Journal of Comparative Economics 28, 716 –738 (2000)

doi:10.1006/jcec.2000.1681, available online at http://www.idealibrary.com on

A Multitask Theory of State Enterprise Reform 1

Chong-En Bai
The University of Hong Kong, Hong Kong, and WDI

David D. Li
Hong Kong University of Science and Technology, Hong Kong, and WDI

Zhigang Tao
The University of Hong Kong, Hong Kong

and

Yijiang Wang
University of Minnesota, Minneapolis, Minnesota 55455, CEPR, and WDI
E-mail: [email protected]

Received May 19, 1999; revised July 27, 2000

Bai, Chong-En, Li, David D., Tao, Zhigang, and Wang, Yijiang—A Multitask Theory
of State Enterprise Reform

During transition, maintaining employment and providing a social safety net to the
unemployed are important to social stability, which in turn is crucial for the productivity
of the whole economy. Because independent institutions for social safety are lacking and
firms with strong profit incentives have little incentive to promote social stability due to
its public good nature, state-owned enterprises (SOEs) are needed to continue their role in
providing social welfare. Charged with the multitasks of efficient production as well as

1
This is a revised version of our earlier working paper titled “State Enterprises in Transition: A
Multi-Task Perspective.” We thank John Bonin, two anonymous referees, and participants of the
William Davidson Institute’s 1997 Workshop and the 1997 AEA Meetings for their many helpful
comments. We also thank Roger Gordon, Yingyi Qian, Gérard Roland, Jan Svejnar, and Chenggang
Xu for earlier discussions on the topic. Partial financial support from the URC research grant of HKU
(Chong-En Bai), DAG99/00/BM24 and the Chinese SOE project of the HLCOR of the HKUST
(David Li), the Research Grants Council of Hong Kong (Zhigang Tao), and the Graduate School’s
Grant-in-Aid, University of Minnesota (Yijiang Wang) are gratefully acknowledged.

0147-5967/00 $35.00 716

Copyright © 2000 by Academic Press
All rights of reproduction in any form reserved.
 MULTITASKS AND STATE ENTERPRISES 717

social welfare provision, SOEs continue to be given low profit incentives, and conse-
quently their financial performance continues to be poor. J. Comp. Econ., December 2000,
28(4), pp. 716 –738. The University of Hong Kong, Hong Kong, and WDI; Hong Kong
University of Science and Technology, Hong Kong, CEPR, and WDI; The University of
Hong Kong, Hong Kong; University of Minnesota, Minneapolis, Minnesota 55455,
CEPR, and WDI. © 2000 Academic Press
Key Words: dual-track reform; multitask; social stability; unemployment.

Journal of Economic Literature Classification Numbers: P31, P23, P20, D21, J23, L20.

1. INTRODUCTION
This paper presents a theory to explain why in almost all transition economies,
reforms of state-owned enterprises (SOEs) have been slow and SOEs continue to
have poor financial performance. The theory starts with the observation that, during
transition, a social safety net has to be established in order to achieve social stability.
Otherwise, social instability caused by mass unemployment would create an unde-
sirable general environment for business and thereby lower the overall efficiency in
the economy. However, at the start of reform, independent agencies specializing in
providing a social safety net are missing because, before transition, SOEs functioned
as the main social welfare providers. It is very difficult to establish quickly an
institution independent of the SOEs to provide the safety net even if the funding is
available. Therefore, during transition, the government chooses to slow down SOE
reform and to keep a certain number of SOEs in order to maintain social stability.
The remaining SOEs continue to be charged with multiple tasks, i.e., the task of
production and that of social welfare provision.
In Section 2, we use a stylized model to illustrate this theory. There are three
groups of results. First, if the existing level of social stability is low, the reformist
government should maintain a certain proportion of SOEs during transition; i.e.,
a complete privatization of SOEs should be delayed. The rationale is that, by
delaying privatization, the multitasked SOEs continue to face low profit incen-
tives and, therefore, have little incentive to divert unobservable effort from social
welfare provision to production. Meanwhile, with profit as their sole objective,
private firms have little incentive to spend resources on social stability due to its
public good nature. Second, as a result of multitasking, SOEs’ financial perfor-
mance is inevitably poor during reform. In particular, the profitability of an SOE
is lower than that of a private firm. This is not only because SOEs have to spend
resources on maintaining social stability but also because they are given low
profit incentive by the government. Third, when the existing level of social
stability is higher or the total factor productivity of an SOE is low relative to that
of a private firm, the reformist government should reduce the proportion of
SOEs, i.e., the process of privatization should speed up.
After illustrating the key insights with a stylized model, we extend our analysis to
two important cases. The first is a reformist government at the beginning of transition
that can directly enhance social stability at a cost, e.g., by using tax revenue collected
718 BAI ET AL.

with a social cost. We show that, when the cost is very high, the government chooses
not to spend sufficient resources itself on social stability. As a result, SOEs are relied
upon to provide needed social stability in transition. Conversely, when the cost is
very low, the government chooses to spend sufficient resources itself on social
stability so that SOEs are not relied upon to maintain stability. Hence, they are
reformed at the beginning of the transition. The other extension is a consideration of
unemployment in the labor market as a source of social instability. All the main
results carry over to the extended models.2
After the theory sections, empirical evidence from China’s reform experience
is presented in Section 5. We argue that the basic premise and predictions of the
theory are supported by evidence from the experience of China’s state enterprise
reform. Although we feel that the theory is also relevant to other transition
economies, future empirical work is needed to confirm this conjecture.
Our theory of SOE reform is based on a second-best argument. The conclusion
that a certain proportion of SOEs should continue to exist during reform is driven
by the condition that the government cannot find other means to provide a social
safety net for unemployed workers. The first-best solution is achievable if an
independent social security system can be established so that this task can be
separated from the SOEs. Then the government can provide strong profit incen-
tives for SOEs, including possibly outright privatization, and thereby improve
their performance. The key reason for this last result is that the poor financial
performance of the SOEs is not simply a reflection of the direct financial
resources that the SOEs spend for social security and welfare purposes. In other
words, it is not simply a result of the redistribution of economic benefits
produced by SOEs. Instead, it reflects also the lower overall efficiency in SOEs
due to weak profit incentives applied to induce them to carry out the multiple
tasks. However, given that an independent institution to provide the public good
does not exist and can not be established quickly, the second-best social optimum
is to keep a fraction of the firms in the economy state-owned and charge them
with the task of producing the public good.
Our theory is not simply one of government intervention to deal with an
externality. Although the externality is a critical element in our theory, at the
center of our work are the questions of how private incentives affect the private
supply of a public good and how the government should design incentives to
achieve the best balance between private profit incentives and the private supply
of public goods. 3 To answer these questions, we derive our results on the optimal
speed of privatization.

