Journal of International Development: Vol. 8, No.

5,683-696 (1996)

THE EFFICIENCY IMPLICATIONS

OF CORRUPTION
MUSHTAQ H. KHAN
Department of Economics, SOAS, University of London, London, UK

Abstract: Corruption has different efficiency effects across countries. Conventional

economic models of corruption are shown to be deficient in explaining these differ-
ences. Instead the article suggests that the distribution of power within the
patron-client networks in which corruption is taking place is an important variable
explaining the differences in the efficiency effects of corruption. Where patrons are
powerful the range of rights transacted is limited and the allocation is likely to be
efficiency maximizing. In contrast where patrons are weak the range of rights trans-
acted is likely to be much wider with the rights allocated according to political calcu-
lations with large efficiency costs.

In late 1995 and early 1996 two parallel corruption scandals rocked India and
South Korea. What is interesting is that although the sums of money involved in
the Indian scandal were relatively small (some $20 million dollars spread over 30
or so politicians over two years) there was a general perception that poverty in
India must have something to do with the corruption of its politicians. In contrast,
in the South Korean scandal the sums of money involved were much larger. In late
1995 ex-president Roh Tae Woo admitted that he alone had accumulated a per-
sonal fortune of around $650 million over a five year period in office. While Roh’s
behaviour had equally serious political repercussions in South Korea, it was diffi-
cult to argue that economic performance had been seriously undermined as a
result.
Does corruption have any effect on economic performance? If so, why did the
presence of wide-spread corruption not have more damaging effects in countries
like South Korea? These are the sorts of questions which the economic analysis of
corruption seeks to answer. In this article we review the routes which contempo-
rary economic theory has followed in providing models of the effects of corrup-
tion. We argue that in general these models are unsatisfactory because they ignore
the distribution of power in the patronxlient networks in which corrupt transac-
tions are conducted. This distribution has an important effect on the efficiency of
corruption by determining which rights are subject to corrupt transactions and the
terms under which they are allocated.

CCC 0954-1748/96/050683-14
0 1996 by John Wiley & Sons, Ltd.
684 M.H. Khan

1 DEFINING CORRUPTION

Corruption has been defined in a number of ways in the literature. The basic dis-
tinction is between normative and positive definitions, summarized in Table 1.
Two variants of normative definitions are provided in definitions 1 and 2 in Table
1. The first looks at acts and is clearly normative while the second looks at the con-
sequences of acts. The second remains a normative definition because the defini-
tion of the public interest may differ across observers. It is also problematic to
define corruption in terms of its consequences because this defines out of existence
cases of beneficial corruption. Economic and sociological comparisons most often
use the third, positive, definition using the standard of legal norms to identify devi-
ations. Thus corruption is defined as deviations from the formal rules governing
the allocative decisions of public officials in response to offers to them of financial
gain or political support (based on Nye, 1967).

Table 1. Definitions of corruption.

Normative definitions Positive definitions
1. Deviations from Ethical Norms 3. Deviations from Legal Norms
2. Actions which Harm the Public Interest

The stipulation that corrupt transactions should violate formal rules rather than
simply ethical norms rules out any disagreements about the appropriate ethical
standards. The additional stipulation that public officials are involved distin-
guishes corruption from thefr which is illegal but which exclusively involves deci-
sions by private individuals. The definition is still open to the problem that formal
rules can vary across countries. A strict application of the definition could lead to
different sets of practices being identified as corrupt. Fortunately, the corrupt
practices which economists have wished to analyse would in fact violate formal
rules in most countries.
The extent of corruption is difficult to objectively measure but it is widespread
in many developing countries. Empirical evidence on corruption comes from jour-
nalistic and sociological case studies as well as cross-country indices of corruption
such as those provided by Business International, Transparency International and
the World Competitiveness Report. A number of empirical estimates of the
effects of corruption have been based on such indices (Ades and Di Tella 1996).
These provide a step forward but the subjectivity of local respondents in reporting
the degree of corruption has to be kept in mind. It is likely that respondents in
poorly performing economies will find even petty corruption more oppressive and
therefore rank the extent of corruption higher. The case study based approach
must therefore complement the index based one wherever possible.
The recent revelations of corruption in South Korea are interesting from this
perspective, A major case which came to the surface in the 1990s was the decision
of the Chun administration to disband the Kukje business group (chaebol) when
its chairman refused to make large transfers to President Chun’s chosen funds. In
1993 a South Korean court found the Kukje break-up illegal opening the way for
further cases to be brought and indicating a shift in the overwhelming power of
 The Efficiency Implications of Corruption 685

