage payment (or the wage premium)
in every period by {1/[(1
a)dq]
1}dqDc. =ðL
EÞ2 :
On the other hand, the average turnover costs
per period on the equilibrium path increase by Thus, an increase in c reduces the wage-
(1
dq)Dc. Overall, the first effect dominates employment elasticity. This is due to the fact
the second one. Intuitively, an increase in turn- that an increase in c reduces the required wage
over costs reduces wage payment in each period, premium, thus making wages less sensitive to
while the increase in average turnover costs per the employment level.
period is small since each job only incurs the
turnover costs occasionally (with probability
1
q) on the equilibrium path. Therefore, B. Welfare Properties
an increase in c shifts the ALC curve down-
ward, leading to an increase in E*. Now, we study how changes in c affect
This result is reversed when c  v/(dq). If c social welfare. In the market equilibrium, each
falls in this region, by Equation (14), an firm’s expected profit is 0 since p(E*) 5
increase in c shifts the ALC curve upward. ALC(E*). The total social surplus S (per
This is because an increase in c cannot reduce period) of the market equilibrium with
the wage payment further (the IRW condition employment E* is
is binding), but it directly pushes up the aver-
ð E
14. For c 2 [0, v/(dq)], the existence of equilibrium E* ð16Þ SðE ; cÞ 5 pðEÞ dE
E ðu þ vÞ
2 (0, L) is guaranteed if p(0) . u + v/(dq). For c  v/(dq), 0
the existence of equilibrium E* 2 (0, L) is ensured if p(L) ,
u + v/(dq) + (1
dq)[c
v/(dq)] , p(0).
ð1
qÞE c:
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 187

The first term is the social value of the total both terms are negative in Equation (17).
output or consumers’ total willingness to pay Therefore, we have the following proposition.
for the total output. The second term is the
total cost of labor, and the third term is the PROPOSITION 3. If c 2 [0, v/(dq)) and Con-
total turnover cost incurred per period. Taking dition (18) is satisfied (the revenue product of
the derivative of Equation (16) with respect to c, labor p(E) is elastic enough), the social surplus
is increasing in c. When c  v/(dq), the social
ð17Þ @S=@c 5 ½pðE Þ
ðu þ vÞ surplus is decreasing in c.

ð1
qÞc@E =@c
Proposition 3 is a surprising result: a little bit

ð1
qÞE : of (exogenous) friction in markets can improve
social welfare. Conventional wisdom tells us
We first consider the case c 2 [0, v/(dq)). that friction in markets is always bad, since it
Substituting @E*/@c from Equation (15) into impedes the smooth functioning of markets.
Equation (17), we get However, our model shows that if there is
a double moral hazard problem in the market,
ð18Þ @S=@c . 05vð1
dÞ=d . a little bit of (exogenous) friction in the market

p9ðE ÞðL
E Þ: can actually make the market function more
effectively. The main reason is that without
exogenous friction, contingent contracts are
Condition (18) is satisfied if |p9(E*)| is not available; thus, to motivate one side of
small enough. In other words, if the revenue the market (workers), a certain amount of
product of labor is elastic enough, then the matching friction has to be created endoge-
social surplus is increasing in turnover costs nously by using efficiency wages. The presence
c. Intuitively, an increase in c has two oppo- of some exogenous friction alleviates the firms’
site effects on social welfare (see Equation moral hazard problem and gives them commit-
17). On one hand, it directly increases the ment power, which makes contingent contracts
total turnover costs, thus reducing social wel- (performance pay) feasible. Contingent con-
fare. On the other hand, it increases the equi- tracts not only reduce the amount of endoge-
librium level of employment, thus increasing nously created friction that is necessary to
the social welfare. If the revenue product of motivate workers but also reduce the total
labor is elastic enough, a small increase in c amount of friction in the market that is neces-
can induce a big increase in the equilibrium sary to motivate workers. Broadly speaking,
employment level, causing the positive effect this result implies that exogenous friction might
to dominate the negative effect and social be more efficient than endogenously created
welfare increases.15 friction in overcoming double moral hazard
When c  v/(dq), the social surplus is problems in markets.
decreasing in c. This is because an increase in
c reduces the equilibrium employment E*, thus