2
An extension to incorporate product market competition can also be made. For details, please see
an earlier version of this paper (Bai et al., 1999).
3
Bai and Tao (2000) take a similar perspective to explain the coexistence of company owned and
independently owned branches in franchising. Their model pays close attention to the context of
franchising and is very different from the one in this paper.
 MULTITASKS AND STATE ENTERPRISES 719

Several important observations are critical for our theory. The first is based on
how social stability is maintained in transition. Unlike most mature market
economies, there are no independent establishments specializing in providing
social services and a social safety net in a transition economy. Moreover, in the
short run, it is very difficult and costly, if not impossible, for the government to
establish such institutions. Existing bureaucracies, e.g., ministries of labor,
finance, and civil welfare, may compete to obtain control rights of the new
institution. Also, due to the stretched political capacity of the reformist govern-
ment, monitoring such new institutions is difficult. Therefore, SOEs should
continue to play the role of providing such services. A typical SOE in transition
economies was established as a minicommunity with all social service facilities,
such as a canteen, a medical clinic, a beauty saloon, a theatre, and shuttle buses.
It is very costly for the government to use tax revenue to replace such functions
of SOEs. 4 Our observation is that SOEs represent the dominant channel through
which social services and the social safety net are provided in transition.
The second observation is that the efforts of SOE managers to provide social
services and to maintain social stability are not observable. Although the gov-
ernment can instruct an SOE to keep a certain number of workers on its payroll,
which indeed has been the case in reality, it cannot observe directly how well an
SOE provides social services to its employees. Therefore, SOE managers must
not face high-powered profit incentives that induce them to cut costs by not
treating surplus workers properly. Similarly, a tax on SOEs for laying off
workers, the so-called unemployment tax, cannot induce SOE managers to
provide social services, since this can at most prevent the SOE from laying off
workers. In China, a well-known tendency is that SOEs treat surplus or unpro-
ductive workers as second-class citizens by depriving them of many benefits,
e.g., denying them health care benefits and delaying the repair work for the
broken windows of their company-provided apartment. This tendency threatens
to stir popular resentment against reform. Based on this observation, our assump-
tion is that an SOE manager can escape government monitoring and divert efforts
from providing social services to employees to seeking higher profits. This is the
basis of our multitask theory.
The third important observation is that obstacles exist in the labor market that
prevent it from clearing during transition. Should the labor market clear, the
social welfare problem on which this paper is based would not exist. Then all

4
In the context of dealing with the problem of social instability associated with unemployment, it
is often suggested that a payroll-type tax be levied so that the employed pay taxes to be used to
subsidize the unemployed as an insurance device. Such a scheme can not substitute for the role of
SOEs because without the social services of SOEs, it is very costly to compensate an unemployed
worker. Given the high cost, we show in Section 3 that the government is likely not to rely on such
taxes. Furthermore, with the incidence of the tax born partly by the enterprise, this actually induces
more unemployment and is counterproductive. Hence, despite such insurance schemes, reforms of
SOEs are still slow in reality.
720 BAI ET AL.

SOEs could be made single-task agents. The government’s concern for income
distribution and its wage regulation to address this concern are a plausible reason
for labor market imperfection in a transitional economy (see Gordon and Li,
1999).
Our analysis of the social safety net departs from those in the traditional
literature that emphasize the trade-off between incentives and insurance for those
covered by a social safety net. 5 A social safety net provides insurance to
economic agents but takes away their work incentives at the same time. This
concern relates to the principal-agent relationship between workers and employ-
ers. Although the trade-off is an important issue, there is another issue that is
unique in a transition economy and is more urgent. This issue involves the
principal-agent relationship between the state and the SOE managers and how to
provide incentives for SOE managers to continue those social services that were
traditionally supported by SOEs if independent institutions providing a social
safety net cannot be established quickly enough. Since social stability is a public
good, the efficiency of the whole economy is affected. This issue becomes
critical to strategies of SOE reform. In this paper, we focus on the public goods
nature of the safety net and show that this factor slows down SOE reform. 6
Our paper is embedded in a lively literature on strategies of economic transi-
tion. 7 Taking a political economy perspective, Dewatripont and Roland (1992a,b)
argue that, when the reformer has to compensate the losers of reform, a gradualist
reform is likely to be optimal because it can isolate the losers at each stage and
minimize government expenditure. Murrell and Wang (1993) discuss how insti-
tutional resources and structural legacies of a communist economy may lead to
a delay in the privatization process. Qian and Xu (1993) provide insights into the
effect of an economy’s prereform organizational structure on the path of reform.
Most closely related to our work are Castanheira and Roland (1996, 2000), Qian
(1996), Li (1997), and Roland and Verdier (1999). In a dynamic neoclassical
Ramsey model of investment, Castanheira and Roland (1996, 2000) show that
too fast closure of SOEs reduces income and savings so that less investment can
be made in non-SOEs. To explain the slow pace of SOE reforms, Qian (1996)
and Li (1997) argue that the lack of large investors or institutions for corporate
governance require continued government control of SOEs in order to curb the
agency costs associated with excessive insider control. Finally, Roland and
Verdier (1999) argue that high unemployment leads to congestion in the labor
market and excessive search costs for new jobs.
Our paper adds to these existing works by providing a new dimension to the
relationship between SOEs and non-SOEs, i.e., the issue of social stability. We

5
We acknowledge an anonymous referee for pointing this out to us.
6
In other words, we discuss the incentives of the providers rather than the recipients of social
welfare.
7
Two excellent surveys on transition are Dewatripont and Roland (1997) and McMillan (1997).
 MULTITASKS AND STATE ENTERPRISES 721

show that SOE managers should be given low-powered incentives. Theoretically,

our model builds upon Holmstrom and Milgrom’s (1991, 1994) multitask theory.
Our innovation lies in emphasizing the difference in the scope of influence of the
tasks. We introduce into the multitask model the possibility that one of the tasks
has a public good nature. As a result, we prove that it is optimal for the
government to maintain a mix of SOEs and private enterprises, even if these
enterprises are assumed to have identical production functions and are managed
by ex ante homogeneous agents.

2. THEORETICAL ANALYSIS
In this section, we set up a highly stylized model to illustrate the main
arguments of the paper. In the model, a reformist government decides on the pace
of SOE reform by choosing a certain proportion of SOEs not to be reformed
during the period. There are three main predictions of this model. First, if the
initial level of social stability is low, the government chooses to keep some
SOEs; these are given low profit incentives and are instructed to devote efforts
to maintaining social stability. Second, as a result, these SOEs are less profitable
than private firms. Third, when either the initial level of social stability is higher
or the total factor productivity of SOEs is lower, the government should choose
to keep fewer SOEs; i.e., the government should speed up reform.
To illustrate the key insights, the stylized model assumes a fixed product price
and does not explicitly consider the labor market. However, the main results
remain unchanged quantitatively when these assumptions are dropped. 8 As an
illustration, in Section 4, we include the labor market by considering employment
as a proxy for social stability. The extension of the basic model to include
product market competition can be found in an earlier version of the paper (Bai
et al., 1999). 9

2.1. The Model

Suppose that there are an infinite number of firms in the economy, populated
in the unit interval I ⫽ [0, 1]. The firms are of the following two types, private
firms, the owner-managers of which keep all the profit, and state-owned enter-

8
We thank an anonymous referee for suggesting the extension.
9
In that paper, we incorporate the interaction between the SOEs and private enterprises via product
market competition. The introduction of private enterprises in transition economies increases product
market competition and thereby affects adversely the performance of state-owned enterprises. The
ensuing question is whether the government should keep any state-owned enterprises in these
circumstances. Presumably, the provision of social stability by SOEs enhances the incentive for effort
of private enterprises due to the complementarity between social stability effort and production effort.
Meanwhile, due to product market competition, the financial performance of the SOEs deteriorates.
We show that, under reasonable conditions, the former dominates the latter and, hence, it is optimal
to keep some SOEs in transition economies.
722 BAI ET AL.

prises (SOEs), the managers of which are paid fixed wages. We use j to denote
a private firm and i to denote an SOE. The proportion of SOEs is p 僆 [0, 1].
The manager of each firm performs two tasks, s and g. Task s is a specific
effort that affects only the revenue of the firm, and g is a general effort that
increases the revenue of all firms of the economy. Here s and g are called
production effort and stability maintenance effort, respectively. The total cost of
the efforts is born by the manager and is given by
c ⫽ s ⫹ g.

