Download Top Down or Bottom Up

Top Down or Bottom Up?
The Spread of Systemic Corruption in the Third World
Feisal Khan
Department of Economics
Hobart and William Smith Colleges
Geneva, NY 14456
[email protected]
Abstract: Recent pronouncements by the World Bank and the IMF have focussed new attention on
systemic corruption in the Third World and its negative effect on economic development. The
traditional approach to modelling corruption in the social sciences has been to see it as a classic
principal-agent problem. While excellent for analyzing individual actions, the principal-agent
approach cannot explain the spread of corruption when society is characterized by a norm of
'non-arm's-length' dealings, or how a superior's corruption affects subordinates' behavior, i.e., it
cannot incorporate social interactions. This paper presents a threshold model of corruption that
shows how, in economies characterized by relatively under-developed formal institutional checks
and balances, 'non-arm's-length' dealings and leadership corruption can rapidly lead to systemic
corruption. The model accommodates the differences seen in corruption levels in the Third World.
Also addressed are some implications of this model for reforming systemically corrupt countries.
The model can be readily adapted to other situations where the attitudes and inclinations of the
leadership are important in determining the final outcome.
December 2001
As this is a work-in-progress, any comments would be greatly appreciated.
1
All models are wrong but some are useful.
George Box
Introduction to the Problem1
In 1997 the fall of Mobutu Sese Seko, for thirty-two years the strongman of Zaire and the undisputed King of
the Kleptocrats, served to underline the importance of corruption as a key explanatory factor in understanding why
some Third World countries' track record on economic development has not been better. According to the Los
Angeles Times, Zaire's already low per capita income of $213 in 1986 had fallen to $148 by the end of 1996. In
contrast to the poverty of his country, Mobutu was estimated to have a personal fortune worth five billion dollars, the
result of "ruthless repression and unbridled corruption."
One estimate of the almost unimaginable level of official corruption in Zaire in the 1970s put the amount of
the annual government operating budget that was "lost or diverted" at 60 percent (Callaghy 1984, 189). More
recently, "presidential appropriations" from the national budget were conservatively estimated at a minimum of 15-20
percent annually. This figure, of course, excludes non-presidential "appropriations." One newspaper report has
Mobutu telling a convention of Zairian public officials to steal but not too much, since then they would be punished.
After thirty-two years of Mobutu, the people of Zaire, potentially among the richest countries in Sub-Saharan Africa,
are among the poorest in Africa. With the economy near collapse and suffering from "endemic lawlessness and
rampant corruption," and Mobutu ousted by a rebel movement, the Times went on to warn that "such chaos is hardly
unique in Central Africa."2
The extent to which widespread and systematic governmental corruption had become a problem for
economic development in the Third World, and for Western policy makers deciding on development assistance to it,
was made apparent by the President of the World Bank, James Wolfensohn. At the 1996 annual meeting of the
International Monetary Fund (IMF) and the World Bank, Wolfensohn stated that corruption was one of the greatest
barriers to economic progress in developing nations and that the Bank would no longer view it with a Nelsonian eye.
This statement was seconded by the Managing Director of the IMF, Michel Camdessus. The magnitude of the
problem, and the limits to which the World Bank was willing to go to ameliorate the situation, was indicated by
Wolfensohn’s statement that if "we [the Bank] find evidence of corruption we will cancel the project." (See the Los
Angeles Times, 4 October 1996, for more details.)
The corruption troubling the Bank and the Monetary Fund was what the coiner of the term "kleptocracy,"
Stanislav Andreski, defined as being indistinguishable from "venality." "Corruption," he wrote, "is... the practice of
using the power of office for making private gains in breach of laws and regulations nominally in force" (Andreski
1968, 92, emphasis added). This definition nicely captures the full flavor of the problem since the "regulations
nominally in force" may be observed only in the breach.
There is considerable evidence to believe that high levels of corruption reduce both economic growth and
reduce overall social welfare: in short, systemic corruption is (rightly) seen as an obstacle to overall economic and
social development (Wei 1999, 1997; Ades and Di Tella 1996; Khan 1996; Mauro 1995; Alam1989; Myrdal [1968]
1989; Andreski 1968; among many others). As such, both the World Bank and the International Monetary Fund,
among other international aid donors, are taking steps to combat it.
I
1
I wish to thank Laurie Brand, Tim Cason, Nora Hamilton, Tolga Koker, Timur Kuran, Vai-Lam Mui, Murat Somer, and
Jennifer Tessendorf for their helpful comments on earlier drafts of this paper. All deficiencies in the text are my sole
responsibility.
2
Preceding data, unless otherwise cited, is from the Los Angeles Times, March 21, 1997, and May 17, 1997; see also
other news reports from early 1997 for further details. See Callaghy (1984) or Gould (1980) for dated yet still comprehensive
analyses of corruption in Zaire and Mbaku (1996), Fatton (1992), Ekpo (1979) for corruption in Africa in general.
2
This is in contrast to the earlier academic views of corruption being either relatively harmless (Leys [1965]
1989) or possibly even beneficial if it helps to overcome structural bottlenecks in the economy (Huntington 19793,
1968; Nye [1967] 1989; Leff [1964] 1989).4 Part of the reason for the change in how academics view corruption
could be due to the realization of how easily petty, or even efficient, corruption could reach systemic levels in many,
especially Third World, countries as the institutional structure decays over time (see, e.g., Shleifer and Vishny 1993)
and how many bureaucrats deliberately increased the level of regulation complexity in order to increase their level of
bribes (Kaufman and Wei 1999).
While now recognizing the negative effects of corruption, the literature lacks a clear enunciation of a
mechanism by which a government becomes systemically corrupt. In short, we do not have a clear enunciation of the
process how an inefficient-predatory outcome could result. The literature also lacks a model that incorporates the
effects of "non-arm's length transactions" being the dominant pattern in the economy. The objective of this paper is to
lay out a simple and intuitively obvious threshold model that shows how a systemically corrupt system of
government, i.e., a true kleptocracy, can come about in a society characterized by non-arm's length transactions. In
addition, some implications of the model regarding the potential outcomes of reform efforts will also be considered.
II
The Basic Principal Agent Model of Corruption
So why do bureaucrats and politicians become corrupt? Accept bribes? Sell off public property for personal
gain? Or act in any of the myriad other ways that fall under the catch-all rubric of corruption? I contend that choosing
corruption over honesty is the optimal choice for a boundedly rational utility maximizer if the benefits from being
corrupt outweigh those from staying honest.
Consider an individual, either a civil servant or a politician, faced with the choice of either behaving in a
corrupt manner or not. The individual contemplating the corrupt act has to weigh the possible costs and benefits of
his actions accruing to him. This is a very important point to keep in mind: the individual considers only his costs and
benefits, not those of society.
Following Becker's (1968) seminal work, the usual treatment of corruption in the social sciences has been to
see it as a classic principal-agent (PA) problem (e.g., Bac 1996; Johnston 1996; Mookherjee and Png 1995; Huang
and Wu 1994; Basu et al 1992; Chander and Wilde 1992; Kiser and Tong 1992; Klitgaard 1991b, 1988;
Rose-Ackerman 1978; Banfield 1975; and many others). Some models of corruption combine aspects of a PA
problem with game-theoretic approaches (e.g., Manion 1996; Besley and McLaren 1993; Rasmusen and Ramseyer
1993; and Macrae 1982), or with dynamic modelling approaches (e.g., Dawid and Feichtinger 1996; Bicchieri and
Rovelli 1995; Feichtinger and Wirl 1994; Andvig and Moene 1990; Lui 1986), but "the standard institutional context
is a two-person hierarchy consisting of a principal and an agent (Bac 1996, 277).5
In the PA approach to modelling corruption, the principal (the supervisor) employs an agent (the subordinate)
to carry out some assigned task. The principal does not know whether or not the agent will perform the task
satisfactorily--in this case the satisfactory performance of the task is that there is no corruption involved. Thus
incomplete information on the principal's part about the agent's intentions may lead to suboptimal outcomes. In
3
Consider this succinct statement of the pro-corruption argument from Huntington (1979, 321), "In terms of economic
growth the only thing worse than a society with a rigid, over centralized, dishonest bureaucracy is one with a rigid, over
centralized, honest bureaucracy."
4
Andreski (1968) is an early exception to this. However, in all fairness to these authors, they were referring to individual
corrupt acts and not to systemic corruption. Paddling against current research, Olsen and Torsvik (1998) argue that, under very
specific circumstances, individual acts of corruption can lead to overall societal gain. However, this is again referring to
individual, i.e., non-systemic, corruption.
5
Sometimes (e.g., Johnston 1996; Klitgaard 1991b) this is described as a principal-agent-client (PAC) problem to denote
the fact that the principal (the department head) employs an agent (the individual bureaucrat) to deal with the client (the private
individual). There is no significant difference between the PA and the PAC approach.
3
addition, the possibility of corruption may also lead to an adverse selection problem as corrupt agents apply for the
government post.
Principals may also be voters who elect a politician (their agent) to govern them. Accordingly, the principal
may employ some other agents to monitor the first agent's work (e.g. external auditors or the General Accounting
Office of the US government), but these may also turn out to be less-than-reliable.... The process of hiring workers to
monitor other workers can go on, in theory at least, indefinitely. A possible solution to the problem outlined here is
to give the external monitors an incentive contract to promote continued diligence, with severe penalty clauses in case
of collusion with corrupt agents. However, this really does not solve the recursive failure problem since collusion
between the corrupt agents and the external monitors is always possible (Besley and McLaren 1993; Basu et al 1992;
Andvig and Moene 1990; Klitgaard 1988; Lui 1986; and many others).
From the perspective of the original agent, the PA model gives a very straightforward analysis of the decision
facing him. The potential payoff can be depicted using a standard economic expected utility model in which the
individual weighs the benefit from corruption against the (net) benefit if he is caught. The expected utility (EU) to an
individual of being corrupt is as follows:
EUC = pBC + (1-p)BP
where p is the probability of not being caught,
BC is the utility from being corrupt,
and BP is the (net) utility after being caught and punished.
(1)
I assume that BC is greater than BP, otherwise it would be irrational to be corrupt unless one knew that he
would never be caught (i.e., p = 1). If BH is the utility from remaining honest (received with a probability of one),
then the individual chooses corruption if his expected utility from that is greater than the utility from honesty: honest
iff BH > EUC.
Klitgaard (1988, 71) puts a slight twist in this standard model by incorporating "the moral satisfaction [the
individual] takes from not being corrupt" into the individual's utility calculations.6 The inclusion of a moral element
in the EU function is a notable refinement since individual utility in this context has at least two distinct components:
that from material benefits and that from knowing that you have done the right thing. 7 This is very important since a
major limitation of the standard PA model is that, although probabilities and payoffs may be the same, different
individuals do not necessarily behave in an identical manner. While one will greedily accept the bribe, the other will
refuse. Some attempts at modelling corruption using the PA form assume a uniform attitude to corruption in order to
keep the model fairly tractable. For example, Mookherjee and Png (1996, 158) voiced a fairly common theme in the
literature when they said that, "Our analysis implicitly assumed that all officials are equally prone to graft." Other
authors realize the excessive oversimplification of this assumption since those officials that have an "ethical
conception of public office and strong moral values may not even reach the point of making a cost-benefit
comparison" (Berkman 1992, 1349). Besley and McLaren (1992, 122) also put great emphasis on the fact that "an
honest person regards his integrity as priceless and thus will not take a bribe." So the explanation for why two
individuals faced with exactly the same payoffs and punishments may act differently lies in the moral character of the
two. We must distinguish between total utility and material utility. Individuals act to maximize their total utility,
which includes both material utility and moral utility. Thus differences in how strongly moral beliefs are held affect
6
Andvig and Moene (1990, 65) agree that "moral costs" may be important to the corrupt agent but do not include a
separate moral cost term in their model.
7
Klitgaard's (1988, 71) expected utility of corruption model is:
EU = U[R(x) + p(x-f) +(1-p)x]
where EU is the expected utility of corruption, x is the payoff from corruption, R(x) is the moral cost to the agent of corruption, p
is the probability of being caught and punished, and f is the size of the penalty for corruption. If the utility from being honest is
less than that from being corrupt, the agent will choose corruption.
4
the level of disutility (cognitive dissonance) from contravening them. That is why there is a distribution of corruption
proclivity and why there will always be some (a very small) number of individuals that are absolutely incorruptible.
Hence the observation of both Ostrom et al (1993, 148) and Klitgaard (1988, 52-55) that there do exist honest
officials in completely (systemically) corrupt organizations.8 Since the prevalent morality in a systemically corrupt
society is not conducive towards ethical behavior, most individuals, taking their behavioral cues from their superiors
and peers, act accordingly.9 Indeed, Huther and Shah (2000) argue that in a systemically corrupt society,
"anti-corruption agencies, ethics offices, and ombudsmen" simply serve to perpetuate the status-quo since they
themselves have been completely corrupted.
Klitgaard's model of expected utility from being corrupt can be modified, the notation made slightly simpler,
and the moral element included, as follows:
EUC
Honesty payoff
and
pBC + (1-p)BP - MC
(2)
BH + MH
(3)
C
P
B > B
MC < MH.