the South Korean state. The Kukje case was only the tip of the iceberg. The reve-
lations eventually led to the admission by ex-President Roh of his personal
involvement in the $650 million scandal.
The opposition leader Kim Dae Jung, one of the harshest critics of the corrup-
tion of the regime admitted to having received 2 1/2 million dollars from Roh. But
he pointed out that this was considerably less than his opponent, the conservative
victor for the presidency, Kim Young Sam, had received (Financial Times 28
October, 1995). In November 1995 24 of the country’s chaebol were implicated
in the scandal including the big four: Hyundai, Samsung, Daewoo and Lucky
Goldstar. The other chaebol named included a ‘roll-call of Korean industry’
(Financial Times 25 November, 1995). It would be difficult to argue that corrup-
tion was either small-scale in South Korea or limited to a few sectors or politi-
cians. The figures are likely to turn out to be the tip of a much larger set of
redistributions which held the Korean industrial policy system together.
This kind of evidence challenges the conclusions of those who see a uniform
negative relationship between corruption and economic performance. The usual
response to the observation of corruption in dynamic countries has been to argue
that the extent of corruption and the presence of countervailing factors varied
across countries. According to this line of argument, the differential performance
is explained by saying that the practices identified as corrupt were (i) not equally
present everywhere, and/or (ii) that in the successful countries corruption was off-
set by countervailing factors. It is important to establish why this response is not
very satisfactory.
While it is virtually impossible to measure precisely the degree of corruption
with any precision, when we compare specific cases of corruption, such as exam-
ples of collusion between government and business it appears that similar prac-
tices have had very different outcomes. Moreover, the countervailing factors often
turn out to be precisely the details of the bargaining between government and
business rather than variables elsewhere in the economy. Thus the evidence sug-
gests that the problem is not just the extent of corruption but also its type. We
should therefore try and see whether there are differences in the types of corrup-
tion which can account for the differences in observed effects.

2 CONVENTIONAL MODELS OF CORRUPTION

Economic models of corruption attempt to identify the efficiency implications of

changes in resource allocation brought about by corrupt transactions. The first
step in this exercise is to identify the ex ante allocation with which the effects of
corruption are to be compared. This is not as simple as it sounds and many of the
controversies in the literature can be traced to ill-defined benchmarks with which
the post-corruption allocation is being compared. Consider figure 1. Suppose we
are at the stage shown in box 5. Do we compare the efficiency implications of this
allocation with the allocations at stage 4,3,2 or l?
The problem is solved in neoclassical models by choosing the perfectly competi-
tive allocation as the benchmark. Suppose this is shown in box 1. Neoclassical the-
ory tells us that welfare is optimized at this allocation and any intervention takes
us away from this optimal level of welfare. Consequently the welfare at stage 2 is
686 M.H. Khan

lntervention
1. Allocation of Rights

4. New Allocation of Rights -4

Intervention CorrYtion
3. New Allocation of Rights
I
Corruption
v
15. New Allocatlon of Rights 1 ---- --*
Figure 1. The allocative effects of corruption: choosing benchmarks,

lower than at 1. The subsequent corruption which takes us to 3 can increase the
welfare loss, as in most rent-seeking models. On the other hand, it can undo some
of the misallocation by allowing resources to be bid back to higher valued uses,
though the level of welfare will still be lower than at stage 1. Thus even with the
neoclassical benchmark, our assessment of the effects of corruption will depend on
whether we compare the efficiency in box 3 with that in box 1 or 2,
Leff's (1964) argument that corruption had beneficial effects in sub-Saharan
Africa because it allowed some entrepreneurs to side-step restrictive monopolies
was effectively comparing a situation such as 3 with the even worse allocation at 2.
In contrast, Myrdal's (1968) argument took the opposite position, arguing that the
possibility of corruption would encourage state officials to deliberately introduce
restrictions which maximized their take. Here the comparison (anticipating subse-
quent rent-seeking arguments) is with a benchmark such as stage 1 which involves
neither the corruption nor the interventions which led to it.
In real-world comparisons where intervention of particular types and corruption
come as package deals, these problems are even more relevant. The perfectly
competitive benchmark may not be useful in such comparisons. Suppose the only
way to get efficiency-enhancing interventions which take us from stage 3 to stage 5
is to accept some of the efficiency loss brought about by corruption at stage 4. We
have to decide whether to choose 3 or 4 as our benchmark. If the corruption is
part of a package deal with particular kinds of intervention, so that stage 4 is not a
feasible benchmark, we should consider the corruption cost as a necessary cost for
attaining the improvement in welfare between 3 and 5. Comparing 5 with 3, cor-
ruption appears as a necessary cost of increasing welfare in the same way as labour
or input costs are necessary costs. If on the other hand it was feasible to attain the
allocation at 4, the corruption which took us to 5 would be an avoidable cost. Con-
sequently by comparing 5 with 4, we would conclude that corruption reduced
welfare. The need to make this type of judgement about feasible benchmarks
makes the analysis of corruption difficult.
The next step in the economic analysis is to model how corruption at different
stages can explain the eventual allocation of rights. The aim is to explain the dif-
ference in welfare levels between the current level and that at some benchmark
allocation. The analysis of corruption has drawn most heavily on rent-seeking
models. These developed to answer the more general question of how the legal
and illegal expenditures on persuading activities affected welfare. While not all
 The Efficiency Implications of Corruption 687