We assume that the levels of the two efforts are not verifiable and hence can
not be the basis of a contract. However, through direct monitoring of managerial
effort, the government can ensure a level of total effort of at least T from each
manager. Given the incentive contracts, the managers choose s and g according
to their self-interest and subject to the constraint that
s ⫹ g ⱖ T.

We assume that the level of social stability G in the economy is given by

G⫽ 冕 I
g共k兲dk ⫹ G 0 ,

where the first term is the integral of the function g over the interval I, or the
aggregate of the stability maintenance efforts of all enterprises, and G 0 is the
initial level of stability plus the government’s effort in providing social stability.
We assume that G 0 ⬎ 0. In our main case, we also assume that G 0 is largely
determined by history and cannot be changed in the short term.
The revenue of each firm is assumed to be
x共k兲 ⫽ ␭ 共k兲 y关s共k兲, G兴 ⫹ ⑀ 共k兲,
where s(k) is firm k’s specific effort, ⑀ (k) is a random variable with mean 0 and
variance ␴ 2, ⑀ is independent across firms, and ␭ (k) is the coefficient of
efficiency of firm k. For a private firm, ␭ ( j) is normalized to be 1. For an SOE,
␭ (i) ⫽ ␭ with 0 ⬍ ␭ ⱕ 1. Thus, ␭ is the total factor productivity of SOEs relative
to private enterprises and is typically less than 1, reflecting lower efficiency of
SOEs even when managerial efforts are the same in the two types of enterprises.
The reason for this lies in various institutional constraints faced by the SOEs. For
example, SOEs are subject to more government intervention than private enter-
prises and this lowers their efficiency. However, the main results of this paper do
not depend on the assumption that ␭ ⬍ 1.
The specification of y assumes that g is a pure public good. Although the
essential point is made most clearly by assuming an infinite number of firms and
 MULTITASKS AND STATE ENTERPRISES 723

the pure public good nature of g(k), the qualitative features of our main results
remain if the economy has a finite number of firms and stability has local effects;
see Section 4 for examples.
Before proceeding to our analysis, we need to make two assumptions on the
production function y(., .). The first one is rather standard:
Assumption 1. Function y is increasing and concave in s and G, and the cross
partial derivative y sG ⬎ 0.
In order to capture the idea that stability is important for a firm’s production,
we assume that when G is very small, y G is very large. We also assume that y s (T,
G 0 ) ⬎ 1, which implies that monitoring is not very effective so that the effort
enforced by monitoring is lower than that induced by proper incentives; see Eq.
(1) below. For the analysis of the benchmark case, i.e., Proposition 1, we assume
further that y G is small when G is very large. Hence, we have a second
assumption:
Assumption 2. We say that lim G30 y G (s, G) ⫽ ⬁ for s ⬎ 0, lim G3⬁ y G (s,
G) ⫽ 0, and y s (T, G 0 ) ⬎ 1, where y G and y s are the first-order derivatives of
y with respect to G and s, respectively.
We assume that managers are risk-neutral to focus on the main reason for the
existence of SOE’s in the model, i.e., SOEs provide public goods, rather than
reduce the risk for agents. 10 The owner/manager of a private enterprise keeps all
the profit of the firm and his utility is
u共 j兲 ⫽ y共s共 j兲, G兲 ⫺ 共s共 j兲 ⫹ g共 j兲兲.
The manager of an SOE receives a fixed wage w i and his utility is
u共i兲 ⫽ w i ⫺ 共s共i兲 ⫹ g共i兲兲.
The manager of a private firm will make no effort to contribute to the public
good G because the marginal cost of providing g( j) is 1 and the marginal benefit
is y G (s( j), G) multiplied by his marginal contribution to G. Since G ⱖ G 0 ⬎
0, y G (s( j), G) ⬍ ⬁. Hence, the marginal benefit of g( j) is very small because
its marginal contribution to G is negligible. It follows immediately that the
manager of any private enterprise will not put in any stability maintenance effort.
In addition, for any given level of general stability, the manager will choose
production effort to maximize his profit. Because y s (T, G 0 ) ⬎ 1 by Assumption
2, optimal s satisfies the first-order condition
y s 共s, G兲 ⫺ 1 ⫽ 0, (1)

10
We do not assume that the workers are risk-neutral. In fact, we assume implicitly that they are
risk-averse and care about the social safety net. We assume only that managers are risk-neutral. Our
results remain valid even if the managers are risk-averse. The proof is available upon request.
724 BAI ET AL.

and s ⫽ s(G) ⬎ T. Assumption 1 implies further that s⬘(G) ⬎ 0. Intuitively,

the owner-manager keeps the profit of his firm and thus has an incentive to
maximize profit. To do so, he allocates no effort to stability maintenance and all
of his effort to production because his stability maintenance effort has only a
small effect on profit while production effort has a nontrivial positive effect.
The manager of an SOE is willing to provide any level of public goods g ⱕ
T because, with a fixed salary, he expends a total level of T effort and is
indifferent to its allocation between production and stability maintenance. The
former is due to the characteristics of government monitoring while the latter
follows because the SOE manager’s payoff does not depend on the profit of his
firm and, consequently, on how his effort is allocated between the two tasks.
Therefore, the manager of any SOE is assumed to do what is requested by the
government so long as s(i) ⫹ g(i) ⫽ T. 11
The fundamental reason for the above results is that stability is a public good,
which implies that the effect of a manager’s stability maintenance effort on the
profit of his firm is generally smaller than that of his production effort. So long
as the manager’s payoff depends on the profit of his firm, he will not fully take
into account the externality of his stability maintenance effort and will provide
too little of it. The government can mitigate the problem of underprovision of
stability maintenance effort only by keeping some SOEs where the managers are
paid fixed wages and are monitored. The results of the above discussion are
summarized in the following lemma.
LEMMA 1. For a private firm, g ⫽ 0 is the optimal choice. For an SOE, the
manager is indifferent to any g in [0, T].
We assume that the objective of the government is to maximize the total social
surplus that equals the unweighted sum of profits of all firms. The timing of
events is as follows. First, the government chooses the percentage of SOEs in the
economy, i.e., p, and makes a uniform effort choice, i.e., g for stability main-
tenance and T ⫺ g for production, for all managers of SOEs, given that the
managers are indifferent to any effort choices satisfying s ⫹ g ⫽ T. Second, the
owner-managers of private firms choose their production efforts, i.e., s.
Recall that Eq. (1) summarizes how the managers of private firms choose s in
response to the stability maintenance level G, which is the second stage of the
game. Now, we analyze the first stage of the game, i.e., how the government
chooses p and g, which jointly determine G, in order to maximize the total social
surplus,

␲ ⬅ p关 ␭ y共T ⫺ g, G兲 ⫺ T兴 ⫹ 共1 ⫺ p兲关 y共s, G兲 ⫺ s兴,

11
Notice that T is the effort level that is exerted by managers without incentive contracts.
Therefore, it is not as high as that of the managers of private enterprises who face high-powered
incentives.
 MULTITASKS AND STATE ENTERPRISES 725

where ␭ y(T ⫺ g, G) ⫺ T is the profit of an SOE, y(s, G) ⫺ s is the profit of

a private enterprise, and G ⫽ pg ⫹ G 0 is the level of social stability.
This section assumes that the owner-managers of private firms keep all the
profits and that the state maximizes the sum of the profits of all enterprises. A
motivation for this assumption is that the government is concerned with main-
taining political control. Social welfare is an index of the average happiness of
people in the society. Higher social welfare helps solidify the government’s
political control. In fact, these assumptions can be relaxed. Results similar to
those in this section can be shown if we assume that the state taxes the profit of
private enterprises and maximizes its revenue, i.e., the sum of taxes from private
firms and profits from state-owned firms.