C. Empirical Predictions and Policy Implications

15. Condition (18) depends on the endogenous vari-
able E*. A more primitive condition cannot be derived Our model generates several testable empir-
unless we impose a specific functional form on p(E). This ical implications. The first implication is about
is because E* cannot be explicitly solved with a general
p(E), which is evident from Equation (12). To derive the forms of employment contracts. Our
a more primitive condition, we assume model predicts that labor markets with differ-
ent turnover costs will use different forms of
pðEÞ 5 u þ v=ðdqÞ þ kðL
EÞ; employment contracts. In particular, occupa-
with k  0 being some constant. Note that the smaller the
tions with high turnover costs are paid a high
k, the more elastic is the revenue product of labor. Now, bonus, and those with low turnover costs are
Condition (18) can be written as: paid a low bonus. The second implication is
v2 ð1
dÞ2 =d2 þ ð1
qÞ½v=ðdqÞ
cð1
dÞ=
about total wage payment. The model predicts
that occupations with high turnover costs are
d . kLð1
qÞ½v=ðdqÞ
c:
paid a low total wage, and those with low turn-
The above condition is valid if k is small enough or the over costs are paid a high total wage. This
exogenous turnover rate 1
q is close to 0. is because high turnover costs lead to high
188 ECONOMIC INQUIRY

bonuses, which reduce the wage premium. Compared to Equation (17), the second
This also implies that workers paid higher term disappears because the government col-
bonuses actually earn less than those paid lects the money. Therefore, the total social
low bonuses.16 welfare is always increasing in t as long as
The third implication is about the relation- c + t  v/(dq). In fact, social welfare is maxi-
ships among turnover costs, bonuses, and mized when t 5 v/(dq)
c. Of course, the
unemployment. The model predicts that downside of a turnover tax is that it may
occupations paid higher bonuses should impede the free relocation of labor when firms
have lower levels of unemployment. How- have different growth prospects, which is not
ever, the relationship between turnover costs modeled in the paper.
and unemployment level is nonmonotonic.
Among the occupations with low turnover
costs, the occupations with higher turnover V. INCORPORATING SEARCH COSTS
costs should have lower levels of unem-
ployment. On the other hand, among the In this section, we extend the basic model to
occupations with high turnover costs, the incorporate workers’ search costs. Search
occupations with higher turnover costs costs are directly borne by workers. More
should have higher levels of unemployment. importantly, a worker incurs search costs as
The final implication is about the sensitivity long as he has actively searched for jobs,
of wage payment to employment levels. Spe- regardless of whether he finds one. Although
cifically, wages are less sensitive to the workers bear search costs directly, firms have
employment level in occupations with high to induce workers to search for two reasons.
turnover costs than in occupations with low First, firms need workers to fill vacancies after
turnover costs. exogenous separations occur. Second, to effec-
By Proposition 4, if the existing turnover tively discipline shirkers, firms have incentives
costs are small in markets, it might be benefi- to reduce the effective job acquisition rate by
cial for the government to tax employers inducing workers to search.
(without compensating employees) whenever We model the unemployed workers’ search
turnover occurs, thus increasing the effective behavior as follows. Anticipating the wage
turnover costs. Actually, a government tax contracts that firms offer, in each period, each
always increases social welfare as long as unemployed worker decides whether and with
the turnover costs c , v/(dq). To see this, sup- what probability to search (here, we allow
pose c , v/(dq) and the government imposes workers to play mixed strategy regarding
a turnover tax t with c + t , v/(dq). Now, search). If a worker searches, he incurs search
the social surplus becomes costs cS . 0 in that period regardless of the
outcome. After unemployed workers make
ð E their search decisions, the effective job acquisi-
SðE ;c;tÞ 5 pðEÞdE
E ðuþvÞ
ð1
qÞE c: tion rate a is determined, which is the ratio of
0 the number of unfilled vacancies to the popu-
lation of unemployed workers who search
The equilibrium E* satisfies actively. Since the environment is symmetric
for all unemployed workers, we focus on sym-
pðE Þ 5 u þ v=ðdqÞ þ ½v=ðdqÞ metric strategies: each unemployed worker

c
tð1
qÞE =ðL
E Þ: searches with the same probability r 2 [0, 1].
Given r, the stationary job acquisition rate is
Note that @E*/@t 5 @E*/@c . 0. Thus,
a 5 minf1; ð1
qÞE=½rðL
qEÞg:
@S=@t 5 ½pðE Þ
ðu þ vÞ

ð1
qÞc@E =@t . 0:
The presence of search costs cS has two
effects on wage contracts. First, it can disci-
pline employed workers: if a worker shirks,
he will be fired and has to incur search costs
16. Of course, this result crucially depends on the risk cS for several periods to find another job. Sec-
neutrality of workers. ond, firms now have to induce workers to
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 189

search, since search costs are directly borne by The new programming problem for each
workers. Specifically, with the presence of cS, firm is as follows, taking the job acquisition
the value functions become the following: rate a as given:

U N 5 ðw þ b
vÞ þ d½qU N þ ð1
qÞU  min fw þ bg

w;b
S
U 5 w þ dU subject to:
U 5 maxfu þ dU ; a  U N U N  U ðIRW1Þ
þ ð1
aÞðu þ dU Þ
cS g:
a½U N
ðu þ dU Þ
cS  0 ðIRW2Þ
Note that U is the maximum of two pay- V 5 V N
c 5 0 ðIRFÞ
offs: the payoff if a worker searches and the U N  U S ðICWÞ
payoff if he does not search, taking a as given.
Firms’ value functions are still the same as V N  V C ðICFÞ:
those in the basic model.
Unlike the basic model, here the effective Compared to the programming problem in
job acquisition rate a depends on unemployed the basic model, the IRW2 condition is added
workers’ search behavior. It is easy to check because firms have to induce workers to
that a reaches its low bound when r 5 1. search. Given the IRW2 condition and the
Define this low bound as ICW condition, the IRW1 condition can be
rewritten as:
aðEÞ [ ð1
qÞE=ðL
qEÞ
U N
ðu þ dU Þ 
cS =ð1
aÞ:
Specifically, a is determined by the following
formulas: The IRW2 condition can be rewritten as:
8
>
> a if a½U N
ðu þ dU Þ
cS . 0 ðC1Þ U N
ðu þ dU Þ  cS =a:
<
0 if U N
ðu þ dU Þ
cS , 0 ðC2Þ
a5 :
>
> c =½U N
ðu þ dU Þ if cS  U N Thus, the IRW1 condition is redundant.
: S

ðu þ dU Þ  cS =a ðC3Þ Getting rid of the value functions, we can sim-
plify the programming problem as follows:
In the first case, each unemployed worker
searches with r 5 1. This is because the return min fw þ bg
w;b
of search is higher than the search costs even if subject to :
all unemployed workers search with r 5 1. In w  u þ v
b þ ð1
dqÞcS =a ðIRW2Þ
the second case, each unemployed worker
w  u þ ðv
b
dqcS Þ=½ð1
aÞdq ðICWÞ
does not search (r 5 0) because the return
of search is too low. In the third case, neither b  dqc ðICFÞ:
Condition (C1) nor Condition (C2) is satisfied.
In this situation, only a mixed strategy equilib- Depending on parameter values, we have
rium exists: each unemployed worker searches two possible scenarios. In the first scenario,
with r 2 (0, 1) such that everyone is indif- dq(c + cS) v, thus efficiency wages are not
ferent between searching and not searching. necessary. Without loss of generality, firms
More explicitly, Condition (C3) and a can set b* 5 v
dqcS. Now, the ICW condition
be written as: becomes wu, and the IRW2 condition
becomes

cS ð1
dqÞ  ðw þ bÞ
ðu þ vÞ w  u þ dqcS þ ð1
dqÞcS =a:

ð19Þ
 ð1
dqÞcS =a
Thus, the ICW condition is redundant if
the IRW2 condition is satisfied. In the
optimal contract, the IRW2 is binding with
ð20Þ a 5 cS ð1
dqÞ=½ðw þ bÞ
ðu þ vÞ: a 5 1. That is,
190 ECONOMIC INQUIRY

w 5 u þ dqcS þ ð1
dqÞcS =a 5 u þ cS : behavior, the wage is set just high enough

to induce the job acquisition rate â 5
^
ð1
qÞE=ðL ^ at which both the ICW

qEÞ,
Under the optimal contract, it can be easily
seen that Equation (19) is satisfied and a 5 1 and the IRW2 conditions are binding.
by Equation (20). This is because under the Following Lemma 2, the ALC curve
optimal contract, unemployed workers will ALC(E) can be easily derived, and the equilib-
adjust their search behavior such that the rium employment level E* is pinned down by
job acquisition rate a 5 1. The ALC curve firms’ free entry condition: p(E*) 5 ALC(E*).
can be written as: We are interested in the wage-employment
relationship. By Equation (22), when E . E, ^
ð21Þ ALCðEÞ 5 w þ b þ ð1
dqÞc the total wage payment w* + b* is increasing
in the employment level E. On the other hand,
5 u þ v þ ð1
dqÞðc þ cS Þ; when E , E, ^ the total wage payment w* + b* is
independent of E.17 This is an interesting
which is independent of the employment result: the wage-employment relationship
level E. changes over the course of business cycle. Spe-
In the second scenario dq(c + cS) , v, thus cifically, wages are rigid during recessions and
efficiency wages are necessary. The optimal are positively correlated with the employment
bonus is b* 5 dqc. Now, the IRW2 condition level during booms. The following proposition
no longer implies the ICW condition. The fol- summarizes the above result.
lowing lemma specifies the optimal w*.
PROPOSITION 4. Suppose dq(c + cS) , v.
LEMMA 2. There is a cutoff E ^ 2 ð0; LÞ such Wages are rigid during recessions and are pos-
that itively correlated with the employment level dur-
8 ing booms. Specifically, the total wage payment
> u þ v
dqc þ ð1
dqÞcS ðL
qEÞ^ is increasing in E when E . E^ and is independent
>
<
=½ð1
qÞ ^
E if E  ^
E of E when E , E. ^
ð22Þ w 5 ;
>
> u þ ½v
dqðc þ cS ÞðL
qEÞ
: ^ If unemployed workers play pure strategies
=½dqðL
EÞ if E . E
regarding search, then the total wage payment
^ is defined by is actually decreasing in E when E , E. ^ This is
where E ^
because when E , E, the IRW2 condition is
^ ^ binding. A decrease in E directly reduces the
ð23Þ v
dqc þ cS ð1
dqÞðL
qEÞ=½ð1
qÞE return to searching, as the job acquisition rate
^
5 ½v=ðdqÞ
c
cS ðL
qEÞ=ðL ^

