bureaucratic corruption and endogenous economic...

Bureaucratic Corruption and EndogenousEconomic Growth

Isaac EhrlichState University of New York at Buffalo

Francis T. LuiHong Kong University of Science and Technology

There appears to be significant diversity in the incidence of bu-reaucratic corruption across countries at different stages of eco-nomic development and under different political and economicregimes. Little theoretical or empirical analysis has been offered,however, on the link between corruption, government, andgrowth. The paper attempts to fill the void through equilibriummodels of endogenous growth. ‘‘Balanced growth’’ is derived as abalancing act between accumulating human capital, which engen-ders growth, and accumulating political capital, which mainly as-sures bureaucratic power. The analysis focuses on the interplay be-tween investment in these two types of capital and its implicationsfor long-term growth under alternative political regimes. Somepropositions are tested and confirmed empirically.

Work on this paper was originally inspired by a question from Gary Becker in aconference on the economic contest between communism and capitalism held bythe Institute for the Study of Free Enterprise Systems (ISFES) of the State Universityof New York at Buffalo in June 1991. In a dinner talk, Becker noted that time-seriesdata generally showed that while per capita income levels in former Soviet bloccountries lagged behind those of developed Western economies, the respectivegrowth rates over a number of decades were similar. This would seem contrary tomany economists’ intuition that free-market systems would outperform commandeconomies. He sought an explanation that does not question the data. Ehrlich isalso indebted to Zheng Yuan for dedicated research assistance and to Sergei Moro-zov and Jinyoung Kim for helpful comments and suggestions. We are responsiblefor the shortcomings.

[Journal of Political Economy, 1999, vol. 107, no. 6, pt. 2] 1999 by The University of Chicago. All rights reserved. 0022-3808/1999/10706S-0011$02.50

S270

bureaucratic corruption S271

I. Introduction

Evidence of bureaucratic corruption exists in all societies, at allstages of economic development, and under different political andeconomic regimes. There appears to be considerable diversity in itsincidence, however, as indicated by some international corruptionmeasures. This diversity and its association with economic growthand development are the focus of this paper.

Recent studies (e.g., Mauro 1995) indicate that countries miredin poverty are inundated by corruption (Zaire, Indonesia, and Haitiare notorious examples), whereas rich countries (Switzerland, Can-ada, and the United States) are less susceptible. There has been littletheoretical or empirical analysis, however, on the link between cor-ruption and long-term growth and development. How corruptionand growth vary with the type of political regime is even less clear.No corruption measures are available for the command economiesof former communist countries. While the reported real per capitalevels of gross domestic product in these economies lagged behindthose of Western democracies in the post–World War II era, how-ever, their long-term rates of economic growth were even higher.1

It is easy to dismiss especially this seemingly puzzling evidence byquestioning the GDP measures in command economies, or the cor-ruption measures in general. Could the apparent facts also be ratio-nalized by some theoretical considerations? Is the relationship be-tween corruption and growth causal, or is it the result of thirdfactors, including government intervention in the economy? Fur-thermore, can a theory of the link between corruption and growthprovide some insights into the growth experience of different politi-cal regimes? This paper explores these issues by applying a modelof endogenous corruption and growth.2

1 The overall growth rate of real per capita GDP (RGDPC) in the sample of 152countries used in the regression analysis of Sec. IIIF was 1.9 percent between 1962and 1992, with a mean RGDPC value of $4,654 in 1980. By contrast, the estimatedrate of growth of RGDPC in the 11 former communist countries included in thatset was 3.7 percent over the same period, with a mean RGDPC value of $3,797 in1980. The corresponding growth rates of RGDPC in the European subset of 29 coun-tries and the eight former communist countries within it were 3.1 percent and 3.7percent, respectively, and the mean values of RGDPC in 1980 in these two groupswere $8,178 and $4,727, respectively.

2 The basic model was first presented in a joint seminar of ISFES and the Institutefor Policy Analysis of the University of Toronto held in Buffalo in May 1992 and inthe session on economic growth and development in the 1992 American EconomicAssociation meetings in Anaheim (Ehrlich and Lui 1992). The models by Murphy,Shleifer, and Vishny (1991, 1993) and Shleifer and Vishny (1993) pursue similarapproaches toward the relationship between corruption and the economy, but themodels are static in nature. They do not systematically model the relationship be-tween corruption, government, and the economy’s growth path.

S272 journal of political economy

The point of departure is that any government intervention inthe economy assigns some resource allocation responsibilities to abureaucratic structure.3 Since the shadow prices generated therebytypically deviate from free-market prices, an incentive arises to closethe gap by various side payments, or bribes. Exercising the opportu-nity to obtain such rents is what is here meant by corruption.

Bureaucratic corruption is thus an inevitable aspect of govern-ment intervention in the economy, whatever its raison d’etre andregardless of the net social benefits it may confer. As an illegal activ-ity, it could be modified through law enforcement, but conventionalenforcement tends to be costly and ineffective against ‘‘victimless’’crimes (Ehrlich 1996). Moreover, as members of the political estab-lishment, bureaucrats have inroads to, and may even be part of, thelaw enforcement apparatus.

Corrupt behavior by itself need not impose a net social cost sinceit involves transfer payments from bribe payers to bureaucrats. More-over, bribes can ameliorate the deadweight cost of government inter-vention by directing scarce resources toward higher bidders (Leff1964; Lui 1985). However, since bureaucratic power holds the prom-ise of economic rents through corruption, individuals have an incen-tive to compete over the privilege of becoming bureaucrats. Existingliterature has referred to such activity as ‘‘rent seeking’’ (e.g.,Krueger 1974). It is here called investment in political capital. Suchinvestment consumes economic resources that could otherwise beused for production or investment in human capital. This is thesource of the social loss from corruption in this analysis.4

As in the recent literature on endogenous growth, investment inhuman capital is identified as the economy’s engine of growth. ‘‘Bal-anced growth’’ thus involves a balancing act between the accumula-tion of human capital, which engenders growth, and political capi-tal, which mainly assures bureaucratic power and thus has little, ifany, beneficial effects on productive capacity.5 The analysis focuses

3 The possibility that bureaucrats will auction off government services, or eventheir own services, in an open market is disregarded since this would make govern-ment transactions indistinguishable from those of the free market.

4 Along similar lines, Murphy et al. (1991, 1993) argue that rent seeking distortsthe allocation of talents away from entrepreneurship and innovation and may there-fore reduce economic growth. But they do not offer an explicit model of the rela-tionship between corruption and long-term growth in an endogenous growth con-text.

5 Some investment in social connections (an aspect of the Chinese guang-xi) mayalso be socially productive, to the extent that it cuts down on information and trans-action costs caused by incomplete markets. What is modeled as socially unproductivehere is investment designed to capture private distributional gains from bureaucraticpower when that power replaces an explicit market-pricing mechanism.


on the interplay between investment in these two types of capitaland its consequences for growth under different political regimes.