rent-seeking is illegal and therefore corrupt, the general models can be amended
to examine the effects of corruption. We argue that the models are indeterminate
in terms of predicting both the extent of corruption as well as its likely social cost.
In the next section we look at some of the factors which might make the analysis
more determinate.
Rent-seeking theories have primarily looked at the determinants and effects of
the magnitude of expenditures on rent-seeking. Rent-seeking describes the activi-
ties and expenditures of individuals who seek to change rights to earn the above
normal profits described as rents. The theories of rent-seeking developed in
response to the observation that measures of the deadweight losses associated
with tariffs and other interventions were relatively small. Buchanan, Posner,
Tullock, Krueger and others argued that the real efficiency cost of intervention
was larger because the artificial rents would persuade other agents to spend
resources trying to acquire rights to these rents till the rate of return had been
equalized across activities (Buchanan et al., 1980).
The first goal of the rent-seeking literature was to identify the magnitude of the
rent-seeking expenditure. While some of this expenditure could be legal, in the
form of lobbying, queueing and so on, much of it is likely to take the form of cor-
ruption in developing countries. The earliest models such as those of Krueger
(1974) and Posner (1975) argued that the rent-seeking expenditure was going to
be equal to the value of the rent being competed for. However, it was soon estab-
lished that the aggregate expenditure is indeterminate in terms of conventional
economic variables.
To some extent, the expenditure does depend on a number of conventional
variables. These include the number of agents competing for rents and the number
of state officials competing to supply the rights demanded. These factors can be
modelled in a way familiar to economists. The structure of demand and supply in
this ‘market’ partially but not fully determine total expenditures. However, total
expenditure also depends on a number of other factors including the determinants
of each individual’s probability of getting the rent in response to expenditures on
rent-seeking (Mueller 1989, pp. 231-235). This probability depends to some extent
on the number of agents competing for the right but it also depends critically on
other features of the political economy which determine the rules governing the
allocation of rights in rent-seeking transactions.
For instance, if there are a small number of agents competing for the rents, they
may end up spending far less than the total amount of the rent if some agents have
a much higher probability of getting the rent. This could be the case if insiders
have an advantage over outsiders (Rogerson, 1982). The outcome here depends
not only on the small number of players but also critically on the enforceability of
the rule determining the probabilities of different agents getting the rent. With the
same number of agents the rent-seeking expenditure could be very large if there
was a reasonable probability for outsiders to become insiders by spending large
amounts of resources on rent-seeking. The analysis is therefore indeterminate
without a description of the political economy which determines the distribution
of probabilities of winning across groups and individuals.
Similarly, attempts to relate the magnitude of rent-seeking expenditures to the
number of state officials offering the rights also proves to be indeterminate. One
of the earliest models of how the agency structure of the bureaucracy can affect
688 M . H. Khan