EUC is the expected utility from corruption,
p is the probability of not being caught and punished,10
BC is the material utility from being corrupt when C is the payoff from corruption,
BP is the net material utility after being caught and punished for corruption when P is the punishment,
MC is the moral cost of being corrupt, possibly the discomfort of having a guilty conscience,
MC=0 is equal to MH, the moral benefit from honesty, which decreases in C,11
BH is the material utility from honesty when H is the payoff from honesty (i.e., wages).
=
=
Naturally, the utility from corruption, BC, has to exceed the (net) utility from being caught and punished, BP.
Furthermore, I assume that the moral satisfaction from honesty, MH, is greater than (i.e., worth more to the individual)
the moral discomfort from being corrupt, MC.12
I argue that the value of MC, the moral cost of corruption, would be extremely important in determining the
EU of corruption for any given individual. And it is also likely that different individuals would have different levels
of moral disutility from corruption. There is no comprehensive explanation for why people have differing degrees of
moral sentiments but, as stated earlier, there does seem to be some common basis for ethical behavior.13
8
Wilson (1993, 103) cites studies indicating that people who exhibited great moral courage (e.g., saving Jews from the
concentration camps in WWII) had parents that stressed the importance of principled behavior and, for want of a better
description, 'doing the right thing.' Similarly, Frank (1988, 248-253) underscored the importance of both parents and societal
institutions in laying down the basic underlying foundations of adult moral behavior; Stretton and Orchard (1994, 174) concur.
In any case, it is an indisputable fact that people have differing degrees of moral sentiments and that societal institutions pay a
large role in determining what individuals will consider "acceptable" behavior (Wilson 1993, 46-48 & 121-122 and Frank 1988,
152). The exact process by which this is done is certainly deserving of study but it is beyond the purview of this work.
9
The preceding refers to the de facto morality, not the de jure one. For example, the news media in Pakistan is full of
government exhortations to behave honestly, to act as a good Muslim should, to report the corrupt, and so on. However, these
entreaties are so far removed from the reality of life in Pakistan that they might as well be in Greek for all the impact they have.
10
Subsumed within p is the probability that if he is caught, he can bribe his way out of punishment.
11
That is, there is a positive moral benefit to not taking a bribe; it decreases with the size of the bribe. I assume that C=0
means that the individual is acting honestly, not that he is acting corruptly without accepting a payoff.
12
Consider the case when the expected material utility from honesty is exactly the same as the expected material utility
from corruption. It seems reasonable to assume that in this case most people would choose to behave honestly. Hence, MH, must
be greater than MC.
13
Van Rijckeghem and Weder (1997) carried out a detailed econometric analysis of whether low wages in civil services
caused corruption. Their results indicate that low wages are not a major source of corruption since the near eradication of
corruption would require wages of anywhere between 2.81 to 7.08 times (depending upon the country) the manufacturing sector
wage. The onus for the cause of corruption, again, falls back on personal morality and the prevailing institutional culture.
5
The individual will choose corruption if his expected utility from corruption (after accounting for a guilty
conscience) is greater than his utility from honesty (after augmenting it for the moral satisfaction from not being
corrupt). Changes in p, the probability of detection and punishment, ceteris paribus, will determine whether the
individual will act corruptly. This is because, at the margin, the individual is indifferent between honesty and
corruption, i.e., the total utility from honesty equals that from corruption, when:
BH + MH = pBC + (1-p)BP - MC
(4)
Solving for p*, the value of p that equates both sides of the equation, gives us:
p* = BH + MH - BP + MC
BC - BP
An analysis of the comparative statics of Equation 5 yields the following results:
i)
dp*
=
1 ;
dBH
BC - BP
ii)
dp*
dMH
=
1 ;
BC - BP
iii)
dp*
dMC
=
1 ;
BC - BP
iv)
dp*
dBP
= BH + MH + MC - BC ;
(BC - BP)2
v)
dp*
dBC
= - (BH + MH + MC - BP);
(BC - BP)2
(5)
The comparative statics lead to some interesting conclusions.14 While these are not necessarily
counter-intuitive, neither are they immediately obvious. In the first three results the size of the effect of a marginal
change in any of the first three variables is dependent upon the value of (BC - BP), i.e., on the difference between the
material utility from corruption and the (net) utility if one is corrupt and caught and punished. When looking at the
remedies for corruption, a seemingly reasonable assumption would be that the effects on an individual's p* of changes
in BH, MH, MC--the material and moral benefits from honest behavior and the moral cost of corruption--would be
entirely separate from the benefits of corruption, BC, or his net utility if he is caught and punished, BP. The
comparative statics of equation 5 show that this is not so.
What the first three results tell us is of course intuitively obvious to all observers: in the absence of a strong
punishment deterrence for corruption, moral and material costs and benefits become extremely important in
determining when a hitherto honest official acts corruptly. The significance of marginal changes in BH, MH and MC
come into their own only if the value of (BC - BP) is extremely low. However, a minimal value of (BC - BP) is not
socially desirable since this would mean that the punishment for corruption is extremely light and, thus, probably has
only a minimal deterrent effect.
Maximizing the value of (BC - BP) would appear to indicate that punishment is more important than moral
and material costs and benefits. This is not necessarily so since we do not know the actual values, in terms of utility,
14
The results here are qualitatively similar to Becker's (1968) seminal paper on the economic approach to crime, except
that he did not include a moral cost term and was more explicitly concerned with the effects upon expected utility of marginal
changes in his variables. Here I am more concerned with how the individual's corruption threshold is affected by marginal
changes in other variables.
6
H
H
C
H
H
C
C
P
of B , M and M , or of their partials. Different values of the variables B , M , M , B and B would lead to
differing answers as to which variable is more important.
The fourth comparative static result, again, underlines the importance of strong punishment as a deterrent for
corruption. The effect of changing the punishment, BP, is of course dependent upon the value of the expression (BH +
MH + MC - BC) but the crucial factor here is the value of the term (BC - BP)2. The interpretation is that, if the
punishment for corruption is already severe (i.e., BC - BP is large), then making punishment more severe will have a
smaller effect than might otherwise be thought. This also indicates that there are declining marginal returns to
progressive increases in the severity of punishment.
The final comparative static result is the most interesting. Again, the key point here is the value of the term
C
P 2
(B - B ) . The counter-intuitive result here is that, as the benefit of corruption increases, the associated drop in p* is
less than might be expected since an increase in the benefit of corruption also increases the value of (BC - BP)2,
thereby reducing the overall size of the effect. This result is counter-intuitive in that we do not expect to see
diminishing decreases in p* as the benefits of corruption increase. We expect to see the opposite.15
It can be safely assumed that most non-systemically corrupt governments would wish to raise p* to the
highest level possible to ensure that hitherto honest officials would require an extremely high (perhaps near unity?)
probability of not being caught before they would act corruptly.16 When corruption is punished relatively leniently,
the role of non-punishment factors becomes more important. Making non-punishment factors more important would
involve actions such as raising the material benefits of honesty (i.e., official salary) and raising the moral cost of
corruption and the moral benefit of honesty. The former is a monetarily expensive proposition while the latter are
likely to be expensive more in terms of time and official commitment than in money. As result five of the
comparative statics shows, the most effective action would seem to be raising the corruption penalties.
In any case, none of these remedies is necessarily cheap or easily implemented. Some of the most effective
remedies (raising the moral cost of corruption) require the longest time to be effective and some of the easiest to
implement (raising the penalties for corruption) may run into political complications (an unwilling legislature or
strong civil service unions). Finally, higher penalties for corruption may have unforeseen consequences as increased
penalties lead to higher bribes being demanded, which in turn may raise the benefits of corruption enough to tempt
hitherto honest officials.
III
Moving Beyond the PA Model
If we consider (1-p) to reflect the degree of honesty to be found in the sub-population that is government
bureaucracies rather than in the population at large, then the probability of corruption being punished is likely to be
quite high for any of the industrialized Western countries. Societal norms of honesty, the punishment for official
corruption, the strength of the media and the independence of the judiciary are strong enough to ensure that the
prevailing bureaucratic culture is predominantly honest. Corruption, while it certainly exists, is usually of the
isolated variety and is often confined to a few minor departments or individuals.
The PA model of corruption, extremely useful in looking at environments where most corruption is of the
isolated kind, has some shortcomings that reduce its explanatory power when it comes to corruption in Third World
15
While the PA model is concerned with individual behavior, an extension to the behavior of the group is possible here.
Ever increasing benefits from being corrupt will result in comparatively few hitherto honest individuals acting dishonestly since
the majority of the officials will already be corrupt, i.e. their p* has already been reached.
16
There is certainly a cost-benefit tradeoff regarding monitoring and enforcement costs on the one hand and reduction in
corruption on the other. Klitgaard (1988) has an extensive discussion of this issue. This aspect of the corruption issue is,
however, beyond the scope of the model presented here.
7
economies. It lends valuable insight into the actions of any given actor but it cannot show--since it was never meant
to--how corruption spreads or how the actions of one individual affect the actions of others.17
The dynamic corruption models mentioned earlier (Dawid and Feichtinger 1996; Feichtinger and Wirl 1994;
Andvig and Moene 1990; Lui 1986) consider cases where the equilibrium changes from one of low corruption to one
of high corruption or vice-versa (Bicchieri and Rovelli 1995). They also consider a situation exhibiting multiple
equilibria with intermediate (but usually unstable) levels of corruption. However, their analysis does not look at the
change in the behavior of individual agents. Furthermore, these dynamic models suffer from an additional drawback:
they do not explicitly consider the effects of leadership behavior on subordinates. This drawback is one shared by
other PA models as well, although Basu et al (1992, 354) do acknowledge the importance of leadership as the
presence of an "incorruptible" elite civil service at the head of the government leads to the norm of honesty. Klitgaard
(1988) and Rose-Ackerman (1978, especially Ch. 9) have the only extensive discussion of the effects of a supervisor's
corruption proclivities on the behavior of subordinates but these do not incorporate threshold effects in their analysis.
Several of the PA models (including the game-theoretic ones and the dynamic ones) incorporate various forms of
threshold effects but of a type somewhat different from the ones in this paper. Some PA models incorporate greater
degrees of institutional refinement and realism than others. For example, Basu et al (1992) and Andvig and Moene
(1990) model the possibility of recursive corruption--the official who catches the corrupt official may himself be
bribable. Dawid and Feichtinger (1996) incorporate the corrupt official's desire to maintain a "good" reputation in
their model. Tirole (1996) allows for the fact that a group's reputation for corruption may affect how individual
members choose to behave. However, what the literature lacks is a model of how a superior's corrupt behavior affects
that of the whole organization and of how the behavior of individual agents change over time.
Why is leadership behavior important? In the real world, who is already corrupt will greatly affect the
decision of the other administrators as to whether or not they will opt for honesty or corruption. Assume, for the sake
of argument, that 10% of the population is corrupt. If this is the bottom 10%, the 1st decile, in terms of power or
influence in society, then this corrupt element is not going to influence significantly large proportions of society since
this is not the reference group on which people are going to pattern their behavior. However, if this is the top 10% of
the population, the 10th decile, in terms of power or influence in society and they are seen to escape punishment, this
will obviously affect large portions of society since many people, at least the next 20 to 30% of society, will use the
top 10% as its reference group. This is a weakness of the PA model since it cannot accommodate disproportionate
changes of this sort. From the PA perspective, marginal changes in p give only marginal changes in outcome.
In another context, the dynamics of revolutions, Kuran (1989, 46) shows the importance of an asymmetric
demonstration effect on individuals and how it can change the situation drastically.18 Similarly, even if individuals are
personally disinclined either to be corrupt or, ceteris paribus, to facilitate corrupt behavior in others, they may well be
forced to act corruptly. In addition, honest individuals may also face persecution by their corrupt colleagues. If their
superiors are corrupt, then non-corrupt individuals may be ordered to facilitate corruption in others. Not complying
with a superior’s request may damage their own careers. It is also possible that after being forced to act corruptly on
the behalf of others, the hitherto uncorrupt individual may decide to act corruptly on his own behalf. In the land of the
blind the one-eyed man may be king; in the land of the corrupt, the honest man is often perceived as merely stupid for
not taking advantage of the available opportunities--even if he may be a highly moral person attempting to set an
example for his fellow citizens. What is needed is a model that can offer some understanding of how systemic
corruption occurs within a society. The level of corruption within a society is usually endogenous to it and thus a
model that helps in better understanding this process is needed. Observers of Third World societies will attest to the
fact that the situation vis-a-vis corruption can change for the worse in the space of a few years. For instance, both
17
The absence of social linkages in the standard PA model led Tanzi (1995, 167) to assert that the Beckerian model is "of
limited applicability" to much of the Third World. This is probably too strong a statement to put forth but it does bring up the
crucial weakness of these type of models.
18
See Section V in this work for a more complete exposition of this.
8
Theobald (1990) and Klitgaard (1988) detail how rapidly, within a generation, systemic corruption replaced isolated
corruption in Nigeria and the Phillippines, respectively.
