Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/emergencepersistOOacem HB31 .M415 i.** Massachusetts Institute of Technology Department of Economics Working Paper Series EMERGENCE AND PERSISTENCE OF INEFFICIENT STATES Daron Acemoglu Davide Ticchi Andrea Vindigni Working Paper 06-32 November 27, 2006 Room E52-251 50 Memorial Drive Cambridge, MA 02142 This paper can be downloaded without charge from the Social Science Research Network Paper Collection http://ssrn.com/abstract=948580 at MASSACHUSETTS INSTITUTE OF TECHNOLOGY JAM 8 2007 LIBRARIES Emergence and Persistence of Daron Acemoglu MIT Inefficient States* Andrea Vindigni Davide Ticchi University of Urbino Princeton November 2006 Abstract Inefficiencies in the bureaucratic organization of the state are often factors in retarding economic development. inefficient organizations has received very Why little certain societies choose or end attention, however. In this paper, simple theory of the emergence and persistence of inefficient states. and poor which We individuals. will show The rich are initially in power, choose redistributive policies. that, may be viewed as important The up with such we present a society consists of rich but expect to transition to democracy, Taxation requires the employment of bureaucrats. under certain circumstances, by choosing an inefficient state structure, the and capture democratic politics. This enables them to reduce the amount of redistribution and public good provision in democracy. Moreover, the inefficient state creates its own constituency and tends to persist over time. Intuitively, an inefficient state structure creates more rents for bureaucrats than would an efficient state structure. When the poor come to power in democracy, they will reform the structure of the state to make it more efficient so that higher taxes can be collected at lower cost and with rich able to use patronage lower rents for bureaucrats. Anticipating this, when the society starts out with an inefficient organization of the state, bureaucrats support the rich, We show that who set lower taxes but also provide enough political support, the coalition of the rich and the bureaucrats may not only choose an inefficient organization of the state, but they may further expand the size of bureaucracy so as to gain additional votes. The model shows that an equilibrium with an inefficient state is more likely to arise when there is greater inequality between the rich and the poor, when bureaucratic rents take intermediate values and when individuals are sufficiently forward-looking. rents to bureaucrats. Keywords: in order to generate bureaucracy, corruption, democracy, patronage politics, political economy, public goods, redistributive politics. JEL Classification: P16, Hll, H26, H41. *We thank Miriam Golden, Guide- Tabellini and seminar participants at IGIER-Bocconi, the Latin American Meetings of the Econometric Society, and Yale for useful comments and suggestions. Acemoglu gratefully acknowledges financial support from the National Science Foundation and the hospitality of the Yale Economics Department. Vindigni gratefully acknowledges the hospitality of Collegio Carlo Alberto and of the Yale Leitner Program in International Political Economy. Introduction 1 There are large cross-country differences efficiency of the state organization (e.g., back at least to Tilly (1990), in the extent of bureaucratic corruption World Bank, 2004). An influential argument, dating maintains that differences in "state capacity" are an important determinant of economic development. 1 The evidence that many less-developed economies in sub-Saharan Africa, Asia and Latin America only have a small fraction of their in tax revenue and the and invested by the government (e.g., GDP raised Acemoglu, 2005) and the correlation between measures of state capacity and economic growth (e.g., Rauch and Evans, 2000) are also consistent with this view. Societies with limited state capacity are also those that invest goods and do not adopt relatively little in public poor. 2 Brazil provides a typical example of a society, where the state sector has been relatively inefficient poor policies that redistribute resources to the (e.g., and democratic politics has generated only limited public goods and benefits for the Gay, 1990, Evans, 1992, Weyland 1996, Roett, 1999). we construct a In this paper, political economy model, which links the emergence and persistence of inefficient states to the strategic use of patronage politics by the elite as a of capturing democratic politics. of public goods Democratic capture enables the and redistribution, but therefore provides a unified answer both to the question of some why societies and why many democracies pursue certain democracies inefficient policies. may why the provision elite to limit at the cost of aggregate inefficiencies. means Our approach inefficient states emerge in relatively pro-elite policies. It also suggests exhibit relatively poor economic performance and adopt various 3 Our model economy consists of two groups, the rich elite and poor citizens. Linear taxes can be imposed on both groups, with the proceeds used to finance public good investments. The rich are generally opposed to high levels of taxes and public good investments. Tax collection requires that the state employs bureaucrats to prevent individuals from evading taxes, but bureaucrats themselves also need to be given incentives so that they exert effort (or do not accept bribes). The efficiency with measure of the organization of the which a central authority can monitor the bureaucrats state. Political competition is modeled either is our by assuming the existence of two parties, respectively aligned with the rich and the poor, or by allowing free 'See, for example, Evans (1989, 1995), Levi (1989), Migdal (1988), Epstein (2000), Herbst (2000), Centeno (2002) and Kohli (2004). See, for example, Etzioni-Halevy (1983) on the importance of state capacity and bureaucratization for the development of the welfare state in the West, and Rothstein and Uslaner (2005) on the importance of state capacity for income redistribution. On the comparative post-war growth performance of democracies, see, for example, Barro (1999). entry into the political arena by citizen candidates (Osborne and Slivinski, 1996, Besley and Coate, 1997). In both cases, there no commitment to is policies before elections and the party that comes to power chooses the policy vector, including taxes, public good provision, and bureaucratic wages, and whether to reform the efficiency of the state institutions. Democratic competition political rich or interesting by the fact that bureaucrats poor parties (candidates) and their support We in made is may be pivotal in the consider two possible organizations of the state: the which bureaucrats "inefficient" inefficient, one in will be detected easily they if which monitoring bureaucrats fail may first is an support either the outcome of elections. "efficient" organization, to exert effort, while the second is difficult. In equilibrium, when the is an state is bureaucrats need to be paid rents in order to induce them to perform their roles of tax collection and inspection. The presence of rents creates the possibility of patronage politics, whereby bureaucrats may support the party that In a society that is will always dominated by the rich maintain the elite or that is inefficient structure. permanently in 4 democracy (with a poor citizen as the median voter), the political process produces an efficient organization and no benefits of bureaucracy, since an inefficient state creates additional costs holding power. Our main the control of the rich beneficial to choose an result elite) and that is is when the power in democracy, there will Therefore, if the rich elite, the current bureaucrats when for the rich elite to capture the find it poor median voter comes be bureaucratic reform, reducing their rents from then on. in power, choose an inefficient organization of the state, —who are receiving rents—prefer to support the with the poor. Consequently, an may the state so as to exploit patronage politics to limit redistribution. In particular, bureaucrats realize that once the to those society starts out as nondemocratic (under expected to transition to democracy, the rich inefficient organization of for inefficient state organization rich rather emerges as a political than vote instrument democratic decision-making process by fostering a coalition between themselves and the bureaucrats. It is also only emerges in equilibrium, but also persists; noteworthy that the when the state is inefficient state inefficient, not the bureaucrats vote for the party of the rich, which chooses not to reform the bureaucracy and continues to maintain the support of the existing bureaucrats and thus Our its political power. analysis shows that patronage politics typically leads not only to the emergence and persistence of an inefficient state apparatus, but also to the overemployment of bureaucrats. 4 In our basic model, the assumption that the main role of bureaucrats is tax inspection is not essential. The important feature is that an inefficient state organization must pay bureaucrats rents in order to provide them with the right incentives. Bureaucrats' role as tax inspectors becomes important in the extension in subsection 5.3, where they can be bribed by producers evading bureaucrats' main role is tax inspection throughout. taxes. We simplify the presentation by assuming that This is because the rich may prefer to hire additional (unnecessary) bureaucrats so as to boost Consequently, a captured democracy will typically feature an inefficient their party's votes. state (bureaucracy), provide relatively few public goods This pattern of bureaucratic inefficiency bureaucrats. corrupt and low-capacity bureaucracies in and Rauch and Evans, 2000). We also many racy and makes is democratic outcomes. An developed countries Geddes, 1991, (e.g., more inefficient state is likely when because greater inequality raises the equilibrium tax rate more appealing it consistent with the stylized view of 5 show that the equilibrium with an greater inequality. This less is and employ an excessive number of for the rich to create an inefficient state also requires inefficient state in there is democ- apparatus to prevent intermediate levels of rents/ "efficiency wages" for bureaucrats; when rents are limited, bureaucrats would not support the rich, while too high rents would make the inefficient state equilibrium prohibitively costly for the rich elite. Finally, an inefficient state is because bureaucrats support the The number rest of the paper is more likely to arise inefficient state in organized as follows. of case studies that illustrate when agents are order to obtain future rents. Section 2 provides a brief discussion of a how patronage politics has been used to limit redis- tribution towards the poor and also discusses the related literature. basic economic and political more forward-looking, Section 3 outlines the environment. Section 4 characterizes the equilibria of the baseline model both under permanent nondemocratic and democratic regimes as benchmarks, and more importantly, under a regime that starts out as nondemocratic and becomes democratic thereafter. We show that state organization and a sufficiently large Section 5 generalizes this framework in a may choose an inefficient in order to create a majority coalition. environment the rich in this last political bureaucracy number elite of directions; in particular, it allows for more general contracts with bureaucrats, considers a citizen-candidate setup for political competition, and allows producers to bribe bureaucrats to evade distinctive implication of our approach sector employees taxes. Section 6 briefly investigates a about the relationship between relative wages of public and the amount of public good provision in democracies. We report cross- country correlations consistent with this implication. Section 7 concludes, while the Appendix contains 5 some of the proofs omitted Even with the overemployment majority in the electorate. methods such as lobbying from the text. of bureaucrats, bureaucrats In practice, the elites may be and the rich elite may not have an absolute able to control the political system using other in addition to the support of the bureaucrats. Here, we isolate our main mechanism by focusing on a baseline model where the rich are able to capture democracy without any lobbying or other non-electoral activities. Motivation and Related Literature 2 In this section, relate our we paper to the existing literature 2.1 Patronage The experiences of patronage number briefly discuss a in political Politics, Inefficient States many societies in Latin and politics, inefficient states Several authors distribution of large (e.g. economy. and Elite Control elite control. Here we inefficient briefly mention three cases. and oversized bureaucracy comes from Gay, 1990, Weyland 1996, Roett, 1999) have argued that the numbers of public jobs, both in the public administration tatal organizations, has created a pattern of patronage politics in Brazil. 6 and The in paras- control over these jobs has enabled traditional elites to preserve their political power and limit the of public good provision and Brazil, elites amount redistribution. have been able to control of repression of redistribution and democratic Patronage and also America, Asia and Africa illustrate the link between Perhaps the most transparent example of Brazil. and of case studies that motivate our analysis amount In fact, despite the high level of inequality in politics for much of the 20 t/l relatively short periods of military rule. and public good provision does not show marked century with only limited Interestingly, the differences amount between military periods. politics has often ensured that even those in poorest neighborhoods of Rio have supported the traditional parties rather than socialist or social democratic parties running on platforms of greater public good provision and redistribution (Gay, 1990). Students of Brazilian politics have noted the role of public sector employees in this process. For example, Roett (1999 p. 91) writes "state company employees emerged supporter of the patrimonial order" pressures from the IMF and as being among the strongest In return, successive governments have withstood external . have not reformed the public sector, despite the "public perception that public-sector workers were overpaid and underworked" (Roett, p. 97). The process of reforming the public sector has started only recently and progressed slowly. Another example of cialiste effective patronage politics is provided by the policies of Parti So- (PS) of President Leopold Sedar Senghor in Senegal. After independence, PS faced increasing challenges from various different opposition groups, including urban workers and farmers. Nevertheless, it managed to preserve its power, with relatively limited amount of some branch of the Brazilian public sector. Evans commonly recognized as a huge cabide de emprego (source of jobs) and In the early 1980s, about 4 million people had a job in (1992) observes that the Brazilian state is remarks that in contrast to the Weberian conception, recruitment of public employees to merit but to political connections. in Brazil is not related repression, largely owing to amount PS of political liberalization also exploited its In fact, Senghor use of patronage politics. its promoted some and allowed the creation of a multi-party system. However, incumbency advantage to manipulate the democratic process an extensive patronage network centered on the state apparatus and parastatal estingly, it was precisely during of sector. Inter- this period of democratization that the size of the public sector grew substantially (Boone 1990, Beck, 1997). Boone (1990 "The strategic allocation by creating government jobs coopted 347) describes this process p. restive intellectuals as: and professionals and incorporated them into political factions anchored in the state bureaucracy." Thanks to this successful implementation of patronage politics, PS much retained of its power following the transition to democracy. A from final Italy. that the A tries. example of the The rise of patronage politics in the face of political competition comes evolution of the Italian bureaucracy in the post- WWII decades demonstrates mechanism that our model identifies significant extension in the Italian Democratic Party (DC) may operate even in relatively developed coun- bureaucracy was initiated by the Italian Christian in the 1950s after the electoral challenge from the Communist Party increased sharply, especially following the 1953 political elections. Until the defeat in War II, as the Italian politics dominant World DC emerged DC created a was dominated by Mussolini's dictatorship. After the war, party. In the 1950s, faced by electoral challenge from the left, highly disorganized and oversized bureaucracy, which subsequently became a natural source of political support and patronage for the party. the expansion of the bureaucracy "The massive system after 1953 was Golden (2003 p. 199) describes the motive for as: of political patronage that the leaders of the their aggregate answer ... DC constructed to enlarging the party's aggregate vote share while protecting the incumbency advantage of individual legislators." In part with the support of the bureaucracy that it created, DCs dominance of Italian politics continued until the 1980s and prevented the formation of a left-wing government. 