the magnitude of corruption was provided by Rose-Ackerman. She considered a

number of alternative bureaucratic structures: fragmented, sequential and hierar-
chical. In the fragmented structure each of a number of agencies provide part of
the approval required, the sequential is similar except that approval is provided in
sequence while the hierarchical corresponds to the command structure usually
assumed in theories of bureaucracy. Rose-Ackerman’s analysis established inde-
terminacy and showed that under particular conditions each of these structures
may dominate in generating the greatest degree of corruption (1978, esp. pp.
176-88).
A similar indeterminacy exists in subsequent rent-seeking models which have
attempted to model how the number of competing officials affects the magnitude
of rent-seeking expenditures. Thus in Congleton’s (1980) model, the competition
amongst officials in a democracy could push the total bribe take down if each offi-
cial was prepared to accept a low bribe. But if the minimum effective bribe was
high, democracy may end up with a higher total bribe take than a dictatorship.
Once again the total rent-seeking expenditure depends on features of the political
economy of the society, here on the determinants of the bargaining power of
bribe-givers and state officials.
While progress was made in pinning down at least some of the determinants of
the magnitude of rent-seeking expenditures, less progress was made in identifying
what part of this total expenditure was likely to be a true social cost. The problem
here includes the one we discussed earlier of agreeing on a benchmark for com-
paring ex post outcomes. In the early rent-seeking models, the entire expenditure
was assumed to be a waste. The justification for this was provided by Bhagwati
(1982) who defined all expenditures on rent-seeking activities as directly unpro-
ductive. The basis for this definition was the assertion that the resources used up in
rent-seeking activities do not produce an output which enters anyone’s utility
function.
However, in terms of neoclassical theory, this definition has the status of an
assertion. It is clearly possible for expenditures, particularly if they are on pet
political causes, to enter into the utility functions of the spenders as well as being
rent-seeking expenditures. Moreover, as Bhagwati himself recognized, if the rent-
seeking expenditures resulted in society moving from a less efficient to a more
efficient position, it would be difficult to say that the expenditures were a waste
from society’s point of view. Recall that a move from stage 2 to stage 3 in figure 1
as a result of corruption could result in an improvement in welfare by reallocating
resources to higher valued uses.
Thus an important source of confusion in the literature is the assertion that the
entire rent-seeking expenditure was a social cost. This implicitly assumes that no
welfare improvement comes about as a result of the reallocations of rights brought
about by the expenditure. To properly identify the social cost associated with any
given magnitude of rent-seeking expenditures, we would have to know the particu-
lar rights which were being reallocated or changed as a result. This is once again
outside the ambit of rational choice models. In no society is every rent-yielding
right up for reallocation or change through rent-seeking. A description of the
political economy which determines which rights are up for negotiation is a critical
precondition for analysing the waste implications of any particular level of rent-
seeking expenditure. If the rights being reallocated result in welfare improve-
 The Efficiency Implications of Corruption 689

ments, the rent-seeking expenditure may be a necessary cost and therefore no

more a waste for society than the expenditure on fuel. If the rights being created
or destroyed result in welfare losses, any level of expenditure on rent-seeking is a
social waste.
A further source of confusion in the literature is a debate about whether bribes
are a social cost if they are simply transferring resources between members of the
same society rather than changing the level of output (Tullock, 1980, Browning,
1980). In theory, if bribes are pure transfers which change the distribution of
income without affecting incentives, they have zero social cost apart from the
transaction cost of organizing the bribes (Varian, 1989). In reality, bribes
inevitably change incentives and therefore have effects on production. Like any
other expenditures on rent-seeking, the efficiency losses due to acts of bribing
have to be compared with the outcomes to assess their social cost.
The incentive effects of bribes have been modelled by Shleifer and Vishny
(1993). Rose-Ackerman’s early work had modelled the magnitude of bribes under
different agency structures of the state. Shleifer and Vishny relate the agency
structure to both the magnitude of bribes and their social cost in terms of lost out-
put for society. Shleifer and Vishny’s three-fold classification is quite similar to
Rose-Ackerman’s. Their first case is a centralized state where a single agency is
the sole supplier of all relevant rights. The second case is a fragmented state where
a number of agencies each provide one of a number of complementary rights. Each
productive user requires a package of all these rights and therefore has to deal
with all the agencies separately. The final case is one where there are a number of
agencies but they can each supply all the relevant rights required by each produc-
tive user.
The outcome in terms of the effect on incentives for the productive sector is
radically different between the three cases. The single agency case is equivalent to
that of the monopoly case in industrial organization. State bureaucrats maximize
their income from bribes by restricting the joint supply of the separate rights to
their profit-maximizing level. Shleifer and Vishny argue that the total bribe col-
lected will be the highest in this case as in a joint monopoly in industrial organiza-
tion where the same firm supplies a number of complementary goods. Since the
rights are over complementary inputs, the level of bribe demanded for each type
of right will not necessarily be maximized.
The fragmented case is a version of the Cournot oligopoly case in industrial
organization. Here a number of agencies each supply a right over a complemen-
tary input. One agency may supply the right to import raw materials, another may
supply the right to set up the factory, and a third may supply access to credit. The
different agencies face a prisoner’s dilemma. In attempting to maximize rents for
itself, each will raise the price of the particular right it supplies so high that overall
activity shrinks and the total bribe collected by all agencies falls. The total bribe
take falls because high individual bribes cause a fall in the level of activity and
therefore in the demand for the rights sold by the state. Output and efficiency are
lower than in the monopoly case.
The final case is a special case of fragmentation where each agency can supply a
package deal to purchasers with all the complementary rights necessary to set up
the business. This corresponds to the competitive case in industrial organization
where a large number of suppliers are supplying the product. In this case the prod-
690 M. H. Khan