As stated before, there are two possible time paths for a transition to systemic corruption: relatively rapid and
comparatively slow. A rapid transformation of the societal norm from isolated corruption (with the isolated incident
of high-level corruption) to systemic corruption lies at the crux of my argument that the transmission mechanism
must be from the top-down. The alternative time path is the comparatively slow one when an entire society's norm
gradually changes from one of honesty to one of corruption. Thus, in the normal course of events, a completely
corrupt leadership would arise "naturally." However, when the transformation takes place within the space of a few
years, there must be some sort of a catalyst at work. This catalyst is, perforce, the inclination of the leadership. So
how, exactly, would these two mechanisms work?
IV
A Threshold Model of Corruption
A heuristically richer model can be built by allowing the criteria for choosing corruption or honesty to be
endogenous to that society. It is obvious that a better approximation of reality would be to have the level of
corruption in a society determined by interactions among the preferences of the individual actors themselves. What is
needed here is a threshold mechanism for changing the prevailing societal norm--honesty or corruption--within the
model. Several of the dynamic PA models described earlier (Dawid and Feichtinger 1996; Bicchieri and Rovelli
1995; Feichtinger and Wirl 1994; Andvig and Moene 1990; Lui 1986) have such a mechanism but they do not
consider the importance of the leadership in setting the norm of the institutional culture.
The PA model of corruption gives us, for every individual in any given population, the value of p--the
probability that he will not be caught if he behaves corruptly--at which he is indifferent between honesty or
corruption. In the same way, the threshold values of p of the entire population can be arranged in ascending order. It
is not necessary to know the threshold of every individual in this society. It is enough for our purposes to know that
(i) every individual has a threshold and (ii) that they are distributed across a range of zero to one. This allows the use
of the important insights gained from the principal-agent model--the concept of an individual's p*--and endogenize
the incentives faced by each individual actor. Once this is done, a more realistic model of corruption, one that
incorporates the importance of social linkages, can be developed.19
Threshold models, also known as bandwagon-effect models, critical-mass models, tipping models, and
cascade models, have a long history in the social sciences. Granovetter's (1978; 1973) and Schelling's (1978) works
are classics in this area and, arguably, the inspiration for an entire generation of models that show how seemingly
"minor" factors can result in "major" results. For example, Schelling (1978) analyzed how separate individual
decisions determined societal outcomes and could lead to overall inefficiencies. Some of the more recent
contributions to the literature have been Hong and Conrad (1998), Corneo and Jeanne (1997), Kuran (1995),
Bikchandani et al (1992), and Witt (1989). While the exact details of these models of social behavior differ, they are
all similar to the threshold model detailed below. The threshold corruption model developed here draws heavily from
Kuran's earlier works (1991; 1989; 1987) and, especially, on chapters three, four and six of his 1995 book, Private
Truths, Public Lies: The Social Consequences of Preference Falsification. The work done here adapts and extends a
great deal of his analytical framework, initially developed to explain "unexpected" revolutions such as the collapse of
Communism in Eastern Europe, to corruption.
19
There is a problem here in that the assumption is that all people are corruptible. But what about the truly incorruptible
ones? We know that there are at least some people that cannot be bought. The importance of personal morality can resolve this
dilemma. Even if no one else can ever detect your corruption, you yourself know what you have done. Hence the corruption is,
in a way, detected. Therefore, if your personal corruption p is one, this means that you will never be corrupt since your
conscience will know what you have done. For a truly religious person, this knowledge might take the form that God will know
your actions, even if no human does.
9
In general, the threshold corruption model developed here will use the same basic premise: seemingly minor
changes in individual behavior lead through a cumulative causation mechanism to major changes in societal
outcomes. Like other such models, the threshold corruption model will show just how the proverbial straw broke the
camel's back.
This model adds to the literature in both fields, corruption and threshold models, in several different ways. It
extends the reach of threshold models to the analysis of societal corruption, a field where such models have been
comparatively little used. In the threshold corruption model, the individual's decision whether or not to act corruptly
is endogenized. That is, the incentives for the individual to act corruptly depend upon the societal value of p, the
probability that corruption will not be detected and punished, and this value is now determined within the model
itself. At the extremes, such incentives do not matter. People who are truly incorruptible or truly corrupt will behave
predictably no matter what the incentives might be. However, the vast majority of the population will fall somewhere
between these two extremes and for them the prevailing societal incentives--the value of p-- will determine their
behavior. Furthermore, the threshold corruption model also accounts for the societal interaction aspect of corruption
since it incorporates the cumulative effects of corrupt behavior by others into its analysis. Moreover, the threshold
model will also show the asymmetric demonstration effect of leadership corruption on subordinates. Finally, the
model will explicitly incorporate socio-cultural effects by showing how prevailing cultural norms that are conducive
to corrupt behavior may reduce each individual's corruption threshold. Thus the threshold corruption model brings
society and culture into the analysis since the actions of the individuals involved change the incentives they face and,
hence, affect their future actions.
In addition, the threshold corruption model will also illustrate the path-dependent nature of the transition to
systemic corruption--it matters who acts corruptly and when. The model will also highlight the imperfect
predictability aspect of the systemic corruption transition--ex ante, we cannot predict, since we lack perfect
information and foresight, who will act corruptly and when. However, the model will allow us to predict, in general,
under what circumstances we can expect to see an increase or a decrease in corruption.
So what is the threshold corruption model? Assume for the sake of simplicity that the total population of a
society consists of 10 persons. These are denoted by the letters a1 through j1 and each individual has to chose between
behaving honestly or behaving corruptly. Each individual faces an infinite number of decision rounds and all
individual decisions are made simultaneously at the start of each round. In order to keep the model tractable, I assume
that, once having made this choice, he cannot change it in that same round. At the start of the next round, the
individual can see the individual decisions of the other members and may then make a new decision.
Whether or not the individual behaves corruptly is determined by his "corruption threshold," the proportion
of the population (or relevant reference group) that has to act corruptly before the individual will. Realizing that
corruption is usually an activity that carries strong societal (and professional) sanctions against it, most people will
act corruptly only if they feel that decision is one that has been made by a sufficiently large number of their
peers--safety in numbers. This is because, as in the principal-agent model, all persons carry out a cost-benefit
analysis that determines at what level of overall societal corruption does the benefit to him of becoming corrupt
outweigh the costs of detection and punishment. Different individuals arrive at different cost-benefit valuations and,
hence, may have differing thresholds.
I assume that p, the probability of corruption remaining undetected and unpunished, is not exogenous to a
society. It seems reasonable to assume that the probability of being caught and punished would tend to decline as the
number of corrupt individuals increases (Lui 1986) since, if the opposite were true, corruption would never increase.
However, this decreasing probability of punishment need not be a linear function of, or directly proportional to, the
number of corrupt individuals. The individual characteristics of different governments and social systems would
determine the value of p at any given moment in time. In general, I shall make the simplifying assumption that the
probability of corruption remaining undetected is a non-decreasing function of the share of corrupt individuals in that
society. In particular, the value of p in this model will be equal to the proportion of corrupt people in the society.
10
With this simplifying assumption, it is now possible to expand from the PA model of individuals to a threshold model
of societal corruption.20
If an individual's personal corruption threshold is 0.20, at least 20% of the people in this society must be
corrupt before he will behave in a corrupt manner (i.e., for him, p = 1 - 0.2 or 0.8). Similarly, a person with a
threshold of 1.0 is incorruptible since all of society must be corrupt before he will be, but this cannot occur since all of
society cannot be corrupt if he is not.
The corruption thresholds of all 10 members of society, arranged in ascending order of their thresholds, are
given in Table I. According to the table, person a1 will always be corrupt, while j1 will never be corrupt. This is
because, with a threshold of 0, person a1 will be the first to be corrupt, even if no other individual in society would
====================================================================
Table I
Individuals
a1
b1
c1
d1
e1
f1
g1
h1
i1
j1
Threshold
0.0 0.2 0.2 0.25 0.35 0.45 0.55 0.65 0.75 1.0
====================================================================
act in a corrupt manner. This may be because a1 is an extreme risk taker, may feel that he will never be caught, has no
moral qualms about violating corruption taboos, and so on. Similarly, person j1 might be extremely risk averse, or
have very high moral standards, or simply place too high a value on his moral comfort to ever be tempted by the
possible corruption payoffs. The overall proportion of corrupt individuals in this society is 0.10 and so p is also 0.1.
This distribution of corruption thresholds is not enough to trigger corrupt behavior in either b1 or c1 since they require
a corrupt proportion of 0.20.
In a situation such as the threshold distribution in Table I, the prevailing norm here is honesty (the number of
honest persons, N, equals 9). However, our individual a1 is still prepared to behave corruptly. But this equilibrium is
vulnerable to a shock since the probability of corruption being detected and punished is less than one. In particular,
any of a number of events can change individual thresholds and certain individual threshold changes can rapidly
change the population equilibrium from “mostly honest” to “mostly corrupt.”
What could such a shock, an exogenous trigger, be in this case? Perhaps a governmental budget crisis means
that fewer resources are being spent on monitoring. Perhaps an individual has unexpected major medical expenses.
Perhaps he is passed over for promotion and so decides to augment his pension by other means. For our purposes,
assume that the shock here is individual b1's daughter’s upcoming wedding.21 All that actually matters here is that he
is now prepared to act corruptly. What possible effect might this have on other individual’s in the same group?
The behavior of any given individual is usually contextual. By this I mean that most individuals do not have
an infallible moral compass to guide their daily actions and, hence, take many of their behavioral cues from the
actions of others around them. Corruptibility, like other forms of human behavior, is dependent upon the relative
corruption of others in society. Individual b1's corrupt behavior now also changes the behavior of persons c1 through
h1 since the higher levels of corruption affect them as well.
If a1 and b1 are corrupt, then c1 will be as well since his threshold is 0.2 and this has been achieved. This in
turn will affect d1, e1, f1, g1, h1 and i1. As shown in Table II, the prevailing norm in this society has now changed to
20
There are several possible reasons for this. First, it is possible that corrupt individuals would not report other individuals
who are observed to act corruptly. Second, collusion is possible in that, for a consideration, monitors do not report corrupt
behavior. Third, it is possible that self selection would ensure that more corrupt individuals are likely to gravitate to the
monitoring branch of the government. However, explicitly incorporating these into the model would make it much less tractable
without adding greatly to its explanatory power.
21
This is, of course, just one of many possible changes. This shock is not as facetious as it may appear. See Wade (1989)
for just such an example.
11
====================================================================
Table II
Individuals
a1
b1
c1
d1
e1
f1
g1
h1
i1
j1
Old Threshold
0 0.2 0.2 0.25 0.35 0.45 0.55 0.65 0.75 1.0
New Threshold
0
0.1 0.2 0.25 0.35 0.45 0.55 0.65 0.75 1.0
====================================================================
to corruption since nine out of ten members are no longer honest.22 Thus N, the proportion of honest individuals in
this society is now 0.1 and there is a bandwagon effect that changes the societal norm from honesty to corruption.
Ordinarily, this bandwagon effect will be slow: the process will not run its course overnight or even in a few weeks.
This is because it may take repeated corrupt acts by a small minority to affect the others23, opportunities for safely
acting corruptly may be relatively limited, or it may even require the induction of new personnel, ones with lower
corruption thresholds, to change effectively the prevailing norm.
I have purposely allowed a1's behavior to affect immediately only b1's behavior (perhaps a1 and b1 are
colleagues that share an office). While an artificial constraint, it is necessary to show the sequential nature of the
spread of systemic corruption. It is entirely possible for one person's corruption to have an immediate effect on two
(or more) colleagues if their corruption thresholds are identical.
The change in the prevailing societal norm will also affect the calculations for determining the expected
benefits of corrupt behavior. Consider Equation 5 discussed earlier:
p* =
BH + MH - BP + MC
BC - BP
(5)
The threshold value of p, p*, is the probability of corruption remaining undetected and unpunished at which a hitherto
honest official is indifferent between corruption and honesty. This transition from honesty to corruption is
determined by the value of p, ceteris paribus. So for the population distribution in Table I, with a 90% honest
population, the relevant value of p is 0.1. Thus, for all individuals except a1,
BH + MH > 0.1BC + 0.9BP - MC.
(6)
and for all individuals except j1,
BH + MH < 0.9BC + 0.1BP - MC.
(7)
Equation 7 represents the situation in Table II. For all individuals except j1, the prevailing p* of 0.9, once the
bandwagon effect of systemic corruption has taken place, is more than the threshold value for any corrupt individual
in the population. The expected payoff from corruption has greatly increased for the average or representative
individual in this society even if the actual values of BC, BP and MC do not change. Corruption, in terms of its
22
This is just one of an infinite number of possible threshold distributions possible. Under different distributions, there
may be no bandwagon effect after a threshold change. Or there may be minor effects such as only one or two more people acting
corruptly. Or the initial distribution may have all members already acting corruptly. I chose a distribution which allows me to
illustrate the specific bandwagon effect I have in mind.
23
The availability heuristic would require that there usually be more than one occurrence before the average observer
decides that the probability of detection is low enough for him to take a chance. See Kuran (1995) for a full exposition of the
effects of heuristics on human behavior.