2.2 Related Literature Our paper is related to a number of different literatures. The first is the political science and sociology literature on the organization of the state and the bureaucracy mentioned above. In contrast, there is relatively little reaucracy in economics. Some work on the internal organization of the state and bu- exceptions include Acemoglu and Verdier (1998), Dixit (2002), Egorov and Sonin (2005) and Debs (2006). None of these papers investigate the relationship between patronage politics and the emergence of the inefficient state as a method of limiting redistribution. Our paper why First, the reason to control political models of elites number also related to a is the power elite in a both As on choose inefficient institutions is such, our paper Baron Grossman and Helpman (1994), Dixit, in which traditional (1997), elites are able to small literature on the inefficiency of the form of redistribution is Becker and Mulligan (2003) and Wilson (1990) argue that to our work. of redistribution are chosen as a way is and Grossman capture democratic of future redistribution. is Nevertheless, there basic Becker-Mulligan- Wilson story, it is bureaucracy is inefficient methods of redistribution (see also Coate not clear chosen by the rich in order to limit the is why also an important distinction; in the the society can commit to the form of and not to the amount of redistribution. In contrast, redistribution also closely related a close connection between this idea and the main our paper, whereby an inefficient state in inefficient amount of limiting the and Morris, 1995, Rodrik, 1995). There mechanism related to other is Acemoglu and Robinson (2006b). politics, e.g., an later democratic regime. and Helpman (1996), and models amount and initially manipulating policies in democratic settings, including lobbying models, such as Austen-Smith (1987), The of other strands of the literature in political economy. in our model the choice of a way of affecting the future political equilibrium so as to bring the party aligned with the interests of the rich to political power, and via this channel, to limit the provision of public goods and taxation. As such, our mechanism is also related to the rationale for inefficient redistribution suggested in Saint-Paul (1996) and Acemoglu and Robinson (2001), where a politically powerful group may push for inefficient forms of transfers in order to maintain its future political power. There is also a small literature on how politicians may distort sons. Papers in this literature include employment) are chosen Shleifer models where policies for strategic rea- inefficient policies (such as excessive state in order to gain votes (e.g., Fiorina and Noll, 1978, Geddes, 1991, and Vishny, 1994, Lizzeri and Persico, 2001, Robinson and Torvik, 2005). papers suggest that inefficient choices (including wasteful investments, large and systems) are inefficient fiscal made Still budget other deficits, in order to constrain future politicians (e.g., Glazer, and Svensson, 1989, Tabellini and Alesina, 1990, Aghion and Bolton, 1990, 1989, Persson Cukierman, Edwards, and Tabellini, 1992). None of these papers feature the mechanism of an elite creating an inefficient state structure to maintain their political power in the face of an emerging democracy. Our model is also related to the literature on comparative politics and public finance (e.g., Persson and Tabellini, 2003, Ticchi and Vindigni, 2005), which investigates sources of differences in fiscal policies among democracies. Our approach suggests an alternative, but complementary, source of variation, related to the desire and the ability of the economic elite to dominate democratic politics, which can generate both differences in the level of public goods provision and Finally, our in the efficiency of the state. approach is related to sociological analyses of "cooptation" in and existing elites in the Marxist sociology political science literatures. In particular, (1980 pp. 228-234) argues that the control of the state apparatus economic elites in hegemony a mechanism through which the in a crucial objective of the Gramsci, 1971). However, this literature neither articulates elite may accomplish such a strategy relative to other options. Basic is Therborn democracy, which they achieve using strategies including cooptation (see also the discussion of 3 democracy by these objectives nor models the costs of 7 Model Description of the Economic Environment 3.1 Consider the following discrete time infinite-horizon economy populated by a continuum 1 of agents, each of which has the following risk-neutral preferences oo BqJ2 P* (4 + Gt-hef), t=o at time Cj > t = 0, where Eo is the expectations at time = 0, all agents, e?t € necessary in some occupations), and h There are two types of agents: n > denote poor agents by the symbol by H, and also use C and {0, 1} is (0, 1) is > is the cost of Ti to denote the set of his skill; AL for is the level of 1 —n are rich (high-skill). (corresponding to low-productivity), and rich agents K of investment in infrastructure, income depending on > effort. 1/2 are poor (low-skill), while L Gt j), the discount factor, the effort decision of the agent (which will be poor and rich agents. There are two occupations: producer and bureaucrat. amount @ 6 denotes the consumption of the agent in question (agent public good enjoyed by We t > 0, is In each period, as long as some undertaken, each producer generates an poor agents and AH > AL for rich agents. If the Another major difference between the Marxist approaches and ours is that in our model bureaucrats can most Marxist approaches, the state apparatus is, ultimately, controlled by the economic elite (e.g., Miliband, 1969, Poulantzas, 1978, Therborn, 1980). In this respect, the notion of bureaucracy and state apparatus in our model is also different from that of Max Weber, which views bureaucracy as an "apolitical" organization, with no goals or interests. See also Alford and Friedland (1985) for a critical discussions of Marxist and non-Marxist theories of the state. side either with rich or poor agents, whereas in investment in infrastructure K is not undertaken at time that period. Producers receive and A set of agents The linear tax rate r t K, is their income net of taxes. denoted by Xt are bureaucrats at time but receive a net wage of wt wage income). consume then no agent can produce within t, > from the government role of bureaucrats 6 [0, 1] is they do not pay taxes on their (i.e., tax collection. on earned incomes These agents do not produce, t. we In particular, same irrespective of whether the individual is rich or poor. To it will This tax rate simplify the discussion, we assume that only poor agents can become bureaucrats. This assumption for the results, since investment in order to finance the infrastructure additional spending on the public good Gt and the wages of bureaucrats. the a will allow for is not necessary be evident below that low-productivity poor agents always prefer bureaucracy more than the high-productivity rich agents (see Remark 2 below). Both and poor agents can try to evade rich taxes. We assume that to evade taxes, he gets caught with probability p(xt), where p : [0, 1] twice continuously differentiable, and strictly concave function with number the Xt = Lfx of bureaucrats exerting positive effort at time e tdj- More t. — p if an individual [0, 1] is (0) an increasing, =0, and x denotes t formally, this This expression incorporates the fact that bureaucrats tries defined as is who do not exert effort are not useful. 8 If an individual simplicity, this is caught evading taxes, is we assume that all of his income during that period not an important assumption). We also assume that there is full not observed. is anonymity in the This implies that future punishments on tax evaders are not possible. Moreover, because of limited > more 0, liability, i.e., serious punishments are not possible. Since effort is costly, their effort decision. probability For income does not accrue to the government either (though this market, so that the past history of individual producers Cj is lost. qt We at time t. bureaucrats will exert effort only assume that If if they do not exert if their effort, compensation depends on bureaucrats are caught with they are not caught, they receive the wage wt, and if they are caught shirking, they lose their wage, but are not fired from the bureaucracy. This assumption simplifies the algebra The and the exposition considerably and probability of detection qt = 1) = 1, so that with an relaxed in subsection 5.1 below. depends on the quality of the organization of the state ("efficiency of the state"). In particular, q{I is efficient we allow two levels of efficiency, It € {0, 1}, such that organization of the state any shirking bureaucrat is Alternatively, instead of inducing bureaucrats to exert effort, it may be important to ensure that they do not accept bribes from the individuals supposed to pay taxes (e.g., Acemoglu and Verdier, 1998, 2000). We investigate a variant of our model with corruption in subsection 5.3. immediately caught, while q (I = 0) = qo < I, so that with an inefficient organization shirking 1=1 bureaucrats are not necessarily detected. To simplify the analysis we assume that no cost relative to / At each = 0. has 9 date, the political system chooses the following policies: • A • The wage • A • The number • A tax rate on all earned income r t 6 € R+. rate for bureaucrats wt level of public [0, 1]. good Gt 6 K+. of bureaucrats hired, Xt € [0, 1]. decision on the organization of the state for the next date, It G {0, 1} of the state at the current date, It-i, is —the efficiency part of the state variable, determined by choices in the previous period. The additional restrictions on these policies are as follows: 1. The government budget 2. If Xt > Xt-i, then decide to quit if constraint (specified below) has to be satisfied at every date. existing bureaucrats cannot be fired (although each bureaucrat can he finds this beneficial). Moreover, are hired and a fraction (Xt — X -\) /X randomly chosen irrespective of We t Xt < Xt-i, then no new bureaucrats of the bureaucrats is fired (those fired = (rt,wt, Gt, being their past history). denote a vector of policies satisfying these restrictions by p t X t , It) 6 71. Description of the Political System 3.2 We t if will consider three different political 1. Permanent nondemocracy: the the rich can vote, and since all environments: rich elite are in rich agents have the available set of policies, the policy vector will power at all dates, same policy preferences over the most preferred by a representative when an We rich elite be implemented. In general, one can imagine that setting up a more efficient state apparatus ditures. meaning that only may involve additional expen- ignore those both to simplify the algebra and also to highlight that inefficient states can arise even efficient organization is costlessly available. Permanent Democracy: the 2. starting at = t regime The paper. way a simple To start with, for at The comparison. t = 0, and third one some of capturing the idea that democracy anticipating that elites, power rich elite are in two environments are It is the majority, are in power at dates all in all future dates, the be democratic with majoritarian elections. will first who form (or at all dates there are elections as described below). Emerging Democracy: the 3. citizens, some point will arrive at we model the democratic system is our main focus in this decisions are originally taken by —in this case right at date = t l. 10 a very simple way, by assuming that there in are two parties, one run by an elite agent and one run by a poor agent, and that bureaucrats cannot run We for office. denotes that party P is P use the symbols elected to office at date to the policies they will implement once they R and to denote these parties Parties are unable to t. come to power. and dt = P make commitments Thus whichever party receives the majority of the votes comes to power and the agent in control of the party chooses the own policy vector that maximizes his Downsian models utility. of political competition the election. Instead, it is This last assumption departs from the standard where parties commit to closer to the literature on citizen-candidate models, which discussed further in subsection 5.2 (see also Alesina, 1988). we will consider candidate, and it is be Specifically, in subsection 5.2, we will show that the same results apply with this richer setup. Nevertheless, two parties to highlight the main forces. Timing of Events 3.3 To will a richer model of democratic politics, where each agent can run as a citizen- useful to start with the simpler environment with only economic their policy platform before recap, the timing of events within each date • The {N, society starts with D} , a set Xt-i some C C the set of bureaucrats political regime, of agents X -\ t efficiency of the state, It-\ € is: who are already bureaucrats (since, must be a subset {0, 1}. nondemocracy or democracy, i.e., st £ by assumption, of the set of poor agents), and a level of Then: 10 In this case, the society is nondemocratic at date t = 0, and we assume that it will become democratic for exogenous reasons at date t = 1. It is possible to model democratization as equilibrium institutional change along the lines of the models of endogenous democratization in the literature (see Acemoglu and Robinson, 2006a, for a discussion and references), but doing so would complicate the analysis without generating additional economic insights in the current context. 10 1. 2. individuals j G In democracy, all j decides vj G {P,R}. In democracy, the elected party or in the policy vector p t 3. Observing = , agent decides G K. is denoted by the same symbol without any risk of confusion). ~ f° r an * ne " cn agents. The number of bureaucrats at J Xtdj \ i-e., the minimum of the number of bureaucrats chosen power and the number of people applying to or remaining in bureaucracy. then min in elite individual and each individual j G Xt-\ decides whether to quit bureau- Naturally, by assumption, \l by the polity i.e., each individual j £ Xt~\ decides whether to apply to become a this vector, cracy, Xf G {0, 1} (which t is or party R, nondemocracy the representative (T t ,w t ,Gt,X t ,It ) bureaucrat, Xt G {0, 1} time P vote for either party [0, 1] < Xt, , This also determines the current set of bureaucrats, Xt. 4. Each bureaucrat decides whether f-p X e{dj, 5. to exert effort, e{ and thus the probability € {0,1}, which determines x t = of detection of individuals evading taxes. Production takes place and each producer decides whether to evade taxes or not, denoted by 4 G {0,1}. 6. A fraction 7. A fraction q t 8. Taxes are collected, remaining bureaucrats are paid their wage, wt, and the public good G t is p (x ( ) of producers evading taxes are caught. = q {It-lj of shirking bureaucrats are caught supplied. Naturally, the society starts with bureaucrats. We also X-\ suppose that I-\ except the actions at time 4 and punished. t = 0, = = 0, i.e., in the initial date there are no incumbent (though this has no bearing on any of our results since the choice of It G {0, 1} is without any costs). Characterization of Equilibria We now characterize the equilibrium of the environments described above. Definition of Equilibrium 4.1 Throughout, we focus on pure strategy Markov Perfect Equilibria (MPE). 