2.2. The Benchmark Case

We first study a benchmark case in which the government is able to rely on a

social safety net to provide social stability rather than solely on SOEs. Alterna-
tively, the benchmark case can be thought of as one in which the initial level of
social stability, i.e., G 0 , is very high so that social unrest is very unlikely. Not
surprisingly, the optimal choice for the government is to have no SOEs, since
private enterprises are more efficient. In other words, the benchmark case
highlights the main reason for keeping SOEs, i.e., the need to provide public
goods.

PROPOSITION 1. When G 0 is sufficiently large, the optimal p ⫽ 0.

Proof. The partial derivative of ␲ with respect to p is

⭸␲
⫽ 关 ␭ y共T ⫺ g, G兲 ⫺ T ⫺ y共s, G兲 ⫹ s兴
⭸p
⫹ ␭ pgy G 共T ⫺ g, G兲 ⫹ 共1 ⫺ p兲 gy G 共s, G兲. (2)

Notice that ␭ y(T ⫺ g, G) ⫺ T ⬍ y(s, G) ⫺ s, since the right-hand side is the

maximal profit of a private firm while the left-hand side is strictly less than the
maximal profit because the optimal s(G) ⬎ T. By Assumption 2, y G (s, G) goes
to 0 when G ⱖ G 0 goes to ⬁. Therefore, the partial derivative of ␲ with respect
to p is negative for all p when G 0 is sufficiently large. QED

2.3. The Main Case

For the main case, we assume that G 0 is determined by history and it is very
costly for the government to change its level in the short term. Furthermore, the
initial level of G 0 is rather low. Then, we show that it is optimal to have some
SOEs in the economy.
726 BAI ET AL.

PROPOSITION 2. When G 0 is sufficiently small, the optimal p ⬎ 0.

Proof. We first prove that the optimal g does not go to zero as p approaches
zero. The partial derivative of ␲ with respect to g is

1 ⭸␲
⫽ ⫺␭ y s 共T ⫺ g, pg ⫹ G 0 兲
p ⭸g
⫹ ␭ py G 共T ⫺ g, pg ⫹ G 0 兲 ⫹ 共1 ⫺ p兲 y G 共s, pg ⫹ G 0 兲. (3)

For g ⱕ T ⫺ ␦ and ␦ ⬎ 0, the first term above is bounded from below by

⫺ ␭ y s ( ␦ , pg ⫹ G 0 ). Because lim G30 y G (s, G) ⫽ ⬁ when p and G 0 are both
sufficiently small, the third term in Eq. (3) dominates and ⭸ ␲ /⭸g is positive. That
is, for sufficiently small G 0 , lim p30 g ⱖ T ⫺ ␦ .
The partial derivative of ␲ with respect to p is given by Eq. (2). To prove that
the optimal p ⬎ 0, we need to show that ⭸ ␲ /⭸p is positive for p ⫽ 0. When p 3
0 and when G 0 is small enough, because lim p30 g ⱖ T ⫺ ␦ , Assumption 2
implies that the last term in Eq. (2) is very large and dominates other terms.
Therefore, the optimal p is greater than 0. QED
The intuition for this result is as follows. When G 0 is small, the marginal
product of G is large and it is desirable to increase the level of G beyond G 0 .
Since it is extremely costly for the government to do, SOEs are used for this
purpose.
An equally interesting question is under what conditions the government
would choose to have some private enterprises. The simple condition is that
either the effort level by an SOE manager (T) or the total factor productivity of
SOEs (␭) relative to private enterprises is low. Under either condition, the
opportunity cost of keeping a large proportion of SOEs, which is measured by the
extra revenue from private enterprises, can be higher than the benefit of gener-
ating high public goods. To be precise, we have the following proposition:

PROPOSITION 3. When T or ␭ is sufficiently small, the optimal p ⬍ 1 for all G 0 .

Proof. When p ⫽ 1,

⭸␲
⫽ ␭关 y共T ⫺ g, g ⫹ G0 兲 ⫹ gy G 共T ⫺ g, g ⫹ G 0 兲兴 ⫺ T ⫺ 关 y共s, g ⫹ G 0 兲 ⫺ s兴,
⭸p

which is negative when T or ␭ is small enough. Therefore the optimal

p ⬍ 1. QED
The next proposition indicates that an SOE is less profitable than a private
firm. This result highlights the idea that SOEs are maintained to promote social
stability despite their low profitability.
 MULTITASKS AND STATE ENTERPRISES 727

PROPOSITION 4. For any p and G 0 , an SOE produces less profit than a private
firm.
Proof. A private firm’s profit is

y共s, G兲 ⫺ s,
where s is chosen to maximize the profit. An SOE’s profit is

y共T ⫺ g, G兲 ⫺ T ⬍ y共T, G兲 ⫺ T ⬍ y共s, G兲 ⫺ s. QED

We note that the basis for Proposition 4 is not the low total factor productivity,
i.e., ␭, of an SOE relative to a private firm. Even if ␭ ⫽ 1, Proposition 4 still
holds. Two factors are responsible for the low profitability of an SOE. First,
private firms free-ride on SOEs for G because g ⬎ 0 in an SOE but g ⫽ 0 in
a private firm. Second, SOEs have lower profit incentives and therefore have a
lower effort level, i.e., T ⬍ s. In other words, the poor financial performance of
SOEs is endogenous.

2.4. Comparative Statics

We examine the effect of changes of ␭ and G 0 on the optimal choice of p and
g. Intuitively, when ␭, the relative total factor productivity of SOEs, is higher, the
opportunity cost of keeping SOEs is lower so that we expect to see a larger
proportion of SOEs. On the contrary, when G 0 , the level of non-SOE-provided
public goods, is higher, the benefit of keeping SOEs is lower and we expect to
see fewer SOEs at the optimum. These intuitive predictions are verified formally
in the following proposition.
PROPOSITION 5. Suppose y s (s, G) ⬍ ⬁. If ␭ and G 0 are sufficiently small, the
optimal p has the following properties:

⭸p ⭸p
共1兲 ⬎ 0; 共2兲 ⬍ 0.
⭸␭ ⭸G 0
Proof. We first show that the optimal p does not go to 1 as ␭ goes to zero. For
p ⱖ 1 ⫺ ␦ and ␦ ⬎ 0, because y GG ⬍ 0, the last term in Eq. (2) is

共1 ⫺ p兲 gy G 共s, G兲 ⱕ ␦ gy G 共s, pg兲 ⱕ ␦ gy G 共s, 共1 ⫺ ␦ 兲 g兲.