Two complementary models of endogenous growth are outlined.In model 1, agents are homogeneous, and the model offers a generalequilibrium solution for the allocation of resources to investmentin bureaucratic and productive pursuits, including work. In model2, heterogeneous agents form separate groups of bureaucrats andworkers as a consequence of different initial endowments and com-mensurate investment strategies. This model provides additional in-sights concerning the covariation of corruption and growth underalternative bureaucratic structures.

II. Homogeneous Agents Case

A. Assumptions

The economy consists of homogeneous agents: each is both a workerand a bureaucrat. Growth is endogenous to the economy, and theengine of growth is human capital. To make the analysis tractable,government plays a neutral role in the economy, its size is exoge-nously determined, and there is no effective law enforcement againstbribes. Human capital is the sole productive asset. Agents thuschoose to invest in human capital or socially unproductive bureau-cratic power (or both), the rent on which is proportional to the ex-tent of government intervention in the economy.

The opportunity set consists of four production technologies. Hu-man capital is generated by

Ht11 5 A(H 1 Ht)ht, (1)

where Ht denotes human capital at period t, H an endowed ‘‘raw’’labor capacity, ht the fraction of time invested in creating future hu-man capital, and A a technological parameter. This is essentially theinvestment technology employed in Lucas (1988), Becker, Murphy,and Tamura (1990), and Ehrlich and Lui (1991). Symmetrically, po-litical capital is generated by

Q t11 5 B(λH 1 Q t)qt, (2)

where Q t denotes the stock of political capital at time t and qt thefraction of time invested generating it. Raw labor is also introducedin equation (2) to permit a possible accumulation of political capitaleven if its initial stock were zero. Since its role seems marginal rela-tive to that of initial, or inherited, political capital, however, λ ishenceforth set equal to zero. The consumption good is producedby


Yt 5 (H 1 Ht)(1 2 ht 2 qt), (3)

where Yt represents gross output or market earnings, the product ofone’s earning capacity (H 1 Ht) and working time (1 2 ht 2 qt).The latter term accounts for the agent’s time constraint; leisure isignored in this model. Finally, the agent’s consumption depends notjust on Yt but also on the net rents or bribes obtained through therelative force of one’s political capital. The consumption, or net in-come distribution, function for individuals is given by

ct 5 31 1 θ ln1Q t

Q *t 24Yt, (4)

where θ represents the portion of all economic transactions that aresubject to government intervention, or the relative ‘‘size’’ of govern-ment in the economy. Net income from bribes is a product of gov-ernment size and a bribe-generating function, ln(Q t/Q *t ), whereQ *t represents the community’s average stock of political capital.6

The idea is that one’s ability to collect rents, or bribes, depends onhow large one’s personal power is relative to that of the averagebureaucrat: agents with Q t/Q *t . 1 would receive net transfers frombribes, whereas those with Q t/Q *t , 1 would pay them. Since allare homogeneous, net transfers per person must add up to zero inequilibrium.

The decision rule for agents is maximization of the utility froman infinite consumption stream, {ct},

U 5 ^∞

t11

βt21 1c 12σt 2 11 2 σ 2, (5)

with β , 1 and 0 , σ , ∞, with respect to the control variables ht

and qt, subject to equations (1)–(4), and the initial conditions forH and Q . An infinite horizon is assumed not just for analytical conve-nience, but also to underscore the enduring impact of a dynastyhead’s investment choices on all future offspring, as indicated byequation (1). The preference parameters β and σ represent the dis-count on future consumption and the inverse value of the intertem-poral elasticity of substitution in consumption, respectively. Agentsalso form homogeneous expectations about the community’s aver-

6 The reward from bribe taking, like crime, is also a function of bureaucrats’ hon-esty and effective law enforcement. The parameter θ represents not just governmentsize but also the probability that bureaucratic power will actually be exploited, which,in turn, reflects the moral impediments against graft and the prospect of punish-ment. Recognizing the role law enforcement may play to suppress corruption wouldnot, however, affect the model’s major propositions.


age political capital {Q *t }, t 5 1, . . . , ∞. While this variable is exoge-nous to individuals, however, it becomes endogenous at the macrolevel after the equilibrium condition Q t 5 Q *t is imposed.

B. Investment Choices

The necessary optimality conditions for h and q are a pair of second-order simultaneous difference equations.7 To economize on space,they are presented below after the equilibrium condition Q t 5Q *t is imposed:

1ct11

ct2

σ

$ βA(1 2 qt11) ; βR h (6)

and

1ct11

ct2

σ

5 β(θMt 1 Nt) ; βR q, (7)

where

ct11

ct

5(H 1 Ht11)(1 2 ht11 2 qt11)

(H 1 Ht)(1 2 ht 2 qt),

Mt ; (H 1 Ht11)(1 2 ht11 2 qt11)qt(H 1 Ht)

,

Nt ;(H 1 Ht11)qt11 Q t11

(H 1 Ht)qt(H 1 Q t11).

For investments in both forms of capital to coexist, the agent’smarginal rate of substitution in consumption must be equal to thediscounted rate of return (βR) on each. There is also the possibilityof zero investment in human, but not in political, capital. Indeed,by equation (7), as long as there is positive government intervention(θ . 0), if investment in political capital approached zero, the rateof return on this investment would approach an infinitely high value.Some investment in Q by any agents would then be optimal to takeadvantage of potential rents flowing from the necessarily positiveproduction by others. A corollary to this proposition can also be

7 The transversality conditions for an optimal accumulation of both human andpolitical capital can be shown to require that limt→∞ βt(ct)2σ Ht 5 0. Under a persis-tent growth equilibrium, this condition can be shown to be fulfilled if β(1 1 g *)12σ ,1, where g * denotes the steady-state growth rate of consumption as t → ∞. Theseconditions assure the boundedness of the objective function as well.


easily proved, namely, that if there is no government intervention,optimal q would be zero. Accumulation of some political capital, hencecorruption, is thus shown to be an inevitable aspect of government interven-tion in the economy.

Although optimal a priori, investment in political capital yieldszero equilibrium net return from bribes to all agents because of theirhomogeneity. This reflects a kind of market failure. All agents wouldbe better off not to invest in Q, but no one can afford doing so. Ifj eschews any investment, ln(Q j/Q*) would become infinitely nega-tive: j would wind up at the bottom of the net income distribution.Unless all agents cooperate to enforce a zero investment policy, eachwould continue to overinvest in political capital, to the detriment ofhuman capital accumulation.