12
expected utility, is now more profitable than ever since the probability of being caught and punished for corruption, p,
is directly dependent upon the proportion of honest people in society. We can see how a simple increase in the
number of corrupt individuals in society can greatly increase the payoff for corruption and tempt even those who have
a relatively high corruption threshold.
Of course if, when he first acted corruptly, a1 was caught and given exemplary punishment, it is entirely
possible that b1's corruption threshold might have been raised from 0.2 to 0.3 instead of being lowered. Then there
would be no need to worry about the effects of corruption on society since the dominant societal norm would continue
to be honesty. So it follows then that strict enforcement of anti-corruption laws and ruthless prosecution of the
corrupt can hold the threshold sequence constant and prevent the vicious corruption bandwagon from starting. A
successful example of just such a strategy is Singapore (Quah 1989), one of the least corrupt countries in the
Transparency International surveys.24
It also then follows that lax enforcement of the anti-corruption laws and/or an inefficient legal system that
cannot swiftly prosecute the corrupt could easily change the existing threshold distribution. This would occur when
hitherto honest officials reevaluate their decision to stay honest in light of the knowledge that the legal system cannot
(will not) easily punish the corrupt.
V
The Importance of Asymmetric Demonstration Effects
So far the assumption has been that every individual's decision carries equal weight in society. This is not a
realistic assumption. The current decisions of some individuals are more important in influencing future decisions
than the decisions of others. I contend that the attitudes of the leadership of society are extremely influential in
determining the behavior of subordinates, particularly with regards to corruption (Basu et al 1992; Whitehead [1983]
1989; Wade 1989; Klitgaard 1988; Rose-Ackerman 1978; and many others). Within the context of a threshold
model, this can be shown by assigning each individual's decision a weight consonant with "his importance and
influence in society" (Kuran 1989, 46). The model then captures the fact that a hierarchical social structure, or even a
non-hierarchical one in which individuals have asymmetrical levels of influence, is very important in bandwagon
effects.
Following Kuran (1989, 46), the effects of a separate weight being assigned to different individuals can be
gauged thus:
n
CT
=
wc
i i
(7)
i 1
where CT is the (weighted) proportion of society that is corrupt;
n is the number of individuals in society;
wi is the weight assigned to each corrupt individual i;
ci is 1 if individual i is corrupt and zero if honest.
All the weights of all the individuals in society sum to 1. Thus if everyone in society is honest then CT equals zero.
Similarly, if only individual i is corrupt but he is of minimal importance in society, then CT is near zero and the society
is still basically honest.
24
The relative accuracy of Transparency International surveys (see, e.g., http://www.gwdg.de/~uwvw/icr.htm) of
comparative international corruption levels have been used questioned by some but their use by agencies such as, e.g., the World
Bank and the International Monetary Fund in their official publications, attest to their general acceptance.
13
====================================================================
Table III
Individuals
a1
b1
c1
d1
e1
f1
g1
h1
i1
j1
CT
Threshold1
0
0.1 0.2 0.25 0.35 0.45 0.55 0.65 0.75 1.0 0.05
Threshold2
0
0.05 0.2 0.25 0.35 0.45 0.55 0.65 0.75 1.0 0.15
Threshold3
0
0.05 0.1 0.25 0.35 0.45 0.55 0.65 0.75 1.0 0.80
Threshold4
0
0.1 0.2 0.25 0.35 0.45 0.55 0.65 0.75 0.0 1.00
Weight
0.05 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.15 0.20
====================================================================
Consider the population in Table III. This is the same population as in Table II and it has the same (new)
threshold distribution, Threshold1. Ignoring the weights gives us the same end result as in Table II. Now consider the
actual weights assigned to our individuals. When a1 behaves corruptly here, the value of CT is 0.05, not enough to
prompt corrupt behavior in individual b1 or, subsequently, any one else. This society is still honest as its CT (i.e.,
weighted corruption level) is only 0.05.25
Suppose that there is a change in individual b1's threshold (for the same reasons as before) and it is lowered to
0.05, i.e., Threshold2 is now in effect. Individual b1 will now also act corruptly and so we have two corrupt
individuals and eight honest ones. However, b1's weight is still 0.1 which, combined with a1's weight of 0.05, gives
us a CT of 0.15--not enough to trigger corrupt behavior in any one else. This is in direct contrast to the situation in
Table II where, given that all individual actions carried equal weight, the change in one individual's threshold was
enough to trigger corrupt behavior in others. The end result here is that we simply have a slightly more corrupt
society but one where the prevailing norm is still essentially honesty (n = 0.8).
Now, if individual c3's threshold changes to 0.1, i.e., Threshold3 is now in effect, the new CT of 0.25 is enough
to trigger corrupt behavior in d3 and every other individual except for j3. We can thus see the importance of weights in
determining whether the eventual outcome is honesty or corruption.
The final example is Threshold4, which replicates the threshold distribution of Threshold1 except that the
corruption threshold of individual i3,, the single most influential individual in the population declines from 1.0 to 0.0.
This, admittedly extreme, change now raises the CT to 0.25, enough to trigger a corruption bandwagon in the rest of
the population. The final CT is now 1.0--complete corruption.
The point of this extreme change is to show that the greater is the individual's weight in society, the more likely
his or her action is to change the status quo. Once again, enforcement of the existing anti-corruption laws and an
efficient legal system is more likely to prevent an undesirable change in the threshold distribution and, hence, prevent
a norm change from honesty to corruption.
VI
Systemic Corruption: Bottom-up and Top-down Paths
All the preceding examples (Tables I, II and III) had a norm of honesty as an initial starting point. But is it
accurate to presume that most non-Western countries started their post-colonial history more honest than they were
corrupt? Based upon what we know about their current corruption levels, very low according to Transparency
International surveys, why presume a relatively honest initial point?
It is my contention that the preceding examples with weights that are skewed towards honesty reflect the
situation in most post-colonial states at the time of independence. Most of these states inherited a bureaucratic
structure that was fundamentally honest--at least as far as the personal enrichment of the ruling colonial elite went
25
There is no objective criteria for assigning these particular weights to these particular individuals. This weight
distribution serves to illustrate the point being made here. A different weight distribution could well give very different results.
14
26
(Hope 1996, 130; Andreski 1968, 103 & 110-115; Myrdal [1968] 1989b, 410). Myrdal buttresses his claims of an
honest British colonial administration by extensively quoting from the post-independence Indian Government's
official Report of the Committee on Prevention of Corruption in 1964, the Santhanam Report. This report grudgingly
admitted that "the higher ranks [of the colonial administration] were comparatively free from this evil [of corruption]"
(Myrdal [1968] 1989b, 418) and that, for the post-independence period, the consensus of testimony to the
Commission was that:
corruption had increased to such an extent that people have started losing faith in the integrity of public
administration. ...corruption has, in recent years, spread even to those levels of administration from
which it was conspicuously absent in the past. (Quoted in Myrdal [1968] 1989b, 417)
It is inconceivable to me that any post-colonial government would have an incentive to paint the colonial rulers as
being more honest than they actually were.
Naturally, this contention of an honest colonial administration holds most strongly for ex-British and then for
ex-French colonies (the bulk of the post-WWII ex-colonial states). Hallet (1965) contends that the much-reviled
Belgian colonial administration in the Congo (later renamed Zaire and now again the Congo) was also extremely
competent and honest. The ex-Portuguese colonies and the remnants of the once vast Spanish colonial empire are a
somewhat different matter but they were numerically fewer than the British and French ones.27
The above statements should not be taken to imply that the source of all corruption is within the Third World
itself. Lambsdorff (1997), after an extensive econometric analysis, compiled an indicator of the bribery potential of
leading exporting countries, i.e., how likely is an exporter to bribe his importing counterpart and how much of an
exporting country's sales abroad are due to bribery. Naturally, Western countries are heavily over-represented in his
"bribery propensity rankings." Similarly, others, e.g., LeVine (1989) and Ali (1985), have remarked upon the role of
multinational (i.e., mainly Western) firms in encouraging corruption in the Third World. In this context, Lambsdorff
(1997) further points out that bribes paid to foreigners by businessmen are tax deductible in Germany and Belgium,
and that the US is the only advanced Western country to currently specifically prohibit bribing foreign
customers--although an OECD internal agreement is supposed to extend this practice to all member countries by the
end of 1998.28 A true chicken-or-the-egg type of question is whether, in the first instance, a bribe was offered to an
official or whether it was demanded by him. However, in either case, the basic point remains that it is an internal
decision (whether to demand or accept a bribe) that starts the rot in the government apparatus, not the decision to offer
it.
Ali takes issue with the assertion of an honest British colonial administration in India. However, his examples
of "corruption" are not convincing and, in a work that has a general anti-colonial tone and includes a scathing
denunciation of neo-imperialism by multinational corporations, even he admits that, while "Colonial rule was not free
26
However, Ali (1985, 194-195), while agreeing with the above contention of an honest British colonial administration,
argues that this honesty in administering "government money" did not extend to their personal dealings with the natives:
accepting "gifts" (e.g., free vacations) was commonplace, as was the custom of officials on the verge of retiring from government
service allowing their wives to accept going-away "presents" from influential natives. Since even these "corrupt" officials are
described as being "scrupulously careful and honest about government money," Ali's implicit argument has to be extortion:
natives who did not give gifts would be punished. However, nowhere does Ali actually state that uncooperative natives were
punished. Corruption in the British administrators, in so far as it existed, would have been of the isolated kind. Collins and
Lapierre (1980 [1975]) also remark on the propensity of the wives of some extremely high-ranking British colonial officials in
India to solicit gifts of expensive jewelry, though they make it clear that this was rare and certainly not the norm in the last
hundred years or so of British rule. In any case, Ali's allegations are flatly contradicted by the post-colonial Indian Government's
own report on corruption.
27
The differences in European colonial policies and administration, and why they existed, is a fascinating dissertation
topic in itself. I have no concise answer for why the differences existed except to say that it probably reflected the metropole
government's own relative honesty levels and attitudes to their people.
28
The actual enforcement of these anti-bribery laws, if the US example of complete laxity after a brief initial enforcement
spurt is anything to go by, will be problematical at best.
15
from its share of corruption, but since independence the shocking increase in graft and bribery has bewildered nearly
every observer (Ali 1985, 157)." So why and how did this "shocking increase" take place?
As Hope (1996), Tanzi (1995), Alexander (1994), Dewey (1991), Ali (1985), Caiden and Caiden (1977),
Jabbra (1976), Wilson (1967), and Banfield (1958), and many others, have argued, much of traditional society (now
found mainly in non-Western countries) is characterized by the belief that objective criteria in public dealings should
not apply to friends and relatives. Thus Tanzi's (1995) norm of "arm's length dealings" does not apply in transactions
involving relatives, clan/tribe members and friends, who are given preferential treatment while non-associates are
discriminated against.
The argument I make here is that the colonial administrators were much less susceptible to pressure from their
family and friends for favorable treatment, if for no other reason than the fact that their families were not in the
colonies and most of their friends were other colonial officials. After decolonization, the natives who assumed full
political power were susceptible to this pressure for favorable treatment and, without the presences of Basu et al's
(1992) "incorruptible civil service elite" to keep them on the straight and narrow, did their duty to their family and
friends rather than to their profession. Given the fact that most of the newly independent states were scarcity
economies, i.e., marked by a relative shortage of modern goods, it would be extremely difficult for government
administrators, who often enjoy considerable discretionary power, to resist the demand of family, friends, and even
perhaps co-ethnics, for favorable treatment in the allocation of scarce resources (Bates 1974). Even if they did not
want to favor friends and relatives, after independence, the administrators no longer have the excuse of saying that the
colonial authorities would not allow them to bypass normal administrative procedures. Thus we see a gradual decay
in administrative standards and a rise in corruption as friends and relatives benefit from their association with those in
power. From this, it is a comparatively easy step to favor complete strangers in exchange for monetary consideration
(Tanzi 1995).
Within the context of the threshold corruption model, this process would work as detailed below. Consider the
new group of government administrators in Table IV at Thresholdt=0 (i.e., the threshold distribution at time t=0).
Assume that this is the situation in a former European colony (e.g., India or Pakistan) immediately after
independence. Corruption is found only at the lowest levels of the administration and the upper ranks are fairly
honest. The only openly corrupt individual here is a2 and society is extremely honest (n=0.9 and CT=0.05). The result
of the removal of an upper echelon not responsive to local pressures is a continual erosion of individual corruption
thresholds--shown here by a 0.03 decrease in each time period. As a result of this continual erosion, at t=2, individual
b2's corruption threshold is reached and the government's weighted corruption level, CT, reaches 0.10. This is not
enough to set a bandwagon effect in motion. We simply have a slightly more corrupt government. At t=4, the
continued erosion in corruption thresholds causes c2 and d2 to behave corruptly as well and the government's CT
reaches 0.30. We still see no bandwagon effect since the erosion in thresholds has not yet reached the critical point.