For sufficiently small ␭, the right-hand side of the above inequality and, thus, the
last term of Eq. (2) are less than T. For all p ⱖ 1 ⫺ ␦ , Eq. (2) is negative if ␭
is sufficiently small because y(s, G) ⫺ s ⬎ 0. Therefore, the optimal p is less
than 1 ⫺ ␦ for small ␭.
Next, we show that, when ␭ and G 0 are sufficiently small, the optimal g ⫽ T.
Since p ⬍ 1 ⫺ ␦ and y s is bounded from above, when ␭ and G 0 are sufficiently
728 BAI ET AL.

small, the third term of the right-hand side of Eq. (3) dominates so that ⭸ ␲ /⭸g is
positive for all g. Therefore, the optimal g ⫽ T.
From Propositions 2 and 3, Eq. (2) holds at the optimal p. Since g ⫽ T is a
constant, the optimal p is determined by Eq. (2) only. Checking the Milgrom–
Wilson condition, we have
⭸ 2␲
⫽ y共T ⫺ g, G兲 ⫹ pgy G 共T ⫺ g, G兲 ⬎ 0. (4)
⭸p⭸ ␭
Therefore, ⭸p/⭸ ␭ ⬎ 0. Also,
⭸ 2␲
⫽ ␭ y G 共T ⫺ g, G兲 ⫺ y G 共s, G兲
⭸p⭸G 0
⫹ ␭ pgy GG 共T ⫺ g, G兲 ⫹ 共1 ⫺ p兲 gy GG 共s, G兲, (5)
where the last two terms are negative by assumption; ␭ y G (T ⫺ g, G) ⫺ y G (s,
G) ⬍ 0 because y sG ⬎ 0. Therefore, ⭸ 2 ␲ /⭸p⭸G 0 ⬍ 0. This implies that
⭸p/⭸G 0 ⬍ 0. QED

3. GOVERNMENT PROVISION OF G
Our analysis has assumed a fixed level of G 0 . In this section, we extend the
model by endogenizing the government’s choice of G 0 . We show that, if the cost
to the government for providing G directly is high, as it might be in the initial
stages of transition, the government will not provide sufficient G and, as a result,
SOEs are maintained to provide G. However, if the cost to the government for
providing G is sufficiently low, it is efficient for the government to be the only
provider of G and SOEs are not needed.
Denote G 0 ⫽ G 1 ⫹ G 2 , where G 1 is the initial level of stability and G 2 is the
government’s effort to augment stability. Assume that the cost of government
effort is ␮␾ (G 2 ), where ␾ is a strictly increasing and convex function. Incor-
porating the government’s choice of G 2 into consideration, the optimization
problem becomes
max␻ ⬅ p关 ␭ y共T ⫺ g, G兲 ⫺ T兴 ⫹ 共1 ⫺ p兲关 y共s, G兲 ⫺ s兴 ⫺ ␮␾ 共G 2 兲
p,g,G 2

s.t. G ⫽ G 1 ⫹ G 2 ⫹ pg
s ⫽ arg max关 y共s, G兲 ⫺ s兴.
The following proposition characterizes the optimal solution:
PROPOSITION 6. Suppose G 1 is positive but sufficiently small. (1) If ␮ is
sufficiently large, the optimal p ⬎ 0. (2) If ␮ is sufficiently small, the optimal
p ⫽ 0.
Proof. The government’s maximization problem can be divided into two steps.
In step 1, for a given value of G 2 , choose ( p, g) optimally subject to the
 MULTITASKS AND STATE ENTERPRISES 729

constraints. Denote the maximum value by ␻ (G 2 ) ⫽ ␲ (G 2 ) ⫺ ␮␾ (G 2 ). By the

envelope theorem, we have

␲ ⬘共G 2 兲 ⫽ p ␭ y G共T ⫺ g, G兲 ⫹ 共1 ⫺ p兲 y G共s, G兲.

In step 2, choose G 2 to maximize ␻ (G 2 ).

By Proposition 1, there exists a G⬘2 such that the optimal p ⫽ 0 if G 2 ⬎ G⬘2 .
Let B 1 ⬎ 0 be the minimum value of ␲ ⬘(G 2 ) over the closed interval [0, G⬘2 ].
If ␮ is sufficiently small so that ␮␾ ⬘(G⬘2 ) ⬍ B 1 , ␻ ⬘(G 2 ) ⬎ 0 for G 2 ⱕ G⬘2 ,
which implies that the government should choose G 2 ⬎ G⬘2 . Given such a choice
of G 2 , the optimal p ⫽ 0. Thus, we have proven part (2) of the proposition.
For G 2 ⬎ G⬘2 , the optimal p ⫽ 0. Then ␲ (G 2 ) ⫽ max[ y(s, G) ⫺ s], which
can be easily shown to be concave in G 2 because y is assumed to be concave in
(s, G). Consequently, ␲ ⬘(G 2 ) decreases for G 2 ⬎ G⬘2 and, therefore, the
maximum value of ␲ ⬘(G 2 ) over the interval [0, ⬁) is the same as that over the
closed interval [0, G⬘2 ]. Because G 1 ⬎ 0, ␲ ⬘(G 2 ) is continuous over the closed
interval [0, G⬘2 ] and, therefore, has a finite maximum value over the interval.
Denote the maximum value by B 2 .
By Proposition 2, there exists a G ⬙2 such that the optimal p ⬎ 0 if G 2 ⬍ G ⬙2 .
If ␮ is sufficiently large so that ␮␾ ⬘(G ⬙2 ) ⬎ B 2 , ␻ ⬘(G 2 ) ⬍ 0 for all G 2 ⱖ G ⬙2 ,
which implies that the government’s optimal choice of G 2 is less than G ⬙2 . Given
such a choice of G 2 , the optimal p ⬎ 0, which proves part (1) of the proposi-
tion. QED
The above analysis shows that, if it is possible for the government to directly
enhance social stability at the beginning of transition by expending tax revenue,
the government may or may not choose to expend a sufficient amount of
resources on maintaining stability. When the government’s cost of providing this
public good is very high, it is not worthwhile for the government to do so even
if it has enough tax revenue. As a result, SOEs are kept unreformed to provide
social stability. Only when the government’s cost of providing the public good
is low enough will SOEs be reformed and social stability be maintained from tax
revenue only.

4. EMPLOYMENT AS A PUBLIC GOOD

In the model analyzed in Section 2, g is the effort of a firm that increases the
revenue of all firms in the economy and it is called the stability maintenance
effort. We gave several examples of this effort, including employment. When
severe unemployment threatens social stability, increasing employment helps to
maintain social stability and, therefore, has a positive effect on all firms in the
economy. In this section, we consider a variation of this model to reflect better
this interpretation of g. For simplicity, we abstract from any direct government
provision of G.
730 BAI ET AL.

Assume that the revenue of each firm is

x共k兲 ⫽ y关s共k兲, l共k兲, G兴 ⫺ l共k兲w ⫹ ⑀ 共k兲,

where l(k) is the labor input of the firm, w is the market wage, G ⫽ 兰 k l(k)dk
is the total level of employment in the economy, and all other variables are the
same as those in the previous model. The function y is assumed concave in (s,
l, G) and s, l, G are pairwise complementary. The economy we consider is one
with significant excess labor supply. The wage does not clear the labor market
and is not affected significantly by labor demand. Therefore, we assume that the
wage w does not depend on G.
In the basic model, a firm’s effort g has the same effect on its own revenue as
it does on other firms’ revenue. In contrast, under the current setup, a firm’s labor
input l has a stronger effect on the firm’s own revenue than on other firms’
revenue; that is, l has a positive local effect. This difference between the previous
model and the current model makes the analysis sufficiently different to warrant
separate treatment.
The manager of an SOE is paid a fixed wage and the wage for employees does
not affect the manager’s payoff. Therefore, the manager chooses s ⫽ T and is
indifferent about the choice of l(k). Then, the manager will choose the labor
input following the government’s instruction. Denote by L the level of labor
input that the government instructs the manager to choose.
The owner/manager of a private firm maximizes profit. We assume that the
government cannot subsidize employment in private firms or force these firms
to keep a certain level of employment. If the government did subsidize
employment or had an employment quota, the firms would have incentives to
create false employment to get the subsidy or to fulfill the quota. Further-
more, rules against the firing of workers do not work because the firm could
make an unwanted worker’s life in the firm so unpleasant that he would quit
instead of waiting for the firm to fire him. Given this assumption, the labor
input l( j) and the effort level s( j) are chosen by the owner/manager to
maximize

y关s共 j兲, l共 j兲, G兴 ⫺ s共 j兲 ⫺ wl共 j兲,

where G ⫽ pL ⫹ (1 ⫺ p)l e and l e is manager j’s expectation of other private

firms’ labor input. The equilibrium conditions are

y s 关s, l, pL ⫹ 共1 ⫺ p兲l e兴 ⫽ 1

y l 关s, l, pL ⫹ 共1 ⫺ p兲l e兴 ⫽ w

l ⫽ l e. (6)
 MULTITASKS AND STATE ENTERPRISES 731

Denote the solution by s( p, L) and l( p, L). From equilibrium conditions (6),

when p ⫽ 0, s(0, L) and l(0, L) do not depend on L. Denote s(0, L) by s 0 and
l(0, L) by l 0 . In this case, G 0 ⫽ l 0 . 12
The government chooses p and L to maximize total social welfare,