C. Multiple Equilibria, or Stages of Development

There is no unique dynamic solution to this system. The interactionbetween political and human capital accumulation produces multi-ple equilibria. A given set of basic parameters—A, B, H, β, σ, andθ—that can support a stable, self-sustaining ‘‘growth equilibrium’’may also support two other stagnant steady states: a stable ‘‘povertytrap’’ and an unstable ‘‘development equilibrium,’’ depending onthe initial values of human and political capital. Above some criticalvalues of H and h, relative to Q and q, investment in human capitalbecomes subject to increasing returns as q is simultaneously falling.The opposite occurs if the relative magnitudes of H and h fall belowthis threshold level. This trade-off between investment in politicaland human capital creates a potential for economic instability incountries that are at a critical takeoff phase: a shock in the systemcould lead to either takeoff toward growth or regression toward pov-erty. These three admissible steady states represent, roughly, threestages of economic development.

Low-level stagnant equilibrium, or poverty trap.—There is no invest-ment in human capital, only in political capital: h 5 0 and qt 5 qt11

5 qs. Earning capacity is at its minimal level Hs 5 H. In this case wehave explicit solutions for qs 5 βθ/(1 2 β 1 βθ) , 1 (since β ,1), and cs 5 Ys 5 H(1 2 qs). Equilibrium is stable. An increase ingovernment size increases optimal investment in Q, causing personalincome and consumption to fall, or dqs/dθ . 0 and dcs/dθ , 0.

Stagnant ‘‘development’’ equilibrium.—The solutions for the controland state variables here are ht 5 ht11 5 hd 5 1 2 [(βA 2 1)/βA][(12 β 1 βθ)/βθ] , 1 and qt 5 qt11 5 qd 5 1 2 (1/βA) , 1, providedthat βA . 1; and Hd 5 HAhd/(1 2 Ahd) 5 constant. The agent investsin both Q and H, but investment in the latter, hd, is insufficient to


effect a takeoff to growth. This equilibrium is found to be unstable;that is, when we start from the equilibrium values of h, q, H, and Q,a shock moving H and h upward would raise the rate of return toinvestment in human, relative to political, capital, thus initiating atakeoff. But the opposite would happen if the shock lowers the initialhuman capital level.

Persistent ‘‘growth’’ equilibrium.—Investment in human capital is suf-ficient to generate self-sustaining growth. Specifically, as t → ∞, wehave ht → hg 5 constant; q → qg 5 constant; and (H 1 Ht11)/(H 1Ht) → Ahg 5 1 1 g, where g . 0 is the asymptotic value of the growthrate of income and consumption. Simulation analyses for parametervalues consistent with a stable growth equilibrium show that in-creases in government size lower investment in human capital whileincreasing investment in political capital, hence corruption; ordhg/dθ , 0 for all values of σ , 1, and dqg/dθ . 0 unless σ 5 0.Moreover, the increase in q exceeds the reduction in hg, so that work-ing time falls, or d(1 2 hg 2 q)/dθ , 0.8 An increase in governmentsize is thus seen to lower both the short-run level and long-run rate ofgrowth of per capita income. Also, any exogenous increase in invest-ment in political capital would necessarily lower the rate of accumu-lation of human capital and economic growth since, by equation(6), (1 1 g)σ 5 βA(1 2 qg), and thus dg/dqg , 0. By contrast, adecrease in time preference for consumption (an increase in β) willincrease both corruption and growth.

D. Corruption and Stages of Development

Comparison of the three alternative equilibria, or stages of develop-ment, produces a monotonic, but opposite, ranking of the level ofincome and the degree of corruption across the three. With the stag-nant, development, and growth equilibria labeled as s, d, and g, re-spectively, the analysis predicts a negative association between thesteady-state levels of investment in political capital (hence corrup-tion) and human capital (hence per capita income), or

qs . qd . qg whereas hg . hd . hs. (8)

8 More specifically, as t → ∞, the implicit solutions for hg and qg in a steady stateof growth are given by (Ah)σ 2 βA(1 2 q) 5 0 and Aσ21β21 hσ 1 θh 2 βθh22σ 1βA12σh12σ 2 h 2 1 5 0, respectively. There is generally no explicit solution for h andq (hence the recourse to simulations), except in the special case in which σ 5 1.In that case, h 5 β/(1 1 βθ) , 1, q 5 θh 5 θβ/(1 1 βθ) , 1, and working timeis given by 1 2 h 2 q 5 (1 2 β)/(1 1 βθ) , 1, as β , 1. Clearly then, an increasein θ lowers h but raises q by a greater amount, so that working time is necessarilylower. By eq. (6), a growth equilibrium may exist; i.e., Ahg 5 1 1 g . 1 only if βAand therefore A are sufficiently larger than one.


The incidence of corruption and its deadweight costs are expectedto be highest in less developed countries and lowest in countriesenjoying persistent per capita income growth, and this result holdsfor any values of the model’s basic parameters.9

E. The Role of Government Interventionand Initial Conditions

As shown in subsection C, an exogenous increase in governmentintervention in the economy (θ) will unambiguously lower both theshort-run level and long-run rate of growth of per capita incomeand consumption if the economy is in a growth equilibrium. Or itwill lower just the level of the two if the economy is in a stagnantequilibrium. Moreover, a large enough increase in θ can cause atransition from persistent growth to an unstable stagnant equilib-rium or a poverty trap. This result stems, in part, from the assumedneutral role of government as a productive agent, which ignores thesalutary effects of such public goods as the rule of law and publicsecurity. It would remain valid, however, even if some degree of gov-ernment intervention were conducive to growth, as in Barro (1990).In that case, the proposition would relate to an exogenous increasein government size above such a critical level.

A related implication concerns the role of initial conditions. Givena set of parameters that can in principle support growth, whetherthe economy converges in practice to one of the two stable steadystates is shown to depend on the initial levels of investment in, andthe stock of, human capital: h(0) and H(0). If the initial humancapital stock and investment flow fall short of (or exceed) their ‘‘de-velopment equilibrium’’ thresholds, an exogenous parametric shockcould nudge the economy toward a stagnant (or growth) equilib-rium, depending on the direction of the shock.

