Post on 06-Apr-2018
8/3/2019 Content Server 7
1/35
INCENTIVES, EQUALITY AND CONTRACT
RENEGOTIATIONS: THEORY AND EVIDENCE IN THECHINESE BANKING INDUSTRY
Hongbin Caiw
Hongbin Liz
Li-An Zhou
Renegotiation plays an important role in contract theory, but theempirical study of renegotiation is almost non-existent in the literature.
Using a unique datasetfrom the Chinese bankingindustry, we find that thelarge majority of managerial incentive contracts are renegotiated afterperformances are realized. We develop a model of contract renegotiationwhere supervisors and managers sign incentive contracts and thenrenegotiate them. In the unique equilibrium of the model, incentivecontracts are almost always renegotiated ex post. Even though renegotia-tion is fully anticipated, incentive contracts affect performance. Thepredictions of the model find strong support from our empirical results.
I. INTRODUCTION
INCENTIVE CONTRACTS IN THE REAL WORLD ARE OFTEN RENEGOTIATED, especially
within firms and organizations. Renegotiation also plays a very important
role in contract theory.1 However, as for contract theory in general, where
THE JOURNAL OF INDUSTRIAL ECONOMICS 0022-1821
Volume LVIII March 2010 No. 1
We thank Masahiko Aoki, Chong-En Bai, Douglas Bernheim, Paul Devereux, AntonioRangel, Minggao Shen, David de Meza, the Editor and the anonymous referees for very helpfulcomments. We are indebted to the William Davidson and the Ford Foundation in Beijing forfunding the survey work in 1998. Cai acknowledges support from NSFC grant (Project No.70573008) and appreciates the financial support of the Project of Chinese Private Economy in the
Center for Research of Private Economy, Zhejiang University. All remaining errors are ours.wGuanghua School of Management and IEPR, Peking University, Beijing, 100871, Chinaand Center for Research of Private Economy, Zhejiang University, 388 Yuhangtang Road,Hangzhou, Zhejiang Province, China.e-mail:hbcai@gsm.pku.edu.cn
zSchool of Economics and Management, Tsinghua University, Beijing, 100084, China.e-mail: lihongbin@sem.tsinghua.edu.cn
Guanghua School of Management and IEPR, Peking University, Beijing, 100871, China.e-mail: zhoula@gsm.pku.edu.cn
1 In moral hazard models, Fudenberg and Tirole [1990] argue that after theagent exerts effortbut before the output is realized, the principal and the agent can achieve mutual gain byrenegotiating the incentive contract to fully insure the agent. In multi-period adverse selection
models, Dewatripont [1988, 1989] argues that if the agents type is revealed in the first period,the principal and the low type agent will want to renegotiate their contract to get rid of anyallocation distortions in the optimal static contract In the same environment, Laffont and
mailto:hbcai@gsm.pku.edu.cnmailto:lihongbin@sem.tsinghua.edu.cnmailto:zhoula@gsm.pku.edu.cnmailto:zhoula@gsm.pku.edu.cnmailto:lihongbin@sem.tsinghua.edu.cnmailto:hbcai@gsm.pku.edu.cn8/3/2019 Content Server 7
2/35
rapid theoretical advances are in contrast with a long lag of empirical work,
there has not been much empirical study on contract renegotiation.2 We are
fortunate to have access to a dataset collected from an on-site survey in 1998
that contains valuable information about contract renegotiation in the
Chinese banking industry. As with the typical incentive schemes introduced
in the course of the reform of Chinas State Owned Enterprises (SOEs),
since the mid 1980s, managers in local bank branches have signed explicit
incentive contracts with their supervisors at the beginning of the year that
specified bonus payments based on their annual performance. Quite
strikingly, these incentive contracts were almost always renegotiated ex
post. The dataset contains information about ex ante incentive contracts,
realized performance, and ex post contracts, as well as detailed information
about branch and manager characteristics, thus opening a rare window intothe process of how incentive contracts are designed and then renegotiated.
In trying to connect theory with evidence in the dataset, one finds that the
basic assumptions of the standard theories of renegotiation do not fit the
reality of the Chinese banking industry. The moral hazard theory of
renegotiation does not apply here because contract renegotiations took
place after the realizations of performance, and thus could not have been
driven by the desire to improve risk-sharing between managers and their
supervisors. The adverse selection theory of renegotiation does not apply
here either because contracts were renegotiatedfor the current year, not for
the future, and thus contracts were not renegotiated with dynamic
considerations in mind. Our dataset is drawn from the Chinese banks, but
the same phenomena are prevalent also in other SOEs in China. Thus the
following questions naturally arise. Why were those incentive contracts
renegotiated? If it was expected that contracts would be renegotiated, what
were the roles of those contracts? What were the effects ofex ante and ex post
contracts on performance?3
In this paper, we develop a simple model of renegotiation to answer these
questions. The model is intended to capture some of the essentialinstitutional features of incentive contracting in Chinas state sector. In
our model, supervisors have strong equality concerns over managers undertheir supervision. Moreover, managers enjoy rents-on-the-job (perks and
career development) which come largely under their supervisors discretion.
renegotiation destroys ex ante incentives: the principal will design a renegotiation-proofcontract that is less efficient than the optimal contract without the possibility of renegotiation.
2 Notable exceptions are Filson, Switzer and Besocke [2005] and Gil [2004], which we discuss
in detail later.3 By renegotiation-proof principle, theoretical analysis of renegotiation typically focuses onoptimal contracts that preempt renegotiations (renegotiation-proof). With such contracts,
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 157
8/3/2019 Content Server 7
3/35
Supervisorsequality concerns and managers rents-on-the-job (which are
subject to manipulation by the supervisors) create opportunities for mutual
gains that constitute the basis for contract renegotiation. In the case of
managers with above average performance, their supervisors will want to
renegotiate their bonus payments downward in exchange for more rents. In
the case of those who have below average performance, their supervisors will
want to pay them higher bonuses than their ex ante contracts specified but
reduce their rents. Ex ante incentive contracts serve as a disagreement point
in the renegotiation process, thus affecting ex post payoffs. Given an ex ante
incentive contract, the manager chooses an optimal effort level anticipating
how the contract will be renegotiated ex post. The supervisor designs an
optimal ex ante incentive contract taking into account the managers best
response and theex post
renegotiation process.We suppose that the benchmark on which supervisors equality concerns
are anchored is endogenously determined. We characterize the unique
equilibrium of the game: the optimal ex ante contracts, the managers best
effort response, performance, and how ex ante contracts will be
renegotiated. The model generates theoretical predictions about how the
endogenous variables ex ante contracts, performance, ex post contracts,
and direction and degree of renegotiation are affected by exogenous
manager and supervisor characteristics and technological conditions. For
example, we show thatex ante
contracts will have higher power andmanagerial performance will be better when the supervisors own incentive
intensity is greater, or the managers value more rents. Moreover, the greater
the supervisors own incentive intensity or the managers responsiveness toincentives, the more likely ex ante contracts will be renegotiated upwards
(higher bonus ex post than specified ex ante) and the greater the degree of
renegotiation (i.e., absolute magnitude of renegotiation) will be.
The predictions of the model find strong support from our empirical results
using the dataset from the Chinese banking industry. In particular, we find that
the supervisors own incentive intensity has positive and statistically significanteffects on both the incentive intensity in the ex ante managerial contracts and
branch performance. Younger managers (who are likely to value more careerdevelopment opportunities) tend to have more highly powered contracts, and
perform better than their older peers. Moreover, when the supervisors own
incentive intensity is greater or the manager is younger, it is more likely that the
bonus payments will be adjusted upwards and the adjustment will be larger.
For example, according to our estimation, if a branch manager is 5 years
younger than the mean, the chance for an upward renegotiation will be 610
per cent higher. The evidence here suggests that our model, with endogenouslydetermined bonus benchmarks, fits the data pretty well.By incorporating some essential organizational features into the standard
158 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
4/35
concerns and managers rents. Incentive contracts affect performance, but
they do so through mechanisms which are more complicated than thestandard
agency model suggests. A number of papers, e.g., Groves et al. [1994], Li
[1997], and Jefferson and Rawski [1999] have tried to estimate whether and by
how much the incentive schemes in Chinese SOEs affected firm productivity
and financial performance. Overall, they found positive correlations between
incentive intensity and firm performance. However, in light of our theoretical
results and empirical findings, it may be an over-simplification to interpret the
positive correlations in the way described in the standard agency model. Given
that incentive contracts were typically renegotiated, the incentive schemes
introduced in Chinese SOE reforms helped improve performance, most likely
because they constrained the renegotiation outcomes and therefore affected
the managers payoffs. Thus, the correlation between performance and actualbonus payments may in fact underestimate the actual incentives managers
could look forward to when deciding howhard to work, because exceptionally
good (bad) performance was more likely to be associated with downward
(upward) renegotiation of bonus payments.