uct is the package of rights necessary for productive activity. Competition between
agencies will in theory push the price of the package of rights to zero and there-
fore the total bribe take to zero as well. The total bribe falls to zero not because
the demand for rights has collapsed but rather because the price of each right has
become zero. The absence of the bribe tax means that output and efficiency are
now at their highest level.
The Shleifer-Vishny policy conclusion is straightforward. Corruption is best
dealt with by increasing the competition between bureaucrats and allowing more
agencies to supply similar rights. The aim is to approximate the third case: the
competitive supply of rights. Total bribes are zero and output is highest. The worst
case is that of competing agencies supplying complementary rights. Total bribes
are not maximized but output is lowest because individual bribes are highest. The
absolute monopoly case with the highest total bribe is in between, with a higher
social output than the fragmented case.
Despite its apparent ability to model the consequences of state fracturing, the
model does not actually fit a casual assessment of the costs of corruption across
countries. As Shleifer and Vishny admit, most successful countries do not resem-
ble the competitive agency structure of their theory. Indeed countries like South
Korea appear to be closer to the monopolistic supply case rather than the compet-
itive supply one. Equally, the classic cases of corruption-led sclerosis do not
exhibit unqualified state fracturing. As the authors point out, countries such as the
Philippines under Marcos were closer to the monopolistic supply case. A charac-
teristic feature of such dictatorships has been that the strongman had a finger in
every pie and could in principle have ensured that the total bribe was maximized
for important packages of rights. On the other hand, Shleifer and Vishny point out
that India approximates the fragmented case but it would be difficult to argue that
the social costs of corruption were greater in India than, say, in the Philippines.
The model may therefore explain a part of the difference in the effects of cor-
ruption across countries in terms of these agency structure differences. But some
of the most successful countries have monopolistic rather than competitive supply
(South Korea, Taiwan). In this they are quite similar to some of the countries
where the social costs of corruption were perceived to be the highest (Philippines
under Marcos, Bangladesh under Ershad). On the other hand, some moderately
successful countries (India) approximate the oligopoly case without collapsing in
the way the model predicts.
One obvious shortcoming of the model is that it is silent about the packages of
rights which are at issue. There is no description of the political economy of the
society which determines which rights are being contested and the rules governing
the resolution of these contests through the mediation of the state. We argue in
the next section that differences across societies in these respects can explain why
the social costs of rent-seeking can differ across institutionally similar state struc-
tures.
In summary, conventional models have limitations in determining both the mag-
nitude of the expenditure on rent-seeking activities as well as in determining the
social cost of these expenditures. The magnitude of expenditures does depend on
some of the variables identified in rent-seeking models such as the number of
competitors for rents and the number of agencies or officials supplying the rights
to these rents. But the magnitude of the expenditure also depends on features of
 The Efjiciency Implications of Corruption 691

the society which determines how the probability of each individual getting the
rent varies with the amount of expenditure she is willing to make. In the extreme
case, the probability may have nothing to with expenditures but may instead be
determined by other political or sociological variables. The models determining
the social cost of these expenditures are even less determinate. To determine
whether expenditures are a social cost or not we need to know which rights are
being contested and the rules determining success and failure for individuals or
organizations in these contests.

3 PATRON-CLIENT NETWORKS AND CORRUPTION

In this section we discuss what can be said in general terms about the factors
which determine which rights are transacted between state and society and the
terms of their exchange. Many of the determinants of these characteristics may be
quite specific to societies but some relevant differences between societies can be
identified at a general level. One of these determinants is the distribution of power
between the state and the organizations in society which are demanding changes
in rights.
Developing country states typically operate through patron-client relationships
with key sections of society. State leaderships operate through these networks to
implement their economic and political strategies and to negotiate changes in
rights. A simple way of capturing the relevant differences in the balance of power
across developing countries is to look at the power relationship between patrons
and clients within such networks (Khan, 1996, and forthcoming).
A comparison of two ideal-typical cases will establish the importance of these
differences. At one extreme we define the patrimonial patron-client network. The
term patrimonial refers to the ability of the state to protect existing rights at low
cost. This is actually implicit in much of economic theory where the state is
assumed to have the power to enforce property rights at low cost. Underlying this is
a distribution of power in society which allows the state to do this. This distribution
of power between the state, right-holders and contestants of rights is in fact charac-
teristic of a relatively narrow range of social structures in developing countries.
Patrimonial patron-client networks, where the state is able to protect the exist-
ing property right structure at low cost, are likely to be unusual in developing
countries. This is because in developing countries current allocations of rights typi-
cally do not have a long history and civic institutions supporting such allocations
are typically underdeveloped. Consequently, patrimonial patronxlient relation-
ships are only likely to develop in a relatively narrow range of developing coun-
tries where the distribution of power between the state and the coalitions
contesting rights is tilted in favour of the state.
At the other end of the spectrum we define the clientelist patron-client network.
State officials as patrons within these networks lack the power to enforce rights.
Property rights within these networks are weakly-defined. Variants of the social
distribution of power which produces this result are the norm in developing coun-
tries. Property rights over valuable resources are newly emerging and the groups
or individuals getting access to these rights typically do not have a long history of
possession. The degree of contestation over rights is consequently much higher.
692 M.H. Khan