At t=5, individual e2's personal corruption threshold is reached and he behaves corruptly as well, raising the CT to
0.40. Finally at t=7, the steady decline in individual corruption thresholds reaches the critical point and a bandwagon
effect is triggered--corruption is now complete as the government's CT reaches 1.
16
====================================================================
Table IV
Individuals
a2
b2
c2
d2
e2
f2
g2
h2
i2
j2
CT
Thresholdt=0
0
0.10 0.20 0.25 0.45 0.60 0.65 0.65 0.90 0.95 0.05
Thresholdt=1
0
0.07 0.17 0.22 0.42 0.57 0.62 0.62 0.87 0.92 0.05
Thresholdt=2
0
0.04 0.14 0.19 0.39 0.54 0.59 0.59 0.84 0.89 0.10
Thresholdt=3
0
0.01 0.11 0.16 0.36 0.51 0.56 0.56 0.81 0.86 0.10
Thresholdt=4
0
0.00 0.08 0.13 0.33 0.48 0.53 0.53 0.78 0.83 0.30
Thresholdt=5
0
0.00 0.05 0.10 0.30 0.45 0.50 0.50 0.75 0.80 0.40
Thresholdt=6
0
0.00 0.02 0.07 0.27 0.42 0.47 0.47 0.72 0.77 0.40
Thresholdt=7
0
0.00 0.00 0.04 0.24 0.39 0.44 0.44 0.69 0.74 1.00
Weight
0.05 0.05 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.20
====================================================================
Table IV illustrates quite nicely how a gradual decay in administrative standards--as shown by the eroding
corruption thresholds29--could lead to a systemically corrupt society. This gradual transition, fueled by a general
societal preference for non-arm's-length transactions, is an inevitable outcome insofar as the rulers of the new state
are unable or unwilling to prevent themselves or their subordinates from using non-arm's-length criteria in their
official dealings. Leadership determined to maintain high societal honesty levels would, of course, obviate the need
for analyses such as this.30 Similarly, an inefficient or non-independent legal system would also help hasten this
transition to systemic corruption if it provided inadequate recourse to those elements of the population who wish to
prevent a non-arm's-length criteria in decision making.
So far the implicit statement in the previous models has been that corruption spreads from the bottom up. This
belief is correct insofar as the stated impetus for corruption came from outside the administrative structure and first
affected the lower ranking members of the government--whose corruption thresholds were lower than those of their
superiors. But, in general, would these corruption thresholds actually be lower for lower-ranking officials? I would
argue, initially at least, yes. This is because natives in the upper ranks of the colonial administration were held to a
higher standard of morality than natives in the lower ranks. Thus, having served in that environment, high-rank
natives who moved into the highest strata of government would, initially at least, retain a comparatively higher
corruption threshold.31 Huntington (1968, 68) is in accord with this belief. This is, however, clearly at odds with the
historical record of rulers such as, for example, Mobutu, Marcos or Duvalier. In these cases, it is not easy to argue
that corruption was a bottom-up phenomenon. The expansion of the threshold corruption model in the next section
shows explicitly how high-level corruption can rapidly change the entire institutional culture from relative honesty to
systemic corruption.
The threshold corruption model shows how the societal norm could gradually change from honest to corrupt.
The model can also easily accommodate cases of very rapid transition. Consider Table V below. This society has the
29
Within the context of the original PA model, the process by which corruption thresholds are lowered would work thus:
continued pressure by family and friends on government administrators marginally lowers their corruption thresholds as their
willingness to accept a slightly higher risk level increases. Some hitherto honest individuals now act corruptly. This has two
effects: (i) some individuals may reevaluate their own assessment of the situation in light of others' successful corruption; and
(ii), family pressure on these still honest individuals increases--"Why can't you do for your family what X did for his?" This
causes a further fall in individual corruption thresholds and more corrupt acts. More corruption leads to a further drop in
corruption thresholds, which, in turn, leads to....
30
See Quah (1989) for an account of how, in Singapore, determined leadership actually raised the societal honesty level to
a point higher than that under the colonial regime.
31
See, for example, Philip Mason's highly readable memoir, A Shaft of Sunlight: Memories of a Varied Life (NY: Charles
Scribner's Sons, 1978), for an account of life in the elite pre-independence Indian Civil Service (ICS) and, especially, the
pronounced differences between native members of the ICS and lower ranking administrative/support staff.
17
same threshold distribution as in Table IV but the weight distribution is changed to reflect its more hierarchical nature
and the greater power of its leadership. Now suppose that there is a change of leadership and individual j3, the most
influential person in this distribution is replaced by k3, the result of, say, a military coup, at the start of time period 1.
Individual k3 has a corruption threshold of 0 and this results in a bandwagon effect which causes the system's norm to
change from honest to corrupt--the new CT is now 1. With such a drastic change in leadership corruption thresholds,
the process which took seven time periods in Table IV is here accomplished in one, thus underscoring the extreme
importance of leadership in determining the speed of the transition.
====================================================================
Table V
Individuals
a3
b3
c3
d3
e3
f3
g3
h3
i3
j3
CT
Thresholdt=0
0
0.10 0.20 0.25 0.45 0.60 0.65 0.65 0.90 0.95 0.05
Thresholdt=1
a3
0
b3
c3
d3
e3
f3
g3
h3
i3
k3
CT
0.10 0.20 0.25 0.45 0.60 0.65 0.65 0.90 0.00 1.00
Weight
0.00 0.00 0.00 0.05 0.05 0.05 0.10 0.15 0.20 0.40
====================================================================
We could add further refinements to the basic model and move closer to the real world. For example, the size
of the population could be increased, enabling the creation of different layers of hierarchy, arranging the population
into peer groups and giving the peer's actions greater weight, and so on. Thus we could show the existence of
(relatively more) honest sub-groups in a (relatively more) dishonest society and vice-versa.32
VII
Incorporating Cross-Peer Linkages into the Slippery Slope of the Culture of Corruption
Several authors have remarked upon the seeming ability of the South Korean and Japanese governments to
keep their economic bureaucracies relatively honest while the political establishment was not (Kong 1996; Khan
1996; Dutt and Kim 1994). How is such an insulation possible? And can the threshold corruption model
accommodate such real-world variations? The answer to the first question is, "Through cross-peer linkages," and the
answer to the second question is, "Yes."
But why are cross-peer linkages important? Such linkages are, of course, common in the real world in that
people in a government department are not hermetically sealed from outside influence. They are aware of what their
peers or near peers in other departments are doing or, what is more important for our purpose, what they are getting
away with. Furthermore, individuals are often transferred across government departments and an individual who
behaves corruptly in one department is not likely to automatically become honest if he is transferred to another: it
depends upon the relative Cs of each department and the individual's own corruption threshold. Individuals thus
respond not only to their leaders' actions but to their peers as well. They take their cross-departmental peers' behavior
32
All societies are sub-divided into myriad sub-groups and, of course, individuals may belong to more than one sub-group.
For example, one sub-group may be individuals working for a particular corporation. Another may be alumni of a particular
university. A third may be members of a particular religious denomination. Any given individual may belong to all three (or
more) of these sub-groups. Many of these sub-groups influence the behavior of members of different sub-groups but this may not
always be so; the issue is one of salience. Your boss' behavior may influence your own but it is unlikely that the behavior of the
boss of a fellow member of the church choir is going to affect materially your own behavior. Sub-groups that are seen as criminal
entities often do not affect persons outside their sub-group. For instance, in the U.S. there exist urban gangs which carry out a
variety of criminal activities but the behavior of these criminal gangs do not (usually) cause individuals in significantly different
sub-urban areas to pattern their behavior after gang members. The behavior of gang members is simply not salient. On the other
hand, gang members' behavior is likely to affect the attitudes and behavior of non-members who live in the same
neighborhood--non-members may wish to emulate them.
18
into account when they decide how to act. It is necessary to acknowledge this important demonstration and
replication aspect of the spread of corruption. Successful corruption by members of one government department acts
in the same way that a bacterial infection in one organ of the body does: if left untreated it can quickly spread to other
organs and then rapidly reach systemic levels. My model can incorporate this quite easily by making just such a
linkage.33
Cross-peer linkages are connections between individuals in different departments, usually but not exclusively
between those who are near or at the same level: for example between c4 and l4 or d4 and n4, in Table 3.6, which
depicts a very simple form of government that has two separate departments, a finance department (FD) and a public
works (project implementation) department (PD). Both departments ultimately report to one man, the Director
General (DG). Each department is composed of nine individuals, excluding the DG, and every individual in the
government has a corruption threshold and a weight. The two-departmental situation can be represented as in Table
VI.34
Both departments have the same DG and he has a corruption threshold of 0.65. Although not incorruptible, he
has the highest corruption threshold in the government. For the sake of preserving the hierarchical nature of the
relationship between individuals, their weights and corruption thresholds are now arranged in descending order. The
separate corruption thresholds and weights of each individual in the two departments give us two different C
(corruption) values (CF and CP, the weighted proportion of corrupt individuals in each department) as an indicator of
their relative degree of corruption. Thus we see that the DG has two honest departments. That is, the (weighted)
norm for each department could not be described as "corrupt." Taking each individual department as a separate entity
(and including the DG as the head of each), their departmental Cs are relatively low (CF and CP of 0.30 and 0.10
respectively). While the corruption thresholds of the DG and the senior-staff of each department are considerably
higher than that of their subordinates, the former are not incorruptible. Under the right conditions, a systemically
corrupt outcome is certainly possible.
33
Granovetter (1973) discusses how "weak ties" between individuals can affect decision-making, I am considering weak
ties between departments here in that individual's do not consider just a peer's behavior but that of the peer's department as a
whole..
34
This type of analysis can be applied equally well to a society that has two ethnic groups or comprises of two regions. It
can also be easily extended to more than two groups or regions.
19
==================================================================
Table VI
Corruption
Director General
Threshold
Weight
x4
0.65
0.20


Finance Dept.
Public Works Dept.
Corruption
Corruption
Individual
Threshold
Weight
Individual
Threshold
Weight
a4
0.60
0.15
j4
0.60
0.15
b4
0.50
0.15
k4
0.50
0.15
c4
0.40
0.10
l4
0.40
0.10
d4
0.35
0.10
m4
0.30
0.10
e4
0.20
0.10
n4
0.25
0.10
f4
0.10
0.10
o4
0.25
0.05
g4
0.05
0.05
p4
0.15
0.05
h4
0.00
0.05
q4
0.00
0.05
i4
0.00
0.00
r4
0.00
0.05
Sum of weights
0.80
Sum of weights
0.80
CF = 0.30
CP = 0.10
=================================================================
The behavior of people in each department determines its C, the level of corruption to be found in it. It follows
then that the individual contemplating corruption would want to know about the overall level of corruption in
government when making his cost-benefit evaluation. So in considering whether the overall governmental C meets
his corruption threshold, the individual uses a weighted average of his department's C and that of the rest of the
government. Using a weighted average is justified because, while a shared corporate culture does shape individual
expectations (Tirole 1996, 3; Huang and Wu 1994, 400), the influence of those individuals with whom one has the
most daily contact is likely to be stronger than that of the entity as a whole. However, the corporate culture of the
entity as a whole cannot be neglected entirely since it has an influence, if not on the rank and file, then at least on the
sub-group's leadership. So the weight assigned by any individual i to the C of the home department must be at least
equal to, if not greater than, the weight assigned by that individual to any other department's C. Equation 8 below
shows such a weighted average, assigning equal weights to both departments.
CT = (0.5CF + 0.5 CP)
(8)
Now consider individual p4 in the PD. His corruption threshold is 0.15 but the departmental weighted
corruption level, CP, is only 0.10. His threshold has not been reached, but this would hold true only if the PD is indeed
hermetic. However, this is not likely. So, in making his corruption decision, p4 uses a cross-departmental weighted
average, CT, in determining whether or not his individual corruption threshold has been reached:
CT
= (0.5*0.30 + 0.5*0.10)
= 0.20 (versus a CP of 0.10).
(9)
Thus p4 corruption threshold has been reached and he acts corruptly at time t=0. The new CP is now 0.15 and
the CT has increased to 0.225. This is still not enough to trigger a systemic corruption bandwagon. However,
consider what would happen if individual o4's corruption threshold drops to 0.20 at time t=1. Using just the
departmental CP of 0.15 is not enough for him to behave corruptly. However, the cross-departmental weighted CT of
0.225 is high enough to meet his corruption threshold. He behaves corruptly. In turn, the resulting higher weighted
20
corruption average is enough to reach n4's corruption threshold of 0.25 at time t=2. This now raises the weighted
corruption level enough to reach m4's threshold of 0.30 at time t=3. This in turn causes d4's threshold to be reached at
time t=4, which in turn.... The final result is systemic corruption as both departmental Cs reach 1.00.
The same results could be obtained in a much shorter time period if a change in government leadership, e.g., a
military coup or successful revolution at time t=1, replaces individual x4 with someone whose corruption threshold is,
say, 0.2. This would set off an immediate corruption bandwagon effect at time t=2 that would convert the entire CT to
1.00. Leadership corruption greatly speeds up the transition mechanism. Thus, under the right circumstances,
systemic corruption can come about in either a top-down or bottom-up manner.