11 We focus on MPE both because bureaucrats more MPE is unique and is relatively straightforward MPE makes the emergence of a coalition between the rich and the in the current context the to characterize and also because the focus on difficult (since there cannot be "commitment" to future rents 11 for bureaucrats). Recall that Markovian strategies condition only on the payoff-relevant state variables (and An MPE on the prior actions within the same stage game). defined as a set of Markovian is strategies that are best responses to each other given every history. In the current S = aggregate state vector can be represented as time political regime at t, I -\ G {0, 1} t previous period, and Xj_i is (s t t , game, the It-\,Xt-\) G S, where s t G {N, D} is the the efficiency of the bureaucracy inherited from the is the size of the bureaucracy inherited from the previous period. 12 Individual actions will be a function of the aggregate state vector St and the individual's identity, in particular, representing whether the individual Thus rich producer or a bureaucrat. which party to vote B} a t G {L, H, G {P,R}, whether to apply for, i.e., v\ Xt G {0, 1}, whether to evade taxes, z\ G {0, 1}, 6 to exert effort, e\ the choice of It G {0, 1}, Tt a An MPE We is : G Xt G [0, 1], strategies can S x {L, H, a mapping a* that B) is a slight abuse of notation, we each individual will decide remain) in bureaucracy, is a producer, and whether Finally, strategies also include M+ when the individual the party is R} x {0, l} 3 x x [0, 1] x R+. [0, n] best response to itself at every possible history. will use v (I H, B} when risk of confusion, no and Gt G a poor producer, be represented by the following mapping -» {P, of group a G {L, is the individual if subcomponents of a rather than the will often refer to there n], [0, at, (or to {0, 1}, if the individual is a bureaucrat. Thus Markovian leader. and as a function of St is \ a) to entire strategy profile, denote the voting strategy of an individual as a function of the efficiency of the state institutions. we will and with Moreover, use the index j to denote individuals or groups interchangeably. 4.2 Preliminary Results We now state a Lemma number 1 If p(xt) <r of results that will be useful t , then z{ ~ for all j £ X t , throughout the analysis. all i-e., Proof. Write the payoff of an individual producer j ^ and the size of (effort-exerting) V} bureaucracy = max { (1 - Tt ) A*, is xt X t producers evade taxes at time at time t when the tax rate is t. rt as (1-p (x t )) A* } + G + /3V->+1 t (a*) , In addition, for each individual we could specify whether the individual is currently a bureaucrat, i.e., whether j 6 Xt -i and whether he is a party leader as part of the individual-specific state vector. Nevertheless, Markovian strategies can be defined without doing this, which simplifies the notation. 12 A3 where is is the productivity of this individual, and a* is the optimal policy, so that The his income first, (1 — Tt) The in taxes. A3 , is what the individual second, (1 — p(x A3 t )) is , his probability (1 —p(xt)), but Limited liability implies c\ PVt+i ( a *) max term §i ven is caught and > home loses all his current and the current behavior has no if he pays a fraction r t of his full productivity A3 with income with probability p(xj). effect on the continuation value the anonymity assumption. This expression immediately establishes that the will pick tax evasion, Since with tax evasion there we need (a*) expected consumption when evading In particular, in this case, the individual takes taxes. consume will t The max incorporates two the discounted optimal continuation value for the individual. terms. 0V 3+1 i.e., is z\ — < iip(xt) 0, Tt, as claimed in the lemma. Lemma no government revenue, implies that in equilibrium 1 to have the following incentive compatibility constraint for producers p{x > t ) n to be satisfied. Alternatively, defining TTtosp-'V), (1) producers' incentive compatibility constraint can be expressed xt > It and is (2) sufficient to ensure that all individuals can be easily verified that since p(-) strictly concave, tt () defined in (1) is is 13 (tj) tt This condition requires the number of bureaucrats exerting This constraint as: effort to be greater than 7r(rt). choose not to evade taxes. strictly increasing, continuously differentiable strictly increasing, continuously differentiable and strictly convex. Lemma 2 // w < -, (3) t It then e\ = for all j G Xt, i.e., Proof. Write the payoff detection probability is bureaucrats will shirk at time of a bureaucrat j G X qt ) w at time t t t. when the wage rate is wt and the q t as V} 13 all - max {w -h,(lt t } + G + PV j+1 t t (a*) , This condition can also be interpreted as a "state capacity constraint" since, given the effective it determines the maximum tax rate. bureaucracy, 13 size of the where a* is the optimal policy, so that for the individual. effort The max operator and receiving the wage PVf+1 (a*) is incorporates two terms representing the payoff to exerting wt for sure, — h, and the payoff to shirking. Since, by assumption, bureaucrats cannot be fired for shirking and limited to shirking that we have Remark — is (1 qt) = e\ Whenever wt < Wt- t caught shirking. Given it is this, h/qt, the as claimed in the 1 In subsection 5.1 below, bureaucracy. In this case, is X £ for all j the discounted optimal continuation value we will allow liability max for > 0, the payoff lemma. bureaucrats caught shirking to be fired from clear that the optimal contract involves firing a bureaucrat the condition in Lemma 2 will have to be forward-looking, imagine a stationary equilibrium, where today and in wage rate c?t operator will pick the second term, so into account the future rents that the bureaucrat will lose to f, the makes sure that bureaucrats is if he taking caught shirking. In particular, future periods the tax rate all if w, and the probability of getting caught is equal then the is q, necessary condition (3) would become w-h (l-f)A L ^ 1-/3 since the left-hand side 1-/3 . „, v y/ V what the individual would is whereas the right-hand side is ^w-h -1-PJ / ' by exerting receive effort at every date, the payoff to deviating for one period, and then switching to exerting effort from then on (implicitly using the one-step ahead deviation principle, see Fudenberg and Chapter Tirole, 1991, getting caught with probability q, 4). In particular, the right-hand side has the individual receiving nothing today from then on, and not getting caught with probability 1 — q, and then receives the discounted version of the left-hand effort). A bureaucrat who and the wage of a in which case he receives side (as he switches loses his job always receives the producer low-skill wage of a w today back to exerting low-skill producer from then on, since along the equilibrium path, there will be no further hiring into bureaucracy. Rearranging terms, the above inequality can be expressed ,< fl(1 .^ + kM. In a stationary equilibrium where bureaucrats are fired replace (3), and when it is compatibility constraint, satisfied, all (5), below as: when caught (4) shirking, condition (4) will bureaucrats will shirk. Correspondingly, the incentive will change to the converse of this condition. We return to a further analysis of this case in subsection 5.1. If bureaucrats are expected to shirk, all individuals will evade taxes and there will be no tax revenues. Consequently, the infrastructure investment 14 K could not be financed and there would be no production. Thus, the society also needs to satisfy the incentive compatibility constraint of the bureaucrats given by w > -, (5) t Qt where qt — q(It-i). This constraint choose to exert is, effort. is necessary and sufficient to ensure that In addition, (poor) individuals must prefer to bureaucrats all become bureaucrats. That the participation constraint w >(l-T )A L + t t h, (6) needs to be satisfied so that bureaucrats receive at least as much as they would obtain in private production. Remark 2 If rich corresponding to agents could become bureaucrats, the equivalent participation constraint, agents would be (6), for rich w > (l-T )A H + t Comparison more of this inequality with condition (6) this inequality The above lemma t, poor agents are always clear that Our assumption that rich agents cannot therefore enables us to avoid imposing explicit conditions to ensure that not satisfied and (6) is it discussion, in particular is. Lemmas 1 and 2, immediately establishes the following (proof omitted) Lemma all makes willing to enter bureaucracy than rich agents. become bureaucrats h. t and 3 In any MPE, = all z\ l for j = conditions (2), (5) and (6) must hold and e\ £ X t and 1 for all j 6 Xt and all t. In other words, in any equilibrium the incentive compatibility constraints of producers and bureaucrats and the participation constraint evades taxes and Prom Lemma all bureaucrats exert of bureaucrats are satisfied, and no producer effort. 3 (and the fact that only poor agents follows that, as long as the constraints (2) and become bureaucrats), (5) are satisfied, it immediately the government budget con- straint can be written as: K+G where the left-hand ture, side is t +w t X t < (1 - n) r t A H+ (n - X t) rt A L government expenditures, consisting of the investment spending on public goods and bureaucrats' wages, while the right-hand side 15 (7) , in infrastrucis government tax receipts collected from rich and poor agents. This expression takes into account that bureaucrats exert effort and no producer evades taxes. Moreover, (7) highlights all that in our model, taxation reduces output through a particular general equilibrium mechanism; the gov- ernment can by hiring bureaucrats and bureaucrats themselves do not produce raise taxes only any output. lemma Finally, the following Lemma 4 Rich agents always always vote for party P, i.e., vote for party R, for all Equilibria under permanent democracy parison to our main political and then transitioning Proposition at each t > =P dt G = GD = t and t is (1 The there exists a unique h, = vh 1, d Ah + = (1 - t d ) AL + [n - 7T the unique solution to the max(l r,G - following results are straightforward: is n) t =^ R, and poor producers > Xt—i, v\ and the following policy vector - — 7i, v\ environment, which involves the society starting as nondemocratic to democracy. It G all j and permanent nondemocracy are of interest as a com- Under permanent democracy, 1 for i.e., C and j ^ & j stated without proof: is Permanent Democracy and Nondemocracy Equilibria under 4.3 immediate and is (t d )) r MPE. implemented X = ir t D AL - [(l (t In this equilibrium, at each t > 0: d ) , - rD ) A L + h]ir (r D - K, ) (8) maximization problem: -t)A l + G (9)' v subject to G = Proof. By (1-ti)tA h Lemma 4, for all + {n-ir(T)]TA L - j G C, v\ vote and form the majority starting at t decisive voter j' G C can be written = = [(1 - t) AL + h]ir {t) - K. P. Under permanent democracy, the poor can 0, thus dt — P for all t. Then the payoff to the as Vf ^(l-^At + Gt+pV^io*),' where again a* is the optimal policy and (3Vt\i value for this individual. The continuation (a-*) 3 is the discounted optimal continuation value /3Vt +1 (a*) 16 is unaffected by current policies, thus the optimal policy can be determined as a solution to the following program: (1-t)A l + G max (10) t,w,X,I,G subject to 7r(T) < X maxi ——— ,(1 -r)A L + h\ < w G < (1 < It is evident that chosen in first all It = 1 t > is slack (since this periods three constraints objective function), so 0. we have X= rium, the following policy vector and t is = 1, w = t -wX-K - X] tA l = 0j so will always be it would allow an increase in w = max {/i, (1 — r) periods where = (t) and r° is all (1 is It G, raising the value of the A L + h\ = (1 — r) A L + 1, i.e., for all t > 0. h. Strict uniquely defined. implemented - rN A L + ) there exists a unique at each t > X = n (t n h, t MPE. In this equilib- 0: ) , G - GN = t 0, the unique solution to the equation [(1 - r) AL + h] n (r) - (1 - re) tA h -'[re -tt Proof. Under permanent nondemocracy, the rich retain payoff to the representative rich individual Vl = where [re G. Proposition 2 Under permanent nondemocracy, It tA h + Moreover, there cannot be a solution in which any one of the then ensures that ir (•) re) relaxes the second constraint relative to it Substituting these equalities yields (9) for convexity of - cr* is (1 j' rA L + political K= power 0. forever. (11) Then the £ 7i can be written as - r) AH + Gt +.0V&1 (O, the optimal policy and /3V/+1 (u*) this individual. (r)] is the discounted optimal continuation value for Because the continuation value /3V^+1 (a*) 17 is unaffected by current policies, the optimal policy can be determined as a solution to the following program: (1- t)A h + max G (12) t,w,X,I,G subject to < 7t(t) max -— Al-T)A UW h + h) < w J G < (1— n) tA h + < It is again evident that be chosen. X= in it (t) It — Moreover, the and program 1 it - X] tA l - [n first wX - K G. relaxes the second constraint relative to w = max {/i, (1 — (12), X = It so will always 0, three constraints must again hold as equalities, so L r) yl + /i} = — r) G = 0, (1 follows immediately that L A + we have Substituting for these equalities /i. and the strict n convexity of () again ensures the uniqueness of the solution to (11). The main conclusion from both benchmark of these political politically decisive agents choose a policy vector consistent always involves an efficient organization of the state, to make the i.e., It environments with their own = 1 for all t > 0. state inefficient. Consequently, both consolidated democratic regimes involve 1 = 1. Moreover, in interests, There is and this no reason and nondemocratic both regimes the capacity of the state in the sense that constraint (2) holds as equality that the is is fully utilized and the minimum number of bureaucrats necessary to prevent tax evasion are employed. It is straightforward to see that the unique solution (r D ,G°) in (9) involves t d since infrastructure spending, > 0, has to be financed (and for the same reason, However, because raising further revenues involves the employment of in Proposition 2). bureaucrats which K > 0, N r > is costly, it is possible that the solution to (9) involves GD — 0. If this were the case, there would be no difference between the political bliss points of poor and rich agents given in Propositions we are more Condition It 1 and 2 and thus no interesting political conflict. Therefore, throughout interested in the case where the following condition 1 The solution can be verified that to (9) involves if GD > the gap between A is satisfied: 0. and AL is small and n' (t) condition will be violated. Therefore, this condition imposes that there inequality in society and raising taxes is is is large, this a certain degree of not excessively costly, so that the poor would a higher level of public good provision than the 18 rich. When Condition 1 is satisfied, it like also follows that r D > t n and the bureaucracy 4.4 more Lemma 5 In N Gt — G = 0, Proposition MPE, an and t n(r N ),G = G N = t = R 6 vector at all dates = 1, change be will elections. efficient state institutions, = R = and It — We i.e., where at date = 0, the lemmas. the rich will choose their political w = (l- t n ) A l + then 1, state," X = h, t = t — tw ) A l + (l meaning that once the party of the poor P, then dt t' > > =P for all t' > t, and we have it (t n ), X = h, t P is being elected ; elected, the results the following equilibrium policy t: t X = ir(T D ), h, (13) t h]7T(T D )-K, given by (9). is 1). Now in (13) is the optimal policy of the citizens in permanent democ- suppose that party P chooses the policy vector specified in the lemma. we start with It that v (I = 1 that v (I = 1 | | = 1 is in power at time Since this includes as part of the payoff-relevant state vector. B) = P. Then party B) ^ P, but X P wins the majority at time < n — 1/2. Then, party P t and suppose that t, It = + 1. it the following 1, Suppose that a* such is Alternatively suppose again wins the majority at time In both cases, repeating this argument for the next period shows that party P t + 1. keeps power at dates and establishes the lemma. To complete the proof we only need X = t start with a series of involves choosing the policy vector wt w = (l-T D )A L + racy (Proposition all the size of the solution to the equivalent of program (12) in the proof of 1, Proof. The policy vector period, n and ) permanent nondemocracy. = G D = (l-n)T D A H +[n-n{T D )]T D A L -[{l-T D )A L + t and r D (t it apply subsequently: Lemma G > ) establishes that the party representing the poor, party an "absorbing If dt d m 0. The next lemma 1 (t given by (11). is Proposition 2 for party of Proposition in n 2: dt if Proof. Given that It It strictly increasing, interesting