EÞ: a decreases. To induce every unemployed
worker to search (with probability 1), the total
wage payment has to increase. As a result, the
Proof. See the Appendix. j wage-employment relationship now assumes
a U-shape. This is a more dramatic result: wages
The intuition for Lemma 2 is as follows. are procyclical during booms and countercycli-
With the presence of search costs, wage con- cal during recessions. In reality, we believe that
tracts need to serve two purposes: motivate unemployed workers are able to adjust their
employed workers and induce unemployed search behavior but not perfectly as our formal
workers to search. Both of them require the model assumed (because it needs perfect coor-
wage to be high relative to unemployment dination). Therefore, more realistically wages
benefit. Moreover, both of them are affected are slightly countercyclical during recessions.
by the employment level. When the employ- The existing theories usually predict a
ment level is high, the resulting high job monotonic relationship between wages and
acquisition rate makes motivating workers unemployment. The efficiency wage model
relatively more difficult. In this case, the wage of Shapiro and Stiglitz shows that there is
should be set just enough to motivate workers. always a negative relationship between wages
On the other hand, when the employment
level is low, inducing search is relatively
17. Note that without search costs, the total wage pay-
more difficult. However, because unemployed ment is always increasing in E when efficiency wages are
workers will endogenously adjust their search necessary.
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 191

and unemployment, while the migration VI. EMPIRICAL EVIDENCE

model of Harris and Tadaro (1970) predicts
Tables 1 and 2 summarize the empirical
that there is always a positive relationship.
predictions of the basic model and the
In contrast, our model predicts that the
extended model, respectively. Note that all
wage-unemployment relationship might be
these predictions are essentially comparative
different in different phases of business cycle.
statics results.
Contractual approaches to wage determi-
nation predict that wages are history depen-
dent. Based on the implicit contract
approach of Harris and Holmstrom (1982), A. Forms of Contracts
Beaudry and DiNardo (1991) showed that There are only a few empirical works on the
wages are downward rigid and are bid up if relationships between occupation and
the market condition improves sufficiently bonuses.19 And in available data, only the fre-
to ensure that workers do not quit.18 Thus, quency of bonuses is reported but not the
wage payments depend on the most favorable amount of the bonuses. Note that the predic-
labor market condition observed since one has tion of our model is about different amounts
begun his job. MM (1993) and Malcomson of merit pay in different occupations. To pro-
(1997) studied the dynamics of fixed wage con- ceed, we simply make the assumption that the
tracts in the presence of holdup problems. frequency of bonuses and the amount of
They show that wages are rigid with respect bonuses are positively correlated.
to shocks of small magnitudes. When the Table 3 is excerpted from table 3B of
cumulative shock reaches sufficient magni- MacLeod and Parent (1999), which is based
tude, the wages are renegotiated either upward on the NLSY data (1988–1990). From the
or downward to reflect the current market table, managers have the highest bonus pay-
condition. Though related, our prediction is ment, food and cleaning service workers have
different from those implications. Specifically, the lowest bonus payment, and professionals
in our model, wages are flexible (both upward and secretaries are paid with medium-sized
and downward) in booms (E . E) ^ but are rigid
bonuses. A similar pattern holds when bonus
^ On the other hand, the
in recessions (E , E). plus promotion is used as the measure of dis-
history-dependent wage-employment rela- cretionary pay.20 This pattern is largely consis-
tionships predicted by the above papers do tent with the predictions of our model. This is
not depend on whether the economy is in because managers usually have high turnover
recession or boom. costs, service workers low turnover costs, and
We are also interested in how changes in professionals and secretaries medium turnover
turnover costs affect the wage rigidity region. costs.
From Equation (23), we can see that an The turnover costs for managers are usually
increase in c or cS reduces E;^ thus, the wage
high for two reasons. First, some firm-specific
rigidity region expands. Intuitively, an knowledge is needed for a manager to be effec-
increase in c or cS reduces the efficiency wage tive in a firm, and it takes time for a new man-
that is necessary to motivate workers, thus ager to acquire this knowledge. Second, each
inducing workers to search becomes rela- managing job may require a different combi-
tively more difficult. As a result, the job nation of skills and personality, so finding
acquisition rate ^ ^ at which both
a (hence E) appropriate candidates for a vacancy takes
the ICW and the IRW2 conditions are bind- a long time and requires substantial effort.
ing increases. By Equation (22), an increase in As a result, the recruiting costs for managers
c or cS makes wages in the wage-procyclical are relatively high. On the other hand, the jobs
region less sensitive to the employment level. for food service and cleaning workers are
This is simply because the required wage pre- fairly standard across firms. Therefore, their
mium is smaller.
19. Throughout this section, both merit pay and
bonuses refer to subjective performance pay. They are
18. In a model of long-term implicit contracts, Harris not objective performance pay, which is conditional on
and Holmstrom (1982) showed that wages are downward contractible performance.
rigid and are bid up when workers’ perceived ability 20. Promotion, which usually is associated with a per-
increases. These results crucially depend on firms’ ability manent wage increase, as a deferred compensation can
to commit to long-term contracts. also motivate workers to exert effort.
192 ECONOMIC INQUIRY