Now assume that we start from a systemically corrupt government, CT of 1.00, with the same individual
corruption thresholds as in Table 3.6. What happens if a new leader whose corruption threshold is 1.00 (the
proverbial incorruptible leader) somehow comes to power? Nothing. The new governmental CT, 0.80, is lower than
before but it is not enough to make any corrupt official change his behavior. Every one other than the leader is still
corrupt and there is no real change in how business is done.35 If the new leader, with his weight still at 0.20, could
also replace the top three individuals in each department with individuals with corruption thresholds above 0.80, then
the resultant change in leadership would swing the pendulum in the other direction--towards honesty. In this case we
are then talking about a wholesale purge of the entire leadership of the government. Of course, if the new leader has
a weight of more than, say, 0.40, then the current threshold distribution would result in a swing back towards a norm
of honesty fairly effortlessly. Unfortunately, the history of much of the developing world does not lend itself to either
proposal. We have seen neither a coup replacing the corrupt by the honest nor have we seen the rise of an honest
Leviathan determined to embark upon the Herculean task of cleaning the Augean stables of government.36
There is nothing sacrosanct about the 0.5 weight assigned to each department. It is simply a convenient number
to use given that the weight assigned to the individual's home department must be at least equal to that assigned to the
rest of the government. It is also a convenient one to use given the individual weights and thresholds used in Table
3.6. Changing this parameter may, under some circumstances give different, either good or bad, results. That is also
acceptable. For example, the British colonial tradition was to isolate the armed forces from the rest of colonial society
as much as possible. Thus a would-be corrupt army officer (at least in the immediate post-colonial era) would assign
a much greater weight to the army's C than to that of any other department. The normal weight assignment would be
to give a substantially higher weight to one's own department than to that of one's peers. Thus, for example, in the
Japanese and South Korean (under General Park) traditions, the economic bureaucracy was kept isolated and shielded
from the political sphere (Kong 1996; Dutt and Kim 1994) while in the Indian case, the once insulated and highly
capable Indian Administrative Service was turned into the civil service wing of the ruling Congress Party (Dutt and
Kim 1994). So for the Japanese and the South Korean economic bureaucracy, the weights assigned by a would-be
corrupt official to the departmental C would be much higher than the weight assigned to the C of the politicians.
Some spillover of corruption is of course likely to occur in even these tightly insulated cases, as shown by recent
reports of corruption in the top bureaucrats of the Japanese Finance Ministry (Efron 1998).
The threshold effect in Table 3.6 did not have a specific time element attached to it. The usual expectation
would be that an overnight change in most governments is not likely. Even in Zaire, Mobutu's reign did not convert
the entire national government into a kleptocracy overnight. We are likely to see a more gradual erosion over time (as
in Table IV), with the rate of corruption spread being determined by several factors (pervasiveness of
35
As an example of this, consider Klitgaard's (1988, 173-175) analysis of the efforts to root out corruption in a government
department in "Ruritania" (actually Pakistan) by a newly appointed "Mr. Clean" that were unsuccessful because he did not have
the full support of his superiors and corrupt departmental officials were able to stymie his efforts. Feichtinger and Wirl (1994,
116-117) argue that this is essentially what happened to Gorbachev's reforms in the Former Soviet Union.
36
Lee Kwan Yew, first Prime Minister of Singapore, could conceivably be cast in the role of an anti-corruption Hercules
but he was not dealing with a systemically corrupt system (see Quah 1989 for more on Yew's anti-corruption strategy).
21
non-arm's-length dealings, the corruption inclinations of the higher leadership, the efficacy of the judicial system, and
so on).
VIII Some Implications of the Threshold Corruption Model
Many, if not all, of the following results are not surprising to any well informed observer of the corruption
scene, whether in the Third World or the First. For example, the importance of leadership is intuitively obvious since
"Do as I say, not as I do" is a notoriously ineffective leadership principle, and the lack of institutional constraints on
the leadership in the Third World has been remarked upon by many (among others, Ades and Di Tella 1996; Diamond
1995; Ostrom 1993; Leff [1964] 1989; Klitgaard 1988; Andreski 1968). However, the threshold model of corruption
described here contributes to the literature on corruption since it provides a mechanism that can actually generate
these results. Since these results can be generated, it is also possible to see under what circumstances these
consequences can be avoided.
The principal-agent model (PA) has no readily generalizable stylized facts for the analysis of systemic
corruption since, as I have noted, it assumes that the leaders--the principals--are honest. By definition, systemic
corruption is the result of a completely corrupt leadership. The basic PA model is also not readily applicable to other
situations which involve system-wide consequences, although it is invaluable for analyzing individual actions in a
wide variety of circumstances. The basic PA model has, as shown in section II earlier and as is apparent from the
model itself, definite implications for the effect on individual behavior of changes in corruption due to changes in the
level of punishment or the probability of detection. The model described here, building upon the PA model, has more
to say about the likely effects of different circumstances on systemic corruption.
i
Variability of the Demonstration Effect
Given the importance of role models (the actions of influential people in society), frames of reference (the
actions of one's own leadership versus the actions of all other individuals in society) and in-group versus out-group
behavior (the actions of people in one's own department versus the actions of others) in determining an individual's
own actions, corrupt leadership is more likely to lead to systemic corruption due to the variability of the
demonstration effect. In plain English, you are more likely to be corrupt if your superior is, since his actions carry
greater salience for you. This fact is captured by the weights assigned to each individual. This, of course,
underscores the importance of leadership behavior in determining the outcome: corruption or honesty.
ii
Centralization of Power
A clear implication of this model is that countries with a highly centralized system of government will be more
vulnerable to systemic corruption. This is because the inclinations of the leadership have a direct effect on a larger
number of civil servants at all levels of government. In short, the network of interactions among individuals in a
centralized system of government allows for the more rapid transmission of corruption. In more decentralized
systems, in a best-case scenario, corrupt leadership in one part of the administrative apparatus may have no spillover
effects since (i) it would not serve as a reference group for many civil servants and (ii) the effects of corrupt leadership
may be countered by other non-corrupt leaders at different levels of government who have independent power bases.
Alternatively, as my interdepartmental weighted corruption model shows, the spread of corruption may simply be
slowed down at best.
iii
Institutional Constraints
An indirect implication of this model is that countries that do not have independent institutions with
countervailing power against the executive, i.e., no formal system of checks and balances based on a separation of
powers system such as that of the US, will be more susceptible to systemic corruption than countries that do. This is
because such a system raises the likelihood that corruption, especially high-level corruption, will be detected and
22
punished (i.e., p is lowered). Furthermore, it also creates independent power bases that the executive has, by
definition, very little control over.
iv
Culture
Culture, i.e., the accepted norms and practices of a society, matters greatly in determining whether the final
outcome is systemic corruption or relative honesty. This is because it is exactly this culture--the environment that the
individual concerned has to live with and in on a daily basis--that provides him with his behavioral and attitudinal
cues. So it is the dominant cultural norms and not some abstract, Weberian ideal of the perfect bureaucrat, that guides
the actions of the vast majority of the decision-makers in any government entity. It stands to reason that societies that
have a norm of "non-arm's-length" dealings, i.e., which are characterized by a relatively strong "traditional "
socio-economic structure, are much more likely to be corrupt than societies that have a relatively weak "traditional"
component to their social fabric. As detailed earlier, the latter tend to have a stronger social safety net and weaker
interpersonal ties between individuals. In this context, the development of strong formal institutions that can
substitute for traditional ones becomes very important if corruption is to be reduced.
v
Predictability
The final state of the system can be only, at best, imperfectly predicted. Imperfect predictability means that, ex
ante, we cannot make authoritative pronouncements about the final outcome no matter how ‘obvious’ it may seem ex
post. This is because so many "small events," many of which are unpredictable and whose importance for the future
cannot be known now, determine the final outcome that seemingly minor changes may have major consequences
down the road.37 In addition, the likelihood of a corruption bandwagon taking place is extremely sensitive to the
distribution of corruption thresholds in any given population or subgroup. Thus a seemingly minor difference may
give us very different results. For example, consider Table 3.6. Here it is individual o4 that sets the corruption
bandwagon in motion when his threshold dropped to 0.20 from 0.25. This minor change was enough to initiate a
sequence that changed the entire status quo. If this change had not taken place, there would be no bandwagon effect.
vi
The Possibility of Reform
The final, and most dispiriting, stylized fact emerging from this model is that minor reforms are generally
useless in a systemically corrupt society. Once the CT of the entire system is close to one, small changes cannot
reverse the status quo. Successful reforms would then have to include a major transformation of the entire governing
apparatus. Simply inducting honest leadership at the very top of the administrative structure will not necessarily give
us an honest outcome--unless the systemic corruption equilibrium is unstable, i.e., sensitive to small changes in
threshold distribution. For example, in Table II, if individual c1's corruption threshold had been 0.25 instead of 0.2, a
relatively minor difference given the population threshold distribution range of 0 to 1, the effect of the drop in b1's
corruption threshold from 0.2 to 0.1 would have been negligible. No bandwagon effect would have taken place.
Unfortunately, given the history of increasing corruption in the developing world, the norm appears to be more
corruption and not less. Successful reform will most likely require a complete reconstruction of the existing
administrative structure--probably the induction of new people with a new threshold distribution.
The PA model holds that corruption reform is relatively simple: raising the probability of detection and
punishment or increasing the severity of punishment will reduce corruption. Such an undertaking, for reasons already
explained, is not likely to be successful. However, as history of the West shows, it is possible to move from a
low-honesty equilibrium to a high-honesty one eventually. The disheartening aspect of this transition in the West,
from the standpoint of today’s Third World countries, was that it took several generations to achieve.
37
Before independence, Joseph Mobutu was a clerk-sergeant in the Belgian Congo's Force Publique. He was selected for
commissioning as an officer by the colonial authorities, an event that paved the way for his eventual coup d'etat. This could in no
way have been foreseen by the board that approved his selection.
23
IX
Conclusions
Corruption is stigmatized by virtually all societies at all times (Ali 1985; Noonan 1984) because, in addition to
the ethical and moral damage it does, it misallocates resources, reduces economic surplus and consequently economic
growth, and degrades the all important link between effort and reward. A corruption free society is as likely as a
crime free one, but there are certainly great benefits from reducing corruption, at least until the point where the
marginal benefit of reduced corruption is equal to its cost.
There are essentially two ways by which a society can become thoroughly corrupted. The first is that, as Table
III showed, for a gradual erosion of corruption thresholds to give us a systemically corrupt outcome. This is a
relatively slow process but, given a societal norm of non-arm's-length dealings and the absence of truly exemplary
leadership determined to prevent it, there is a certain inexorable inevitability to it. The second is through a much
more rapid transformation of existing societal mores due to the rise to power of a thoroughly corrupt leadership. This
is the model laid out in Table V of this work. I would argue that the presence of both a non-arm's-length dealings
norm and leadership corruption, acting as a mutually reinforcing relationship, is responsible for the speed of the
transition to systemic corruption in the developing world.
However, as has been shown, corruption has been universally condemned by all societies. If the entire societal
norm is one of non-arm's-length dealings and the leadership is no different than the population, then why condemn
them? Because it is clear that systemic corruption has many of the characteristics of a Prisoner's Dilemma game. If
everyone is honest, then we all gain. If all are honest save I, then my corruption has a much larger reward accruing to
it than my honesty. However, if all are corrupt, then we all lose. Thus a universal condemnation of corruption serves
an important societal function. Its aim is to prevent, at the systemic level, self-interested (i.e., profit-seeking)
behavior from taking the next step and turning into opportunistic (i.e., corrupt) behavior. The latter, of course, serves
no good societal purpose.
More effective corruption reduction is possible only if we can better understand its spread. The complete
eradication of isolated corruption is probably neither feasible nor desirable from a cost-benefit perspective. The
removal of systemic corruption is both feasible and desirable, though neither easy nor inexpensive. The traditional
approach to modelling corruption in the social sciences has been to view it as a principal-agent problem. While
invaluable in understanding the individual's decision to act corruptly or not (i.e., what is the individual's corruption
threshold), the PA model does not increase our understanding of how corruption spreads in a society where there are
few effective institutions to temper the power of the executive. In the atomistic PA model, the individual changes
only his own behavior, without reference to that of any other person. My threshold model underscores the importance
of feedback effects on the spread of corruption. Specifically, it incorporates the effects of high level corruption on the
overall corruption level in society, an important social dynamic that many other theories of corruption miss. Thus this
model includes important sociological effects that a simple PA model cannot adequately handle: the importance of
role models, particular frames of reference and in-group versus out-group behavior in determining an individual's
own actions. Finally, as Tanzi (1995) and others have argued (persuasively) that societies without a norm of
non-arm's-length dealings are much more likely to be corrupt than societies that do have this norm (as shown in Table
IV). In such a situation, the character of the leadership assumes even greater importance. When the natural
predilection of the vast majority of the bureaucracy is to favor their own friends or relatives, it requires an iron-willed
executive indeed to enforce the rules. A corrupt leadership is far less likely to exhibit such resolve than an honest one.