In the clientelist case, state officials as patrons within patron-client networks

are unable to monopolize the protection of rights. Mafia-like groups are likely to
emerge engaged in private protection activities for patrons both in the private sec-
tor and in the state. In turn these ‘clientelist’groups organize their own challenges
on rights. This difference in the distribution of power within patron-client net-
works has important implications for the efficiency of corruption. On the one
hand, the range of rights being contested varies greatly between clientelist and
patrimonial networks. In addition, the factors determining success also differ
between these networks with important allocative implications.

3.1 The Range of Contested Rights

The standard assumption in rent-seeking models is that the rent-seeking expendi-

tures are being targeted towards a limited range of rights. This is consistent with
the existence of patrimonial patron-client networks. It is assumed that rent-seek-
ing does not affect the entire range of rights collectively. If all rights were being
contested, the social allocation of rights would collapse without protection mecha-
nisms (such as clientelist coalitions or mafias). In addition, the implicit assumption
in conventional rent-seeking models is that while the state can limit the range of
rights being negotiated, it only wishes to create (and trade in the creation of)
socially harmful rights. It is not clear why the state should have such an objective
function. One reason could be that only harmful rights have rents as in the stan-
dard neoclassical model. But most contributors to the rent-seeking literature
would admit that in reality socially beneficial rents frequently exist (as when
scarce resources or innovations have to be protected from free access).
Indeed, in developing countries, where capital, entrepreneurial skills and techni-
cal knowledge are in scarce supply, we would expect positive rents pervading the
modern sector. If the distribution of power within patron-client networks was of
the patrimonial type, state officials would be secure in their ability to allocate or
protect such rents. This ability would enable them to capitalize (by appropriation)
a share of the streams of future rents. The net present value of their streams of
future income are maximized if productive rights are created and protected since
over time productive rights are likely to create larger streams of appropriable
income than redistributive rights. Corruption in this case may not be socially
harmful because state officials have an incentive to create and protect productive
rights.
The combination of economic dynamism and high levels of corruption in some
East Asian countries can be explained in terms of the emergence of a distribution
of social power which sustained patrimonial patron-client networks. Such a distri-
bution of power emerged in South Korea under Park and under the KMT in Tai-
wan (Amsden, 1989; Wade, 1990). During their high-growth period, states in these
societies could effectively enforce rights and could reallocate or change them at
low cost. Long-run profit maximization by state officials dictated that productive
rights were protected and changes were in efficiency-enhancingdirections.
Under a clientelist political settlement, neither top state officials nor any other
single group has the ability to define which rights will be protected and which
changed. The range of rights which are being created or reallocated depends on
 The Efficiency Implications of Corruption 693

the objective functions of a large number of clientelist groups and on their relative
power. In a situation of instability no group is likely to have a long-term view and
rights which maximize long-run profits are not likely to be created. Instead the
rights which are likely to be created and reallocated are rights which generate
rents over short time horizons. This does not mean that powerful individuals and
groups within the state will not be getting very large benefits in the form of bribes.
They will, but only by virtue of belonging to one or more of the clientelist organi-
zations competing over rights.
It has often appeared to observers that dictators such as Ershad of Bangladesh
or Marcos in the Philippines had the power to create productive rights but chose
to create short-term rents instead. Their failure to create efficiency-enhancing
rights is clear. The problem is to explain it. If we maintain the assumptions of the
patrimonial system, we have to explain why a state which feels secure in its ability
to selectively create rights will nevertheless choose to create and share socially
damaging short run rents which do not maximize its profits. This can only be
explained by persistent cognitive failure or a very short time horizon. It is interest-
ing that new institutional economists have begun to stress differences in mental
models and ideologies across countries (North, 1995; pp. 22-4). However, we
would argue that it is possible to explain to a large extent the relative performance
of states without recourse to cognitive problems by looking at differences in the
distribution of power across countries (see also Khan, 1995).
The time horizon explanation is weak because these and other dictators wanted
to be in power for very long. The cognitive failure explanation is also not satisfac-
tory because dictators in these countries have occasionally tried to limit the effects
of clientelist competition and failed. Marcos, for example, made an unsuccessful
attempt in 1975 to crack down on the decentralized appropriation which we have
described as clientelist contestation (Klitgaard, 1988; pp. 13-97). A series of
unsuccessful attempts by successive rulers in Pakistan and Bangladesh to combat
clientelism is described in Khan (forthcoming). The problem lay not in the cogni-
tive models of successive state leaderships but the social distribution of power
which prevented them from defining which rights were to be created and pro-
tected.
Despite these occasional attempts at central control, a wide range of rights were
being continuously contested by powerful clientelist groups with unplanned and
unpredictable consequences. It was difficult in this context for the top state offi-
cials to create and allocate the rights which would maximize long-run rents. Given
this constraint, the next best strategy was the one which was eventually followed.
This was for the state leadership to organize the largest clientelist group, partici-
pate in the contestation of rights and use superior organizational power to grab
most of what was available. This was a perfectly rational response and not the
product of cognitive failure. The social consequences of such corruption were,
however, large and negative.