␲ ⫽ p关 y共T, L, G兲 ⫺ T ⫺ wL兴
⫹ 共1 ⫺ p兲关 y共s共 p, L兲, l共 p, L兲, G兲 ⫺ wl共 p, L兲 ⫺ s共 p, L兲兴,

where G ⫽ pL ⫹ (1 ⫺ p)l( p, L). For simplicity, we have assumed

implicitly that the only social benefit of employment is through the produc-
tion function. In reality, there are other direct benefits, especially when the
labor market does not clear so that there is unemployment. We omit these
direct benefits to focus on our main point. When these direct effects are
included, there will be a stronger case for our main result that it is optimal to
retain some SOEs under certain conditions. Specifically, our main results can
be stated as in Proposition 7:

PROPOSITION 7. The optimal p is positive provided that y G (s 0 , l 0 , G 0 ) is

sufficiently large. In this case, SOEs employ more workers and generate less
profits than private firms.

Proof. To facilitate exposition, we first introduce the notation that

V共G兲 ⬅ max y共s, l, G兲 ⫺ s ⫺ wl,

s,l

which is an increasing function of G, because the Envelope Theorem implies that

V⬘(G) ⫽ y G (s, l, G) ⬎ 0. With this notation, the total social surplus when there
are no SOEs is

␲ 共0兲 ⫽ V共G 0 兲,

and the total social surplus when p ⬎ 0 and L ⫽ l 0 ⫹ l 1 for some l 1 ⬎ 0 is

␲ 共L, p兲 ⫽ p关 y共T, L, G兲 ⫺ T ⫺ wL兴 ⫹ 共1 ⫺ p兲V共G兲,

where G ⫽ pL ⫹ (1 ⫺ p)l( p, l 0 ⫹ l 1 ). Due to the complementarity between

s, l, and G, it can be shown straightforwardly that l( p, l 0 ⫹ l 1 ) ⱖ l 0 for any
positive p and l 1 , provided that the solution to the equilibrium conditions in
(6) is unique and stable. Then, G ⱖ pL ⫹ (1 ⫺ p)l 0 ⫽ G 0 ⫹ pl 1 . Take the
difference between ␲ (L, p) and ␲(0), we have

12
The notation G 0 here has a slightly different meaning from that in Section 2. In both places, G 0
is the level of stability when p ⫽ 0, but here it is not the initial level of stability as defined in
Section 2.
732 BAI ET AL.

␲ 共L, p兲 ⫺ ␲ 共0兲 ⫽ p关 y共T, L, G兲 ⫺ T ⫺ wL ⫺ V共G 0 兲兴

⫹ 共1 ⫺ p兲关V共G兲 ⫺ V共G 0 兲兴
ⱖ p关 y共T, L, G 0 兲 ⫺ T ⫺ wL ⫺ V共G 0 兲兴
⫹ 共1 ⫺ p兲关V共G 0 ⫹ pl 1 兲 ⫺ V共G 0 兲兴
⫽ p关 y共T, L, G 0 兲 ⫺ T ⫺ wL ⫺ V共G 0 兲兴 ⫹ 共1 ⫺ p兲V⬘共G 2 兲 pl 1 ,
where the inequality results from G ⱖ G 0 ⫹ pl 1 ⬎ G 0 and the fact that y and
V increase with G. The last equality is a result of applying the Mean Value
Theorem, with G 2 being some number between G 0 and G 0 ⫹ pl 1 . By the
Envelope Theorem, V⬘(G 2 ) ⫽ y G (s 2 , l 2 , G 2 ), where s 2 and l 2 are the optimal
choices by the owner/manager of the private firm for the given G 2 . Suppose
y G (s 0 , l 0 , G 0 ) is sufficiently large so that
y G 共s 0 , l 0 , G 0 兲 ⬎ l 1⫺1 关V共G 0 兲 ⫹ T ⫹ w共l 0 ⫹ l 1 兲 ⫺ y共T, l 0 ⫹ l 1 , G 0 兲兴
for some positive l 1 . Because 0 ⬍ G 2 ⫺ G 0 ⬍ pl 1 and (s 2 (G 2 ), l 2 (G 2 )) are
continuous in G 2 , for sufficiently small p, (s 2 , l 2 , G 2 ) is very close to (s 0 , l 0 ,
G 0 ). Consequently, (1 ⫺ p) y G (s 2 , l 2 , G 2 ) is very close to y G (s 0 , l 0 , G 0 ) so that
共1 ⫺ p兲 y G 共s 2 , l 2 , G 2 兲 ⬎ l 1⫺1 关V共G 0 兲 ⫹ T ⫹ wL ⫺ y共T, L, G 0 兲兴.
Then,
␲ 共L, p兲 ⫺ ␲ 共0兲 ⬎ 0.
Therefore, the optimal proportion of SOEs, p, is positive.
The profit produced by an SOE is ␲ 1 ⫽ y(T, L, G) ⫺ T ⫺ wL and that
produced by a private firm is ␲ 2 ⫽ V(G) ⫽ V[ pL ⫹ (1 ⫺ p)l( p, L)]. Since
␲ 2 is the value of y(s, l, G) ⫺ s ⫺ w when s and g are optimally chosen while
␲ 1 is the value of the same function when s and g are suboptimal, ␲ 2 ⬎ ␲ 1. The
derivative of the total social surplus, ␲, with respect to p is
⭸␲ ⭸G
⫽ ␲ 1 ⫺ ␲ 2 ⫹ 关 py G 共T, L, G兲 ⫹ V⬘共G兲兴 .
⭸p ⭸p
Because the optimal p is positive, ⭸ ␲ /⭸p ⱖ 0 at the optimum. Then ⭸G/⭸p ⬎
0 at the optimum; that is, total employment increases with the proportion of
SOEs. This can only happen when L ⬎ l( p, L) or, in other words, when each
SOE employs more workers than each private firm. QED