To the best of our knowledge, Filson, Switzer and Besocke [2005] and Gil
[2004] are the only papers that investigate empirically contractual renegotia-
tion. Both consider revenue-sharing contracts between movie theaters and
distributors: the former is about the United States and the latter is about
Spain. Both papers argue that renegotiation in the movie industry does not fitthe standard contract theory of moral hazard or adverse selection. Filson,
Switzer and Besocke [2005] suggest that risk aversion and transaction costs
may be the forces driving renegotiation. Using a large dataset, Gil [2004]shows that renegotiation in the Spanish movie industry is quite common
(55%) and is always one-sided. He argues that dynamically adjusting prices
as more information is available can be rationalized in a setting of incomplete
contracts and learning. In the context of the Chinese banking industry, our
paper shows that renegotiation is pervasive and two-sided. More impor-
tantly, renegotiation in our context is about modifying theexpost
contractualterms that were agreed upon at the beginning of the current period, instead of
adjusting future contractual terms. We argue that equality concerns of thesupervisors are the driving force of contractual renegotiations in our context.
A large body of literature has recently emerged about the economic
implications of fairness, inequality aversion, and other social psychological
considerations (see, e.g., Fehr and Schmidt [1999]).4 For example,
Englmaier and Wambach [2006] and Rey Biel [2007] extend the standard
moral hazard model by considering inequality averse agents, and show,
among other things, that incentives tend to be flatter and team-based
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 159
8/3/2019 Content Server 7
5/35
incentives are more likely to be used than in the standard model. These
papers do not consider contractual renegotiation. Unlike these papers, our
purpose is to explain contractual renegotiations observed in our data. For
this purpose, we analyze a situation in which the principal rather than the
agent has equality concerns.
Besides the renegotiation literature, our paper also contributes to the
literature on incentives within firms.5 In particular, the spirit of our paper is
closely related to that of Prendergast [2002a]. Because of our different
objective, our model differs from Prendergast [2002a] in that: (i) whereas in
our model, supervisors care about income equality among their subordi-
nates, Prendergast [2002a] supposes that supervisors will display favoritism
towards individual subordinates; (ii) whereas in our model, supervisors
design incentive schemes for their subordinates, in Prendergast [2002a]bonus schemes are decided by firms; and (iii) a feature that is crucial in our
model but is absent in Prendergast [2002a] is that supervisors in our model
have control over their subordinates rents-on-the-job.
The rest of the paper is structured as follows. The next section outlines the
institutional backgrounds of incentive contracting and renegotiation in the
Chinese banking industry. In Section 3, we present the model and derive its
theoretical predictions. Section 4 then describes the data, and Section 5
presents the empirical results. In Section 6, we discuss possible extensions of
the model, robustness and alternative explanations of our empirical results.Section 7 contains concluding remarks.
II. INSTITUTIONAL BACKGROUNDS IN THE CHINESE BANKING INDUSTRY
In this section, we briefly describe the institutional backgrounds of the rural
financial institutions in China: the Agricultural Bank of China (ABC) and
the Rural Credit Cooperatives (RCCs). The discussion below serves twopurposes. First, it familiarizes readers with the institutional context of our
dataset, which was collected from an on-site survey of these two financial
institutions. Second, it provides motivations and justifications for some key
elements of our model.
The ABC is one of the four specialized state-owned banks in China.6 The
RCCs were formerly under the supervision of ABC, but they attained
independence as financial institutions in 1994. The ABC, with headquarters
in Beijing and branch offices at locations at every administrative level, servesclients both in urban and rural areas.7 The RCCs, which collectively make
5
For surveys on incentive provision within firms, see Prendergast [1999] and Gibbons andWaldman [1999]
6 The other three are the Industrial and Commercial Bank of China, the Construction Bank
160 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
6/35
up the Federation of RCCs (xin yong lian she), or the headquarters at the
county level, only provide financial services to rural areas. In general, both
the ABC and the RCCs establish branches in each township, paralleling the
territorial structure of the government system in order to minimize
overlapping within the same institution. The ABC and RCCs currently
dominate the formal financial system in rural China. As of the late 1990s,
they accounted for nearly 80 per cent of total rural deposits and loans (Parket al. [1997]).
In terms of internal management and operation, the ABC and the RCCs
are very much alike, because the RCCs were formerly a part of the ABC and
they operate in similar regulatory, institutional and business environments.
The main difference between these two financial institutions lies in the
ownership structure. The ABC is state-owned, while the RCCs arecollectively owned by member rural households. The collective ownership
of the RCCs and its less hierarchical internal structure (only two
management levels: town branches and the county headquarters, under
the board of trustees) probably lead to a more equitable culture than in the
ABC.8
In the pre-reform era, Chinese state-owned banks, like other SOEs, did
not give any monetary incentives to managers and workers. Their wages
were fixed by certain pre-determined formulae linking wages to worker
characteristics such as age, seniority and position, but not to performance.Since the early 1980s, as part of the enterprise reform programs, China has
sought to reform its banking sector. As a major reform initiative to improve
the performance of state-owned banks, and in particular to increase depositsand reduce non-performing loans, the government introduced a bonus
system into the sector in the mid 1980s (Dipchand et al. [1994]). Typically,
supervisors sign a so-called responsibility contract with each individual
branch manager. The contract, normally signed on an annual basis, specifies
a formula tying the branch managers monetary rewards to branch
performance. For both the ABC and the RCCs, county-level banksupervisors determine incentive contracts for township branch managers.
While the details of the incentive system (such as the kinds of contracts andthe autonomy allowed to supervisors by government regulations) has
evolved over the years of the reform, its basic structure has remained more or
less the same. Moreover, while the economy has grown rapidly with market
forces playing an increasingly important role, the internal organization
structure of the large SOEs and state banks has remained quite stable. In
particular, in the late 1990s, there was still no active outside managerial
market, hence managers careers were largely confined within the enterprise
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 161
8/3/2019 Content Server 7
7/35
or bank that first employed them. It remains difficult to dismiss or lay off
managers and workers.
In order to assess how and how well the incentive system introduced in the
Chinese SOE reforms performs, it is important to understand how the
incentive schemes operate to impact performance in the institutional
environment of the Chinese SOEs and state banks. One puzzling
phenomenon is that actual bonus payments to managers at the end of the
year systematically differ from those agreed upon at the beginning of the
year. In our dataset, contract renegotiation occurred in more than 90 per
cent of the cases. Such contract renegotiation is common in other SOEs.
Typically, good-performing SOE managers get downward revisions of
promised bonuses, while bad-performing ones are not punished finan-
cially, as they are supposed to be according to their original contracts (see,e.g., Shirk [1993]).