It follows that a given level of government intervention will bemore harmful to persistent growth prospects in human capital–poor

9 Proof. By eq. (1), Ah 5 Ht11/(H 1 Ht). Therefore, in state s, hs 5 0; in state d,Ahd 5 Hd/(H 1 Hd) , 1; and in state g, Ahg 5 Ht 11/Ht . 1. It follows that hs 5 0,hd , 1/A, and hg . 1/A, or hg . hd . hs. Also, from eq. (6), (ct 11/ct)σ $ βA(1 2q). Rewriting eq. (6) for the three states, in state s we obtain 1 5 (ct 11/ct)σ .βA(1 2 qs); in state d, 1 5 (ct 11/ct)σ 5 βA(1 2 qd); and in state g, 1 , (ct 11/ct)σ 5βA(1 2 qg). Since βA(1 2 qg) . βA(1 2 qd) . βA(1 2 qs), it follows that qs . qd

. qg. Because of the negative correlation between h and g across the alternativestages of development, the ranking of working time, 1 2 h 2 q, across these stagesis generally ambiguous. However, if σ 5 1, the closed-form solutions for h and qindicate that working time in the developed countries under the stable growth equi-librium, (1 2 β)/(1 1 βθ), is necessarily lower than working time under the stablepoverty trap, (1 2 β)/(1 2 β 1 βθ), as is indeed the case empirically.


countries than in human capital–rich ones. The former can affordfewer ‘‘errors’’ in government policies. This implication seems con-trary to conventional theories of development, which typically rec-ommend more involvement of government in less developed coun-tries.

III. Heterogeneous Agents Case: Workersand Bureaucrats

Missing from the homogeneous agents case are two possible typesof specialization among agents. Bureaucrats (B) and workers (W )are typically separate groups. Also, each may invest exclusively ineither human or political capital rather than necessarily in both. Theheterogeneous agents model accounts for both types of specializa-tion, as well as for the distributive effects of corruption across groups.

Added assumptions.—Each group of agents is homogeneous inter-nally. They differ only in their initial endowments of political capital:Q b(0) . Q w(0). The advantage of B over W may originate from en-dowed differences in such attributes as ethnicity, language, religion,or ideological cohesion. Whatever its source, the gap between Q b(0)and Q w(0), and the exogenously determined group size of W relativeto B, α 5 Nw/Nb, must be sufficiently large to eliminate the possibil-ity of ‘‘jumps’’ from W to B or vice versa.10 A key assumption concernsthe structure under which bureaucrats operate: ‘‘competitive’’ or‘‘monopolistic.’’ The first is a case of decentralized bureaucracy,where bureaucrats compete over relative personal power. The sec-ond involves a centralized bureaucracy in which bureaucrats’choices are controlled by a central authority. By contrast, workersare assumed to behave as strictly independent agents.

A. The Competitive Case

The opportunity set underlying production in the economy remainsformally the same as in Section II, except that equations (1), (2),(3), and (5) are now specified separately for W and B by labelingthe corresponding control and state variables by the subscripts wand b. The major substantive change modifies the net income (con-

10 While the underlying sources of heterogeneity may be contested over time, theanalysis implies that they are likely to exert long-term influence as long as there areno changes in group-specific basic parameters because of the initial endowments’impact on agents’ optimal investment strategies. Jumps from B to W are preventedby the equilibrium condition that B’s net income is at least as high as W ’s. In turn,this requires α to be sufficiently high, as indicated in subsection B below.


sumption) distribution functions to account for the distributionaleffects of corruption across groups:

cw(t) 5 {1 2 γ[1 2 Q(t)]}Yw(t) (9)

for workers and

cb(t) 5 51 1 ln3Q b(t)Q *b (t)46{Yb(t) 1 αγ[1 2 Q(t)]Yw(t)} (10)

for bureaucrats, where Q ; Q w(t)/Q b(t) denotes the ratio of work-ers’ to bureaucrats’ political capital, and γ denotes the portion ofthe economy under government influence, which is conducive toextracting rents (the analogue of θ in Sec. II). In this model, B windsup with positive net bribes, paid by W, and the the bribes are higherthe lower Q and the higher α 5 Nw/Nb and γ. Since bureaucrats arehomogeneous, however, in equilibrium each winds up with the samepolitical capital as the average, Q b(t) 5 Q *b (t), and neither reapsany ex post gains by competing over relative power.

The idea behind these specifications of the net income functionsis that a bureaucrat’s rent depends on his relative political powerboth within the B group and across the B and W groups. By contrast,a worker’s net income depends just on her relative power to avoidhigh payments to B or receive some favors as well. The basic choicefor both remains their optimal investment strategies.

The decision rule is given by equation (5), except that it appliesseparately to W and B. As in Section II, the optimization analysis foreach bureaucrat takes as given the group’s average political capitalQ *b (t), but the solution is then modified by the equilibrium condi-tion Q b(t) 5 Q *b (t).

B. Competitive Solutions: Multiple Equilibria

The interaction between the investment strategies of members of Wand B increases the number of admissible steady states. As in SectionII, however, there are just three development stages: poverty trap, stag-nant development equilibrium (which is, again, unstable), and self-sustaining growth equilibrium. These are supported by (a) zeroinvestments in H by both W and B (a poverty trap); (b) positive in-vestments in H by either group or both (at equal rates), but sufficientto support just a stagnant (development) equilibrium; or (c) largeenough investments in H by either group or both (at equal rates)that yield persistent and balanced growth. Each stage of develop-ment, in turn, is attainable through two categories of admissible

bureaucratic corruption S281TABLE 1

Competitive Equilibria

DevelopmentPoverty Trap (Unstable) Persistent Growth

Set I: B and W Invest Equally in Political Capital(qw 5 qb . 0 or Q ; Q w/Q b 5 Constant)

hw 5 hb 5 0 hw, hb . 0 hw 5 hb 5 h . 0Low-level stagnant Hw, Hb 5 constant Ahw 5 Ahb . 1

equilibrium Unstable, stagnant cw, cb grow by the same percentageequilibrium

Set II: Only B Invest in Political Capital (Qw 5 0 or Q ; Q w/Q b 5 0)

hw 5 hb 5 0 hw 5 0, hb . 0 a. hw 5 0, hb . 0qb 5 [δ(1 1 γ)β]/ Hb 5 constant Ahb . 1 (inefficient)

(1 2 β 1 δβ) Unstable, stagnant b. qb 5 1; Hb 5 0Low-level stagnant equilibrium 0 , hw 5 β12σA(12σ)/σ , 1

equilibrium Ahw . 1 if βA . 1cw, cb grow by the same percentage

(full specialization)

steady states of balanced growth: either both W and B invest equallyin political capital or only B invest in such capital (see table 1). Theadmissible steady states must be consistent with the equilibrium con-dition that B ’s net income is at least as high as W ’s, or cb(t) . cw(t),so that B do not switch to W.

The multiplicity of admissible steady states reflects the influenceof two types of externalities inherent in the competitive case: withinthe B group and across groups. A bureaucrat acting independentlylacks the incentive to take into account the effects of his own invest-ment in bureaucratic power on similar investments by competingbureaucrats, or on workers’ incentive to invest in productive humancapital.