case with regime shows that with bliss point as in is is power and from then on there first result to office (•) Equilibrium with Regime Change look at the rich are in Our n permanent democracy than larger in is Political We now since , > n— to rule out the case where v (I = 1 | B) = R and 1/2 (the proof to eliminate the case where bureaucrats randomize between the two 19 way parties in a to bring party is costly for the rich (recall It = Then from Lemma 1. R to program power t bureaucrat utility of the VB (R) is the same as a poor producer. = 1 The | B) =R V B (d), - t n ) A L + PVJ < (1-t d )A l + G d + will = choose fact that the last also the utility that Then denoting (a*) I3V3{cj*) (P), a bureaucrat term the maximal utility of a poor is will receive when party P is in power, cannot be a best response, completing the proof of the lemma. intuition for this result as follows. is Once the party choose their preferred policy vector, which includes will R and 1 we have (1 where the inequality follows from the v (I =R B) t = = VB is | X = r(T N ), G = G N = 0. the utility of a bureaucrat supporting party d by Since this 1 5, t agent. = (12) in the proof of Proposition 2), party w =(l-r N )A L + h, This implies that the Since v (I identical). is It of the poor wins = 1, an and given an election, they efficient state, bureaucrats will have no reason to support the rich party and the poor will continue to win future periods and the organization of the state will continue to be efficient. elections in all An organization of the state ensures that bureaucrats receive no rents and receive the efficient same payoff as poor producers. Thus they point of the poor will be implemented in it-i is = 1 — i-e., when the state all is efficient will also support party P, and the political future periods. This —the rich will not lemma also implies that bliss when be able to win a majority. This related to the basic idea of our approach: the rich can only convince bureaucrats to vote for their party by committing to giving them rents and organization of the state We is inefficient, i.e., — next investigate whether or not the rich vote for their party starting with It -\ of the rich, The It -\ next Lemma when lemma in power, will — 0. this can only be achieved when the 0. may be Since there is able to convince the bureaucrats to no commitment to choose policies in line with its characterizes these policies starting with It-\ 7 Suppose that It-\ = 0, then wt = h/qo. 20 Moreover, policies, the party (the rich agents') preferences. = 0. if dt — R, then Gt =G = 0, and dt if = P, then Gt = GD given by the solution to the following maximization program: max (1 K t,G AL + G r)' (14)' v subject to G = (l-n)TAH + [n-ir(T)]TAL -.—ir(T)-K. Qo Proof. That any party, when follows immediately Lemma from power and inheriting in The follows immediately choose will GD from the program as in (14), additional restriction that Remark involve rich = u>t imposing u>t and the poor is The next lemma provides necessary — Lemma MPE, d = R, 8 In an when as there, sufficiently high, the solution to the starting with It -\ K h/qo. To h/qo cannot be = GE = see that party P with the (10), m h/qo. With the same reasoning 0. = = io t choose Gt will go back to the maximization problem 3 As with the solution to the maximization problem GD = R fact that party in (12) after suffices to it choose 0, will investment in infrastructure 3 (otherwise, the financed and there will be zero production). — It~\ the solution to (14) (9), may the level of inequality between the program (14) will involve GD > 0. conditions for the party of the rich to win an election 0: i.e., t (1 - q ) the rich will -> win the election - rD ) A L + GD + (1 at ^G time t, if It-i = 0, D (15) , P <?0 and X >n-\, (16) t where GD Proof. is given by Lemma (8), GD is given by (14), and t d 6 establishes that It-i consider the scenario in which party bureaucrat will be v\ = R fired). in given by necessary. — and (9). Now suppose that It-i X X -\ € X t > which individual j Then, his net per-period payoff will t t is be Wt (so that = and no current pivotal and chooses — h = (1 — qo) h/qo, lifetime utility of Vi In contrast, is chooses It Consider the case in all future periods. and give him a R = is if j & Xt were - -i-il^^. 1-p to choose v\ Vi = P when pivotal, his value = ±-h+G D +pvi+1 21 (17) go . would be (18) when party P last expression incorporates the fact that if V +l 3 where This t the continuation value is bureaucracy, setting It The comparison = 1, of (17) and that / and it R power from then on, given by in the poor are in power, they reform the an absorbing state. —as a weak inequality—as a necessary condition when they Condition (16) are pivotal. were violated, bureaucrats would not be pivotal and party half of the votes even with both (15) and all = = of bureaucrats voting v\ (16) are necessary. sufficient to ensure dt Moreover, (15) — also necessary is R would receive less than R. This argument establishes that as a strict inequality R, since when both of these conditions hold, strategy for bureaucrats to vote for party and the 1 is (18) gives (15) for bureaucrats to support party since, if — is R whenever It-i = —and (16) are also a weakly dominant it is and the coalition of bureaucrats rich have a majority. Lemma which the bureaucrats 8 determines the conditions under and (a rich agent running for office) will will support party R be numerous enough to give them the majority. Condition (16) requires the size of the bureaucracy to be sufficient to give the majority to party R when all bureaucrats. vote with the rich. Nevertheless, X— size of bureaucracy. In particular, at To ensure that Let us first it n— w— we need to consider two cases separately. r E as the tax rate that party R would choose does, define E h/qo- Clearly, r it K, given (r E ) is - - (1 - n) r E A H - necessary to avoid tax evasion, first X= it [n (t e ) , and the The second that bureaucratic wages are equal - R is vr (r E when )} r the E AL +K= 0. (19) minimum number of bureaucrats are employed. case corresponds to the one where optimal size of bureaucracy for party satisfied as its unconstrained optimal given by the unique solution to the equation In other words, t e balances the government budget The actual government budget may not balance. 1/2, the policy to finance the investment in infrastructure, to n — 1/2 may not be the it {t e > n— ) also sufficient to 1/2, so that the unconstrained make sure that condition (16) is rich have a majority. case applies when this inequality does not hold, i.e., when -it {t e ) < n— 1/2. In this case, the unconstrained optimal policy for the rich would not satisfy (16), and party R cannot win the election with the win an election only if X >n— 1/2, minimum number and with of bureaucrats. Instead, party this larger size of bureaucracy, 22 R can budget balance S.E given by the solution to requires the greater tax rate f l\h 2/ It a A" AH - =-t* !-B A h + {l-n)f-E rich to 7r (t- 5 t ), E < te go The either t The the maximum max is <r E is given by (l tt f , (t e ) and n — 1/2, , and whenever and correspondingly, }. be able to generate for the rich to remains to check whether the rich inefficient state in order to (t and , td is _ 0) given by maintain political this question: te (9), — (I d - te ) ^ ^_ Under the is inefficient). - r D ) AH + Ah to set It >G d = for all t if the given by (19), and f is B) | —R rule of party (that E is given by (20). they will vote for party is, P, the per period return of the , , R ^ ^ . For the second part, note that party , (l-f E )A H > _ T D^ A H + G D^ AH + p G D When t e > f E party R can remain period return (l — r E ) A H which establishes and obtain the per r£ > also have f in power. It and , (8), whenever the state is we ), lemma answers following Proof. Suppose that bureaucrats play v rich E of {'r and commit to an {1 GD (20) is satisfied: e > fE orr E R (r 9 Suppose that condition (15) holds. Then the rich prefer following condition where > n 1/2 from the bureaucrats to remain in democracy. Lemma is needs to set prefer to pursue this strategy power — have provided the necessary conditions results so far sufficient votes R 0. This implies that the size of the bureaucracy necessary for the . form a winning coalition the tax rate that party K= 2 can be verified that whenever n n — 1/2 < :. * can always choose in power by choosing the its first 7" = part (21). myopic optimum when in power. This will give a representative rich agent utility VR = Here (l - t e ) Ah is (1 - rE ) A H + -L- current consumption, and (3 [(1 - td ) AH + G D ] [(l - td ) Ah + GD ] . / (1 — /?) uation value, which follows from the observation that since, by assumption, r E n — 1/2 > 7r E (r implies that party X= and thus party ) P will win n — 1/2 and guarantee the higher rate f . This all R will lose the election at the elections in to be in will give all power next date. forever, (1 - if) 1-/3 23 the contin- < t E we have , Then Lemma future dates. Alternatively, party 6 R can choose but at the expense of taxing the rich at a representative rich agent utility yn is A" ' Comparison of VR Remark 4 Condition If were not 1 the conditions in satisfied, 1 does not hold, that neither part of condition (21) could hold. This system of patronage inefficient Now putting these all (15) It and (21) = for > R the rich party 0, be 9 could never and t d = r E sat- so , a direct consequence of the fact that a is necessary for the rich to set up obtain: environment with emerging democracy. the political hold, then there exists a all t (21). politics. lemmas together we Proposition 3 Consider is Lemma we have G D — between the rich and the poor significant conflict in policies an second part of in the previous expression gives the when Condition In particular, isfied. VR with unique MPE. In If conditions this equilibrium, the rich elite choose always remains in power and the following policies are implemented: w t — X = , t = max{7r(r £ ) ,n- 1/2} , go G where t e is given by (19) and f the unique MPE unique policy vector is 0, andr = max {t E ,t e } t , given by (20). involves It = 1 in the initial period, and for all t > 1 dt , and {l-r D )A L + X = n(T D ), = G t = G D = {l-n)r D A H +[n-n(T D )}T D A L -[(l-T D )A L + h}n(T D )-K, given by is Lemma 9, h, t (9). first part of the proposition follows immediately from combining which provide the conditions, summarized by (15) and (21), the rich relative to living under the rule of party P. P is in is 8 desirable for (15) or (21) does not hold, then power and the second part of the proposition follows immediately from and Proposition Remark When Lemma under which the party of the rich, R, can convince the bureaucrats to vote for them, and this party the is t Proof. The 6 =P w and t d and = GE = on the other hand, one or both of conditions (15) and (21) hold with the reverse inequal- If, ity, t Lemma 1. 5 Proposition 3 does not cover the case in which one of conditions (15) and (21) holds as equality; in this case the MPE is no longer unique. 24 It is straightforward to see that poor party could receive the majority of the votes, or the in such a case, either the rich or the rich could be indifferent between maintaining an inefficient and an efficient state. We do not describe the equilibrium in these cases to avoid repetition and to save space. Remark 6 It can also be verified that the set of parameter values where It equilibrium in Proposition 3 high values of j3 is our major first result. are in power temporarily at time and a straightforward way emerges as an of doing this is to consider as in the proof of Proposition 5 in the Appendix. Proposition 3 who A nonempty. is — t large (inefficient) bureaucracy as a establishes the possibility that the rich elite, It = 0, way may choose an inefficient state organization of credibly committing to providing rents to This enables them to create a majority coalition consisting of themselves and bureaucrats. This coalition implements policies the bureaucrats, and thus capture democratic politics. that support low redistribution and low provision of public goods, but creates high rents for more bureaucrats. Perhaps interestingly, after inefficient state institutions persist fact that at efficient. any date these The reasoning bureaucrats in the by keeping an is and the t = 1, even when the society This rule of the rich continue. inefficient institutions is democratic, the is in spite of the can be reformed at no cost and made more related to the formation of the coalition between the rich first place. The rich realize that they will inefficient state structure and creating and the be able to maintain power only sufficient rents for bureaucrats. If these rents disappear, bureaucrats will ally themselves with the poor, since their net income will be the same as the net income of poor producers (recall Lemmas 5 and 6). It is precisely the presence of inefficient state institutions creating rents for the bureaucrats that induces them to support the policies of the Recognizing rich. this, when in power the rich choose to maintain the inefficient state structure. At the next date, the party representing the rich receives the support of the bureaucrats and the continues. The model and the persistence state than they therefore generates a political of inefficient state institutions. economy theory for both the emergence 14 would have been under permanent democracy (recall Proposition 1 inefficient and Lemma the size of the bureaucracy can be greater than under permanent democracy. This could be the case 14 consequently, the rich remain in power and the cycle noteworthy that even though taxes are lower in the equilibrium with It is also 9), rich; when The nature the rich elite hire more bureaucrats than necessary for preventing tax evasion from the persistence of policies arising in Coate and Morris (1999), is not on persistence of a certain set of collective decisions within a given institutional framework, but on the persistence of the inefficiency of state Hassler et al. of persistence here (2003), or is different Gomes and Jehiel (2005), because the focus institutions. 25 — in order to create a majority in favor of the persistence of the inefficient state X where > tt (t £ ) In particular, note that bureaucracy will be . control of the elite than in i.e., in the case more numerous under the democracy whenever tt (t d ) <n- Since in this case equation (21) implies that t e 1/2. < t d we must also have , n (r E ) < tt (t d < ) n — 1/2 and thus X > n (t e . ) Consequently, the rich not only choose an inefficient state organization, but they also choose overemployment of bureaucrats, number the of bureaucrats in the sense that strictly greater is capture of democratic politics by the rich bureaucracy is now than that necessary elite therefore creates unnecessarily large and for tax inspection. an inefficient state, The with poorly monitored and overpaid bureaucrats, and also leads to a situation in which the capacity of the state is not fully utilized. These inefficiencies imply that the allocation of resources in a captured democracy is worse than in a nondemocracy (or than in a perfectly functioning democracy). Naturally, these number inefficiencies have a political rationale, which is to increase the of bureaucrats that will vote for the party aligned with the rich, so that the rich can maintain political power in the future. Interestingly, because creating efficient one (which is an inefficient bureaucracy is more costly than creating an smaller and gives bureaucrats no rents), the citizens are worse off in a nonconsolidated (emerging) democracy, where they are taxed at rate max {t e ,t \, than they would be under a consolidated nondemocracy, where they are only taxed at rate t n max {r ,t }. Moreover, the rich are also worse off in this < equilibrium than they would be in a permanent nondemocracy, since they are paying higher wages to bureaucrats and possibly employing an excessive number of them. 4.5 We Comparative Statics next investigate the conditions under which the equilibrium involves the emergence and persistence of inefficient state institutions. The following proposition establishes that a certain degree of inequality between the poor and the rich high discount factor, /?, (i.e., and intermediate bureaucratic a high level of rents, (1 — A /A L ), a sufficiently qo) h/qo, are necessary for the emergence of inefficient state institutions. Proposition 4 Consider an economy characterized and the function p (•) . Holding all L H by the parameters (f3,n,A ,A ,K,h,qo) other parameters constant, we have: 26 1. there exists a > 1 2. there exist a' > 1 3. there exists such that and 9>0 Proof. For the E (21) A H /A L < that if (1 AL — A H > A H jA L > O , is necessary for It G D / (1 — GD > necessary for a', is O ). = sum 4; = 1; ft < implies It = 1; It = l. be violated and the conditions 1 will and Proposition 9 Suppose this is the case. some minimal there exists such that O , a' It efficient, i.e., inspection of the maximization problem Lemma of the other terms < this implies that there exists always h/qo £ (H,6), then 3 that 0, is A H /A L > Remark Since G D > 0. Since the 0. go) Condition cannot hold. Then the result follows from For the second part, recall from (15) — Remark part simply recall first then the state a, such that as long as (0, 1) and 9 such immediately shows that as (9) if £ O on the level of inequality, say such that hand right < for all 3. From Proposition (0, 1) in (1 3, — side of (15) (15) will condition go) is h/qo = positive, be violated and thus = l- It — For the third part, note that bureaucratic rents are equal to h/qo — h (1 — go) h/qo, which needs to be greater than or equal to the right hand side of (15). Let this right hand side be denoted by 9 (and note that This implies that we need value of (1 — (21) cannot (1 — 0). If (1 satisfied, thus It = 1. — > h/qo go) say #0i such that go) h/qo, be > 9 h/qo go) < then (15) 9, will 9 > 0. Next observe from te = 1- It is evident that This implies that for /( = 0, be violated and It = 1. (19) that there exists a when te — 1, we need h/qo < 9q condition and thus (l-qo)h/qo<9. U The first part of the proposition implies that a certain level of inequality the emergence of an inefficient state. not be redistributive and will it This is is necessary for with limited inequality, democracy intuitive; not be worthwhile for the rich to set up an inefficient will bureaucracy in order to keep the poor away from power. The second part implies that the high discount factor is also necessary for the because bureaucrats vote the future. Instead, if GD > 0. If their R as an "investment" , that inefficient state. is, This follows to obtain higher returns in they deviate and vote for party P, in the current period they receive both the same high wages P, for party emergence of the (since It = 0) and the positive discount factor were very small, to convince bureaucrats to support their party. 