This kind of renegotiation of incentive contracts is puzzling, because it
does not fit in with the standard explanations. One obvious candidate for an
explanation would be that incentive contracts are not carried out as they
were agreed upon because they cannot be enforced by the shaky Chinese
court system. However, this cannot answer the question, why are those
contracts written at all, if they are expected to be non-binding ex post? And
why do they seem to have significant impacts on performance, according to
various empirical studies (e.g., Groveset al
. [1994]; Li [1997]; and Jeffersonand Rawski [1999])? The standard theories of renegotiation based on
adverse selection and moral hazard cannot explain contract renegotiations
in the Chinese banks and SOEs. Renegotiations of incentive contracts in theChinese banks and SOEs take place after performance is realized. This is
inconsistent with renegotiation in moral hazard models, where renegotia-
tion can only happen after the agent exerts effort but before the output is
realized. It is also different from renegotiation in multi-period adverse
selection models in that contracts can be renegotiated from the current
period to future periods, but the contracts of the current period are binding.To understand incentive contracting and renegotiations in the Chinese
banks and SOEs, we need to incorporate important features of theinstitutional environment into the standard agency model. First, supervisors
themselves are employees in state banks and SOEs which are burdened with
a rigid institutional system, and whose objective functions are quite different
from that of the principal in the standard agency model. Specifically, they
themselves come within the same incentive system as the managers under
their supervision, and hence care about the overall performance of branches
they are responsible for. However, they are not mere profit or outputmaximizers. As employees of the state banks and SOEs, supervisors have totake into account the goals and objectives of their organizations The SOEs
162 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
8/35
paternalism referred to by Kornai [1980, 1992] when he describes the
typical social relation between superiors and subordinates in SOEs. More
generally, supervisors in large organizations may care about income equality
among their subordinates because they want to minimize costly influence
activities arising from unequal distribution of wages and bonuses (Milgrom
[1988]), or to avoid being accused of favoritism (Prendergast and Topel
[1996]). Supervisors equality concerns can also result from considerations
that unequal distribution among managers may make them less cooperative
towards one another or may directly reduce their utilities if managers have a
desire for fairness, inequality aversion or have other social psychological
considerations such as envy (Fehr and Schmidt [1999]). Moreover,
supervisors in the Chinese state banks and SOEs probably have strong
tendencies toward centrality bias and leniency bias that are welldocumented in the personnel literature for firms in the developed
economies.9 Combining these reasons, supervisors in the Chinese banks
and SOEs can have quite a strong concern for equality to maintain balance
among their managers in terms of wages and bonuses.
Secondly, as is typical in the strictly hierarchical Chinese banks and
SOEs, supervisors have a great deal of discretion over the well-being and
careers of managers under their supervision. Supervisors control evaluation,
recommendation and other decisions contributing positively or negatively
to managers promotion prospects. Supervisors have discretion overmanagerial perks (e.g., company apartments, cars, etc.) and whether and
how far to intervene in the management of local branches (e.g., on personnel
decisions, task assignments) so much as to affect managers rents-on-the-job. A supervisors discretion over managers rents can come in a variety of
forms, many of which are very hard to specify in advance and extremely
difficult for outsiders to verify (e.g., participation in a delegation that will
visit the United States on a business trip). Thus, such rents cannot be
specified ex ante in incentive contracts, and are at the discretion of the
supervisorsex post
.
10
Thirdly, supervisors tend to care more about income equality than rent
equality among their managers for the following reasons. Some rents, suchas opportunities to enhance promotion chances (e.g., slots in training
programs) are intrinsically unequal, so supervisors are not expected (and
will not be able) to act equally with regard to promotion prospects among
9 See, for example, Murphy and Cleveland [1991] and Coens and Jenkins [2000]. Fordiscussions and references on leniencybias and centrality bias, see Prendergast [1999] section
2.2.10 The broad discretion of supervisors over subordinates, such as promotions, taskassignments, and housing allocations, has been thought of as a salient institutional feature
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 163
8/3/2019 Content Server 7
9/35
managers under their supervision. In addition, while managers incomes can
always be compared on a dollar to dollar basis, it can be very difficult to
compare various kinds of perks and favors and career development
opportunities each manager receives. Moreover, it can be very hard to
discern whether supervisors discretionary decisions over managers career
issues are purely favors (which would require justification from the equality
perspective) or purely business decisions. Therefore, relative to income
equalization, supervisors can find it harder to equalize rents on the one hand
and feel much less pressured to do so on the other.
In the next section, we build a simple model to capture these important
features of incentive contracting and renegotiations in the Chinese state
banks and SOEs, and show that they give rise to a renegotiation process
different from those described in the standard models of renegotiation.
III. THE MODEL AND THEORETICAL ANALYSIS
In our model, a supervisor S (county branch manager) supervises a numberof ex ante identical managers (township branch managers). Let M be a
representative manager. M is expected to exert some minimally acceptable
effort, and has a guaranteed employment with a fixed wage w0. The wage w0
is fixed in two senses. First, it is independent of the managers performance(unless in extreme events such as criminal acts, in which case he will be
demoted or fired). Second, the wage formula is fixed by the bank
headquarters in Beijing, so it is out of the control of S. Suppose M can
exert extra effort e to improve the performance of his branch: x5 e y,where x is a performance measure and y is a random variable that is
independent across managers. For simplicity, and to ensure non-negative x,
we assume y is uniformly distributed on [0, Z]. We will call Ms extra effort
beyond the minimal acceptable level simply his effort. Ms private effort
cost is 0.5ge2
, whereg4
0 is a cost parameter.To motivate M to exert effort, at the beginning of the year S offers M a
linear incentive contract: w5w0 ax, where the choice variable a ! 0measures the ex ante incentive intensity of the contract. After signing theincentive contract with S, M chooses an effort level e. The performance
measure x is then observed by both M and S near the end of the year. Our
model departs from the standard moral hazard model in the final contract
execution stage. Instead of enforcing the ex ante contract based on the
observed performance x, at the end of the year, M and S can renegotiate to
reach a different agreement. Note that since performance is realized andknown to both parties, renegotiation at this stage cannot be driven by risk-
sharing concerns (even if M is more risk-averse than S) and thus is not the
164 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
10/35
simplicity, and fits well with the reality of the Chinese state banks and firms
since the supervisor has authority over the managers.
During the renegotiation stage in our model, S will choose the amount of
rents that she will give to M, and also adjust the ex ante bonus rate a to a0,
which represents the ex post incentive intensity. As discussed earlier, the
rents S allocates to M during the renegotiation stage represent the
supervisors ex post discretion over Ms conditions of employment, such as
whether and how much to intervene in Ms management of his branch (e.g.,
on personnel decisions), and other tangible and intangible perks and career
opportunities (e.g., training opportunities). These kinds of rents are suitable
for renegotiation instruments, because they are difficult to specify ex ante
and hence cannot be easily included in the ex ante contract.11
LetdA
[0,D
] be the amount of rents S gives M during the renegotiationprocess. We suppose that M is risk neutral, thus his payoff function isu5w0 ax bd 0.5ge
2, where b4 0 is a parameter measuring how much
M values rents.
We suppose that the supervisors payoff function from supervising a
representative manager M is given by
v A 0:5d2 r ax 0:5lax b2
where the interpretation of each term is as follows:
The first term A is a positive constant. The second term 0.5d2 is the cost to S of giving rents dto M. The third term (r a0)x reflects the fact that S benefits from good
performance by M (where the parameter r4 0 measures how much S
values Ms performance) and that a bonus payment to M reduces her
payoff. As discussed in the preceding section, the supervisor herself is an
employee like the managers, except that she is one level above M in thebanks hierarchical structure. Thus S is evaluated and rewarded by her
own superior based on the aggregate performance of all branches under
her supervision. The parameter r depends on how strong the supervisors
own incentive intensity is. The last term 0:5lax b2 represents the supervisors equality
concerns, where b denotes the bonus benchmark which the supervisor
uses to compare Ms bonus payment, and the parameter l measures how
strongly the supervisor cares about income equality. As discussed earlier,supervisors in the Chinese banking sector and SOEs have quite strong
equality concerns regarding their subordinates and thus we expectl tobe
large. How b is determined will be specified later. Roughly speaking, b
can be thought of as the average bonus among branch managers,
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 165
8/3/2019 Content Server 7
11/35
hence ax b measures how far a particular Ms bonus is from theaverage.
Our model has symmetric and complete information, thus we use
backward induction to solve for the symmetric subgame perfect equilibrium(henceforth equilibrium) of the game in which ex ante all managers are
treated equally.
III (i). Renegotiation Stage: Ex Post Incentives
In the renegotiation stage, given the ex ante incentive scheme a and the
realized performance measure x, S chooses dand a new bonus scheme a0 to
maxd;a0
v A 0:5d2 r a0x 0:5la0x b2
subject to Ms participation constraint:
w0 bd a0x 0:5ge2 ! w0 ax 0:5ge
2
which can be simplified as a0x bd ! ax.Solving for the supervisors optimization problem, we have (all technical
proofs are in the Appendix)
Lemma 1. When ax ! B, S will lower a to a0 a 11lb2 Bx lb2
1lb2, and in
exchange offers M rents d lbaxB1lb2
. When axoB, S will increase a to
a 05B/x, and gives Mno rents.