As in the homogeneous agents case, a sufficiently large degree ofgovernment size may not be consistent with growth equilibrium;hence the general expectation that a higher γ would be associatedwith a lower level of per capita income. Similarly, under two of thepersistent growth equilibria (sets I and IIa in table 1), a higher γ isexpected to lower the short-run level as well as long-run rate ofgrowth of income for W or B or both. In contrast, greater preferencefor the future (β) will raise both income growth and corruption forW or B or both. An exception applies under the full-specializationequilibrium case (set IIb) in which B specialize just in accumulatingpolitical capital and W specialize in production and investment in


human capital. This results in a higher growth rate of output percapita relative to the other admissible growth equilibria in table 1.11

Here the equilibrium level of investment in human capital can beshown equal to hw 5 β1/σA(12σ)/σ , 1, whereas the growth rate ofworkers’ output, and thus per capita income and consumption(given α), converges on

1 1 g ; Ahw 5 (βA)1/σ . 1, (11)

provided that βA . 1.

C. Monopoly Equilibrium: ‘‘Organized Corruption’’

Suppose now that bureaucrats are endowed not just with superiorpolitical capital, but also with a tightly controlling political organiza-tion—an autocratic regime—in which a powerful, but rational, lead-ership is capable of imposing its will on all members. Politically dis-enfranchised workers, in contrast, continue to behave competitively(Ns is too large or weak to organize). In such a regime, the leadercan arrive at an ‘‘organized corruption’’ scheme that simulates anefficient monopoly solution.

The opportunity set and optimal choices.—The only change in thetechnology underlying the monopoly versus competitive case is themodified net income distribution function for bureaucrats:

cb(t) 5 Yb(t) 1 αγ[1 2 Q(t)]Yw(t), (12)

where Q(t) ; Q w(t)/Q *b (t). The solution for bureaucrats’ invest-ment strategy now internalizes the two types of externalitiesweighing on the competitive structure: (a) a contest over relativebureaucratic influence, which results in overinvestment in politicalcapital; and (b) a lack of individual incentive to consider the impactof the group’s average bureaucratic power on the behavior of work-ers—especially their investment in human capital. These externali-ties are internalized if the party’s autocratic (but informed) leader-ship chose directly the average stock of bureaucratic power for thegroup and enforced it on each member. This could be achievedthrough a tight hierarchical structure, fortified by harsh penaltiesfor noncompliance with the leadership’s imposed limits on accumu-lating political power for graft.

Permissible steady states.—Remarkably, the modified investmentstrategy in political capital now yields only three admissible steady-state equilibria for the economy.

11 By eq. (6), which here applies to workers, as t → ∞, Ahw ; 1 1 gw 5 βA(1 2qw). Since qw 5 0, gw, which here represents the long-term growth rate of per capitaincome for both workers and bureaucrats, is maximized.


i) Unique poverty trap. Investments in political and human capitalmust be identical for both B and W, or hw 5 hb 5 0 and qw 5 qb 5q . 0 , so Q ; Q w/Q b remains constant. There is a similar solutionunder competition, except that q is lower here, as competitionamong bureaucrats is eliminated.

ii) Stagnant development equilibrium. For a given value of Q ;Q w/Q b, equilibrium is similar to that under competition, or hw .0, hb . 0, and qw 5 qb 5 q . 0. Investment in human capital justsupports stagnant levels of Hw and Hb. This steady state may not exist.The solution is underdetermined because optimal Q in this case maybe incompatible with the initial endowment ratio Q(0).

iii) Unique growth equilibrium. The only admissible steady stateis a ‘‘specialization solution’’: Q w 5 0, 0 , q*b , 1 (i.e., Q 5 0); hb

5 0; and 0 , hw 5 β12σA(12σ)/σ , 1. The steady-state growth rate thusconverges on the same level achievable under the full-specializationcompetitive growth equilibrium (eq. [11]). The solutions are identi-cal except that, in the monopoly case, B’s investment in politicalcapital need not be at its upper level (qb 5 1). Since W refrain fromany investment in political capital, or Q w(t) 5 0, B can maintainQ b(t) at a level just sufficient to ensure that Q(t) 5 Q w/Q b 5 0 inequation (12), which provides maximal distributional rents for B.By devoting some of their time to work, bureaucrats can now achieveadded earnings in the amount of Yb(t) 5 H[1 2 qb(t)], which repre-sent an added constant monopoly rent.12

D. Monopoly versus Competitive Solutions

Under the competitive political market structure, it was possible fora specific set of parameters (α, γ, β, σ, A, B, and H) to support anyone of the three stages of development, including growth equilib-rium, depending on the initial capital endowments of W and B. Un-der the monopolistic structure, in contrast, such a parameter set ispredicted to yield only the growth equilibrium steady state, regard-less of initial conditions. Put differently, if the production technol-

12 Shleifer and Vishny (1993) reach a somewhat similar conclusion when compar-ing the static efficiency of a monopolistic relative to a monopolistically competitivesystem of rent seekers essentially because of the added waste involved in competitionamong bureaucrats. This analysis addresses the effects of alternative market struc-tures on both the level and rate of growth of personal income. Note that per capitaoutput and consumption levels under monopoly do not necessarily exceed thoseunder the analogous competitive solution since the relative magnitudes depend onthe ratios of B to W in the two regimes, α 5 Nw/Nb, the initial levels of humancapital in each, and the direct costs of enforcing the monopoly regime, which theformal analysis ignored for simplicity.


ogy and preferences are consistent with growth, the monopoly struc-ture guarantees its eventual occurrence.

Proof. Suppose that the monopoly were to find itself initially inone of the admissible stagnant equilibria solutions outlined in sub-section C above (cases i and ii), whereby qw 5 qb 5 q and hw 5 hq $0. The monopoly’s leadership would then realize that by raising qb

just a trifle above qw, it could eventually reap very large distributionalgains at minimal cost. The reason is that if qb . qw, then over timeQ w/Q b ; Q will tend to zero. But the optimality condition for invest-ment in political capital by workers can be shown to imply that ifQ 5 0, workers would lose any incentive to invest in political power,or qw 5 0, which contradicts the requirement that qw 5 qb . 0 in thestagnant or development steady states. Thus, if technically feasible,only the growth equilibrium would emerge in practice.