If for no other reason than the fact that a dishonest leadership has already demonstrated that its strength of character is
lacking.
Historical data from India appears to support this contention. The norm in Indian/South Asian society is
non-arm's-length dealings (Tanzi 1995; Alexander 1994; Dewey 1991; Bhatia 1967). The Indian government's own
Santhanam Report indicated that, seventeen years after independence, leadership corruption was now widespread
(Myrdal [1968] 1989b, 417). More recent reports from India indicated that the country was rapidly reaching systemic
24
levels of corruption (Tharoor 1997; Economist March 2 1996, 36; Dahlburg 1996; Los Angeles Times, 31 October
1996, pA3).
The falsification of at least a part of the model would be if corruption among the leadership did not cause
increased (overall) administrative corruption. South Korea and Japan, two post-WWII economic success stories,
have had extremely high levels of leadership corruption. However, many analysts (e.g., Khan 1996; Kong 1996)
argue that these countries were unique in their ability to keep the administrative bureaucracy separate from the
political leadership. Unlike most Third World states, the administrative bureaucracy was not turned into a de facto
extension of the ruling party apparatus, one whose purpose was not policy implementation but to ensure the
continuation in power of the ruling establishment. Thus, using the analysis from equation 8 of the model, the weight
assigned by a Japanese or South Korean bureaucrat to the corruption of the political establishment would have been
very low. There would then be minimal spillover corruption effects. Unfortunately, this kind of
compartmentalization of the administrative (i.e., policy implementation) apparatus is very rare in the Third World.38
Dutt and Kim (1994, 196) describe how the Indian Administrative Service, "virtually unparalleled in the Third
World" in its professional competence, was turned into a de facto extension of the ruling political party.
Finally, what does the threshold corruption model add to the methodological literature on corruption? While
neither the concept of using a Principal-Agent (PA) approach to model criminal behavior nor that of using a
bandwagon/threshold approach to model social situations is new, what is an original contribution to the literature is to
combine both of these methodological approaches into a cohesive whole that increases our understanding of how
systemic corruption can come about.
The PA model gives invaluable insight into individual decision-making by examining the incentives faced by
each individual as he or she is trying to decide whether to act corruptly or not. The threshold model brings the social
interaction aspect of the corruption decision to the fore since corruption is a social act and no analysis of systemic
corruption can ignore this aspect of it.
The threshold corruption model utilizes both of these methodological approaches to help explain systemic corruption.
The model's major innovation is in endogenizing individual corruption thresholds, and hence individual incentives,
and then showing how systemic corruption can come about.
Furthermore, the threshold corruption model can account for cultural effects in the systemic corruption
phenomenon. This is done by specifically incorporating the effects of a norm of "non-arm's-length" transactions into
the analysis of how the transition to systemic corruption occurs. Finally, the assigning of a weight, consonant with his
standing in society, to each individual incorporates the asymmetric demonstration effect, i.e., the importance, of
leadership behavior into the analysis of both how a systemically corrupt society comes about and of any efforts to
reform it. These three main points, I contend, make the threshold corruption model a valuable addition to both the
corruption and threshold literatures.
This threshold corruption model, of course, would not have been possible without the work of Becker (1968;
1974), Granovetter (1973; 1978), Schelling (1978) or Kuran (1995; 1991; 1989; 1987). It takes its basic threshold
sequence and general representation format from Kuran and its underlying logic from Granovetter and Schelling. It is
thus a logical extension of these works as well as an original contribution/extension of their societal analysis.
In conclusion, let me reiterate that the model is useful in understanding how corruption spreads and achieves
systemic levels. The model gives a more formal exposition of the old Chinese adage that fish (and nations) rot from
the head down. Well over two millennia ago, a Chinese sage wrote lines that indicate that there is truly nothing new
under the sun.
The key to success or failure in government lies in the ruler.
If the inkling line is properly set above,
38
However, reports (e.g., Efron 1998) seem to indicate that corruption has spread to the civil service mandarins of the
Japanese finance ministry. Thus in the long run, even in Japan, successful compartmentalization is not necessarily possible.
25
the wood will be straightened beneath it.
It is not that the inkling line does anything in particular to the wood,
but rather that the disciplining of the wood in following the inkling line makes it so.
Thus if the ruler is truly upright,
honest officials will be commissioned and villainous persons will hide themselves;
but if the ruler is not upright,
the wicked will get on in the world and loyal subjects will withdraw into retirement.
(Translation of section 7 of The Hua Nan Tzu, Book Nine: The Art of Rulership from Ames 1983,
182-183, in Huang and Wu 1994, 390.)
In accord with The Hua Nan Tzu, the model suggests that the single most important determinant of whether
isolated corruption becomes systemic is the honesty level and determination of the leadership to prevent systemic
corruption from becoming the norm. From this, it is a natural conclusion that higher level corruption should be
punished more severely than lower level corruption. This is because especially severe sanctions against higher level
officials will carry a much greater demonstration effect, as well as reduce the public perception that corruption
punishments are suffered only by the unimportant and unconnected. This is unfortunately, easier said than done. As
von Klausewitz said, winning a war is actually a very simple matter but, in war, it is the simplest things that are the
most difficult.
Even more unfortunate is the end result of systemic corruption: states like Pakistan, India, Zaire, Kenya, and
virtually any other Fourth World and many Third World countries that one cares to name. The state becomes
incapable of meeting any of its people's needs or of sustaining economic development. The end of the Cold War has
made it possible for international aid donors (both Western governments and international organizations) to insist on
higher ethical standards in recipient countries. During the height of the Cold War, "national security considerations"
caused them to overlook, and at times actively assist in, the looting and impoverishment of Third World countries.
Many of the most corrupt countries are either large aid recipients (e.g., India, Pakistan and Kenya) or are extremely
dependent on Western markets for their products (e.g., Nigeria). If international aid donors are interested in
improving the living standards of the poor of the world, the best first step is to insist on some level of accountability
for the rulers of the Third World.
26
BIBLIOGRAPHY
Ades, Alberto, and Rafael Di Tella. 1996. "The Causes and Consequences of Corruption: A Review of Recent
Empirical Contributions. Institute of Development Studies Bulletin 27 (April):6-11.
Alam, M. Shahid. 1995. "A Theory of Limits on Corruption and Some Applications." Kyklos 48(3):419-435.
---. 1989. "Anatomy of Corruption: An Approach to the Political Economy of Underdevelopment." American
Journal of Economics and Sociology 48 (October):441-456.
Alexander, K. C. 1994. The Process of Development of Societies. New Delhi: Sage Publications.
Ames, Roger T. 1983. The Art of Rulership: A Study in Ancient Chinese Political Thought. Honolulu: University of
Hawaii Press.
Anderson, Annelise. 1995. "The Red Mafia: A Legacy of Communism." In Lazear 1995.
Andreski, Stanislav. 1968. The African Predicament: A Study in the Pathology of Modernisation. NY: The
Atherton Press.
---. 1966. Parasitism and Subversion: The Case of Latin America. NY: Pantheon Books.
Andvig, Jens Christian, and Karl Ove Moene. 1990. "How Corruption May Corrupt." Journal of Economic
Behavior and Organization 13 (January):63-76.
Bac, Mehmet. 1996. "Corruption, Supervision, and the Structure of Hierarchies." Journal of Law, Economics and
Organization 10 (October):277-298.
Bandow, Doug, and Ian Vasquez. 1994. Perpetuating Poverty: The World Bank, the IMF, and the Developing
World. Washington, D. C.: The Cato Institute.
Banfield, Edward C. 1958. The Moral Basis of a Backward Society. New York: The Free Press.
Barkow, Jerome H., Leda Cosimides, and John Tooby, editors. 1992. The Adapted Mind: Evolutionary Psychology
and the Generation of Culture. New York: Oxford University Press.
Basu, Kaushik, Sudipto Bhattacharya, and Ajit Mishra. 1992. "Notes on Bribery and the Control of Corruption.
Journal of Public Economics 48 (August):349-359.
Bates, Robert H. 1974. "Ethnic Competition and Modernization in Contemporary Africa." Comparative Political
Studies 6 (January):457-484.
Baumol, William J., and Kyu Sik Lee. 1991. "Contestable Markets, Trade, and Development." World Bank
Research Observer 6(January) :1-18.
Becker, Gary. 1968. "Crime and Punishment: An Economic Approach." Journal of Political Economy 76
(March/April):169-217.
Becker, Gary, and George Stigler. 1974. "Law Enforcement, Malfeasance, and Compensation of Enforcers."
Journal of Legal Studies 3 (January):1-19.
Berkman, Umit. 1992. "Bureaucracy and Bribery: A Conceptual Framework. International Journal of Public
Administration, 15(6):1345-1368.
Besley, Timothy, and John McLaren. 1993. "Taxes and Bribery: The Role of Wage Incentives." The Economic
Journal 103 (January):119-141.
Bicchieri, Cristina, and Carlo Rovelli. 1995. "Evolution and Revolution: The Dynamics of Corruption." Rationality
and Society 7(April):201-224.
Bhatia, Krishan. 1967. The Ordeal of Nationhood: A Social Study of India Since Independence. New York:
Atheneum.
Bikchandani, Sushil, David Hirshleifer; and Ivo Welsh. 1992. "A Theory of Fads, Fashions, Custom, and Cultural
Change as Informational Cascades." Journal of Political Economy 100 (October):992-1026.
Bok, Sissela. 1978. Lying: Moral Choice in Public and Private Life. New York : Pantheon Books.
27
Buchanan, James. 1975. The Limits of Liberty: Between Anarchy and Leviathan. Chicago: University of Chicago
Press.
Caiden, Gerald E., and Naomi Caiden. 1977. "Administrative Corruption." Public Administration Review
37(May-June): 302-309.
Callaghy, Thomas. 1984. The State Society Struggle: Zaire in Comparative Perspective. New York: Columbia
University Press.
Chander, Prakash, and Louis Wilde. 1992. "Corruption in Tax Administration." Journal of Public Economics 49
(December):333-349.
Cheung, Steven N. S. 1996. "A Simplistic General Equilibrium Theory of Corruption." Contemporary Economic
Policy 14 (July): 1-5.
Clarke, Michael, editor. 1983. Corruption, Causes, Consequences and Control. London: Pinter.
Corneo, G., and O. Jeanne. 1997. "Snobs, Bandwagons, and the Origins of Social Customs in Consumer Behavior."
Journal of Economic Behavior and Organization 32 (March):333-347.
Dahlburg, John-Thor. “India’s Diary of Corruption.” Los Angeles Times, 1 March 1996, p.A1.
Dawid, Herbert, and Gustav Feichtinger. 1996. "On the Persistence of Corruption." Journal of Economics
64(2):177-193.
de Leon, Peter. 1993. Thinking About Political Corruption. Armonk, NY: M. E. Sharpe.
de Soto, Hernando. 1989. The Other Path: The Invisible Revolution in the Third World. New York: Harper and
Row.
Diamond, Larry. 1995. "Democracy and Economic Reform: Tensions, Compatibilities. and Strategies for
Reconciliation." In Lazear 1995.
Dutt, Amitava Krishna and Kwan S. Kim. 1994. "Market Miracle and State Stagnation? The Development
Experience of South Korea and India Compared. In Dutt, Kwan, and Singh1994.
Dutt, Amitava Krishna, Kwan S. Kim, and Ajit Singh, editors. 1994. The State, Markets and Development: Beyond
the Neoclassical Dichotomy. Aldershot, U.K.: Edward Elgar.
Easterly, W. 1993. "How Much Do Distortions Affect Growth?" in Journal of Monetary Economics 32
(November):187-212.
Economist. 1997. "Crime and Parliament." 27 September, 1997, 45.
---. 1996. "Muck and Money." 15 June 1996, 57.
---. 1996. "Muck Spreads." 2 March 1996, 36.
Efron, Sonni. 1998. "Corruption Charges Taint Agency Responsible for Regulating Banks and Extricating Nation
From its Fiscal Crisis." Los Angeles Times, 28 January.
Ekpo, Monday U., editor. 1979. Bureaucratic Corruption in Sub-Saharan Africa: Toward a Search for Causes and
Consequences. Washington, D.C.: University Press of America.
Fatton, Robert, Jr. 1992. Predatory Rule: State and Civil Society in Africa. Boulder, C.O.: Lynn Rienner
Publishers.
Feichtinger, Gustav, and Franz Wirl. 1994. "On the Stability and Potential Cyclicity of Corruption in Governments
Subject to Popularity Constraints." Mathematical Social Sciences 28 (October):113-131.
Fiorentini, Gianluca, and Sam Peltzman, editors. 1995. The Economics of Organised Crime. Cambridge: The Cambridge
University Press.
Gallagher, Mark. 1991. Rent Seeking and Economic Growth in Africa. Boulder, C.O.: Westview Press.
Giglioli, P. P. 1996. "Political Corruption and the Media: The Tangentopoli Affair." International Social Science
Journal 48 (September):381-391.
Goettingen, University of, and Transparency International. 1996. Internet Corruption Rankings. June, online
document at http://www.uni-goettingen.de/~uwvw/icr.htm.