Lemma 1 is easy to understand. The variable B b 1=l can be thoughtof as an adjusted bonus benchmark. When the bonus payment to M in
accordance with the ex ante contract is larger than the benchmark B,Swould
like to renegotiate it down in exchange for more rents to M. In this case,
Lemma (1) says that the ex post bonus paymenta0x is a linear combination ofthe ex ante bonus payment ax and the benchmark B. Consequently, the ex
post incentive intensitya0 is increasing in the ex ante incentive intensitya and
decreasing in the realized performance measure x. On the other hand, when
the bonus payment to M according to the ex ante contract is smaller than the
benchmark B, S does not need to provide M with any rent (d5 0), but her
equality concern leads her to offer a higher bonus to M.
III (ii). Managerial Optimal Effort Choice
Given an ex ante incentive intensitya, if M exerts effort e, his expected payoffcan be calculated as follows If ax5 a(e y) ! B or if y ! ye B=a e
166 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
12/35
w0 a0x 0.5ge25w0 B 0.5ge
2. Therefore, Ms expected payoff is
1
Eu w0 0:5ge2 B
y
Z
ZZy
ae y
Zdy
w0 0:5ge2 B
y
Z
a
ZeZ ey 0:5Z2 0:5 y
2
Using dy=de 1, solving from the FOC for Ms optimal effort gives
2 e Za B
Zg a
For the second order condition to hold, it must be that aoZg. For e to be
non-negative, it must be that a ! B/Z. Otherwise, ifao
B/Z, then e5
0.Intuitively, since M expects to receive the bonus benchmark B no matter how
low his performance is, exerting a small amount of effort does not increasehis actual income when aZoB (note that M receives B ex post as long as
a(e y) a(e Z)oB). But effort is costly. Thus M will exert effort only ifaZ ! B. When Equation (2) gives Ms optimal effort, it must be thatZ2g4 aZ ! B. It follows that
3de
da
Z2g B
Zg a
2>0
Lemma 2. If the ex ante incentive intensity a ! B/Z, then M will chooseeffort in accordance with Equation (2); ifaoB/Z, Mwill choose e5 0. Ms
optimal effort is increasing in a.
Lemma 2 shows that the ex ante contract can provide incentives for
managerial effort even if it is anticipated that it will be renegotiated. The
reason is that the bonus and rents M will receive ex post is determined by his
ex ante contract when y !
y, thus, greater ex ante incentive intensity inducesgreater managerial effort. Moreover, higher managerial effort reduces y and
hence increases the likelihood that M gets a higher payoff, which provides an
additional incentive for M to exert more effort.
III (iii). Optimal Ex Ante Incentive Intensity
Ex ante, S is to choose an incentive intensity a to maximize her expected
payoff in the whole game. Once a is chosen, S can correctly anticipate Ms
optimal effort choice given by Lemma 2. Since managers are ex anteidentical, S will choose the same a for all managers; and managers willchoose the same effort level under the same contract (at least in the
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 167
8/3/2019 Content Server 7
13/35
8/3/2019 Content Server 7
14/35
and when performance is bad, i.e., y ! y 0:5Z 1=al, a0 is
7 a0 a
1 0:5Z 1=al y
x
Proposition 1 characterizes the equilibrium of the whole game. In deriving
the results in Proposition 1, since the exact solution is not easy to obtain, for
simplicity we used an approximation by assuming thatl2 is sufficiently large
relative to other parameters of the model (see the proof in the appendix for
details). This approximation does not affect the qualitative properties of the
solution and has no effect on the comparative statics results we derive below.The institutional environment in the Chinese state banks also fits well with
the assumption of large l, where supervisors are paternalistic towards their
subordinates (Kornai [1980], [1992]).
III(iv). Theoretical Predictions of the Model
From Proposition 1, it is easy to derive the following result regarding the
comparative statics of our model.13
Proposition 2. The optimal ex ante incentive intensity (a) and the
managerial performance (x) are increasing in the supervisors own incentive
intensity (r) and the managers preferences for rents (b), and decreasing inthe effort cost parameter (g) and the riskiness of the environment (Z).
While the effects ofr and g can appear from the standard principal-agent
models, the effects of b and Z arise because of the new features of
renegotiation in our model. The intuition is as follows. When the manager
has stronger preferences for rents (greater b), the supervisor can use rents
more effectively to reduce his ex post bonus payments if he achieves above
average performance. Since expected ex post bonus payments will be lower
and hence the cost of providing incentives is lower, the supervisor can usehigher a to motivate the manager to work harder, leading to greater
expected performance. When the environment is more risky (greater Z), the
manager exerts less effort for a given incentive intensity a, because it is more
likely that his performance falls below the bonus benchmark, and hence he
gets paid the benchmark bonus independent of his effort (Equation 4). This
increases the cost of providing incentives, reducing the optimal incentive
intensity and performance.14
13 For Z, since l is assumed to be relatively large, the term 2/(Zl) in the denominatorof Equation (5) will be relatively small.
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 169
8/3/2019 Content Server 7
15/35
We next derive implications on contract renegotiation. Let Da a0 a
be the difference between ex post and ex ante incentive intensity. Then
Da4 0 if y
8/3/2019 Content Server 7
16/35
For largel, the above expression is increasing ina because the denominator
is decreasing in a, and the numerator is not much affected by a. Thus, the
degree of renegotiation Y is also increasing in r and b for downward
renegotiation. As above, the comparative statics with respect to g and Zare
ambiguous. Summarizing, we have
Proposition 4. The degree of renegotiation, namely, the absolute value of the
difference between ex post and ex ante incentive intensities, Y5 |Da|, is
increasing in r and b.
Proposition 4 states that the degree of renegotiation will be greater when
the supervisor faces stronger incentives herself (larger r), or when the
manager has stronger preferences for rents (largerb
). The reason for theseresults is as follows. For larger r or larger b, the optimal ex ante incentive
intensitya and the managerial effort e will be greater. As a result, the bonus
benchmark will be higher. If a managers performance falls below this
benchmark, his bonus payment will be adjusted upwards to the benchmark
during the ex post renegotiation stage. In this case, the higher the
benchmark, the greater the degree of renegotiation. If a managers
performance exceeds the bonus benchmark, his bonus payment will be
reduced to a linear combination of the ex ante bonus payment and the bonus
benchmark during theex post
renegotiation stage. In this case, the degree ofrenegotiation, namely the amount of adjustment, will be greater if the ex ante
bonus payment and the benchmark are both larger. Therefore, for both
downward and upward renegotiations, we obtain the same comparativestatics results about the amount of bonus adjustment with respect to r andb.
These predictions cannot be easily anticipated if one is considering a model
with renegotiation using an exogenously fixed bonus benchmark.
IV. DATA
The data were collected by one of the authors and colleagues from rural China
during the summer of 1998. The survey randomly sampled 59 townships in 15
counties in Jiangsu and Zhejiang, two of Chinas most developed coastalprovinces, one north and the other south of Shanghai. As discussed earlier, the
hierarchical structure of Chinas banking system resembles the governments
administrative system; thus, each township surveyed has one ABC branch and
one RCC branch. In only three instances was a bank branch missing from the
survey, and in total, 57 ABC and 58 RCC branches were sampled.
The bank survey had two main components. The first component,conducted through interviews with township branch managers, acquired
detailed information about branch managers incentive contracts actual
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 171
8/3/2019 Content Server 7
17/35
of the bank branches provided the second part of the data by filling out a set
of financial tables from their branchs financial accounting records. For
various reasons outside our control, there are a certain number of obser-
vations with missing values on performance and contract data.17 Since our
study requires information on performance and contractual terms, we end
up with between 79 and 115 observations. To maximize the sample size for
the analysis, we allow the number of observations to vary for different
regressions. Despite the small sample size, this dataset is still quite valuable,
given the rich information about contracts and renegotiations it contains.
For both ABC and RCC branches in our dataset, a typical incentive
contract for branch managers is constructed as follows. First, the contract
specifies a number of performance measures: deposit growth, loan
performance (i.e., percentage of performing loans), bank safety, and otheradministrative and party duties. Each performance measure is scaled to
points which typically range from zero to 100. The better the performance,
the higher the points obtained. If a branch manager performs very poorly, he
might get zero points. In any case, the points cannot be negative. To get a
measure of overall performance, the contract uses a weighting system,
assigning relative weight to each performance measure. For example, if
deposit growth, loan performance and all other measures have weights of
25%, 25%, and 50%, and a branch manager receives 90, 80, and 100 points
on the three measures, then his overall performance amounts to a weighedsum of three performance points, which equals 92.5 points. Finally, the
contract specifies the amount of bonus or reward each performance point is
worth. In the previous example, if each performance point is worth 50yuan,
then the branch manager should get 4,625 yuan, according to the contract.