15 level of public it good provided by party would be impossible for rich agents Finally, the third part of the proposition implies that bureaucratic rents need to take intermediate values. If bureaucratic rents are very 10 Robinson (2001) and Acemoglu and Robinson (2006b) also obtain the result that higher discount factors lead to greater inefficiencies. However, in these models the source of inefficiency is very different. In particular, inefficient political equilibria arise when pivotal agents elites or rulers are sufficiently patient and may — thus take inefficient actions in order to secure their future political survival. 27 — small, bureaucrats would not support the party of the they are very large, rich. If it becomes prohibitively costly for the rich to control democratic politics. While Proposition 4 shows that a certain degree of inequality is necessary for It = 0, it does not establish that inequality has a monotonic effect on the likelihood of an inefficient state. The next distribution in the economy, Y= income, — n) (1 Proposition 5 Suppose 1 — 7T (t < ) Then 1. a simultaneous increase in i.e., A H + nA L that (r) ix there exists the inefficient state equilibrium, AH and decrease of the in AL income such that remains constant. log-concave in r and t is /3 restrictive assumptions. we mean a mean-preserving spread In this proposition, by greater inequality mean somewhat more proposition establishes this result under G such that for (0, 1) i.e., It = more 0, d all j3 given by (9) satisfies r D > greater inequality /3, < makes likely. Proof. See the Appendix. Remark 7 The condition that n p (x) that takes the power function form, this condition. violated, The condition that we would have that the non-increasing in A L (see the (r) is log-concave td < i.e., 1 — p tt utility of the (x) is = Pqx 01 D (r ) < poor in not very restrictive. For example, any Po for 1 is also > and a G natural; democracy (l if (0, 1), satisfies this condition were — t d ) A l + G d would be Appendix). In addition to generalizing the part of Proposition first 4, this result implies that taxes (and public spending) can be higher in more equal societies, because unequal societies are more likely to create inefficient bureaucracies to limit taxation and public spending. This result therefore presents an alternative explanation to the often-discussed negative cross-sectional correlation between inequality and redistribution 5 Perotti, 1996, Benabou, 2000). Extensions In this section, we discuss a bureaucrats to be fired straint of bureaucrats will lose when he ment where each number when they is Third, of extensions of our benchmark model. First, we allow are caught shirking, so that the incentive compatibility con- forward-looking and takes into account the rents that a bureaucrat gets caught not exerting effort. Second, we allow a richer political environ- individual can run for office (form a party) as a citizen-candidate, so that bureaucrats can also form their rich. (e.g. we consider the own party and compete against the party of the poor and the case where the moral hazard problem of bureaucrats arises from their temptation to accept bribes that might be offered by taxpayers. 28 Equilibrium 5.1 The main When Bureaucrats Can Be Fired result of the previous section, Proposition 3, when they bureaucrats cannot be fired was derived under the assumption that are caught shirking. This simplified the analysis by enabling us to write the incentive compatibility constraint of bureaucrats in the simple form As discussed of condition (5). Remark in when they allow bureaucrats to be fired optimal. is The this was mainly discussion in Remark firing is clear that the incentive compatibility constraint of bureaucrats, the equivalent of be: Given this condition, all of the results ifications. In particular Lemma any Gt 10 Consider MPE, = GE if a\ = 0, =R are caught - p{1 (1 (5), in this case, would -^ h (22) . from the previous section apply with appropriate mod- we have (proof omitted): environment where bureaucrats can the and It -i where f = we have 0, w = t fi (l - fE) Then be fired for shirking. L A + - (1 (3 (1 - q )) h/q in and the solution to is (±-e(i-«o))h - Hl _ f E )AL+ A, from the viewpoint when they bureaucrats now establishes that, in a stationary equilibrium, 1 w>P(l-r)A L + We for expositional reasons. are caught shirking. It which commits to of discouraging shirking, a contract shirking 1 (1 - n) t e A h - [n - Xm f ] EAL +K= (23) 9o = max {n where X m (f E ) , n — 1/2} . Moreover, we have the following generalization of Lemma Lemma 9 (proof omitted): 11 Consider the environment where bureaucrats can any MPE, It-i = the rich will win the election at time t only if there 0; if bureaucrats prefer to support the party of the rich, P (1 _ rE) AL + ^LlM-M h > (J be fired for shirking. an is i.e., rich-bureaucrat coalition has the majority, if the _ TD) A L + G D + i.e., inefficient state, D -^G P l (24) , if * >n-i (25) t where f is given by (23), GD is given by (8), td i.e., if <70 and Then in is given by (9), and GD is given by the solution to max(l-r)^l L + G T,G subject to {l-n)rA H +[n-n(T)]TA1 /3 ( 1 .r E )AL + (l-l3(l-qo))<7o 29 n (r) - K. this equilibrium Furthermore, the rich prefer (t and choose d - rE AH > ) GD It = at time 3: the political environment with emerging bureaucrats can be fired caught shirking. MPE one in which the rich is elite = Tt — f where f , involves is h/qQ )) , conditions (24) and (26) hold, the unique in the initial period X = max {n t E ,n) (f and for all t thereafter, implemented the following policies are at all dates: G = GE = 1/2}, t 0, It — 1 in the initial period, (1 and for all t > 1 a\ , = P and t t Combining Proof. Lemma under which the party of the 10 and rich, Lemma immediately from Lemma 6 -n (t d )} t d Al -K- and (26), R, can convince the bureaucrats to vote for them, and this P do not hold, then party [n MPE the unique policy vector 11 provides the conditions, (24) When desirable for the rich relative to living under the rule of party P. or both, and given by (23). - t d ) A L + h, X = vr (t d ), G = G D = (1 - n)r D A H + - t d ) A L + h] n (t d ), and t d as given by (9). = [(1 is q ) = democracy and suppose that one or both of conditions (24) and (26) hold with the reverse inequality, the unique If wt - f E A L + (1-[3{1- (1 [3 It power and the rich party always remains in Wt choose if if (26) Proposition 6 Consider Then, only . These two lemmas give the following analogue to Proposition if t is in instead (24) or (26), power and the second part of the proposition follows and Proposition 1. Proposition 6 demonstrates that the main results from Proposition 3 generalize to the environment where bureaucrats can be worth noting, however. factor, is makes the emergence /?, can be but main One important caught shirking. analysis, Proposition 4 of an inefficient state more difference is showed that a higher discount likely. Instead, when bureaucrats the relationship between the discount factor and the emergence of inefficient states fired, more complex. Higher rents, In our fired if (5 again increases the importance that bureaucrats attach to future also reduces the level of rents, because being fired it from bureaucracy becomes more costly. 5.2 Political Equilibrium Citizen- Candidates The previous between P analysis limited the political system under and R, the two parties representing the justified this democracy interests of the by assuming that bureaucrats are not allowed are not allowed to run for office, it is to a two-party competition to run poor and the for office. Even if rich. We bureaucrats possible that a party representing their interest might 30 form. If such a party forms, bureaucrats may vote for that party, and the coalition between the rich and the bureaucrats, choosing low public good provision and low taxes, We now materialize. investigate whether in general we expect this to may not be the case or not when multiple parties can enter the political system. We follow Osborne and Slivinski's (1996) and Besley and Coate's (1997) citizen-candidate model, where each individual agent can run as a candidate and upon election chooses his mostpreferred policy vector. This setup make not The problem with the office. we used above, and the policy vector was chosen credible policy promises was elected quite similar to the one is since parties could after a politician (party) citizen-candidate models in general is that when more than two parties compete, coordination among the citizens regarding which party has a chance to win the election voting stage. is important for the outcomes and typically lead to multiple equilibria in the To avoid these problems, we consider the following modification of the standard citizen-candidate model: Each individual can decide 1. is taken to be small utility and run to form a party (in particular, we for office, and where e \ 0). will consider the case from coming to power, but simply benefit from being in this has cost e, which Individuals derive no power by implementing policies that are in line with their interests. 2. Given ferable votes, all parties that are running for office, individuals vote using ballots with trans- meaning that each individual ranks all parties in strict order of preference. particular, the vote of individual j can be represented as \\ distinct elements of {R, P, according to the rule, first B}, e.g., vJ — RBP. In the t Then preferences of the voters. the party that gets the lowest fraction of votes to the second-ranked choice of the voters who had is = i^h, first stage, as is where i\,i2 and In £3 -are parties are allocated votes standard with this type of voting eliminated, and its votes are allocated originally voted for this party. This process continues until one of the parties has a majority. To simplify the discussion, in this section we assume that bureaucrats cannot be caught shirking, so the incentive compatibility constraint for bureaucrats though this has Given no effect this setup, the notion of analysis in this case Lemma on any of the is still given by (5) results in the section. Markov Perfect Equilibrium is is modified accordingly. The a weakly dominant strategy for each individual. Proof. The transferable votes imply that is if tractable thanks to the following series of lemmas: 12 Truthful ranking individual is fired pivotal, has a choice at any stage of the elimination process, either an between two options, and thus 31 is better off ranking his more preferred outcome above a any choice Lemma of the is 13 In any MPE, there same two outcome. Alternatively, the individual a best response. This establishes that truthful ranking group. Proof. The (or less preferred Thus the will never be maximum number two bureaucrats) of parties be will not pivotal, weakly dominant. more than one party operated by an individual result follows since the policies chosen rich agent or is is is three. by two parties run by two poor agents Moreover, from identical. agent ranks parties truthfully, thus the addition of a new party will Lemma 12, each not change the equilibrium probability that a party run by a poor individual, a rich individual or a bureaucrat wins the election. Thus conditional on a party run by a poor agent other poor agent to incur the cost e Lemma > and form a existing, there is no point for any party. 13 then enables us to simply look at the (truthful) preference ranking of each individual over at three parties {P, R, B}, corresponding to parties run by a poor individual, a rich individual and a bureaucrat (there is no source of confusion in this notation, since there can at most be one party run by a poor agent, one run by a rich agent, and one run by a bureaucrat). To do types of parties. parties P and R of bureaucrats). bureaucrat. Lemma The Our previous We Gt 0, know the policies that will be chosen by the three R trying to is come to office by attracting the votes therefore only need to look at the policy choice of a party run by a following — to analysis already establishes the policies that will be chosen by (provided that party 14 Taking future policy vector: we need this, lemma characterizes this choice: election results as given, the party d t X = min [X -i,n (t b )}, wjb — arg would choose the following and (t b ,w b ) such t t =B that max w subject to mm{X ^,ir(T)}w + K t Proof. < (1 - tA h + n) [n - min {X t _i,7r (r)}] tA l . This immediately follows by writing the program to maximize the return to a bureaucrat (without allowing firing of existing bureaucrats): w+G max T,w,X ,I,G t 32 subject to min{X _i,7r(r.)