The rationale can be stated more intuitively: bureaucrats’ rentsare proportional to workers’ productivity, which is the bureaucracy’sbread ticket. To the extent that the system’s technological con-straints are consistent with self-sustaining growth in workers’ produc-tivity, decisions on the accumulation of political capital by an effi-ciently organized bureaucracy would strive to maximize workers’productivity. Indeed, this solution obtains when workers refrainfrom any investment in political capital.13

E. Implications for the Growth Experience of DifferentRegimes

The preceding propositions provide some insights about the empiri-cal evidence that motivated this paper. The analysis shows that effi-cient ‘‘monopolistic’’ systems may produce the same asymptoticlong-term growth rate of per capita income as ‘‘competitive’’ sys-tems, or even higher. The growth rates approach equality when com-petitive systems are in growth equilibrium involving a full-specializa-tion solution similar to the monopoly case (eq. [11]). It is temptingto equate these competitive versus monopolistic equilibria meta-phorically with developed democratic countries versus autocraticbut efficiently managed (at least over some lengthy period) auto-

13 As for the ranking of investments in political and human capital across thesteady-state equilibria that exist under the monopoly setup, it is easy to show that,for workers, it is the same as the ranking proved in the homogeneous agents case: thetwo are negatively correlated along the development path. For bureaucrats, however,investment in human capital is nil in the growth equilibrium whereas investmentin political capital may not be the lowest. Under the competitive setup, the sameranking as in the homogeneous agents case clearly applies in all cases of nonspecial-ization equilibria, where both W and B invest equally in political capital.


cratic command economies. On this interpretation, the steady-stategrowth rates of per capita income in the autocratic regimes may behigher than, or converge on the same level as, those in successfuldemocratic systems, although the level of per capita income in theformer regimes may lag behind that in the latter (see n. 12).

Moreover, under the full-specialization equilibria, the growth rateis not expected to be affected by exogenous changes in the size ofgovernment (γ) or the relative sizes of B and W (α). The reasonis that workers are specialized in accumulating human capital, andbureaucrats devote full time to accumulating political capital (in thecompetitive case) or investing in it an amount just sufficient to en-sure their edge over workers (in the monopoly case), while re-fraining from investment in human capital.

To be sure, this result holds only within certain bounds: if govern-ment size (γ) were to exceed a critical level, say γ*, the model’s basicset of parameters would no longer be supportive of persistent growthequilibrium.14 However, under competitive political structures,growth may result from a nonspecialization equilibrium wherebyboth workers and bureaucrats invest equally in human and politicalcapital (set I of table 1). This case is analogous to the homogeneousagent case of Section II. In this case, an increase in γ lowers theoptimal investment in human capital, as well as working time, andhence unambiguously both the level and rate of growth of per capitaincome.

The model further suggests that under all stable stagnant equilib-ria, greater government intervention, which may be especially highin a command economy, would cause a fall in the level of humancapital accumulation. The likely delay in the economy’s takeoff froma stagnant to a growth equilibrium as a result (see Sec. II) impliesthat real per capita output and consumption levels in all growthequilibria would also be lower the bigger government size (γ), re-gardless of its impact on growth (g).15

The preceding implications emphasize the interaction betweengovernment and growth through corruption. Since corruption and

14 An efficient monopoly is therefore expected to limit the degree of governmentintervention in the economy below such a critical level, and the same can be ex-pected of an efficient democratic regime, where competing pressure groups mayhold down the deadweight costs of government (see Becker 1983).

15 Average output, or per capita income, equals Y * 5 Yw µ 1 Yb(1 2 µ), whereµ 5 α/(1 2 α), and α 5 Nw/Nb. If larger government or a command economy isassociated with a larger share of the labor force in government, then governmentsize would affect per capita output directly: Since Yw $ Yb in all stable equilibria, ahigher α will lower Y *. Since B’s contribution to output is measured as governmentspending in national income accounts, however, this effect may not be captured byconventional GDP per capita measures.


growth are endogenous variables, there is no causal relationship be-tween the two. Under a competitive bureaucratic structure, however,exogenous changes in corruption (given γ) would reduce both thelevel and rate of growth of per capita income in the nonspecializa-tion growth equilibrium, as was true under the homogeneous agentscase. In the full-specialization case, the long-run growth rate is inde-pendent of both government size and the bureaucratic corruptionlevel. An exogenous increase in corruption lowers working time,however, and hence per capita income. But since the long-run equi-librium values of workers’ investment in human capital remain un-changed, the fall in per capita output may be followed by an increasein its short-run growth rate over the period of the economy’s returnto its equilibrium growth path.

F. Some Supportive Evidence

Table 2 presents the results of a regression analysis using Summersand Heston’s (1991, 1995) panel data concerning real per capitaGDP and the government share of GDP (G). The full sample con-tains 152 countries over the period 1960–92, although data for indi-vidual countries are often available for shorter periods. The depen-dent variable is the log of real per capita GDP (LRGDPC), and allregressions are based on a fixed-effects regression model. The basicspecification is

LRGDPC 5 a0 1 a1T 1 a2T 3 IG 1 a3T 3 COMM(13)

1 a4LIG 1 a5COMM.

In (13), a0 is a vector of country-specific dummy variables (they aresuppressed in table 2). The average growth rate of real per capitaincome over the sample period is estimated by the regression coeffi-cient of LRGDPC on chronological time, T. How government inter-vention (G) affects the growth rate is estimated via an interactionterm of T and the log of the initial level of G (LIG) in each country.Initial rather than current log value of G (LG) is emphasized (butboth are used) because the latter is likely to be simultaneously deter-mined with current levels of per capita income. The variable COMMis a dummy variable distinguishing 11 communist countries (COMM5 1) from other countries in the sample (see n. 17 below). Thesecountries represent empirically a distinct set of autocratic commandeconomies over the sample period. The interaction term T 3COMM thus estimates the difference between the growth rate incommunist relative to noncommunist countries. The variables LIGand LCOMM are also introduced in the regression to account for

TA

BL

E2

Go

vern

men

t,Po

liti

cal

Reg

ime,

and

Gro

wth

Reg

ress

ion

s,19

60–9

2

Dep

ende

nt

Var

iabl

e:L

og(R

GD

PC)

Eu

rope

anSe

tO

EC

DC

ou

ntr

ies

Incl

ud

ing

8Pl

us

6E

uro

pean

Eu

rope

anC

om

mu

nis

tC

om

mu

nis

tO

EC

DC

ou

ntr

ies

Co

mm

un

ist

Full

Sam

ple

(152

)C

ou

ntr

ies

(29)

Co

un

trie

s(3

0)(2

4)C

ou

ntr

ies

(6)

(N5

4,34

7)(N

588

7)(N

594

7)(N

578

9)(N

515

7)

Mod

el1

Mod

el2

Mod

el1

Mod

el2

Mod

el1

Mod

el2

Mod

el1

Mod

el2

Mod

el1

Mod

el2

Var

iabl

e(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)(1

0)

T.0

37.0

63.0

28.0

35.0

21.0

30.0

30.0

37.0

15.0

28(2

1.94

)(2

9.31

)(9

.45)

(10.

50)

(6.0

7)(8

.13)

(6.8

5)(8

.43)

(2.1

0)(3

.28)

T3

CO

MM

.017

.013

.008

.006

.005

.005

(10.

11)

(7.9

0)(6

.52)

(5.1

7)(4

.24)

(4.1

3)T

3L

IG2

.007

.000

5.0

022

.001

.006

(211

.27)

(.43

6)(1

.652

)(2

.75)

(2.4

7)T

3L

G2

.015

2.0

022

.001

2.0

04.0

01(2

20.6

2)(2

1.33

)(2

.63)

(22.