Supervisors at county banks determine the weighting system and the
incentive intensity (i.e., how much each performance point is worth) for each
branch manager under her supervision. The weights assigned for each
branch in a county are not exactly the same, but very similar. The variations
in weights mainly come from the cross-county differences.In our empirical study, we focus on deposit growth and loan performance
and ignore other performance measures. Deposit growth and loanperformance are two key measures of bank performance. To achieve a high
growth of deposit, a branch manager must try very hard to develop ties with
township and village enterprises and convince them to deposit their funds in
his branch. Achieving high loan performance is obviously not an easy matter
in rural China. Just around the time when our survey was underway, Chinese
state banks accumulated so many bad loans that Chinas policy makers were
afraid of a potential financial crisis. To improve loan quality, Chinese banks
172 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
18/35
introduced strong incentive mechanisms in China, and some even let bank
managers hold lifetime responsibility for bad loans. Understandably, with
strong emphasis on loan performance, bank managers became fairly
conservative or risk-averse in their lending.
The importance of deposit growth and loan performance is reflected in
their relatively high weights among all performance measures in the
weighting system, each with an average weight of 26% (Table I). In some
cases, the weight is as high as 75% for deposit growth, and in others, loan
performance accounts for an even greater weight. Other measures,
accounting for less than 50% on average in the weighting system, are
mainly non-business oriented (e.g., bank safety, administrative and party
duties) and hard for outside observers to interpret. Profit does enter into the
contracts of some branch managers, but its weight is generally low (less than10% on average), and nearly half of the branches have zero weight for profit.
This is understandable because it is hard to measure, and bank branches
have little control over many of the variables (e.g., interest rates, wages, and
numbers employed) that affect profits.
Table I
SumaryStatistics
VariablesNumber of
Observations MeanStandardDeviation Min Max
Branch performance measureDepoist growth (a) 82 0.24 0.29 0.41 1.45Performing loans/all loans (b) 93 0.83 0.24 0.36 1.00Weight on deposit growth (c) (%) 80 25.60 15.07 0 75.00Weight on loan performance (d) (%) 80 25.73 12.48 0 78.79Weighted bank performance measure(ac bd)
80 0.59 0.24 0.36 1.37
Wage and bonus reward schemesFixed wage (A)(1,000 yuan) 111 9.48 2.73 3.48 18.00Ex ante reward per point (e) (yuan/point) 115 35.17 36.87 4.50216.00Ex post reward per point (f) (yuan/point) 109 48.51 44.36 2.94259.26Actual total bonus reward (1,000 yuan) 109 4.37 3.20 0.20 18.00
Branch incentive measureEx ante incetive intensity (B5 ce de) 80 19.89 22.24 0 118.8Normalized ex ante incentive intensity (B/A) 80 2.22 2.27 0 9.93Ex post incentive intensity (C5 cf df) 79 22.56 21.56 0.38108.79Normalized ex post incentive intensity (C/A) 79 2.56 2.58 0.08 13.89
Contract renegotiation measuresProportion of contracts renegotiated 79 0.94Upward renegotiation (I5upward,05otherwise)
79 0.32
County banks performance incentives(bonus/wage ratio)
101 0.77 0.36 0.18 1.67
Branch mangers age (years) 115 39.90 7.60 28 60Branch mangers education (years) 113 12.63 1.66 9 19Branch mangers years of residence in thetownship
115 17.92 19.67 0 60
Range of industry growth rate during 19947 105 0 55 0 37 0 1 61
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 173
8/3/2019 Content Server 7
19/35
Focusing on deposit growth and loan performance, we construct a
weighted branch performance measure that equals the weighted average of
deposit growth and percentage of performing loans, where the weights are
their respective incentive weights. This will be used as the empirical measure
of the overall branch performance x in our theoretical model.18 Table I
reports the summary statistics of the performance measures and other
variables. In our sample, bank deposits on average grew 24 per cent, similar
to the macro-level deposit growth rates in both Jiangsu and Zhejiang
provinces in the same year (1997). Percentage of performing loans was on
average 83 per cent of all the loans in the sample branches, somewhat higher
than the national average.19 Both performance measures, as well as the
weights, exhibit a great deal of variation across bank branches in the sample.
The weighted branch performance measure we construct has a mean of 0.59and a standard deviation of 0.24, with a minimum of 0.36 and a maximumof 1.37.
Information on contract terms is also reported in Table I. On average,
branch managers in our sample get paid a fixed wage of 9,480 yuan, and a
year-end bonus of 4,370 yuan, or about 46 per cent of the fixed wage. This
indicates that incentive schemes are important for branch managers since
bonuses are quite a significant portion of their total income. It can also be
seen that both the fixed wage and the bonus have quite large variations in the
sample. In theex ante
contracts, the bonus reward for each performancepoint has a mean of 35.17 yuan and a standard deviation of 36.87, with a
minimum of only 4.5 yuan and a maximum of a hefty 216 yuan. Since we
focus on deposit growth and loan performance, we define the ex anteincentive intensity corresponding to a in the model as the bonus reward per
performance point specified in the ex ante contracts weighted by deposit
growth and loan performance. Specifically, the ex ante incentive intensitya is
defined as the total weights on deposit growth and loan performance
multiplied by the bonus reward per performance point and then divided by
the fixed wage (in thousand yuan). We use the fixed wage as a normalization,because fixed wage variations largely reflect the differences in standard of
living across locations. Our empirical results are qualitatively unchanged if
18 Here we do not consider issues of effort allocations related to multi-tasks (Holmstrom andMilgrom [1991]). Multitasking is an important topic that deserves thorough investigation, butis beyond the scope of this paper. Moreover, it appears that the relative weights of differentperformance measures were very close within each county, and thus probably reflected mostlycounty supervisors preferences. The composite measure of performance is sufficient for ourpurposes for two reasons. First, contract renegotiations were about incentive intensities (i.e.,
bonus payment per performance point) but not the weighting system. Second, we get similarregression results by using deposit growth and loan performance separately as managersperformance measures.
174 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
20/35
we drop this normalization. As can be seen from Table 1, before
normalization, the ex ante incentive intensity has a mean of 19.89 yuan
per weighted performance point and a standard deviation of 22.24. After
normalization, it has a mean of 2.22 and a standard deviation of 2.27.
However, managers usually did not get paid the bonus rewards exactly as
specified in the ex ante contracts. Renegotiations of incentive contracts were
prevalent in our dataset. On the whole, 94 per cent of incentive contracts
were ex post renegotiated to varying degrees, of which 25 per cent were
renegotiated upward, and the rest renegotiated downward. Contract
renegotiations took the form of adjusting the reward per performance
point. From Table I, the ex post reward per point has a mean of 48.51 yuan, a
large increase from the ex ante reward per point of 35.17 yuan. As with the ex
anteincentive intensity, we define the
ex post incentive intensitycorrespond-ing to a0 in the model as the bonus reward per performance point actually
paid to the managers weighted by deposit growth and loan performance.
The ex post incentive intensity has a mean of 22.56 yuan per weighted
performance point and a standard deviation of 21.56. After being
normalized by the fixed wage, it has a mean of 2.56 and a standard deviation
of 2.58.20
Figure 1 presents a scatter plot of ex ante and ex post incentives, both
normalized by wage.21 Consistent with Lemma 1, there is a strong positive
correlation betweenex ante
andex post
incentives. More importantly, theslope is clearly less than 45 degrees. Without renegotiation, all the
observation points should be on the 45 degree line. The fact that the slope
is below the 45 degree line implies that managers with good performance arelikely to receive bonuses less than their ex ante contacts specified.
One unique prediction of our theory, that there is a positive relationship
between the supervisors incentives and the renegotiation adjustment (the
absolute value of the difference between ex post incentive intensity andex ante incentive intensity), is confirmed by the data. Figure 2 shows that the
renegotiation adjustment (with normalization) indeed increases withthe county bank incentive intensity, i.e., the supervisors bonus/wage ratio.
We will test this relationship more rigorously in Section V.Table I also reports the summary statistics of other variables used in the
empirical analysis. We use the ratio of bonus to fixed wage of county bank
employees, termed the county banks performance incentive, to measure
the supervisors incentive intensity, r, in the model. This ratio is on average
0.77, significantly higher than that of township branch managers. Note that
20 In the subsequent analysis in Section 5, we mainly use the normalized incentive measuresfor branch managers. However, our results remain qualitatively unchanged if we only use
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 175
8/3/2019 Content Server 7
21/35
each county corresponds to one supervisor of either the ABC or the RCCs.