} < X -r) A L + < to G < (1 t in;is -ttt,(1 ! /i I •iVtT < constraint. Notice that Lemma n) rA H + n -X] tA l [ wX - K G. would maximize Intuitively, bureaucrats - their wages subject to the government budget 14 applies taking the results of future elections as given. If the current bureaucratic government could influence the outcome of future elections, this could be beneficial for it only by increasing Xt above min {Xt_i,7r (t government budget constraint) make this policy vector even )}, which would (from the less attractive to poor and rich agents. The key government to the results in this section is the following observation: because a bureaucratic maximize wages paid to bureaucrats (and provide no public goods), will a lower utility to poor agents than a rich government would do. As a result, To show bureaucratic government will never get elected. vote of individual j at time means that the party t by Vj which , is this more we it will see that a formally, let us denote the = PRB a ranking over {P, R, B}. For example, vj individual ranks the poor party first, the rich party second in the bureaucratic last. We now Lemma 15 have the following rankings IfjEH, If.j € £ and If j 6 X, thenv{ Proof. We j <£ Proposition then v£ for individuals: = RPB. X, then w\ = PRB. = BRP. have already established that voters rank parties truthfully, so that rank their own party first. (3) to attract Assuming that party the bureaucrats, the poor as second choice; hence, if to the bureaucrats since neither of j offers wB party B get the rich are in power. Hence, chooses a wage for voters implements the policy characterized we have that bureaucrats indeed € X, then Vj them R all = BRP. in prefer the rich to Moreover, the poor prefer the rich any public good, but the rich tax less than the bureaucrats. This follows since both the rich and the bureaucrats choose to finance if yields K, and bureaucrats higher than the wage h/qo that the bureaucrats if j G C and 33 j £ X, then v^ = PRB. Finally, the second choice of the rich is for the poor, both because the poor would provide a positive amount of the public good rather than zero as the bureaucrats would, and because the poor would tax then the bureaucrats given that the marginal cost of taxation for the poor less zero for the bureaucrats (who are not taxed by assumption). vf It follows positive, is that and j £ 7i, then if = RPB. n Lemma 15 implies that the poor, when they cannot have a majority by themselves,' will support the rich party, thus as long as the bureaucrats are not in majority by themselves, X t < 1/2 and the rich pursue the policy in Proposition 3, we have dt will = i.e., R. This implies that the rich can continue to use same political strategies as in the previous section to control democracy. political decision-making in Now combining the previous lemmas, we have the following proposition, which mirrors Proposition 3. Proposition 7 Consider environment with emerging democracy and free the political entry by citizen candidates. Suppose e [ 1/2, where t e game, only a party run by a rich agent — by It for 0, all t, If w — where f is X — t , max {71- ,n- (r^) MPE The unique equilibrium policy vector 1/2}, is is active, G = GE = t MPE = 1 t case (21) hold. Then, from Lemma 15, when The assumption that X-\ = assumption that tt (t e ) < 1/2 ensure X never win an election. X-i < 1/2, when and the the rich party crat incurs the cost e j argument as is in power. Therefore, to form a party, the policy in Proposition 3 (if [ to when < for it (t d < ) [n First, - and for all t, it (t d )] t d Al all - suppose that con- 1/2, the bureaucratic party implies that in the initial period that X < the rich party 1/2 continues to be the is in power, and thus bureaucrats support party in the proof of Proposition 3. Next, no poor agent incurs the cost s election, given is involves only a party run by a poor Proof. This proposition can be proved by backward induction. and < = max {t e ,t e }, and r t 0, the unique equilibrium policy vector involves It t ditions (15) ) given by (20). defined by (9), then the unique and 5 (t- of the citizen- candidate political t>l,VH= {l-r D )A L + h, X = n(T D ), G = G D = {l-n)T D A H + [(l-T D )A L + h]n{T D )-K. will it one or both of conditions (15) and (21) holds with the reverse inequality and d 1/2, where t agent h/q t is active. and (21) and that conditions (15) Then, in any defined by (19) above. is and that political R no bureau- by the same knowing that bureaucrats support party R, compete against party R as long as party R is choosing they did deviate from this policy, then a poor party can win an and thus a poor agent will find it beneficial to enter 34 and form a party since e I 0; thus despite the fact that party P R has would not be running, party vector as in Proposition 3). Finally, since e cannot be an equilibrium J 0, it same policy to adopt the for no rich agents to form a party, since such a party would create strictly positive gains for each rich agent, the cost of creating a party The proof inequality is of the cases e J 0. is where one or both of conditions and (15) (21) hold with the reverse similar. This proposition therefore shows that our main results regarding the use of an way by the state as a The is enriched to allow free entry by citizen-candidates of any additional insights that interesting in this case is producers prefer to support the party of the B, since the inefficient rich elite to control the democratic political process continue to apply even when the political structure occupation. and is that when the poor R, rather than the party of the bureaucrats, rich, would impose high taxes and provide no public goods (spending latter all the proceeds on bureaucratic wages). Bureaucratic Corruption 5.3 We now briefly discuss an extension of our basic the side of bureaucrats is not related to their model effort, in which the moral hazard problem on but to whether or not they accept bribes from producers evading taxes. This source of moral hazard problem we as the effort choice of bureaucrats. Moreover, will see below that is arguably as important it leads to an interesting pattern of de facto regressive taxation as a result of successful patronage politics by the rich elite. The economic and a two-party system. political The only environment difference is similar to the baseline version of the is that the bureaucrats no longer have an effort choice. > Instead, producers that have evaded taxes can pay a bribe b them in order to avoid corruption inefficient is if 6 {0, 1}. we allow When / = and corruption is two for 1, detected with probability q(I there = 1) levels of is = an 1. First, When = loses the bribe after = described by organization of the state and = qo 0, < the state organization 1. is We make a number of is limited to only one period; the but receives no other punishment. Second, payments and other income confiscated are we assume that 0) 1 efficiency, we assume that a bureaucrat caught accepting bribes loses his wage and the bribe, but the punishment producer paying the bribes monitoring efficient detected with probability q (I assumptions to simplify the exposition. constraint. Third, to the bureaucrat inspecting paying taxes. Similar to the baseline model, the state variable model with lost all and thus do not enter the government budget matching with a bureaucrat, the producer has 35 bribe all the bargaining power and makes a take-it-or-leave-it bribe offer to the bureaucrat. assumptions can be relaxed without changing our main there the concave while x/ is is a range for x G minimum (1 [0, — x) is convex and both are equal to zero x m such that p{x) ] > x/ that the region where p(x) — (1 x) is from The policies of the rich. (6), since > x/ (1 — We x — assume that x m there is inefficient so that q is 0. is — Therefore, lower than and characterize the go participation constraint of the bureaucrat no cost of t is slightly that effort. It requires w >(l-T )A L The x). for irrelevant for the equilibrium. Let us start with the case where the state different Note that the function x). bureaucracy necessary to finance the infrastructure K, which ensures size of the most preferred < x/ (1 — parameter values, p(x) that, for the relevant part of the p{x) results. that each bureaucrat can be matched with at most one producer and we assume Finally, All of these (27) . t incentive compatibility constraint for bureaucrats (5) is now replaced by the following "no bribe constraint" w > t where b t is (1 - go) [wt + bt ) (28) , the bribe offered to the bureaucrat by a producer. Intuitively, the right hand side of (28) represents the expected return of a bureaucrat that accepts a bribe b u given by the sum of the wage and the condition (28) does not hold, by the probability of not being detected. bribe, weighted it is not possible to prevent the corruption of bureaucrats by producers. 16 Condition (28) implies that, given the public sector wage than a threshold b (w t ) will what maximize follows, we drop time when they do not pay (z = 0) taxes (ii) bribing is When X, G, and the in power, the rich decision variable z 6 whether to pay taxes. The expected ^— w, = p (x) max j A H Expression (30) incorporates the following (x); only bribes higher utility is [ with probability p , (29) subscripts to simplify notation. their per-period utility with respect to r, w, uH t = -^—w. {0, 1}, which, as before, designates their decision of of the rich w be accepted, where b(w) In If 1 - <?o facts: 1 + [1 (i) G. (30) producers are inspected by a bureaucrat detected with probability :6 - p (s)] A H + J qo; (iii) the bribe offered by the Van Rijckeghem and Weder (2001) provide evidence that higher public sector wages relative to manufacturing wages reduce the scope for the corruption of the public administration. 36 rich to bureaucrats (29); and (iv) when Expression (30) is p equal to the lowest acceptable bribe b(w) is inspected, the income of a rich producer maximized subject Or) is = qow/ max {A (1 — qo) defined in — qow/ (1 — go) , 0}. to the following constraints max / A L - -S°— w ,o\ + 1-90 I J xw + G + and subject to the participation constraint [1 - p (x)] A L < K <{u-x)tA l (1 - r) AL (31) , (32) , Constraint (31) requires (27) of the bureaucrats. that the poor prefer to pay taxes to tax evasion. This constraint has to be satisfied since at least one class must pay taxes, otherwise investment, K (this is because, it would not be possible to finance the infrastructure the poor prefer to evade taxes, the rich will do so a fortiori). if Constraint (32) implies that the government budget constraint the fact that public revenues Lemma 16 Suppose come from the taxation that the rich prefer not to pay GE = and for some r G = taking account of of the poor only. taxes. W E =(l-qQ )A L /qQ 'p(x) is satisfied, Then their optimal policies involve (33) , T, (34) [0, 1]. Proof. See the Appendix. We rich. will next show that given (33)' Substituting for these expressions, and (34), the equilibrium involves tax evasion we obtain the utility of the rich by the when they evade taxes as uH In contrast, when a (z = 0) = (1-t)A h + t (A h - A L (35) . ) rich agent pays the tax rate (while all others evade taxes) his utility would be uH {z = l) = (1-t)A h < (1-t)A h + t(A h = uH (z = Next let f (36) , -A L ), 0). denote the unique value of t satisfying the government budget constraint (32), at the candidate equilibrium with the rich agents evading taxes „(^Y*—to AL + K=[n-ir(f E )]TJS A L Qo 37 , (37) where As is 7r () is again defined in main in the (1). two cases to consider depending on whether n — n (t e ) analysis, there are greater than or less than 1/2. Here many there are sufficiently majority, i.e., n— tt (t simplify the analysis by focusing on the case where bureaucrats so that, together with the rich, they are the absolute < ) we The converse 1/2. case with n — (r tt > ) 1/2 necessitates that the rich create an inefficiently large bureaucracy in order to win the election. Since the results we do not in this case are again similar, Lemmas 5 and 6 continue apply ever vote for the poor, there is them discuss in this extension. in this modified environment. In particular, if bureaucrats a permanent transition to an equilibrium with an efficient state lemma with the poor in power within one period from the election. The following characterizes the policy vector that the poor would implement in the period they win the election existing organization of the state when the Lemma state is - (1 inefficient and the also the policy vector that they will choose efficient. 17 Suppose max is when r) =P that d± and consider the following maximization program: AL + G T,G,W subject to G = z{l-n)rA H + and (1 Then is [n-Tr(T)]TA L - (r)) max ^A L - (tt (t)) max j A H - t) Al < p (tt p — JZ^m^ o| + y w, where z 6 ' w, {0, 1} the policy vector that the poor + 1 [1 [1 - p (tt (r))] A L < - p (tt term in the last = 1, is would choose when It-i given by Tt two inequalities The penultimate is = t , Gt constraint When It -\ is — 1 = = q$. = rt 0, r) AH — f , Gt , , q (I) program in this lemma two inequalities becomes = and z 1, or z = GD w = w D , t , = 1 and the first last constraint allows the rich. represents the "no program Notice that the penultimate one will also be satisfied automatically and the state , The policy vector that the poor would no tax evasion constraint of the satisfy the - (1 AL r) G D w — w D when inequality in the maximization satisfied, AH < - equal to zero. tax evasion" constraint for the poor, while the whether or not to (r))] (1 denotes the decision of the rich whether to pay taxes. given by the solution to this program when q (I) choose when It-\ K vm (t) - is efficient, bribery zero. 38 is not possible and the to choose if this last H > A L ). (since A max term in the last = 0, The next two lemmas Lemmas are the analogues of under which the bureaucrats are willing to vote in which they are Lemmas similar to those of Lemma Itr-l = power to the one in 8 and for the rich, which the poor are and the in rich prefer the allocation power. Since their proofs are they are omitted. 9, 18 In an MPE, the rich in and 9 and determine the conditions 8 win the election will at time t (i.e., dt = R) if only if 0. 1 -<7o AL > (1 - (wD + P) GD) + [(1 - t d ) A L + GD (38) ~\ , <?0 where w D GD , and f , GD are defined in Lemma 17. Condition (38) implies that the bureaucrats prefer to be in an inefficient state under the rule of the rich, given the poor. In fact, whereas if "wage policy" that they vote for the is optimal for the for one period (while the state than voting the bureaucrats obtain a wage equal to (1 rich, they vote for the poor, they obtain a wage of if rich, rather is inefficient), wD — go) for the A L /qo, and public with provision of GD and subsequently a payoff equal to the payoff of the poor under an efficient state. Lemma 19 Suppose the following condition _ [I where f Proof. with an It is p (jr and G D A H + £D] < are defined in immediate that (f ) as in ) Lemma (39) is (^))] fr 17, and f , AH + p ^ is sufficient to the tax rate f to living under Lemma _p [j (-*)) is all t if The ( A H _ AL) (39) ensure that the rich prefer to be in power evade taxes and pay bribes equal to AL with democracy with taxes and public good provision given 17. lower than the expected payoff that they get they are in power). for given by (37). Condition (39) states that the payoff to the rich when the state are in power) = is satisfied inefficient state, set probability by t $J>) GD and Then, the rich prefer to set It that condition (38) holds. is efficient when the state latter payoff reflects the following facts: only the payers are inspected with probability p (ir Or )) = f , and the is (and the poor inefficient (and poor pay taxes, tax rich offer a bribe equal to AL to the bureaucrat inspecting them. The its following proposition characterizes the equilibrium with bureaucratic corruption. Since proof follows that of Proposition 3 closely, it is 39 omitted. ditions (38) and (39) MPE hold, the unique and for the initial period Then, if con- the political environment with emerging democracy. Proposition 8 Consider the rich party all t thereafter, following policies are implemented at = Wt one in which the rich is R elite choose It = in always remains in power and the all dates: tf* lzM A\ = Xt = *(**), <7o G where t e pay a is 0, = te and r t = A L when inspected. on the other hand, one or both of conditions (38) and (39) hold with the reverse inequal- MPE the unique vector involves It = for 1 and for all t, all t > 1, dt — P and the unique policy is where t D =(l-¥>)A L X = n{TD ), w t = G t = G D = (l-n)f D A H +[n-7T(f D )}f D A L -{l-T D )A L 7r{f D )-K, and ti G Proof. The are defined in first remains to be proved inequality, the poor t , Lemma 17. part of the proposition follows from is when one that will be in power. To see evading taxes. Moreover, as before, if rich would have rich under It ~i p is (tt (t)) max ( A H this constraint is democracy to setting up an the state (1 [l — p (tt (v is efficient with the reverse elite (i.e., It = 1), inefficient state and the poor will be in would not prefer an inefficient state since the "no tax evasion constraint" for the , ^—w, ol + 1-90 some tax to rich agents, which ))] than f violated (since straint will be satisfied at — t') A h (39) hold part that is I , and The only straightforward since to prevent tax evasion by themselves, the to set a higher tax rate = 16-19. note that these conditions (with the reverse this, power. Therefore, we only have to show that the rich and no tax evasion. This Lemmas or both of conditions (38) inequality) are sufficient for the rich to prefer At f , given by (37). Moreover, only the poor pay taxes, while the rich evade taxes and bribe equal to b If, ity, = GE = t AH . is - p (tt (t)')1 A h < [1 (1 - r) AH . J it rate r' satisfied as equality for is ). Thus, E which would give a per-period E max [A H — qow/ (1 — than (n (r > strictly less AL f , p )) Therefore, the rich are always better off evading taxes when This establishes that conditions (38) and (39) holding with reverse inequality are the equilibrium with the poor in power to emerge. 