5. EMPIRICAL EVIDENCE
Since our research is motivated by China’s experience of state enterprise
reform, our empirical evidence comes from China. Future empirical research on
other transition economies is needed in order to confirm the more general
relevance of our theoretical analysis. Some of the evidence presented below
 MULTITASKS AND STATE ENTERPRISES 733

justifies the assumptions of our analysis, i.e., social stability is important to

economic growth and it is a public good. Other evidence validates the main
prediction of the model, that is, the reformist government made a conscious
choice of slowing down SOE reform and instructed SOEs to continue contrib-
uting to social stability, which led to the poor performance of SOEs during the
transition.
The premise of our theory is that social stability is intrinsically important for
economic growth and that, during reform, there are threats to social stability.
Investors, especially foreign investors, tend to be very sensitive to social stability.
In surveys, Western investors in emerging market economies often list social
stability as the leading factor affecting their choice of investment location.
Among the threats to social stability in transition economies, massive unemploy-
ment is one of the most prominent. In the Chinese context, the unemployment
problem is particularly acute. As labor market mobility improved during the
reform, approximately 100 million Chinese peasants left their rural residences to
search for jobs in cities. Meanwhile, efforts to improve SOEs’ efficiency led
invariably to large layoffs, not only because SOEs have had large amounts of
surplus workers traditionally but also because many SOEs were making un-
wanted products and should have been liquidated. Some studies estimate that 20
to 30% of the total labor supply in China’s SOEs is redundant (Wang, 1996).
China’s reformist leaders understand the importance of maintaining social
stability for the success of reform. Specifically, they have made it very clear that
they do not see mass layoffs as an appropriate solution to the surplus labor
problem in SOEs. In an important speech on SOE reform in 1997, then Vice
Premier (and now Premier) Zhu Rongji stressed that excessive workers in SOEs
should not be resolved through massive layoffs. Also, cases of bankruptcy of
SOEs should be dealt with more aggressively by using mergers to minimize the
social impact of unemployment that bankruptcy will cause. Zhu Rongji also
instructed the state banks to reduce interest payments for those SOEs searching
for new ways to reemploy their excessive labor. Vice Premier Wu Bangguo, who
is in charge of SOE reform, proclaimed that the government should take it as an
extremely important responsibility that workers under furlough from financially
distressed SOEs are reemployed and that their basic livelihood is guaranteed
(Wu, 1997).
However, unlike the case of developed market economies in which designated
government agencies provide social welfare services, it was the SOEs that had
been the main providers of most social welfare programs before reform. During
reform, they continue to play the same role: “Many Chinese state enterprises
have long operated as social-welfare institutions, with canteens, hospitals and
schools all operating on one budget” (Asian Wall Street Journal (or AWSJ), July
13, 1998, p. 3.) The establishment of separate government agencies to provide a
social safety net can not be accomplished overnight. As early as 1990, when the
Chinese government was drafting the eighth 5-year plan, it emphasized the
734 BAI ET AL.

TABLE 1

Reasons for Producing Money-Losing Products

(Units: Percentage of Responses by Enterprise Directors)

Not imp. Important Very imp. Score Ranking

“Forced” 31.3 23.5 24.8 1.53 0.692

“Lack of tech.” 36.0 33.1 12.7 1.40 0.627
“Subsidy” 40.4 19.8 11.9 1.16 0.600
“Employment” 20.8 31.8 30.4 1.76 0.675
Weight 1 2 3

Note. The table is based on answers to the survey question “why do you produce money-losing
products?” The survey was conducted with 769 SOEs from 1980 to 1989. See Li (1997) for details.
The choices are “Forced” ⫽ “forced by the government,” “Lack of tech.” ⫽ “lack of technology,”
“Subsidy” ⫽ “according to current government policies, we benefit from producing these products,”
“Employment” ⫽ “in order to provide jobs for surplus workers.” Enterprises directors were asked to
evaluate each choice (“Forced” etc.) by using a mark, such as “Not important.” The score is defined
as each choice’s weighted average mark across all enterprises, with the reported weight in the table.
The ranking is the frequency with which an enterprise director chose this option as the highest mark
among all choices, although ties were possible.

importance of establishing a social security system. Toward the end of the

decade, an AWSJ article commented that “Beijing has yet to roll out a nationwide
pension system, leaving many workers laid off without adequate medical and
housing benefits” (AWSJ, Aug. 6, 1998, p. 5). In a front-page editorial, the
People’s Daily said, “The speedy liquidation of state enterprises has generated
more layoffs than the government re-employment programs can handle, . . . . In
some areas, this has led to serious social problems” (AWSJ, Aug. 6, 1998, p. 5).
Our theory predicts that private enterprises do not have any incentives to
employ surplus labor to help maintain social stability, and this is indeed the case
in China. As noted in a report by China’s State Economic and Trade Commis-
sion, local governments had overemphasized selling SOEs as a tool of reform.
The practice has exerted negative impact on local economic development and
social stability because “the first thing private owners typically do to turn around
unprofitable enterprises is to cut loose their bloated work-forces” (AWSJ, July 13,
1998, p. 3). As could be anticipated, “Mounting layoffs have resulted in scattered
labor protests throughout the country” (AWSJ, Aug. 6, 1998, p. 5).
A main prediction of the theory is that SOEs are instructed to maintain
employment so as to help maintain social stability, and, as a result, the
performance of SOEs suffers. In a survey of 769 Chinese SOEs in 1990,
managers were asked why they continue to produce products that lose money.
More than 60% of those managers who responded listed direct employment
considerations for workers as “important” and “very important” reasons (see
Table 1). The other reasons for overproduction, “forced by the government”
 MULTITASKS AND STATE ENTERPRISES 735

TABLE 2

Total Employment of China’s State Sector (Units: Million)

Year 1980 1981 1982 1983 1984 1985 1986

Employment 80.19 83.72 86.30 87.71 86.37 89.90 93.33
Year 1987 1988 1989 1990 1991 1992 1993
Employment 96.54 99.84 101.08 103.46 106.64 108.89 109.20
Year 1994 1995 1996 1997
Employment 108.90 109.55 109.49 107.66

Source. China Statistic Year Book 1995 and China Statistic Year Book 1998.

and “getting subsidies from the government,” are also employment-related.

To provide employment, many Chinese SOEs continue to produce apparently
obsolete, uncompetitive and money-losing products. Over 30% of state
enterprises are taking financial losses and relying on government subsidies
(Bai and Wang, 1998).
A rather creative way for Chinese SOEs to help maintain social stability
while undergoing restructuring is called xia gang. For cases in which paying
workers for doing nothing is more beneficial than continuing to keep them in
productive service, workers are paid an amount for living expenses by the
SOE to “step down from their post” (xia gang). A worker in xia gang status
still enjoys all the standard welfare provisions of the SOE. In 1996, the
number of SOE workers in xia gang status reached nine million, which
accounts for roughly 15% of total employment in SOEs. The meager living
expense paid to xia gang workers is only a small fraction of the cost to
support these workers because of the high benefits that they receive, such as
heavily subsidized company housing and full or almost complete medical
coverage. In a typical Chinese SOE, the ratio of welfare-provision assets,
e.g., company housing, company-run child care and schools, and sports
facilities, to total assets reaches 35 to 40%. Welfare benefits in monetary form
reach more than 50% of the total wage bill (see Liu, 1995).
Overall, with various measures taken by the government during the reform era,
Chinese SOEs have not only avoided major layoffs but have also increased total
employment until very recently. As the figures in Table 2 show, total employ-
ment in the SOEs increased by more than 25% between 1980 and 1992 and
stayed stable between 1992 and 1997. In recent years, despite the significant slow
down of SOE output growth, the total employment of SOEs has not decreased
much. The SOEs have made a major contribution to keeping surplus labor in the
economy and off the streets.
The comparative statics in Proposition 5 predict a positive correlation
between the existing level of social instability and the speed of privatization
but a negative relationship between the relative efficiency of the SOEs and
736 BAI ET AL.

the speed of privatization. This is consistent with recent observations of the

development of privatization in China. 13 From 1993 through 1997, the Chi-
nese economy experienced a period of declining inflation rates and high real
growth. During this time, SOEs became increasingly inefficient (Bai and
Wang, 1998). In September 1997, the Fifteenth Party Congress called for
speeding up the reforms of SOEs. However, the adverse impact of the Asian
financial crises threatened China’s export markets and deflation replaced
inflation in domestic and foreign markets. Economic growth slowed and the
pressure of unemployment, and therefore of social instability, mounted. Not
surprisingly, the government took measures to slow down privatization of the
SOEs (AWSJ, July 13, 1998, p. 3).