We use age to proxy for the managers preference for rents, b. Younger
managers have longer career horizons and value more career development
opportunities (e.g., slots in training programs), and thus have higher b than
older managers. In our sample, branch managers are about 40 years old onaverage, with the youngest at 28 and the oldest at 60. Education and years of
residence in the township are used as proxies for g, because managers withmore education and living in the locality longer (i.e., with more local
experience and connections) have a lower effort cost (smaller g). Finally, the
riskiness of the environment, Zin the model, is proxied by the range of the
annual growth rates of per capita real industrial GDP in the townships
during the period 19941997.22
Expostincentives
Ex ante incentives
Ex post incentives Fitted values
0 9.93104
.079657
13.8961
Figure 1
A Scatter Plot ofEx ante and Ex post Incentives
The ex ante incentives are defined as the total weights on deposit growth and loan performance
multiplied by the bonus reward per performance point specified in the ex ante contracts and then
divided by the fixed wage. The ex post incentives are defined in the same way as the ex ante
incentives except that the bonus reward per performance point is the actual one.
22 For example, if the annual growth rates of per capita real industrial GDP in a township
176 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
22/35
V. EMPIRICAL TESTS
In this section we present the results testing the theoretical predictions of the
model regarding the ex ante incentive intensity and branch performance
(Proposition 2) and renegotiations (Propositions 3 and 4).
V(i). The Ex Ante Incentive Intensity and Bank Performance
Proposition 2 states that the ex ante incentive intensity (a) increases with the
county supervisors performance incentives (r), the township managers
preference for rents (b), and responsiveness to incentives ( g), but decreaseswith the riskiness of the environment in the township (Z). The linear
regression for testing Proposition 2 can be expressed as follows:
10 a a0 a1 r a2 age a3 education a4 years a5 Z e
where r is the county bank managers bonus wage ratio to measure the
supervisors incentive intensity age education and years (of residence in
Renegotiationadjustment
County bank incentive intensity
Renegotiation adjustment Fitted values
.176471 1.66667
0
3.45238
Figure 2
A Scatter Plot of Renegotiation Adjustment against County Bank Incentive Intensity
Renegotiation adjustments refers to the absolute value of the difference between ex post incentive
intensity and ex ante incentive intensity, and country bank incentive intensity refers to the ratio of
bonus to fixed wage of county bank employees.
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 177
8/3/2019 Content Server 7
23/35
potential difference between ABC and RCCs, we also include a bank type
indicator (ABC5 1, RCCs5 0) in some specifications. Proposition 2
predicts that a14 0, a2o 0, a34 0, a44 0 and a5o 0.
In Column 1 of Table II, we report the regression result testing the main
predictions of Proposition 2 regarding the determinants of the ex ante
incentive intensity. The major findings are the following. First, consistent
with the theoretical prediction, the county banks performance incentives
have a positive effect on the township managers ex ante incentive intensity.
It is significant at the five per cent level. The estimate of the coefficient is
1.959, which means that an increase in the county bank performance
incentive (bonus wage ratio) by one standard deviation (0.36) will increase
the township branchs ex ante incentive intensity by 32 per cent of its mean.23
Second, consistent with the prediction that theex ante
incentive intensity isdecreasing in the responsiveness to incentives, managers with more
education and local experience have ex ante contracts of greater incentive
intensities, although only the coefficient of education is significant. Third,
consistent with the prediction that the ex ante incentive intensity increases
in b, the branch managers age has a negative effect on the ex ante in-
centive intensity, although it is not statistically significant at the conven-
tional level.
Finally, the estimated effect of the range of township industrial growth
has the expected sign, though not significant. The insignificance of thisvariable as well as others is probably a consequence of the relatively small
number of observations in our sample. In addition, this can also be caused by
other counterweighing forces such as those summarized by Prendergast[2002b], i.e., incentives and risk are positively correlated. Prendergast [2000,
2002a, b] provides several theoretical explanations as to why, contrary to the
prediction of the standard moral hazard model, incentives can be positively
correlated with risks. These explanations include (i) supervisor favoritism;
(ii) costly investigations; and (iii) firms trade-off between direct monitoring
and delegation. These effects may operate in our settings to cause thepositive correlation between incentives and risks (proxied by the range of
township industrial growth).Proposition 2 also states that bank performance increases with the county
supervisors performance incentives and the township managers human
capital, but decreases with the riskiness of the environment in the township.
We test this part of Proposition 2 by using the weighted performance
measure as the dependent variable with the same regression specification as
above except that the dependent variable is now replaced by the weighted
performance measure.
178 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
24/35
Tabl
eII
RegressionResults D
ependentvariable
Exanteincentiveintensity
(normalized)
Weightedbranch
per
formance
Upward
renegotiation
Upward
renegotiation
Upward
renegotiation
Degreeof
re
negotiation
Fullsample
Fu
llsample
Fullsample
RCCs
ABC
F
ullsample
OLS
OLS
Probit
Probit
Probit
OLS
(1)
(2)
(3)
(4)
(5)
(6)
banks
performanceincentives
wagera
tio,r)
1.959
0.150
0.853
0.566
1.812
0.904
(2.58)
(2.43)
(3.93)
(1.87)
(3.23)
(1.74)
anagersage(
b)
0.073
0.006
0.028
0.041
0.002
0.087
(1.65)
(1.28)
(2.61)
(3.15)
(0.13)
(1.23)
anagerseducation(
g)
0.483
0.011
0.072
0.054
0.165
0.187
(2.27)
(0.76)
(1.69)
(1.02)
(2.53)
(2.55)
anagersyearsofresidenceinthe
p(
g)
0.026
0.003
0.010
0.007
0.010
0.031
(1.64)
(2.21)
(2.36)
(1.25)
(1.54)
(1.56)
ofindus
trygrowthrateduring1994
0.162
0.145
0.146
0.240
0.263
1.671
(0.24)
(2.81)
(0.78)
(1.05)
(0.94)
(1.68)
pe(RC
C5
0,ABC5
1)
0.448
0.014
0.279
0.982
(0.92)
(0.27)
(20.9)
(1.82)
ations
78
66
79
41
38
67
ed
0.22
0.16
0.23
heexant
eincentiveintensityisdefinedasthetotalweightsondepositasthegrowth
andloanperformancemultipliedby
thebonusrewardperperformancepointspecifiedin
ntecontractsandthendividedbythefixedwa
ge.Upwardrenegotiationisadumm
yvariablewhichequalsoneiftheex
postincentiveintensityexceedstheex
anteincentive
.Thedegreeofrenegotiationisdefinedasth
eabsolutevalueofthedifferencebetweenex
postandex
anteincentivein
tensities(normalizedbyfixedwage).Countrybank
eintensityreferstotheratioofbonustofixedwageofcountybankemployees.Num
bersinparenthesesaret-ratios.Forcolumns(3)-(5),marginaleffects(dF/
dx)ratherthan
ficientsarereported.Significancelevel0.1,0.05and0.01arenotedby,
,and,
.
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 179
8/3/2019 Content Server 7
25/35
8/3/2019 Content Server 7
26/35
opportunity to test this prediction. It is likely that the county supervisors of
RCCs have a larger equality concern for two reasons. First, RCCs are
collectives, while ABC is a state-owned bank, and thus RCCs may care
more about equality. Second, the headquarters of the RCCs are at the
county level, and thus the supervisors as the top managers are likely to care
about the equality of the whole organization. In contrast, the ABC county
branches are only a bottom level of the hierarchical structure of a large state
bank with the headquarters in Beijing, which may not care as much about the
equality of its managers as is the case with the RCCs. Because RCCs have
stronger equality concerns, we expect that the impact of the county
supervisors incentives on the probability of upward renegotiation will be
smaller.
Regression results reported in Columns 45 of Table II indeed show thatthe impact of the county supervisors incentives on the probability of
upward renegotiation is smaller for RCCs than ABC. The impact for the
ABC is 1.812, which more than triples that for RCCs (0.566). This result
provides strong support for our model, which predicts that the impact of the
county supervisors incentives on upward renegotiation will be smaller when
they have stronger equality concerns.