40 this conutility of qo) , 0} + in power. sufficient for The most mocratic interesting result in Proposition 8 is do not pay any taxes at politics, the rich L equal to the tax burden on poor agents, A that, all. when they Instead, they (sometimes) pay bribes This implies that patronage politics turns de jure . proportional taxation into a de facto regressive one. In other words, able to set up an inefficient state and receive the support to limit redistribution and public are able to capture de- when the rich elite are of bureaucrats, they are not only able good provision, but they are also able to shift most of the burden of taxation to the poor. Consequently, the tax rate faced by the poor may be higher when corruption An 6 A is possible than in the baseline model where both Empirical Implication and distinctive empirical implication of our model and public good provision. wages when the elite and poor pay taxes. Some Evidence is that democracies where relative wages of bureaucrats are high should provide fewer public goods. This crats are paid higher relative rich because, is all else equal, bureau- use patronage politics to limit redistribution In contrast, a naive intuition may suggest that relative wages of bureaucrats and public good provision should be correlated positively, either because there is greater provision of public goods, more when activities are entrusted to bureaucrats and they need to be paid more, or because countries with a greater willingness to tax will spend more both on public employment and on public good We provision. 17 next look at the cross-country correlation between the relative wages of bureaucrats and public good provision among democracies. Our measure crats is average wage of public-sector employees relative to for 1991-2000. Our main measure of public GDP good provision GDP for 1991-1998, and we also of GDP as an alternative dependent is wage of the relative of bureau- per capita from World total (central) Bank government ex- penditure as a fraction of look at social services and welfare spending as a fraction variable. 18 Both of these variables are from the IMF's International Financial Statistics (see the details in the Appendix). focus on democracies, than or equal to we limit the To sample to countries with an average Polity score greater 5 over the period 1991-1998, which corresponds to "stable democracies" Persson and Tabellini, 2003). Our baseline sample contains 51 observations. Figure 1 (see shows the correlation between the relative wages of bureaucrats and government expenditure share of 17 An may be that, with a fixed government budget, higher public sector wages would government to reduce the rest of public good expenditures. In practice, there is considerable variation in the level of government budgets, and we will see below that same results apply with a measure of spending on social services and welfare. We choose the total government expenditure as our main measure, both because wc have more observations on this variable and also because the alternative, social services and welfare spending share of GDP, is heavily influenced by the age structure of the population. See below. alternative intuition force the 41 GDP. A strong negative relationship is with most European countries hav- visible in the figure, ing lower relative wages for bureaucrats and government expenditures than in Latin and Asian countries (there are few African countries The regression corresponding to Figure The bust standard errors in parentheses. government expenditure share of approximately 5. Column GDP 1 is in our sample). shown column in 1 of Table with the ro- 1, wages and correlation between relative bureaucratic is statistically highly significant, 2 of the table controls for more on public goods and income per capita GDP with a t-statistic of per capita. Richer countries spend also correlated with relative bureaucratic wages. is This regression shows that log income per capita is indeed significant, but the relationship be- tween relative bureaucratic wages and government expenditures remains strong (the ficient of democracy score, which and has insignificant is coefficient Column declines from -4.96 to -3.63, which continues to be significant at less than 1%). controls for the Polity American little effect 3 also on the coef- the relative wage of bureaucrats. Figure 2 shows the conditional correlation between relative bureaucratic 3 of Table The same negative 1. Column wage and government expenditure share relationship as in Figure GDP corresponding to column of 1 is again visible. 4 controls for the age structure of the population, in particular the fraction of We the population between the ages of 15-64 and the fraction over the age of 65. expect the age structure of the population to have a direct effect on Social Security spending and thus also on total government expenditures. The results in column 4 show that controlling for the age structure variables significantly reduces both the coefficient estimate of the relative The bureaucratic wage and the standard errors. standard error of 0.85, which Columns percentage of 5-8 repeat the GDP as the is still same significant at coefficient estimate now is -1.82, with a 5%. regressions using social services and welfare spending dependent variable. The results are similar to those for as a government expenditure and typically stronger, except when we control for the age structure variables. In particular, in columns 5-8, relative bureaucratic wage is significant at less than 1%. Once we include the controls for the age structure of the population, however, the relationship between the relative bureaucratic wage and social services and welfare spending is no longer significant; the coefficient estimate declines significantly and the standard error doubles. This result might reflect the fact that social services of the population and there and welfare spending are is little closely related to the age structure cross-sectional variation left once we control for the age structure variablesas. In addition to the results shown in Table 1, we have also "semi-stable" democracies (those with Polity scores between 42 experimented with including and 5). The results are similar but slightly weaker. The results are also similar when we construct the sample using Freedom House measures of political various other controls. The and We civil rights. also checked the robustness of the results to results are broadly similar when we control for the legal origin of the country, for parliamentary versus presidential systems, for majoritarian versus proportional democracies, and for the age of democracy. Nevertheless, the results are significantly weakened or disappear when we since, as Figures 1 it of continent full set and 2 show, the American and Asian Overall, control for a dummies. This results reflect the contrast of is not entirely surprising, European countries countries. appears that there is a significant negative relationship between government expenditure and the relative wages of bureaucrats, which becomes weaker for the age structure of the correlation is to Latin when we control population and for continent dummies. While this cross-country not as robust as we would like it to be, it is nonetheless encouraging for our approach, since the negative relationship between relative wages of bureaucrats and government expenditure is a counter-intuitive implication of our model and a naive intuition would have suggested the opposite relationship between these two variables. 7 Concluding Remarks Inefficiencies in the bureaucratic organization of the state are often factor in retarding economic development. opment emphasize the Many viewed as an important sociological accounts of comparative devel- role of state capacity (or lack thereof) in explaining are able to industrialize and modernize ficient state organizations (e.g., why some Evans, 1995, Migdal, 1988). societies In addition, inef- appear to coincide with limited amounts of public good provision and redistribution towards the poor. Existing approaches do not address the question certain societies choose or end up with such inefficient organizations and do not of why clarify the relationship between inefficient state organizations and limited redistribution. We presented a simple theory of the emergence and persistence of inefficient states, in which the organization of the public bureaucracy redistributive politics. In particular, is manipulated by the rich by instituting an them little We public good provision and there also to limit the extent of and public good provision. Captured democracies not only but also create a number of major distortions: the structure of the state too showed that an may be overemployment inefficient state creates its 43 order to influence inefficient state structure, the elite are able to use patronage and capture democratic politics. This enables redistribution elite in limit redistribution, is inefficient, there is of bureaucrats. own constituency and tends to persist over time. would an he will an Intuitively, inefficient state structure creates When the median efficient state structure. reform the structure of the state to make it more (poor) agent comes to power in democracy, more efficient so be collected at lower cost (especially in terms of lower rents this, when the organization of the state rents for bureaucrats than is inefficient, that the higher taxes can for bureaucrats). bureaucrats support the Anticipating rich, who set lower taxes but pay high wages to bureaucrats. In order to generate enough political support, the coalition of the rich and the bureaucrats the state, but they may The model shows further expand the correlated because higher inequality an of elite control in inefficient state organization tures in a number size of bureaucracy so as to gain additional votes. when democratic interesting implication of this result The pattern not only choose an inefficient organization of that an equilibrium with an inefficient state greater income inequality and is may is is more likely when there An taxes are anticipated to be higher. may be that inequality and redistribution makes the capture of democratic politics democracy based on patronage politics more negatively likely. and the emergence of bears some resemblance to the inefficient bureaucratic strucIn addition to these case studies, of countries. we provided cross-country correlations consistent with a distinctive implication of our model, that among democracies there should be a negative relationship between the relative wages of state employees and the amount The of public good provision. general message from our analysis some democracies may become will is that "not all democracies are created equal" while ; adopt policies that redistribute to poorer segments of the captured by traditional elites. These captured democracies not only choose low levels of redistribution, but, as part of their political rationale for survival, create a range of inefficiencies. Our model suggests that to the relatively poor performance of a Analyses of the effect of methods via which the rich number they also typically these inefficiencies might be related of democracies in Latin America and Asia. 19 such policies on economic growth and investigations of other may limit the are interesting areas for future work. amount of redistribution in democratic politics Another interesting area careful empirical analysis of the relationship between the variation expenditure, relative wages of state employees and potential 1 society, others elite for further study in the extent of is a more government capture of democratic politics. poor economic performances of Latin American democracies is beyond the scope of the current paper. Nevertheless, it may be conjectured that the political environment may be more conducive to populism when the majority of the population fare relatively badly under democracy (see Acemoglu, 2007) and the type of democratic capture studied in this paper is likely to limit the benefits of democracy for the majority of the population. Another potential "populism". political factor in the Why some countries pursue populist policies is 44 Appendix A: Omitted Proofs Proof of Proposition 5 7.1 Consider changes in inequality that keep mean pre-tax income, Y= — (1 n) A H + nA L , constant. This implies the following simple relationship between the pre-tax incomes of rich and poor agents: Y-nA — —n „„ AH = L , , 40) . 1 To prove the desired when to hold there is we need result, to greater inequality, show that and (15) when A L i.e., (21) in lower (and is Lemmas AH is 8 and 9 are more likely given by (40)). Let us rewrite condition (15) as —h > <7o-i-goL that more generally X= ~ ~j^G D l ' Lemma Next, consider condition (21) in _ r D )/_A L + G D + " (1 v 9. max {n (t e ) n — , Suppose first that (41) P te > r i.e., , when X= -n (r E ) (recall 1/2}). Then, combining the government budget constraint with (40) gives (19) 7T „s " (t e K h ) Y-n{T E )A L + v Y q wL _,, e g )A l -tt(t ' ( 42 ) e Substituting for t from this expression, condition (21) can be rewritten as h Instead, and when r E < f , G E \ Y-n(r E )A L f the size of the bureaucracy K X=n— is _ 1/2 = A. Solving for f from (20) (40) as +K Y-\A L Xh/qp E the relevant part of condition (21), (l - fE\ AH > (1 - (3) [ ' ' (l - te ) Ah + - t d ) A h + GD [(l , ] can be expressed as - < \ - { [(1 13) t e + (t These three expressions define 9, 6 and 8 be created under two different scenarios: 1. if X= 7r (t e ) and if . Now if X=n— 1/2 = will A and if increase with income inequality. 1 to 2 or vice versa in a first inefficient state will 6; h/q <T. prove that higher inequality makes the inefficient state equilibrium more likely by showing decreasing in the level of inequality We (45) conditions (41) and (45) are satisfied, which requires that the upper thresholds (9 in case from -K}=T. summarizing our analysis, an < h/q < 9< We (Y - \A L ) conditions (41) and (43) are satisfied, which requires 6 2. G D /A H )} d - We way 1 and 9 in case 2) are increasing and the lower threshold, —these naturally imply that the intervals A = 8 — 9 and A* = 8 will also to make show that an increase the inefficient state less establish an intermediate result: 45 in inequality likely. 9, is —9 does not cause a switch Claim We 1 have 8r D 1 _ dAL ~ ~ tt" and 8A L (r D ) program first-order condition of first-order condition (48) therefore YThe solution for t d that t d < <0, (46) AiLL interior td - 1 tt (t t dY - (48) . 1 (t tt d (A l + ) h) is (49) . becomes tt' (t d (A l + h) ) always positive since is d < ) K- (47) is the equilibrium level of the public good (8) provided by the poor , GD = The ) + h) an (9) for 8t (40) p (A L dG D Using (t tt' T E TT (r E ) >0. E E Y-ir'{T ){T A L + h/q )-Tr{T B )A L dr 1 Proof. The + ensures that r D - AL = 0. (50) K > 1. Differentiating (50) gives (46). < needs to be financed. Moreover, the assumption Next, differentiating the government budget constraint (19) and using (40) gives (47), where the e is always to the left of the peak of the Laffer curve. is positive since t denominator Next, given the definition of 8 in (41) we obtain that 89 1 SUA*" where the expression i- _ rD _ ^ [1 ( T D)] >Q (51) 90 for the limit uses the fact that dG D (A L ) jdA L exists, is finite and is independent and the inequality again follows from the assumption that t d < 1 — tt (t d ) < 1. This inequality implies that for sufficiently high /?, the inefficient state becomes more attractive to bureaucrats as the on /3, level of inequality increases. We now show that higher by a decrease inequality, represented in AL with the inefficient state also more attractive to the rich by increasing 9 and 6 AH given by (40), makes . Consider two cases: Case From 1: X = tt(t e (49), ). we have T DY = Gd + K + tt (t d (A l + ) h) (52) Substituting (52) into (43) and some algebra gives Ah tx{t b ) Ah tt(t e ) v (53) ' Next, note that d [{A H - A L /A H ) 8 [A L /A H ] dA L 8 [1/tt (r g )] _ dA L 8 [tt (t p ) ] AH dA L /tt 8A L (t e )} _ w' (t [tt ( (t 46 3t 1 ) r B)] d 8t d ) tt{t e ) 8A l tt' e l-n(A H y 2 tt 8A L (t d ) <0 <0 n' (r [k(t e )Y g dr 1 ) 8A L <0 and from (49), using (50), 8G L di 8A L 8A L AL - n (r D < ) 0. (54) Differentiating (53), in turn, gives 8B nir' dA L (r E 8t e A h - A L ) u nD G +TT n A H - Al [T z A" dAL A L -n A H dA L tt(t e ) Al -tt t d (t ,' A^_ ' r AH tt{t e ) A" EtfdAL fr( T l + ) („-D d {t ) tt(t e ) l-n(A H AH dr D GD + 2 ) l-n(A H n (t 8A l d ) tt' (r \tt{t e )) 2 GD + ) g 8t e ) dA L 2 {A L + h) 8A l which can be rewritten as x' 89 8A L e dr E A H + dA L A 1 [tt{t e ) tt D - (r ) E TT (T L TT (T E ) n(r°) AH J d 8t d tt' (t ) e tt{t ~8A^~ (t d ) tt' (t \tt{t e )} ) + G A" [7r{T Ef d A^ + -A L ^ D 1-rE (t ) e 8t e ) 2 + n AH - AL AL -r D + + A" A" 1 tt(t e ) (A L dA^ GD + AH + - n (A H (55) ) h) ' ) Since tt d (t /tt ) 8t e /8A l > e (t ), 0, 8t d /8A l < are negative. Furthermore, tt(t e ) 8A^ A +h + where the right and side of concavity of tt and n > > (56) (t) implies that (t e all ), the terms in (55) except the last tt(t e ) tt(t e ) (r -tt' 1 D ) + tt' tt" {t d ) {t d ) + tt It 1- (t)] > tt" (t) (t) tt and is (56) sufficient to ensure that (56) is negative. the terms in (55) including tt {t d ) /tt (t e ) are negative, and therefore as desired. Note that this conclusion holds irrespective of the value of /3. This implies that line, we have obtained using the expression for 8r D /8A L in (46). Now, the log- is [tt' tt all Case 2: X = n - 1/2 = A. The proof parallels that of case (1/2)A L we obtain 1. From (45), using (52) and the fact that Y — XA L = 89/8A L < (1 — n) A H + , = j\(1-P)t e {Y-\A l )+P tt (r D {A L + ) L h)+K- Xt d A l + G d - (l-"M* + (l/2M Gg (57) Note that 8[{{\-n)A H + {\/2)A L )/A H 8A L 47 Y ] 2(1 HY -n){A H ) (58) Differentiating (57) and using (54), (58) and the 8r E /8A L fact that and exists is independent of /?, we have /3-i 8A L A ; v v ' 1 Y 8A L v dA L ' (l- n )A H + AH nD 2(1 -n){A H f /o\ (l i (1/2) AaL dA L v ' ^V-w.' dA L ' which can be rewritten as 3-i '!.") K <M'") 8A L ' v A ' \ ,--(-1' dA L " '"' v ' — \A H -A L 8t d ..L ( «--1 A L \ -A 2 i" 3A 1 2 AH V Since 8t d /8A L < 0, d > ) (t tt' n > 1/2 and ^ L < 0, AH , negative. Therefore, a sufficient condition to ensure that (59) *' (r D ) L §g (A where the right hand side of (60) + h) + * is negative for sufficiently high The proof ft = -«> (r°) 7r(r all is D . n){A nf w G + ' terms than n (t d ) are in (59) other negative is 1 ±^ff +n (r°) < (60) 0, obtained using the expression for dr D /8A L in (46). This condition equivalent to (56) and the log-concavity of is (t°) . . 