---. 1995. Internet Corruption Rankings. April, online document at http://www.uni-goettingen.de/~uwvw/icr.htm.
28
Gould, David J. 1980. Bureaucratic Corruption and Underdevelopment in the Third World: The Case of Zaire.
New York: Pergamon Press.
Gould, David J, and Jose A. Amaro-Reyes. 1983. The Effects of Corruption on Administrative Performance:
Illustrations From Developing Countries. World Bank Working Paper No. 580. Washington D. C.: The
World Bank.
Granovetter, Mark. 1978. "Threshold Models of Collective Behavior." American Journal of Sociology,
83(6):1420-1443.
Granovetter, Mark. 1973. "The Strength of Weak Ties." American Journal of Sociology, 78(6):1360-1380.
Haggard, Stephan, and Steven B. Webb. 1994. Voting For Reform: Democracy, Political Liberalization, and
Economic Adjustment. Washington, D.C.: The World Bank.
Hallet, Jean-Pierre. 1968 [1964]. Congo Kitabu. Reprint, NY: Random House.
Heidenheimer, Arnold J., Michael Johnston, and Victor T. LeVine. 1989. Political Corruption: A Handbook. New
Brunswick, NJ: Transaction Publishers.
Hong, C. S., and K. A. Conrad. 1998. "Bandwagon Effects and Two-Party Majority Voting." Journal of Risk and
Uncertainty 16 (May-June):165-172.
Hope, Sr., Kempe Ronald. 1996. Development in the Third World: From Policy Failure to Policy Reform. Armonk,
N. Y.: M. E. Sharpe.
Huang, Peter H, and Ho-Mou Wu. 1994. "More Order Without More Law: A Theory of Social Norms and
Organizational Culture." Journal of Law, Economics and Organization 10 (October):390-406.
Huntington, Samuel P. 1979. "Modernization and Development." In Ekpo 1979.
---. 1968. Political Order in Changing Societies. New Haven, CT: Yale University Press.
Huther, Jeff, and Anwar Shah. 2000. Anti-Corruption Policies and Programs: A Framework for Evaluation. World
Bank Working Paper No. 2501. Washington D. C.: The World Bank.
Jabbra, J. G. 1976. "Bureaucratic Corruption in the Third World." Indian Journal of Public Administration 22
(October-December):673-91.
Johnston, Michael. 1996. "The Search for Definitions: The Vitality of Politics and the Issue of Corruption." International
Social Science Journal. 48(September):321-335.
---. 1982. Political Corruption and Public Policy in America. Monterey, CA: Cole.
Kaufmann, Daniel and Shang-Jin Wei. 1999. Does Grease Money Speed Up the Wheels of Commerce? World Bank
Working Paper No. 2254. Washington D.C.: The World Bank.
Khan, Mustaq H. 1996. "A Typology of Corrupt Transactions in Developing Countries." Institute of Development
Studies Bulletin. 27 (April):12-18.
Kiser, Edgar, and Xiaoxi Tong. 1992. "Determinants of the Amount and Type of Corruption in State Fiscal
Bureaucracies: An Analysis of Late Imperial China." Comparative Political Studies 25 (October):300-331.
Klitgaard, Robert. 1991a. Adjusting to Reality: Beyond "State versus Market" in Economic Development. San
Francisco: ICS Press.
---. 1991b. "Gifts and Bribes." In Zeckhauser 1991.
---. 1988. Controlling Corruption. Berkeley: University of California Press.
Kong, Tat Yan. 1996. "Corruption and its Institutional Foundations: The Experience of South Korea. Institute of
Development Studies Bulletin. 27 (April):48-55.
Krueger, Ann O. 1974. "The Political Economy of the Rent-seeking Society." American Economic Review, 64
(June):291-303.
Kuran, Timur. 1995. Private Truths, Public Lies: The Social Consequences of Preference Falsification.
Cambridge, MA: Harvard University Press.
---. 1991. "Now Out of Never: The Element of Surprise in the East European Revolution of 1989." World Politics.
44 (October):7-48.
29
---. 1989. "Sparks and Prairie Fires: A Theory of Unanticipated Political Revolution." Public Choice. 61
(April):41-74.
---. 1988. "The Tenacious Past: Theories of Personal and Collective Conservatism." Journal of Economic Behavior
and Organization. 10 (September):143-171.
---. 1987. "Preference Falsification, Policy Continuity and Collective Conservatism." Economic Journal. 97
(September):642-665.
Lambsdorff, Johann Graf. 1998. "An Empirical Investigation of Bribery in International Trade." European Journal
of Economic Development. 10 (June):40-59.
Lazear, Edward P., editor. 1995. Economic Transition in Eastern Europe and Russia: Realities of Reform. Stanford,
CA: The Hoover Institution Press.
Leff, Nathaniel. 1989 [1964]. "Economic Development Through Bureaucratic Corruption." American Behavioral
Scientist. 8 (November):8-14. Reprinted in Heidenheimer et al 1989.
Leys, Colin. 1989 [1965]. "What is the Problem About Corruption?" Journal of Modern African Studies. 3
(1965):215-230. Reprinted in Heidenheimer et al 1989.
Los Angeles Times. 1996. "World Bank Steps Up Attack on Global Corruption." 4 October, page D8.
Lui, Francis, T. 1986. "A Dynamic Model of Corruption Deterence." Journal of Public Economics. 31
(Nov):215-236.
Macrae, John. 1982. "Underdevelopment and the Economics of Corruption: A Game Theory Approach." World
Development. 10 (Aug):677-687.
Manion, Melanie. 1996. "Corruption by Design: Bribery in Chinese Enterprise Licensing." Journal of Law,
Economics and Organization. 12 (April):167-195.
Mauro, Paulo. 1995. “Corruption and Growth." Quarterly Journal of Economics, 3 (August):681-712.
---. 1994. The Composition of Government Expenditure: The Good Guys and the Bad Guys. Harvard: Harvard
University, unpublished manuscript.
Mbaku, John Mukum. 1996. "Bureaucratic Corruption in Africa: The Futility of Cleanups." Cato Journal. 16
(Spring):99-118.
Meny, Yves. 1996. "'Fin de Siecle' Corruption: Change, Crisis and Shifting Values." International Social Science
Journal. 48 (September):309-320..
Monshipouri, Mahmood, and Amjad Samuel. 1995. "Development and Democracy in Pakistan: Tenuous or
Plausible Nexus?" Asian Survey. 35 (November):973-989.
Mookherjee, Dilip, and I. P. L. Png. 1995. "Corruptible Law Enforcers: How Should They Be Compensated?" The
Economic Journal. 105 (January):145-159.
Mueller, Dennis C. 1993 [1989]. Public Choice II: A Revised Edition of Public Choice. Reprinted, Cambridge:
Cambridge University Press.
Myrdal, Gunnar. 1989a [1968]. "Corruption: Its Causes and Effects." In Asian Drama: An Enquiry into the Poverty
of Nations. Vol. 2. New York: The Twentieth Century Fund, pp.937-951. Reprinted in Heidenheimer et al
1989.
---. 1989b [1968]. "Corruption as a Hindrance to Modernization in South Asia." In Asian Drama: An Enquiry into
the Poverty of Nations. Vol. 2. New York: The Twentieth Century Fund, 951-958. Reprinted in Heidenheimer
et al 1989.
Noonan, John T., Jr. 1984. Bribes. New York: Macmillan.
North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance. Cambridge, UK: The
Cambridge University Press.
Nye, J. S. 1989 [1967]. "Corruption and Political Development: A Cost-Benefit Analysis." American Political
Science Review, (June):417-27. Reprinted in Heidenheimer et al 1989.
30
Olsen, T. E. and G. Torsvik. 1998. Collusion and Renegotiation in Hierarchies: A Case of Beneficial Corruption.
University of Bergen Economics Department Working Paper. Bergen, Norway: University of Bergen.
Olson, Mancur. 1982. The Rise and Decline of Nations. New Haven: Yale University Press.
Ostrom, Elinor, Larry Schroeder, and Susan Wynne. 1993. Institutional Incentives and Sustainable Development:
Infrastructure Policies in Perspective. Boulder, C.O.: Westview Press.
Pavarala, Vinod. 1996. Interpreting Corruption: Elite Perspectives in India. New Delhi: Sage Publications.
Pritchett, Lant. 1995. Where Has All the Education Gone? World Bank Discussion Paper.
Quah, Jon S. T. 1989. "Singapore's Experience in Curbing Corruption." In Heidenheimer et al 1989.
Rasmusen, E, and J. M. Ramseyer. 1994. "Cheap Bribes and the Corruption Ban: A Coordination Game Among
Rational Legislators." Public Choice. 78 (March):305-327.
Root, Hilton. 1996. "Corruption in China: Has it Become Systemic?" Asian Survey, 36 (August):741-757.
Rose-Ackerman, Susan. 1996. "Democracy and 'Grand' Corruption". International Social Science Journal. 48
(September):365-380.
---. 1978. Corruption: A Study in Political Economy. New York: Academic Press.
Sandler, Todd. 1992. Collective Action: Theory and Applications. Ann Arbor: The University of Michigan Press.
Schelling, Thomas C. 1978. Micromotives and Macrobehavior. New York: W. W. Norton & Co.
Scott, James C. 1985 Weapons of the Weak: Everyday Forms of Peasant Resistance. New Haven, C.T.: Yale
University Press.
---. 1972. Comparative Public Corruption. Englewood Cliffs, N.J.: Prentice Hall.
Shah, Anwar. 1997. “Fostering Responsive and Accountable Governance: Lessons from Decentralization
Experience." Paper presented at the World Bank Conference on Evaluation and Development: The
Institutional Dimension in Washington.
Shleifer, Andrei, and Robert W. Vishny. 1993. "Corruption." Quarterly Journal of Economics. 108
(August):599-617.
Stevens, Mike, and Shiro Gnanaselvam. 1996. "The World Bank and Governance." Institute of Development Studies
Bulletin. 26 (April):97-105.
Sturzenegger, Federico, and Mariano Tommasi. 1994. "The Distribution of Political Power, the Costs of
Rent-Seeking, and Economic Growth." Economic Inquiry. 32 (April):236-248.
Tanzi, Vito. 1995. "Corruption: Arm's-Length Relationships and Markets." In Fiorentini and Peltzman 1995.
Tanzi, Vito, and Hamid Davoodi. 1998. Roads to Nowhere: How Corruption in Public Investment Hurts Growth.
Economic Issues #12. Washington, D.C.: The International Monetary Fund.
Tharoor, Shashi. 1997. India: From Midnight to the Millennium. New York: Arcade Press.
Theobald, Robin. 1990. Corruption, Development and Underdevelopment. Durham, N.C.: Duke University Press.
Thompson, D. F. 1993. "Mediated Corruption: The Case of the Keating Five." American Political Science Review.
87 (June):369-381.
Tirole, Jean. 1996. "A Theory of Collective Reputations (With Applications to the Persistence of Corruption and to
Firm Quality). Review of Economic Studies. 63 (January):1-22.
Tollison, Robert D.1982. "Rent Seeking: A Survey." Kyklos. 35(4):575-602.
Tullock, Gordon. 1996. "Corruption Theory and Practice." Contemporary Economic Policy. 14 (July):6-13.
---. 1967. "The Welfare Costs of Tariffs, Monopolies and Theft." Western Economic Journal. 5 (June):224-232.
Van Rijckeghem, Caroline, and Beatrice Weder. 1997. Corruption and the Rate of Temptation: Do Low Wages in
the Civil Service Cause Corruption? International Monetary Fund Working Paper #WP/97/73. Washington
DC: International Monetary Fund.
Wade, Robert. 1989. "Politics and Graft: Recruitment, Appointment, and Promotions to Public Office in India." In
Ward 1989.
31
---. 1985. "The Market for Public Office: Why the Indian State is not Better at Development." World Development
13 (April):467-497.
Ward, Peter, editor. 1989. Corruption, Development and Inequality: Soft Touch or Hard Graft? London:
Routledge.
Waterbury, John. 1973 [1989]. "Endemic and Planned Corruption in a Monarchical Regime." World Politics
25(4):533-555. Reprinted in Heidenheimer et al 1989.
Wei, Shang-Jin. 1999. Corruption in Economic Development: Beneficial Grease, Minor Annoyance, or Major
Obstacle? World Bank Working Paper No. 2048. Washington D. C.: The World Bank.
---. 1997. Why is Corruption so Much More Taxing Than Tax? Arbitrariness Kills. National Bureau of Economic
Research Working Paper #6255.
Wilson, James, Q. 1993. The Moral Sense. New York: The Free Press.
Wilson, Peter J. 1967. A Malay Village and Malaysia. New Haven, CT: Hraf Press.
Whitehead, Laurence. 1989 [1983]. "On Presidential Graft: The Latin American Evidence." In Clarke 1983.
Reprinted in Heidenheimer et al 1989.
Witt, Ulrich. 1989. "The Evolution of Economic Institutions as a Propagation Process." Public Choice. 62
(August):155-172.
Zeckhauser, Richard, editor. 1991. Strategy and Choice. Cambridge M.A.: The MIT Press.