TABLE 1 rates and merit pay correspond to efficiency

Empirical Predictions I wages and subjective performance pay, respec-
tively, in theoretical models. This empirical
Low Turnover High Turnover
Costs Costs
evidence is consistent with the prediction of
our model: merit pay can reduce the amount
Contract form Mainly efficiency Big bonuses of efficiency wages, so workers paid by stan-
wages
dard rates enjoy a higher wage premium, hence
Total wage payment High Low
earn more than workers with merit pay.22
Unemployment High Low
One implication of our model is that
Wage-unemployment High Low
elasticity
the level of equilibrium unemployment is a
decreasing function of the usage of bonuses.
Based on the 1990 British WIRS data, MM
found that there is a negative correlation
turnover costs are usually low. It is also rea- between the percentages of workers with merit
sonable to think that the turnover costs for pay and unemployment rates among occupa-
professionals and secretaries are lower than tions. The study by MacLeod and Parent
those of managers and higher than those (1999) further supports this result. Using data
of service workers, since their jobs usually from the NLSY 1988–1990, they showed that
have a firm-specific component which is there is a strong negative relationship between
smaller than managers’ but bigger than service the use of discretionary bonuses and local
workers’. unemployment.
A similar pattern emerges from the 1990
British Workplace Industrial Relations Survey
C. The Wage Curve
(WIRS), which is reported in Table 4. Manual
workers have the smallest bonus, while manag- As in Solon, Barsky, and Parker (1994),
ers have the highest bonus. Moreover, among most of the empirical works just test whether
manual workers, the incidence of bonuses is the real wage is procyclical but do not esti-
decreasing in their skills. This is also largely mate the whole wage curve. Fortunately,
consistent with our model, since turnover costs a small but important literature initiated by
in an occupation are roughly increasing in Blanchflower and Oswald (1994, BO hereaf-
required skills: the more skills a job requires, ter) does estimate the whole wage curve.
the more firm-specific skills are involved, hence Using the U.S. General Social Survey
the higher the turnover costs. (GSS) data (1974–1988), they estimated the
Though this empirical evidence is largely industry wage curve (wage as a function of
consistent with our model, it is not a test of the unemployment rate in industries) and
our theory.21 We hope that some carefully the regional wage curve (wage as a function
designed empirical work can be done in the of the regional unemployment rate). Fig-
near future to directly test our model. ure 3, copied from BO (p. 107), illustrates
their estimation result. Both curves are ini-
tially downward sloping and then become
B. Wage Differentials and Unemployment upward sloping. Both wages are minimized
Using data from the Industrial Wage Sur- at an unemployment rate of approximately
vey, Brown (1992) conducted an empirical 6%–8%. These wage curves are consistent
study on the relationship between wage levels with the empirical predictions of our ex-
and methods of pay. He found that workers tended model. The upward-sloping portion
paid by standard rates on average earn a higher
wage than those with merit pay. Standard 22. Combined with the information given in Table 3,
the prediction that occupations receiving higher bonuses
would have a lower level of total compensation implies that
21. One may think that the pattern given in Table 3 is managers would have lower total compensation on average
also consistent with firms using bonuses to select more than cleaning service workers. This seems to contradict the
able workers, which may matter more in managerial occu- fact that managers earn more than cleaning service workers.
pations and for professionals. However, bonuses cannot However, in Brown’s empirical analysis, he controlled for
be enforced by the court. Therefore, how much bonus will human capital. The fact that managers earn more than
be posted does not only depend on how much bonus firms cleaning workers is simply because the former have more
are willing to post but also depend on how much firms can human capital. After controlling for human capital, clean-
credibly post. ing workers actually earn more than managers.
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 193