20)

(.49

)C

OM

M1.

186

.935

2.9

582

1.03

32

1.25

42

1.32

0(1

6.36

)(1

6.93

)(2

28.7

4)(2

30.0

0)(2

42.0

6)(2

40.4

0)L

IG2

.105

2.2

242

.249

21.

655

2.3

12(2

1.41

)(2

6.69

)(2

7.60

)(2

15.1

4)(2

6.09

)L

G.1

502

.295

2.2

652

.141

2.4

06(8

.82)

(27.

45)

(26.

67)

(22.

73)

(25.

64)

Adj

uste

dR

2.9

61.9

635

.965

.968

4.9

507

.962

6.9

56.9

582

.878

6.8

968

No

te.—

Th

era

nge

ofob

serv

atio

ns

vari

esfo

rdi

ffer

ent

coun

trie

s.N

umbe

rsin

pare

nth

eses

are

esti

mat

edco

effi

cien

tsov

erth

eir

stan

dard

erro

rs.


their effects on the RGDPC level across countries. This specificationhelps separate between the effects of G and COMM on the levelversus rate of growth of RGDPC.

The full-sample regressions are summarized in columns 1 and 2of table 2. The effect of government size, as captured by IG, is shownto have a significant adverse effect on the growth rate of RGDPCand a negative, albeit insignificant, effect on its level. The commu-nist subset in this sample is shown to have a higher level as well asrate of growth of per capita GDP relative to noncommunist countries(model 1). That COMM raises even the RGDPC level is attributableto the inclusion of poor countries in the sample, which may not haveyet joined a persistent growth regime.

To isolate better a set of developed countries that arguably experi-enced persistent growth, the same regression model is estimated incolumns 3 and 4 of table 2 just for the 29 European countries inthe sample, eight of which are former communist countries.16 In col-umns 5 and 6, the model is also estimated for the OECD countries—the richest group in the sample—including and excluding the rela-tively richer six European communist countries.17 The results aregenerally consistent with the predictions of subsection E above. BothG and COMM now exhibit a negative and significant effect on thelevel of per capita GDP. The communist subset is still showing arelatively higher growth rate, but the discrepancy over the developedcountries’ growth rate is much narrower than in the full sample.Moreover, government size has an insignificant effect on the growthrate of the more developed countries, and this is true for the higher-income communist countries as well. The results provide some sup-port for the proposition that the long-term growth rate in advanceddemocratic or efficient autocratic regimes may be little affected byeither the size of government or the nature of the political regime.

Table 3 presents some complementary results concerning the rela-tionship between some available corruption measures and growth.Two corruption indices, based on surveys of the business communityconducted by correspondents of Business International (see Mauro

16 European countries include Austria, Belgium, Bulgaria, Cyprus, Czechoslovakia,Denmark, Finland, France, East Germany, West Germany, Greece, Hungary, Ice-land, Ireland, Italy, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal,Romania, Spain, Sweden, Switzerland, Turkey, United Kingdom, USSR, and Yugo-slavia.

17 The OECD countries over the sample period include, in addition, Australia,Canada, Japan, New Zealand, and the United States but exclude Cyprus and Maltaand all former communist countries. The six communist countries added to this setin cols. 5 and 6 are the eight listed in the European set less Bulgaria and Romania,whose RGDPC is below the group mean of the 11 communist countries in the fullsample, which includes China, Laos, and Mongolia.

TA

BL

E3

Co

rru

ptio

nan

dG

row

thR

egre

ssio

ns,

1981

–92

(N5

763)

Dep

ende

nt

Var

iabl

e:L

og(R

GD

PC)

Sam

ple

of

68C

ou

ntr

ies

Var

iabl

eM

odel

1M

odel

2M

odel

3M

odel

4M

odel

5M

odel

6

T2

.006

.009

.039

.063

.047

.067

(21.

99)

(2.5

3)(4

.06)

(6.5

9)(5

.22)

(7.3

7)T

3IR

ED

TA

PE.0

03.0

02.0

02(5

.97)

(4.7

9)(4

.63)

T3

ICO

RR

UPT

.000

42

.000

1.0

001

(.86

)(2

.28)

(.29

)T

3L

IG2

.015

2.0

18(2

4.99

)(2

6.07

)T

3L

G2

.018

2.0

21(2

6.31

)(2

7.18

)IR

ED

TA

PE.1

73.2

12.2

12(1

0.42

)(1

2.70

)(1

2.86

)IC

OR

RU

PT.0

98.1

20.1

15(7

.60)

(9.3

8)(9

.35)

LIG

26.

532

4.29

6(2

45.1

7)(2

33.7

2)L

G.3

27.3

96(3

.61)

(4.3

1)A

djus

ted

R2

.988

4.9

878

.988

8.9

884

.989

5.9

892

No

te.—

See

not

eto

tabl

e2.


1995), are explored: REDTAPE, measuring ‘‘the regulatory environ-ment foreign firms must face when seeking approval and permits—the degree to which it represents an obstacle to business’’; and COR-RUPT, measuring ‘‘the degree to which business transactions in-volve corruption or questionable payments.’’ Both are ranked in de-scending degree of corruption severity from most (0) to least (10)corrupt. Mauro collected and used the Business International indi-ces for 1981–84 to measure their impact on the rate of growth ofRGDPC over the same four-year period.18 Since corruption is an en-dogenous variable, however, in this analysis only the initial (i.e.,1981) levels of the indices (prefixed by I) are used in the regressions.They are intended to estimate the impact of predetermined corrup-tion measures on the level and rate of growth of RGDPC over 1981–92, using the same Summers and Heston data as in table 2. Theregression model is essentially the same as (13), with the levels andtime interaction terms of the two indices added to the regression.The sample size, consisting of a panel of only 68 countries with par-tial or complete corruption data over this period, is much more lim-ited, however, than that used in the full set of table 2. Also, the aver-age level of RGDP in this set was over 30 percent higher than theaverage in the full sample of table 2 (both circa 1980), and no datawere available for communist countries.