3.2 The Allocation of Rights

Apart from the range of rights which are contested, the basis on which rights are
allocated between agents is also likely to be different between these networks. The
694 M. H. Khan

assumption in conventional rent-seeking models is that rights to rents are allo-

cated in response to expenditures on rent-seeking. This assumption is more plausi-
ble in the patrimonial patron-client network. If the state officials acting as patrons
can appropriate part of the rents, they are likely to allocate the right in response to
bids being made by competing agents who wish to acquire the rent. This allocative
procedure can be efficiency-enhancingbecause it favours the allocation of scarce
resources to users who are most likely to maximize returns.
This is not necessarily the case in the clientelist patron-client network. Here the
state officials who are patrons in particular transactions are themselves being chal-
lenged by state officials and private agents in competing clientelist coalitions. The
allocative rule for patrons in this network is likely to be different. State officials
are likely to look at a combination of economic and political rewards in making
allocations. Rights will not be created or allocated to agents who are the highest
bidders. Instead clients or clientelist groups with superior organizational power
are likely to get payoffs for their support in the form of allocations of rights. This
means that rights will not always (or indeed very rarely) go to individuals or
groups who are the highest value adding users. This too contributes to the ineffi-
ciency generated by the clientelist system.
An important manifestation of a clientelist process is that while it may be rela-
tively easy to create new rights, it is very difficult to change or transfer existing
rights if this hurts powerful constituencies who already possess them. This means
that even if some agents are higher value-adding users who are willing to bribe to
change the structure of rights, the state may be politically unwilling to consider
reallocations of rights. In contrast, in patrimonial networks or in the models con-
sidered in conventional economics, allocative decisions are based on calculations
of economic gain by state officials.
If existing rights cannot be changed, competition results in the creation of new
rights. One manifestation of this is excessive entry into new industries, excessive
employment creation for powerful white collar workers and so on (Bhaskar and
Khan, 1995). The proliferation of new rights dissipates rents even when their
preservation is socially valuable. The multiple sources of power in the case of
clientelist competition may give the appearance of the fragmented state machinery
discussed by Rose-Ackerman and Shleifer and Vishny. But the causes are quite
different. The problem is not that the institutional structure of the state is frag-
mented, preventing potential coordination. In fact in many countries with corrupt
leadership, the supreme leader is nominally in a monopolistic position. Instead the
problem is due to the proliferation of competing clientelist organizations many of
which may include as key players powerful individuals within the state. The appar-
ent similarity of clientelist competition with the disarticulated supply side struc-
ture of state agencies is misleading. The solution here is not further competition
between bureaucrats. Competition between clientelist organizations which include
important state bureaucrats and political elites is already high and is the source of
the problem.
Let us return to our examples of the corruption scandals in India and South
Korea. In both countries the magnitude of the transfers themselves should only be
a starting point in an enquiry into the networks of patron-client exchanges. In
South Korea our argument would be that the networks were patrimonial. The
rights created and altered by the state were efficiency-enhancing.The profit maxi-
 The Eficiency Implications of Corruption 695

mizing strategy for state officials in patrimonial networks is usually to create the
most productive rights and cream off as much of the rents created as possible. In
the Indian case, the transfers were located within clientelist networks. Bribes paid
by businessmen went into organizing political factions for their patrons who were
clients of higher level patrons. What businessmen eventually get from their expen-
ditures in clientelist networks depends very little on the rents they can potentially
generate from productive enterprises created with state help. It depends rather
more on the relative power which their chosen faction turns out to have relative to
other clientelist factions. The cause of sclerosis is not the size of the expenditures
on rent-seeking but rather the distribution of power between clientelist coalitions
which prevents any group in society from pursuing long run profit maximization.