TABLE 2 than 12% and the wage curve based on hourly

Empirical Predictions II wage exhibits a U-shape with the minimum
Procyclical in booms,
wage reached at an 11.5% unemployment rate.
rigid in recessions Though there are some minor differences,
all the above estimation results show that
The wage Low Turnover High Turnover
Costs Costs
the wage curve either flattens out or becomes
curve
upward sloping at fairly high unemployment
The wage Big Small rates. They are largely consistent with our
rigidity region empirical prediction that wages are procyclical
Wage-unemployment High Low during booms and either rigid or countercycli-
elasticity
cal in recessions.

of the wage curve suggests that unemployed D. Wage-Unemployment Elasticity

workers are not able to adjust perfectly their BO also estimated the wage-unemployment
search behavior. relationship for different occupations using
BO also estimated the wage curve based on data from the 1990 British WIRS. The results
the U.S. Current Population Surveys (1964– are reported in Table 5. Unskilled manual
1991). The results are slightly different from workers have the largest wage-unemployment
those from the GSS: the wage curve is signif- elasticity, supervisors the lowest elasticity, and
icantly downward sloping when the unem- clerical workers some medium elasticity. As
ployment rate is low, and it flattens out as we argued before, turnover costs are increas-
the unemployment rate increases, but there ing in the order of unskilled manual workers,
is no upward-sloping portion of the wage clerical workers, and supervisors. Therefore,
curve. A similar estimation result holds for the pattern of wage-unemployment elasticity
data from the British Social Attitude Surveys is largely consistent with the predictions of
(1983–1989): the wage curve flattens out when our model. Unskilled manual workers have
the unemployment rate is greater than 13%. the lowest turnover costs; hence, their methods
Using the General Household Surveys’ data of pay are mainly efficiency wages. As a result,
(1973–1977) from Britain, BO found that their wages are more procyclical. On the other
the wage curve has a U-shape with the turning hand, supervisors (managers) have high turn-
point occurring around an unemployment rate over costs; thus, their methods of pay are
of 4.5%. Based on the International Social mainly subjective performance pay, which
Survey Program data (1986–1991) from West leads to low wage-unemployment elasticity.
Germany, the estimation of BO shows that the Table 6 reports the results of two other
wage curve flattens out around an unemploy- studies: BO on the British General Household
ment rate of 11%. Bratsberg and Turunen Surveys 1973–1977 and Kennedy and Borland
(1996) estimated the U.S. wage curve for (2000) on the Australian Bureau of Statistics
young workers using the 1979–1993 waves Income Distribution Survey 1982–1994. Again,
of the NLSY. According to their study, the managers have the lowest wage-unemployment
wage curve based on annual earnings flattens elasticity, and manual workers or clerks have
out when the unemployment rate is higher the highest elasticity.

TABLE 4
TABLE 3 Bonuses across Occupations II
Bonuses across Occupations I Occupations Incidence of Bonus (%)
Bonus +
Professional and managerial 35
Occupations Bonus (%) Promotion (%)
Supervisors 32
Managers 28.46 47.37 Clerical, administrative, 30
Professionals 15.46 29.22 and secretarial
Secretaries 11.60 25.60 Skilled manual 22
Food service workers 7.49 18.66 Semiskilled manual 16
Cleaning service workers 7.43 17.33 Unskilled manual 11
194 ECONOMIC INQUIRY

FIGURE 3 TABLE 6
The Wage Curves Wage-unemployment Elasticities II
Occupations British GHS data ABS IDS data
Manual
0.0721
Clerks
0.0896
Professionals
0.0631
0.0224
Managers
0.0497
0.0198

Notes: GHS 5 General Household Survey; ABS IDS 5

Australian Bureau of Statistics Income Distribution
Survey.