Finally, we test Proposition 4, which predicts that the degree of
renegotiation increases with the county supervisors incentive intensity
and the branch managers preferences for rents. The dependent variable, thedegree of renegotiation, is defined as the absolute value of the difference
between ex post and ex ante incentive intensities (normalized by fixed wage),
and independent variables are the same as before. The regression result(Column 6, Table II) indeed shows that the county supervisors incentive
intensity has a positive and significant effect on the degree of renegotiation,
which is consistent with the predication of Proposition 4. The bank
managers age has a negative effect on the degree of renegotiation, as
predicted by the model, though it is insignificant.
To summarize, the empirical test results generally support the theoreticalpredictions of our model. The key variables have the expected effects
just as the model predicts, and are both statistically and economicallysignificant in most cases. Although the empirical results on Proposition 2 are
just standard outcomes from agency models, the evidence in support of
Propositions 3 and 4 is unique for the test of our theoretical model, which
does not follow from simple intuitive arguments based on conventional
agency models or renegotiation models. These predictions are mainly driven
by the trade-off between the supervisors equality concerns and the
managers preference for rents. Incentive contracts affect performance,but they do so through mechanisms more complicated than the standardagency model suggests Institutional environment and organizational
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 181
8/3/2019 Content Server 7
27/35
VI. DISCUSSIONS ON ALTERNATIVE EXPLANATIONS
VI(i). Manager Heterogeneity and Relative Performance
One concern about our analysis is how robust our results are if we take into
account differences in managerial ability. In our model, we assume that allbranch managers in a county are ex ante identical, but our data indicates
clearly that managers differ in their age, education, experience, etc.
Obviously, our symmetry assumption is made to greatly simplify the
analysis by reducing the problem of multiple agents into that of a single
agent. However, we believe this modeling device is for simplicity and is not
the driving force of the main results. When managers are observably
different (e.g., age, education, experience), the supervisors equality
concerns will most likely allow for a certain amount of income differences
among managers. For instance, if everyone in an organization feelsthat more senior or experienced managers should be paid more, then
equalizing income among managers will be considered as unequal. If
the heterogeneity of managerial ability can be fully taken into account
in the supervisors equality concerns, then our analysis does not need to
change much because the residual income inequality after taking into
account the explicit differences will still be symmetric among managers.
If the supervisors equality concerns do not account for the observable
differences of managers, then the supervisor, in the renegotiation stage,
will allocate rents and re-adjust ex post incentive intensities exactly asin the current model. However, managers with different abilities will have
different best responses to the same incentive contract and the supervisor
will need to design different incentive contracts for different managers.
While the analysis will unavoidably become very cumbersome, the
main points of the current model should still be valid. Empirically,
managers observable personal characteristics are all controlled for in ouranalysis.
A further extension of our analysis is to consider the situation in which
managers abilities may not be observable to supervisors, that is, thesituation with adverse selection. With adverse selection, there could either be
a separating equilibrium (though the state bank many not be allowed to offer
a menu of contracts) or a pooling equilibrium in which all managers choose
a same incentive contract. In a separating equilibrium, the managers
abilities are revealed through choosing different contracts, then the rest ofthe game is played as a complete information subgame just as in our model.
In a pooling equilibrium, managers with the same incentive contract and
different abilities will choose different effort levels, but the supervisor will
allocate rents and adjust bonuses based on the observable performances ofthe mangers. In either case, the qualitative results of our model regarding
182 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
28/35
Finally, since the situation we analyze has multiple agents, another
question naturally arises: are relative performance contracts used?
Theoretically, relative performance contracts are useful when there is some
common shock in the agents performance measures. However, they are not
useful in our model as y is assumed independent across branches.
Empirically, our survey of the managers compensations does not show
any evidence of relative performance evaluations.
VI(ii). Subjective Performance and Relational Contracts
One may wonder whether contractual renegotiations observed in our data
can be explained by relational contracting. In the literature on relational
contracting (e.g., MacLeod and Malcomson [1989]; Baker, Gibbons andMurphy [1994]; and Levin [2003]), firms use discretionary and non-binding
bonus (based on subjective performance measures) promises to motivate
managers to exert efforts, while the bonus promises are made credible
through repeated interactions (relational capital). Take the model of Baker,
Gibbons and Murphy [1994] as an example. They show that under certain
conditions, bonus payments based on both subjective and objective
performance measures can be optimally combined to provide proper
incentives. Since subjective performance measures are not observable to
outsiders, theex post
bonus payment will be different from the bonuspayment according to the ex ante contract based on objective performance
measures.
Relational contracts, however, do not seem to be able to explain the
pattern of renegotiations in our data. As shown before, the ex post bonus
adjustments in our context are negatively correlated with the objective
performances. That is, managers with good performances are more likely to
see their actual bonuses reduced relative to the ex ante contracts. To be
consistent with Baker, Gibbons and Murphy [1994], one must assume that
the subjective performance measures are negatively correlated with theobjective performance measures. This is highly unlikely in our context.
VI(iii). An Alternative Explanation Based on Profit-Sharing
One possible explanation of the positive correlation between the county
supervisors incentive intensity and the likelihood of upward renegotiation is
profit-sharing within a firm. That is, if the firm does well, everyone in the firm
gets a greater bonus that year, including the county supervisor as well as
branch managers. In this case, the county supervisors incentive intensity, asmeasured by the actual incentive pay of the county supervisor, may have
picked up the impact of profit sharing
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 183
8/3/2019 Content Server 7
29/35
that the degree of upward renegotiation (i.e., ex post incentives minus ex ante
incentives) should be positively correlated with the county supervisors
incentive intensity. However, both our theory (Proposition 4) and empirical
tests (Column 6 of Table II) show that the county supervisors incentive
intensity is positively correlated with the degree of renegotiation (i.e., the
absolute value of the difference between ex post incentives and ex ante
incentives).25 Secondly, the organizational difference between the ABC and
RCCs is also inconsistent with the profit-sharing hypothesis. Profit sharing
would suggest a stronger correlation between the county supervisors bonus
income and township managers bonus income for the RCCs than for the
ABC because RCCs are cooperatives and hence more likely to share profits.
But our results in Columns 4 and 5 in Table II show the opposite.
VII. CONCLUSIONS
Motivated by the observations coming from the Chinese banking industry,
we build a model of incentive contracting and renegotiation thatincorporates some of the essential organizational features of the Chinese
state banks into the standard agency framework. In our model, incentive
contracts are almost always renegotiated in equilibrium, yet they still affect
performance even though renegotiation is fully anticipated. Contract
renegotiations result from the trade-off between supervisors equalityconcerns and managers rents-on-the-job that come under their supervisors
discretion. We then use a dataset from the Chinese banking industry to test
the predictions of the model. Overall, our theory of renegotiation fits theevidence quite well.
Our theory can be applied to contract renegotiations in other contexts.
For example, consider the so-called fiscal contracting system adopted in
19801993 in China (Shirk [1993]). Researchers have long noticed the high
and pervasive levels of complaints among provincial officials about the
central governments discretion in the enforcement of intergovernmentalfiscal contracts. For instance, the central government created numerous
means to level up revenue-submission ratios for those provinces exhibitinggood fiscal performance, including forced borrowing or reclaiming central
ownership of lucrative local firms (Wong et al. [1995]; Ma [1997]). However,
there is also evidence that despite these enforcement problems, provincial
officials still responded to fiscal contract incentives (Jin et al. [2005]). These
observations can be easily explained by our theory. Our model also suggests
that it may be an over-simplification to draw conclusions about the effects of
25 We run an OLS regression where the degree of upward renegotiation is the dependent
184 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
30/35
the incentive system on provincial fiscal performance simply by regressing
performance on the incentive intensity of the agreements.
As a final note, we would like to point out two limitations of our study.
One is the relatively small sample size, which causes some of the parameter
estimations to be statistically insignificant. Another limitation is that our
theoretical model is static, and correspondingly our empirical analysis is
cross-sectional due to data limitations. In future research, it would be
interesting to study the dynamics of contracting and renegotiating when the
principal has equality concerns.
APPENDIX
Proof of Lemma 1: If Ms participation constraint is binding, then substituting
a0x5ax bd into the objective function (and ignoring constant terms) gives0:5d2 bd 0:5lax bd b2. Solving for the first order condition, we get
11 d lb1=l b ax
1 lb2
lbax B
1 lb2
where we define B b 1=l. We will assume that l is sufficiently large so that B4 0.d ! 0 requires that ax ! B b 1=l. When axoB, then d50. The second ordercondition is clearly satisfied.