4 CONCLUSIONS

The efficiency effects associated with corruption depend on the way in which the
rights to be transacted are selected and the terms under which the bargaining over
their allocation happens. Conventional models analysing the efficiency implica-
tions of corruption are deficient because they ignore how rights are selected and
the political constraints on their allocation process. We have argued that one of
the important factors affecting these processes is the distribution of power
between the state and the organizations competing over the creation or realloca-
tion of rights. A number of mechanisms were suggested which would explain why
corruption in countries with patrimonial patronxlient networks may not be effi-
ciency retarding. In contrast corruption in countries with clientelist patronxlient
networks may be associated with structural sclerosis, the proliferation of rights
and slow growth. In each case performance is related not to the extent of corrup-
tion but rather to the political structures which sustain different processes through
which rights are created and reallocated.

REFERENCES
Ades, A. and Di Tella, R. (1996). ‘The causes and consequences of corruption: a review of
recent empirical contributions’,IDS Bulletin 27 (2).
Amsden, A. (1989). Asia’s Next Giant. Oxford: Oxford University Press.
Bhagwati, J. (1982). ‘Directly unproductive, profit-seeking (DUP) activities,’ Journal of
Political Economy 90 ( 5 ) .
Bhaskar, V and Khan, M.H. (1995). ‘Privatization and employment: a study of the jute
industry in Bangladesh’, American Economic Review 85 (1).
Browning, E. (1980). ‘On the welfare cost of transfers’. In Buchanan, J. Tollison, R. D. and
Tullock, G. (eds.) Towards a Theory of the Rent-Seeking Society. College Station: Texas
A&M University Press.
Buchanan, J. Tollison, R. D. and Tullock, G. (1980). (eds) Towards a Theory ofthe Rent-
Seeking Society. College Station: Texas A&M University Press.
Congleton, R. (1980). ‘Competitive process, competitive waste and institutions. In
Buchanan, J., Tollison, R.D. and Tullock, G. (eds.), Towards a Theory of the Rent-
Seeking Society. College Station: Texas A&M University Press.
Khan, M. H. (forthcoming) Clientelism, Corruption and Capitalist Development.Oxford:
Oxford University Press.
696 M.H. Khan

Khan, M.H. (1996). ‘A typology of corrupt transactions in developing countries’, IDS

Bulletin 27 (2).
Khan, M. H. (1995). ‘State failure in weak states: a critique of new institutionalist explana-
tions’. In Harris, J., Hunter, J. and Lewis, C. (eds) The New Institutional Economics and
Third World Development. London: Routledge.
Klitgaard, R. (1988). Controlling Corruption. Berkeley: University of California Press.
Krueger, A. 0. (1974). ‘The political economy of the rent-seeking society’, American Eco-
nomic Review 64.
Leff, N. (1964). ‘Economic development through bureaucratic corruption’, American
Behavioral Scientist, (reprinted in Ekpo, M. U. (ed.) (1979). Bureaucratic Corruption
in Sub-Saharan Africa: Toward a Search for Causes and Consequences. Washington:
University Press of America).
Mueller, D. C. (1989). Public Choice II: A Revised Edition of Public Choice. Cambridge:
Cambridge University Press.
Myrdal, G. (1968). Asian Drama: An Inquiry into the Poverty of Nations. New York:
Pantheon.
North, D. (1995). ‘The new institutional economics and Third World development’. In Har-
, Hunter, J. and Lewis, C. (eds) The New Institutional Economics and Third World
r i s ~J.,
Development. London: Routledge.
Nye, J. S. (1967). ‘Corruption and political development: a cost-benefit analysis, American
Political Science Review, 61 (2).
Posner, R. A. (1975). ‘The social costs of monopoly and regulation’, Journal of Political
Economy 83.
Rogerson, W. P. (1982). ‘The social costs of monopoly and regulation: a game theoretic
analysis, Bell Journal of Economics, 13.
Rose-Ackerman, S. (1978). Corruption: A Study in Political Economy. New York:
Academic Press.
Shleifer, A. and Vishny, R. W. (1993). ‘Corruption’, Quarterly Journal of Economics, 108
(3).
Tullock, G. (1980). ‘The cost of transfers’. In Buchanan, J. Tollison R. D. and Tullock, G.
(eds) Towards a Theory of the Rent-Seeking Society. College Station: Texas A&M
University Press.
Varian, H. (1989). ‘Measuring the deadweight costs of DUP and rent seeking activities’,
Economics and Politics 1(1).
Wade, R. (1990). Governing the Market: Economic Theory and the Role of Government in
East Asian Industrialization. Princeton: Princeton University Press.