or countercyclical in recessions. Our model

generates rich empirical implications. The
forms of wage contracts and total wage pay-
ments are different in occupations with dif-
ferent turnover costs. Occupations using
more bonus payments have lower total wage
payments and lower unemployment rates.
VII. CONCLUSIONS Occupations with high turnover costs have
We studied contract selection between effi- low wage-unemployment elasticity. Some
ciency wages and subjective performance pay empirical evidence is consistent with these
to motivate workers in a labor market setting. predictions.
Though subjective performance pay is cheaper Though our model is couched in a labor
than efficiency wages, it is limited by the firms’ market setting, it can also be applied to other
incentive to renege. The presence of turnover markets where both parties in a relationship
costs borne by firms reduces firms’ incentives have moral hazard problems and both are
to renege, thus making implicit bonuses cred- able to change partners in markets. For exam-
ible to some extent. In the optimal contracts, ple, consider buyer-seller relationships in
the amount of the bonus is positively corre- a market setting, in which the quality of goods
lated and the amount of wage premium nega- is observable but not verifiable. To motivated
tively correlated with the turnover costs borne sellers, buyers can either offer higher fixed pri-
by firms. Up to some threshold, an increase in ces (analogous to efficiency wages) or post
turnover costs effectively reduces the total some bonuses (which have to be self-enforcing),
wage payment and total labor costs, thus tying payments to the quality of goods. Turn-
increasing the equilibrium employment level over costs will generally affect the optimal
and social welfare. contracts and have welfare implications, sim-
The extended model incorporates workers’ ilar to those shown in the labor market model.
search costs. In this setting, the wage- The model can be extended in several
unemployment relationship turns out to be directions. First, in the model, we have
different during booms and recessions: wages assumed that the turnover costs are exoge-
are procyclical in booms and are either rigid nous. When firms are able to choose turnover
costs within some range, they might have
incentives to choose the level of turnover
costs that minimizes their ALCs. Second,
TABLE 5
in the model, we have assumed that workers
Wage-unemployment Elasticities I are homogeneous. Yang (2005) studied rela-
Occupations Coefficient tional contracts with heterogeneous workers.
Unskilled manual
0.0916
Now moral hazard interacts with firms’
Skilled manual
0.0325
learning about workers’ types, which results
Clerical
0.0434
in nonstationary relational contracts. This
Supervisors
0.0048
helps to explain why contractual terms
change as the length of a relationship
 YANG: EFFICIENCY WAGES & PERFORMANCE PAY 195

increases. Third, it is also interesting to ð29Þ ^

ð1
qÞE=ðL ^ [ â:

qEÞ
model relational contracts in a setting where
the demand for labor is fluctuating over time. Note that E^ is unique and E ^ 2 ð0; LÞ. Combining Equa-
Such a model enables us to see more explicitly tions (28) and (29) gives rise to Equation (23).
how wages and contract forms change during ^ Then,
First, consider the case E . E.
the course of business cycles. This is left for
future research. ^
aðEÞ 5 ð1
qÞE=ðL
qEÞ . ð1
qÞE=
^ 5 â:
ðL
qEÞ
APPENDIX
Proof of lemma 1. Thus, the binding condition is the ICW condition, and
After some algebra, the programming problem can be firms will induce unemployed workers to search with prob-
simplified as the following: ability 1 with the effective a equaling the low bound aðEÞ.
As a result, the optimal w* is given by:
min fw þ bg
w0; b0
w 5 u þ ½v
dqðc þ cS ÞðL
qEÞ=½dqðL
EÞ:

ð24Þ wþbuþv ðIRWÞ Now, consider the case E  E.^ In this case, firms will just
induce â by paying the following wage:

ð25Þ w  u þ kðv
bÞ ðICWÞ w 5uþv
dqcþð1
dqÞcS ðL
qEÞ=½ð1
qÞ

^ ^
E:

With the wage specified above, it can be verified that Equa-

ð26Þ b  dqc ðICFÞ tion (19) is satisfied and that from Equation (20), the effec-
tive a is
where k 5 1/[(1
a)dq] . 1. Condition (25) can be further
reformulated as: a 5 cS ð1
dqÞ=½ðw þ dqcÞ
ðu þ vÞ
^
5 ð1
qÞE=ðL ^ 5 â:

qEÞ
ð27Þ w þ b  u þ v þ ðk
1Þðv
bÞ:

From Equation (27), we see that if b , v, then only Equa-

tion (25) is binding and the total compensation w + b is REFERENCES
decreasing in b; so b should be set as high as possible sub-
ject to condition (26). Condition (24) is binding if and only Akerlof, G., and L. Katz. ‘‘Workers’ Trust Funds and The
if b v. Actually, Condition (24) specifies the low bound of Logic of Wage Profiles.’’ Quarterly Journal of Eco-
the total wage payment, u + v. Therefore, when c , v/(dq), nomics, 104, 1989, 525–36.
the optimal contract is b* 5 dqc and w* 5 u + k(v
b*). Baker, G., R. Gibbons, and K. Murphy. ‘‘Subjective Per-
When c v/(dq), the optimal contracts are b* 2 [v, dqc] formance Measures in Optimal Incentive Con-
subject to w* 5 u + v
b* . 0. j tracts.’’ Quarterly Journal of Economics, 109, 1994,
1125–56.
Beaudry, P., and J. DiNardo. ‘‘The Effects of Implicit
Proof of lemma 2. Contracts on the Movement of Wages over the Busi-
ness Cycle: Evidence from Micro Data.’’ Journal of
Note that the right-hand side of the IRW2 condition is Political Economy, 99, 1991, 665–88.
decreasing in a and goes to infinity as a goes to 0. On the Blanchflower, D., and A. Oswald. The Wage Curve.
other hand, the right-hand side of the ICW condition is Boston, MA: The MIT Press, 1994.
increasing in a and goes to infinity as a goes to 1. Thus, Bratsberg, B., and J. Turunen. ‘‘Wage Curve Evidence from
there is an â 2 ð0; 1Þ such that the right-hand sides of both Panel Data.’’ Economics Letters, 51, 1996, 345–53.
conditions are equal. That is,
Brown, C. ‘‘Wage Levels and Method of Pay.’’ Rand Jour-
nal of Economics, 23, 1992, 366–75.
Bull, C. ‘‘The Existence of Self-Enforcing Implicit Con-
ð28Þ v
dqc þ ð1
dqÞcS =â tracts.’’ Quarterly Journal of Economics, 102, 1987,
5 ðv
dqc
dqcS Þ=½ð1
âÞdq 147–59.
Farrell, J., and E. Maskin. ‘‘Renegotiation in Repeated
Games.’’ Games and Economic Behavior, 1, 1989,
Therefore, if a . â, then the IRW2 condition is redundant 327–60.
and the ICW condition is binding. If a  â, then the ICW Green, F., A. Felstead, K. Mayhew, and A. Pack. ‘‘The
condition is redundant and the IRW2 condition is binding. Impact of Training on Labor Mobility: Individual
Moreover, the fixed wage w is minimized when a 5 â. and Firm-Level Evidence from Britain.’’ British
^ such that
Define E Journal of Industrial Relations, 38, 2000, 261–75.
196 ECONOMIC INQUIRY

Harris, M., and B. Holmstrom. ‘‘A Theory of Wage ———. ‘‘Motivation and Markets.’’ American Economic
Dynamics.’’ Review of Economic Studies, 49, 1982, Review, 88, 1998, 388–411.
315–33. MacLeod, B., and D. Parent. ‘‘Job Characteristics and the
Harris, J., and M. Tadaro. ‘‘Migration, Unemployment Form of Compensation.’’ Research in Labor Eco-
and Development: A Two-Sector Analysis.’’ Ameri- nomics, 18, 1999, 177–242.
can Economic Review, 60, 1970, 126–42. Malcomson, J. ‘‘Contracts, Holdup, and the Labor Markets.’’
Hashimoto, M., and B. Yu. ‘‘Specific Capital, Employ- Journal of Economic Literature, 35, 1997, 1916–57.
ment Contracts, and Wage Rigidity.’’ Bell Journal Salop, S. ‘‘A Model of the Natural Rate of Unemploy-
of Economics, 11, 1980, 536–49. ment.’’ American Economic Review, 69, 1979, 117–25.
Kennedy, S., and J. Borland. ‘‘A Wage Curve for Australia.’’ Shapiro, C., and J. Stiglitz. ‘‘Equilibrium Unemployment
Oxford Economic Papers, 52, 2000, 774–803. as a Worker Discipline Device.’’ American Economic
Lang, K. ‘‘Persistent Wage Dispersion and Involuntary Review, 74, 1984, 433–44.
Unemployment.’’ Quarterly Journal of Economics, Solon, G., R. Barsky, and J. Parker. ‘‘Measuring the
106, 1991, 181–202. Cyclicality of Real Wages: How Important is Com-
Levin, J. ‘‘Relational Incentive Contracts.’’ American Eco- position Bias.’’ Quarterly Journal of Economics, 109,
nomic Review, 93, 2003, 835–57. 1994, 1–25.
MacLeod, B., and J. Malcomson. ‘‘Implicit Contracts, Stiglitz, J. ‘‘Alternative Theories of Wage Determination
Incentive Compatibility, and Involuntary Unem- and Unemployment in LDCs: The Labor Turnover
ployment.’’ Econometrica, 57, 1989, 447–80. Model.’’ Quarterly Journal of Economics, 88, 1974,
———. ‘‘Investments, Holdup, and the Form of Market 194–27.
Contracts.’’ American Economic Review, 83, 1993, Yang, H. ‘‘Nonstationary Relational Contracts.’’ Work-
811–37. ing Paper, Ohio State University, 2005.