Then the optimal ex post bonus scheme is
12 a0 a bd=x a 11 lb2
Bx
lb2
1 lb2
In this case, it can be calculated that Ss payoff is n5A n2, where
13 v2 rx B 0:5
l
0:5lax B2
1 lb2
If Ms participation constraint is not binding, then S will set d50. Solving for a0
from the first order condition, we get
14 a0
b 1=lx
Bx
The second order condition is clearly satisfied. In this case, it is easy to show that Ss
payoff is n5A n1, where
15 v1 rx B 0:5=l
So the lemma is proven. Q.E.D.
Proof of Proposition 1:
Since y B=a e 0:5Z 1=al, Ss expected payoff can be written as
Ev A
Zyv1=Zdy
ZZv2=Zdy
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 185
8/3/2019 Content Server 7
31/35
Rewriting Equation 15 gives
v1 re ry ae 0:5Z 0:5=l r ae 0:5Za ry 0:5=l
Since ax b ae y ae 0:5Z ay 0:5Z, by Equation 13, we have
v2 v1 0:5lax B2
1 lb2 v1
0:5lay 0:5Z 1=l2
1 lb2
Plugging in the values ofn1 and n2 into En gives
Ev A r ae 0:5Za 0:5rZ 0:5=l 0:5l
Z1 lb2
ZZy
ay 0:5Z 1=l2dy
A 0:5rZ 1=l r ae 0:5Za l
6aZ1 lb20:5Za 1=l3
C r ae 0:5Za l6Z1 lb2
0:5Z3a2 30:5Z2a=l 1=al3
where Cis a constant.
The FOC with respect to a is
r ade
da e 0:5Z
l
6Z1 lb220:5Z3a 30:5Z2=l 1=a2l3
0:5r a
g
1
Zgl 0:5Z
1
6Z1 lb2lZ3a=4 3Z2=4 1=al2
0
Simplifying it, we have
3rZ1 lb2
g
61 lb2
gl 3Z2lb2 1:25
1
al2
alZ3=4 6Z1 lb2
g
It is easy to check that Ss objective function is globally concave in a, and thus the first
order condition gives the optimal effort choice.The exact solution to this equation is not easy to obtain. However, the solution is
easy to visualize. The right hand side is linear in a, and can be thought of as the
marginal cost ofa. The left hand side can be thought of as the marginal benefit ofa.
It consists of several constant terms plus 1/(al2), and is hence strictly decreasing in a.
From the shapes of the marginal cost and benefit of a, clearly there is a unique
solution to the first order condition. As an approximation, we suppose that l2 is
sufficiently large relative to the other parameters in the model and relative to the
feasible range ofa, so that the term 1/(al2) on the left hand side is sufficiently small and
can thus be omitted. This approximation makes the marginal benefit curve flat and
shifts it down, resulting in an underestimation of a. However, the properties of thesolution, in particular the comparative statics with regard to other parameters, are not
affected by this approximation That S cares strongly about wage equality (large l) is
186 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
32/35
With this approximation, we can solve for a from the first order condition:
16 a ffi3rZ 6
l 3gZ21 1
41lb2
lgZ3
41lb2
6Z
By Equations 4 and 5, we get Ms effort choice as
17 e 0:5a=g 1=Zgl ffi3rZ 6l 3gZ
21 141lb2
lg2 Z3
21lb2 12gZ
1
Zlg
Ms performance is simply x5e y.
By Equations 12, 14 and 5, we obtain the ex post incentive intensity as follows. When
y ! y, or ax ! B, the ex post incentive intensity is given by
18
a0 a1
1 lb2
ae y a0:5Z y 1=l
x
lb2
1 lb2
a1 y 0:5Z 1=al
x
lb2
1 lb2
When y
8/3/2019 Content Server 7
33/35
Fudenberg, D. and Tirole, J., 1990, Moral Hazard and Renegotiation in AgencyContracts, Econometrica, 58, pp. 12791320.
Gibbons, R. and Waldman, M., 1999, Careers in Organizations: Theory and Evidence,in Ashenfelter, O. and Card, D. (eds.), Handbook of Labor Economics, 3B (North-
Holland, Amsterdam, The Netherlands).Gil, R., 2004, Renegotiation and Relational Contracts, Unpublished paper, University
of California at Santa Cruz, U.S.A.Groves, T.; Hong, Y.; McMillan, J. and Naughton, B., 1994, Autonomy and Incentives
in Chinese State Enterprises, Quarterly Journal of Economics, 103(4), pp. 183209.Holmstrom, B. and Milgrom, P., 1991, Mutlitask Principal-Agent Analyses: Incentive
Contracts, Asset Ownership and Job Design, Journal of Law, Economics andOrganization, 57, pp. 2552.
Jefferson, D. and Rawski, H., 1999, Ownership, Performance and Economic Reforms(Oxford University Press, Oxford, England).
Jin, H.; Qian, Y. and Weingast, B., 2005, Regional Decentralization and Fiscal
Incentives: Federalism, Chinese Style, Journal of Public Economics, 89, pp. 17431762.Kornai, J., 1980, Economics of Shortage (North-Holland, Amsterdam, The Netherlands).Kornai, J., 1992, The Socialist System: The Political Economy of Communism (Oxford
University Press, Oxford, England).Laffont, J. and Tirole, J., 1988, The Dynamics of Incentive Contracts, Econometrica, 56,
pp. 11531175.Levin, J., 2003, Relational Incentive Contracts, American Economic Review, 93,
pp. 835857.Li, W., 1997, The Impact of Economic Reform on the Performance of Chinese State
Enterprises, 19801989, Journal of Political Economy, 105, pp. 10801106.Ma, J., 1997, Intergovernmental Relations and Economic Management in China (St.
Martins Press, New York, New York, U.S.A.).MacLeod, B. and Malcomson, J., 1989, Implicit Contracts, Incentive Compatibility, and
Involuntary Unemployment, Econometrica, 57, pp. 447480.Milgrom, P., 1988, Employment Contracts, Influence Activities, and Efficient
Organization Design, Journal of Political Economy, 96, pp. 4260.Murphy, K. R. and Cleveland, J., 1991, Performance Appraisal: An Organizational
Perspective (Allyn and Bacon, Boston, Massachusetts, U.S.A.).Park, A.; Brandt, L. and Giles, A., 1997, Giving Credit Where Credit is Due: The
Changing Role of Rural Financial Institutions in China, mimeo, William DavidsonInstitute, University of Michigan, Ann Arbor, Michigan, U.S.A.
Prendergast, C., 1999, The Provision of Incentives in Firms, Journal of EconomicLiterature, 37, pp. 763.
Prendergast, C., 2000, What Trade-Off of Risk and Incentives?, Amercian EconomicReview, 90, pp. 421425.
Prendergast, C., 2002a, Uncertainty and Incentives, Journal of Labor Economics, 20,pp. 115137.
Prendergast, C., 2002b, The Tenuous Trade-off Between Risk and Incentives, Journal ofPolitical Economy, 110, pp. 10711102.
Prendergast, C. and Topel, R., 1996, Favoritism in Organizations, Journal of PoliticalEconomy, 104, pp. 958978.
Rey Biel, P., 2007, Inequality Aversion and Team Incentives, forthcoming, Scandinavian
Journal of Economics.Shirk, S., 1993, The Political Logic of Economic Reform in China (University of CaliforniaPress, Berkeley, California, U.S.A.).
188 HONGBIN CAI, HONGBIN LI, LI-AN ZHOU
8/3/2019 Content Server 7
34/35
Wei, F., 2000, Chinas Financial Sector Reform in the Transition to a Market Economy:Key Issues and Policy Options (LIT, Hamburg, Germany).
Wong, C.; Heady, C. and Woo, W., 1995, Fiscal Management and Economic Reform in thePeoples Republic of China (Oxford University Press, Hong Kong).
The World Bank, 1997, China 2020Development Challenges in theNew Reforms, A WorldBank Country Study (The World Bank, Washington, D.C.).
CONTRACT RENEGOTIATIONS IN THE CHINESE BANKING INDUSTRY 189
8/3/2019 Content Server 7
35/35
Copyright of Journal of Industrial Economics is the property of Blackwell Publishing Limited and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.