6. CONCLUDING REMARKS

This paper presents a multitask theory of state enterprise reform. SOEs are
charged with not only the task of efficient production but also the task of
social welfare provision. The theory starts with the observation that, during
transition, maintaining employment and providing a social safety net to the
unemployed are important to social stability, which in turn is crucial for the
productivity of the whole economy. However, independent institutions for a
social safety net are lacking and firms with strong profit incentives have little
reason to promote social stability due to its public good nature. Therefore, it
is inevitable that SOEs continue to play their multitask role during the
transition. Accompanying the multitask objectives of SOEs, low-powered
profit incentives continue to be imposed on the SOEs so that SOE restruc-
turing is delayed.
The multitask theory has two simple implications. First, it implies that SOE
reform and the economic performance of the SOEs cannot be satisfactory unless
their social welfare task is taken over by independent social welfare institutions. 14
Second, the desirable pace of SOE reform has to take into account the speed of
emergence of supporting institutions, in particular, those replacing the functions
of the SOEs that are unrelated to efficient production. 15

13
Another interesting episode occurred in the early 1980’s when tens of millions of urban high
school graduates who had been sent to the countryside during the Cultural Revolution returned to the
cities, waiting for urban employment. The mounting potential for social instability prevented any
major SOE reform from being implemented. Indeed, large-scale SOE reform started only after the
mid-1980’s, when this social stability issue subdued.
14
This view predicts poor economic performance of SOEs during reform, contrary to what
Jefferson and Rawski (1994) have argued. For the debate on China’s SOE reform, see Woo et al.
(1993), Woo et al. (1994), and Bai et al. (1997).
15
See Kornai (1992) for a philosophical discussion of this view. Murrell and Wang (1993) provide
a general analysis.
 MULTITASKS AND STATE ENTERPRISES 737

REFERENCES
Bai, Chong-en, Li, David D., and Wang, Yijiang, “Enterprise Productivity and Efficiency: When Is
Up Really Down?” J. Comp. Econ. 24, 3:265–280, June 1997.
Bai, Chong-en, Li, David D., Tao, Zhigang, and Wang, Yijiang, “A Multi-Task Theory of the State
Enterprise Reform,” unpublished manuscript. The University of Hong Kong, Dec. 1999.
Bai, Chong-en, and Tao, Zhigang, “Contract Mixing in Franchising as a Mechanism for Public Good
Provision.” J. Econ. Manage. Strategy 9, 1:85–113, Spring 2000.
Bai, Chong-en, and Wang, Yijiang, “Bureaucratic Control and the Soft Budget Constraint.” J. Comp.
Econ. 26, 1:41– 61, Mar. 1998.
Castanheira, Micael, and Roland, Gérard, “Restructuring and Capital Accumulation in Transition
Economies: A General Equilibrium Perspective,” Discussion Paper 1372. Centre for Eco-
nomic Policy Research, Apr. 1996.
Castanheira, Micael, and Roland, Gérard, “The Optimal Speed of Transition: A General Equilibrium
Analysis.” Int. Econ. Rev. 41, 1:219 –239, Feb. 2000.
Dewatripont, Mathias, and Roland, Gérard, “Economic Reform and Dynamic Political Constraints.”
Rev. Econ. Stud. 59, 4:703–730, Oct. 1992a.
Dewatripont, Mathias, and Roland, Gérard, “The Virtues of Gradualism and Legitimacy in the
Transition to a Market Economy.” Econ. J. 102, 411:291–300, Mar. 1992b.
Dewatripont, Mathias, and Roland, Gérard, “Transition as a Process of Large-Scale Institutional
Change.” In David M. Kreps and Kenneth F. Wallis, Eds., Advances in Economics and
Econometrics: Theory and Applications, Volume 2, pp. 240 –278. Cambridge/New York/
Melbourne: Cambridge Univ. Press, 1997.
Gordon, Roger H., and Li, David D., “The Effects of Wage Distortions on Transition: Theory and
Evidence from China.” Europ. Econ. Rev. 43, 1:163–183, Jan. 1999.
Holmstrom, Bengt, and Milgrom, Paul, “Multitask Principal-Agent Analyses: Incentive Contracts,
Asset Ownership and Job Design.” J. Law, Econ. Organ. 7, Special Issue:24 –52, 1991.
Holmstrom, Bengt, and Milgrom, Paul, “The Firm as an Incentive System.” Amer. Econ. Rev. 84,
4:972–991, Sept. 1994.
Jefferson, Gary, and Rawski, Thomas, “Enterprise Reform in Chinese Industry.” J. Econ. Perspect.
8, 2:47–70, Spring 1994.
Kornai, János, The Socialist System: The Political Economy of Communism. Princeton, NJ: Princeton
Univ. Press, 1992.
Li, David D., “Government Control During Transition: Theory and Evidence from China,” unpub-
lished manuscript. Ann Arbor, MI: Department of Economics, University of Michigan, 1997.
Liu, Shijing, “Zhong Guo Guo You Qi Ye de Xing Zhi Yu Gai Ge Luo Ji” (“The Nature of SOEs
and the Logic of Reform”). Jing Ji Yan Jiu (Econ. Res.), 4, 29 –36, Apr. 1995.
McMillan, John, “Markets in Transition.” In David M. Kreps and Kenneth F. Wallis, Eds., Advances
in Economics and Econometrics: Theory and Applications, Volume 2, pp. 210 –239.
Cambridge/New York/Melbourne: Cambridge Univ. Press, 1997.
Murrell, Peter, and Wang, Yijiang, “When Privatization Should be Delayed: The Effect of Commu-
nist Legacies on Organizational and Institutional Reforms.” J. Comp. Econ. 17, 2:385– 406,
June 1993.
Qian, Yingyi, “Enterprise Reform in China: Agency Problems and Political Control.” Econ. Tran-
sition 4, 2:427– 448, 1996.
Qian, Yingyi and Xu, Changgang, “Why China’s Economic Reforms Differ: The M-Form Hierarchy
and Entry/Expansion of the Non-State Sector.” Econ. Transition 1, 2:135–170, 1993.
Roland, Gérard, and Verdier, Thierry, “Transition and the Output Fall.” Econ. Transition 7, 1:1–28,
1999.
Wang, Cheng, “Zhong Guo Jiu Ye Zhuan Yi: Zun Yi Xin Shi Ye, Jiu Ye Bu Zu Dao Xiao Lui Xing
738 BAI ET AL.

Jiu Ye” (“The Transformation of Employment in China: Implicit Unemployment, Underem-

ployment, and Efficient Employment”). Jing Ji Yan Jiu (Econ. Res.), 5, 38 – 46, May 1996.
Woo, Wing Thye, Fan, Gang, Hai, Wen, and Jin, Yibiao, “Efficiency and Macroeconomic Conse-
quences of Chinese Enterprise Reform.” China Econ. Rev. 4, 2:153–168, June 1993.
Woo, Wing Thye, Hai, Wen, Jin, Yibiao, and Fang, Gang, “How Successful Has China’s State
Enterprise Reform Been?” J. Comp. Econ. 18, 3:410 – 437, June 1994.
Wu, Bangguo, “Making Tireless Efforts to Provide Re-employment and Reduce Unemployment-
Related Poverty.” People’s Daily (Overseas Ed.), May 29, 1997.
Zhu, Rongji, “How to Resolve the Difficulties of the SOE?” People’s Daily (Overseas Ed.), Jan. 9,
1997.