Both IREDTAPE and ICORRUPT (the initial values of these vari-ables) appear to exert a significant adverse effect (as indicated bytheir positive signs) on the level of per capita GDP. The variableIREDTAPE also exerts a significant adverse effect on the rate ofgrowth of RGDPC, whereas ICORRUPT has an insignificant effecton it, perhaps because bureaucratic red tape is a better measure ofcorruption opportunities created through government interventionin the economy. The analyses in Sections II and III indicate thatshifts in some unmeasured exogenous factors, such as time prefer-ence for consumption, may even cause productivity growth andcorruption to shift in the same direction, especially in the moredeveloped countries, where the long-term growth rate may be inde-pendent of the incidence of corruption. The estimated effects ofgovernment conform even more sharply to those obtained in thefull-set regressions of table 2. When treated as a predetermined vari-able, initial government size, IG, is found to have an adverse, andsignificant, effect on both the level and rate of growth of RGDPC.The combined results of tables 2 and 3 are generally consistent with

18 Both are ranked in descending degree of severity of corruption from most (0)to least (10) corrupt. We are indebted to Paulo Mauro for providing the electronicdata files.


the implication of model 2, that exogenous changes in both govern-ment size and corruption would adversely affect the level of per cap-ita income, but not necessarily its long-term growth rate in the moredeveloped countries.

IV. Concluding Remarks

The two models outlined in this paper are subject to a number oflimitations. Government is taken to play a neutral role in the econ-omy, and the role of law enforcement against corruption is ignored.It is recognized implicitly only in the autocratic (monopoly) model,where the leadership is assumed to have tight control over excessiveaccumulation of power by bureaucrats. The heterogeneous agentsmodel abstracts from a determination of the equilibrium distribu-tion of agents between worker and bureaucrat groups, pinning it ondifferences in inherited political capital by members of each group.This problem is resolved, however, by the general equilibrium analy-sis of Section II, yielding generally consistent inferences. The analy-sis recognizes just two extreme bureaucratic structures—competitiveand monopolistic—abstracting from the possibility of intermediatestructures. Also, the assumption that agents make investmentchoices as dynasty heads for the benefit of all future offspring maybe more questionable for the monopoly case. Since the right of suc-cession to such a leadership position of power is uncertain undermonopoly, the leadership in such a regime is more likely to operateon a shorter, and thus less efficient, horizon than competitive struc-tures.

Yet these simplified models produce some testable insights intothe relationship between corruption, government, and growth. Amajor implication of the analysis is that, in contrast to the case ofconventional felonies in which wealth and crime are positively associ-ated (see Ehrlich 1996), corruption and per capita income level areexpected to be negatively correlated across different stages of eco-nomic development. The difference is that corruption depends oninvestment in political capital as a ticket for entry to the bureaucraticrank, unlike entry to many criminal activities, which requires littleskill. Such an investment has repercussions on the incentive of pro-ductive agents to invest in human capital. The relationship betweencorruption and the economy is thus explained as an endogenousoutcome of competition between growth-enhancing and socially un-productive investments and its reaction to exogenous factors, espe-cially government intervention in private economic activity.

The analysis also indicates that the relationship between govern-ment, corruption, and the economy’s growth is nonlinear. Govern-


ment intervention in private economic activity hurts most in thepoorest countries and those at a critical takeoff level. This may ex-plain the prevalence of corruption in countries trapped in poverty,such as Zaire and Haiti. It can also explain the unstable growth expe-rience in countries such as Bangladesh, India, and Mexico, wherea high degree of past government intervention has contributed tounstable growth experiences. The evidence presented in Section IIIFlends support to the hypothesis that large government size will havean adverse effect on the level of income at any stage of developmentand generally, but not necessarily, on the rate of growth of incomein all countries. The analysis and findings indicate that governmentsize may have little impact on the economy’s growth rate in the moredeveloped countries. Likewise, the impact of exogenous corruptionshocks is expected to be nonlinear, having an adverse effect on thelevel but not necessarily the rate of growth of per capita income inthe more advanced economies.

Perhaps the most intriguing result of the analysis concerns thepossibility that autocratic regimes, such as the command economiesof the former Soviet bloc, could in principle achieve a rate of growthequal to or higher than decentralized democracies, albeit not ahigher level of per capita income. These economies can be success-ful as long as an informed leadership is operating to maximize thelong-term growth potential of productive agents and constrains bu-reaucratic corruption to a degree commensurate with this objective.This explains why economically successful autocratic regimes oftenresort to forceful anticorruption campaigns and why corruption of-ten intensifies when the leadership loses its grip on power. At thesame time, the analysis also anticipates the potential failure of auto-cratic regimes because of the leadership’s inability to focus on long-term goals, its susceptibility to ideologically induced policy errors,and the general deadweight costs associated with protecting an auto-cratic regime by brute force.

References

Barro, Robert J. ‘‘Government Spending in a Simple Model of EndogenousGrowth.’’ J.P.E. 98, no. 5, pt. 2 (October 1990): S103–S125.

Becker, Gary S. ‘‘A Theory of Competition among Pressure Groups for Po-litical Influence.’’ Q.J.E. 98 (August 1983): 371–400.

Becker, Gary S.; Murphy, Kevin M.; and Tamura, Robert. ‘‘Human Capital,Fertility, and Economic Growth.’’ J.P.E. 98, no. 5, pt. 2 (October 1990):S12–S37.

Ehrlich, Isaac. ‘‘Crime, Punishment, and the Market for Offenses.’’ J. Econ.Perspectives 10 (Winter 1996): 43–67.

bureaucratic corruption S293Ehrlich, Isaac, and Lui, Francis T. ‘‘Intergenerational Trade, Longevity, and

Economic Growth.’’ J.P.E. 99 (October 1991): 1029–59.Krueger, Anne O. ‘‘The Political Economy of the Rent-Seeking Society.’’

A.E.R. 64 ( June 1974): 291–303.Leff, Nathaniel H. ‘‘Economic Development through Bureaucratic Corrup-

tion.’’ American Behavioral Scientist 8 (November 1964): 8–14.Lucas, Robert E., Jr. ‘‘On the Mechanics of Economic Development.’’ J.

Monetary Econ. 22 ( July 1988): 3–42.Lui, Francis T. ‘‘An Equilibrium Queuing Model of Bribery.’’ J.P.E. 93 (Au-

gust 1985): 760–81.Mauro, Paolo. ‘‘Corruption and Growth.’’ Q.J.E. 110 (August 1995): 681–

712.Murphy, Kevin M.; Shleifer, Andrei; and Vishny, Robert W. ‘‘The Allocation

of Talent: Implications for Growth.’’ Q.J.E. 106 (May 1991): 503–30.———. ‘‘Why Is Rent Seeking So Costly to Growth?’’ A.E.R. Papers and Proc.

83 (May 1993): 409–14.Shleifer, Andrei, and Vishny, Robert W. ‘‘Corruption.’’ Q.J.E. 108 (August

1993): 599–618.Summers, Robert, and Heston, Alan. ‘‘The Penn World Table (Mark 5):

An Expanded Set of International Comparisons, 1950–1988.’’ Q.J.E. 106(May 1991): 327–68.

———. ‘‘Penn World Tables, Mark 5.6.’’ Cambridge, Mass.: NBER, 1995.www.nber.org.

bureaucratic corruption and endogenous economic...

Documents