2(1 7r (t) is ensure sufficient to it. This establishes that 88 as desired. so far has established that the lower threshold 8 declines as inequality increases the upper thresholds 8 and 8 is /8A L increase as inequality increases. To complete the proof of the and that proposition, we need to ensure that there would be no switch from the wider to the smaller interval, which could happen if we have a switch from t e > f to t e < f or vice versa. However, as /3 — 1, we have that > , (21) is equivalent to if if rE >f E & X = max{^(T B ),n-l/2}= W (r B te <t e & X = max so that for (5 {tt {t e ),ti- 1/2} ) =n- 1/2 = A sufficiently large, at the point of a possible switch, d - te Ah > ) and (t and (r D te = f , GD - f£) AH > GD we have 8 = 8. This completes the proof. Proof of 7.2 Lemma 16 Suppose that the bureaucratic wage is given by wE in (33). (28) of the bureaucrats inspecting low-skill producers as large as their bribes, income AL . Holding r is fixed, a decrease in and thus reduce government revenues to Then the incentive compatibility constraint even when the producers offer a bribe from w E will allow bureaucrats to accept satisfied w zero. Therefore, it cannot be optimal. Increasing w is also not beneficial for the rich. Condition wE , (31), on the other hand, ensures that the poor choose to pay taxes. Holding increasing taxes would induce the poor not to pay taxes and taxes is the rich also not beneficial. is Given (33) and (34), decreasing in G, so that this variable is it is is w fixed at therefore not beneficial. Reducing also straightforward to verify that the utility of set equal to zero. This argument shows that (33) and (34) gives a stationary point of the optimization problem of the the rich will not find it beneficial to change either one of x or w by itself. To complete the rich, since 48 we need to show that it is also not beneficial to change x and w simultaneously. We will do this by showing that the payoff function of the rich is strictly quasi-concave. Consider the problem of the rich if they do not pay taxes. Clearly, constraints (31) and (32) in equilibrium hold as equalities. We can thus solve out for the tax rate from the government budget constraint as proof, T = xw + - (n K+G x) AL we obtain Substituting this expression into (31), K+G w= [Qo/{l-<lo)](n-x)p(x)-x' > (1 — qo) A L /qo (which will be true in equilibrium). Now, substituting (61) in the provided that = 0, we can represent the objective function of the rich (30), and observing that this is maximized at w G problem of the rich as the following single dimensional maxt/ If this unique problem is = AH - strictly quasi-concave, (n-z) - (l-q )x/q p(x) it must have a unique ,,,/ (X) = U ,, U (x) to be strictly = 0. For this, it is For 1 + (1 - go) - /q p (x) [(n-x) quasi concave, sufficient for - it is (1 concave pay in x. - 2 go) \p' jx) x] /q - qo)x/q \p jx)} \p (re)] + (1 - -"• p(x)} sufficient that its 2 I (1 2 second derivative q )p(x) /q — (1 - go) \p' is negative (x) x] /go [ when to be can be easily verified that this is always the case since p (x) is increasing and This completes the proof that (33) and (34) is optimal for the rich when they prefer not strictly decreasing in x. It to Corresponding to this solution. be unique levels of r and w, since these variables are defined uniquely by the To check that this function is indeed strictly quasi-concave, note that x, there will previous equalities. U' (x) (x) maximization problem taxes. 49 Appendix B: Data Sources and Definitions Our dataset builds on the cross-country dataset compiled by Persson and Tabellini (2003) (henceforth PT). Our sample of "stable democracies" consists of countries with an average Polity score greater than or equal to 5 over the period 1991-1998. We also report results with a sample containing all democracies over the period 1991-1998). With the (defined as countries with average Polity score greater than exception of the average government wage to per capita GDP (which comes from the World Bank), all our variables are from PT's dataset. These variables are the following: Central government expenditures as a percentage of GDP: constructed using the item Government Finance-Expenditures in the IFS, divided by GDP at current prices and multiplied by 100. IMF-IFS CD-Rom and IMF-IFS Yearbook. Consolidated central government expenditures on social services and welfare as percentage of GDP: from the GFS Yearbook, divided by GDP and multiplied by 100. Source: IMF-GFS Yearbook 2000 and IMF-IFS CD-Rom. Log GDP per capita: per capita real GDP defined as real GDP per capita in constant dollars (chain index) expressed in international prices, base year 1985. Data through 1992 are taken from the Penn World Table 5.6, while data on the period 1993-98 are computed from data taken from the World Development Indicators, the World Bank. These later observations are computed on the basis of the latest observation available from the Penn Word Tables and the growth rates of GDP per capita in the subsequent years computed from the series of GDP at market prices (in constant 1995 U.S. dollars) and population, from the World Development Indicators. Sources: Penn World Tables - mark 5.6 (PWT), available on Source: http://datacentre2.chass.utoronto.ca/pwt/docs/topic.html. The World Bank's World Development Indicators, available on http://www.worldbank.org. Polity: score for democracy, computed by subtracting the AUTOC score from the DEMOC score, and ranging from +10 (strongly democratic) to -10 (strongly autocratic). (DEMOC) is the index of autocracy (democracy), derived from codings of the competitiveness of political participa- AUTOC tion, the regulation of participation, the openness and competitiveness of executive recruitment, and constraints on the chief executive. Source: Polity IV Project (http://www.cidcm.umd.edu/inscr/polity/index.htm). Age structure variables: percentage of population between the ages of 15 and 64 in the total population and percentage of population over the age of 65 in the total population. Source: World Development Indicators CD-Rom 1999. GDP: mean value of the average government between 1991 and 2000. It is computed as the average of the two data points available for the periods 1991-95 and 1996-2000. When data for one of the two periods are not available, only the available time period is used. The variable is calculated by dividing the average government wage by the GDP per capita figure. The average government wage is calculated as the total central government wage bill divided by the number of employees in total central government. The total central government wage bill is the sum of wages and salaries paid to central government employees, including armed forces personnel. The number of employees in total central government is the sum of total civilian central government and the Armed Forces. Source: Schiavo-Campo, de Tommaso and Mukherjee (1997), Table A-3, Average government wage wage to per capita relative to per capita GDP http://www-wds.worldbank.org/external/default. 50 References 1. Acemoglu, Daron (2005) of Monetary Economics, "Politics and Economics in Weak and Strong States," Journal 52, 1199-1226. 2. Acemoglu, Daron (2007) "An Economic Model of Populism," 3. Acemoglu, Daron and James A. Robinson (2001) in progress. "Inefficient Redistribution," American Political Science Review, 95, 649-662. 4. Acemoglu, Daron and James A. Robinson (2006a) Economic Origins of Dictatorship and New Democracy, 5. Acemoglu, Daron and James A. Robinson (2006b) "Persistence of Power, Elites and Institutions," 6. York; Cambridge University Press. NBER Working Paper, No. 12108. Acemoglu, Daron and Thierry Verdier (1998) "Property Rights, Corruption and the Allocation of Talent: A General Equilibrium Approach," Economic Journal, 108, 1381- 1403. 7. Acemoglu, Daron and Thierry Verdier (2000) "The Choice Between Market Failures and Corruption," American Economic Review, 90(1), 194-211. 8. Aghion, Philippe and Patrick Bolton (1990) "Government Domestic Debt and the Risk of Default: A Political-Economic Model of the Strategic Role of Debt," in Capital Markets and Debt Management, R. Dornbusch and M. Draghi, 9. Alesina, Alberto (1988) "Credibility and Political eds., Cambridge, Convergence in a MA: MIT Press. Two-Party System with Rational Voters," American Economic Review, 78(4), 796-805. 10. Alford, Robert R. and Democracy, 11. York; Cambridge University Press. Voting," Public Choice, 54, 123-139. Baron, David American 13. New the State Austen-Smith, David (1987) "Interest Groups, Campaign Contributions, and Probabilistic 12. and Roger Friedland (1985) Powers of Theory. Capitalism, P. (1994) "Electoral Competition with Informed and Uninformed Voters," Political Science Review, 88, 33-47. Barro, Robert (1999) Determinants of Economic Growth: Study, MIT Press, Cambridge MA. 51 A Cross-country Empirical 14. Beck, Linda J. (1997) "Senegal's "Patrimonial Democrats": Incremental Reform and the Obstacles to the Consolidation of Democracy," Canadian Journal of African Stud- ies/Revue Canadienne des Etudes Africaines, 31(1), pp. 1-31. 15. Becker, Gary S. and Casey Mulligan (2003) "Deadweight Costs and the Size of Govern- ment," Journal of 16. Benabou, Roland (2000) "Unequal Besley, 46, 293-340. Societies: American Economic Review, tract," 17. Law and Economics, Income Distribution and the Social Con- 90(1), 96-129. Timothy and Stephen Coate (1997) "An Economic Model of Representative Democracy," Quarterly Journal of Economics, 108, 85-114. 18. Boone, Catherine (1990) "State Power and Economic Crisis in Senegal," Comparative Politics, 22(3), 341-357. 19. Centeno, Miguel (2002) Blood and Debt: War and Nation-State in Latin America, Uni- versity Park, PA.; Pennsylvania State University Press. 20. Coate, Stephen and Stephen Morris (1995) of Transfers to Special Inter- Journal of Political Economy, 103(6), 1210-1235. ests," 21. "On the Form Coate, Stephen and Stephen Morris (1999) "Policy Persistence," American Economic Review, 89(5), 1327-1336. 22. Cukierman, Alex, Sebastian Edwards and Guido Tabellini (1992) "Seigniorage and Political Instability," 23. American Economic Review, 82(3), 537-555. Debs, Alexandre (2006) "The Wheel of Fortune: Agency Problems in Dictatorships," MIT mimeo. 24. Dixit, Avinash (2002) "Incentives and Organizations tative Review," Journal of 25. Dixit, Human in the Public Sector: An Interpre- Resources, 37(4), 696-727. Avinash K., Gene Grossman and Elhanan Helpman (1997) "Common Agency and Coordination: General Theory and Application to Government Policy Making," Journal of Political 26. Egorov, Economy, 105(4), 752-769. Georgy and Konstantin Sonin (2005) "Dictators and Their Problems in Dictatorships," New Economic 52 School, Moscow, mimeo. Viziers: Agency 27. Epstein, Stepan (2000) Freedom and Growth. The Rise of States and Markets in Europe, 1300-1750, London; Routledge. 28. Etzioni-Halevy, 29. Eva (1983) Bureaucracy and Democracy, London; Routledge. Evans, Peter B. (1989) "Predatory, Developmental, and Other Apparatuses: Economy parative Political A Com- Perspective on the Third World State," Sociological Forum, 4(4), 561-587. 30. Evans, Peter B. (1992) "The State as Problem and Solution: Autonomy, and Structural Change," 31. in Predation, Embedded Stephan Haggard and Robert R. Kaufman Economic Adjustment: International Constraints, The Politics of and the State, Princeton; Princeton University Press. (eds.), Distributive Conflicts Evans, Peter B. (1995) Embedded Autonomy, Princeton; Princeton University Press. 32. Fiorina, Morris P. and Roger G. Noll (1978) "Voters, Legislators and Bureaucracy: American Economic Review, stitutional Design in the Public Sector," Game 68(2), 256-260. MIT 33. Fudenberg, Drew and Jean Tirole (1991) 34. Gay, Robert (1990) "Popular Incorporation and Prospects for Democracy: Theory, Cambridge; In- Press. Some Impli- cations of the Brazilian Case," Theory and Society, 19(4), 447-463. 35. Geddes, Barbara (1991) "A American racies," 36. Glazer, Game Theoretic Model of Reform in Latin American Democ- Political Science Amihai (1989) "Politics Review 85(2), 371-92. and the Choice of Durability," American Economic Re- view, 79(5), 1207-1213. 37. Golden, Miriam A. (2003) "Electoral Connections: The Effects of the Personal Vote on Political Patronage, Bureaucracy and Legislation Political Science, 33, pp. 38. On Postwar Italy," British Journal of 189-212. Gomes, Armando and Philippe Interactions: in Jehiel (2005) "Dynamic Processes of Social and Economic the Persistence of Inefficiencies," Journal of Political Economy, 113(3), 626-667. 39. Gramsci, Antonio (1971) Selections from the Prison Notebooks, Publishers. 53 New York; International 40. Grossman, Gene M. and Elhanan Helpman (1996) "Electoral Competition and Special Interest Politics," Review of Economic Studies, 63(2), 265-286. 41. Hassler, John, Jose V. Rodriguez Mora, Kjetil Storesletten and Fabrizio Zilibotti (2003) "The Survival of the Welfare 42. Herbst, Jeffery I. State," (2000) States and American Economic Review, Power in Africa: 93(1), 87-112. Comparative Lessons in Authority and Control, Princeton NJ; Princeton University Press. 43. Kohli, Atul (2004) State- Directed the Global Periphery, 44. Levi, New Development: Political Power and Industrialization in York; Cambridge University Press. Margaret (1989) Of Rule and Revenue, Berkeley and Los Angeles CA; University of California Press. 45. Lizzeri, Alessandro and Nicola Persico (2001) "Provision of Public Goods under Alter- native Electoral Systems" American 46. Migdal, Joel (1988) Strong Societies Economic Review, 91(1), 225-239. and Weak States: State-Society Relations and State Capabilities in the Third World, Princeton; Princeton University Press. 47. Miliband, 48. Ralph (1969) State Osborne, Martin J. and Al. in Capitalist Society, Slivinsky (1996) New York, Basic Books. "A Model of Political Competition with Citizen-Candidates," Quarterly Journal of Economics, 111(1), 65-96. Roberto (1996) "Growth, Income Distribution, and Democracy: 49. Perotti, Say," Journal of 50. Economic Growth, What the Data 1(2), 149-87. Persson, Torsten, Gerard Roland and Guido Tabellini (2000) "Comparative Politics and Public Finance," Journal of Political Economy, 108(6), 1121-1161. 51. Persson, Torsten and Guido Tabellini (2003) The Economic Effect of Constitutions, bridge MA; The MIT 52. Persson, Run a Press. Torsten and Lars E. O. Svensson (1989) Deficit: "Why a Stubborn Conservative would Policy with Time-Inconsistent Preferences," Quarterly Journal of Eco- nomics, 104(2), 325-345. 53. Poulantzas, Cam- Nicos (1978) Political Power and Social Classes, London, Verso. 54 54. Rauch, James and Peter B. Evans (2000) "Bureaucratic Structure and Bureaucratic Performance in Less Developed Countries," Journal of Public Economics, 55. Robinson, James A. (2001) "When 75, 49-71. a State Predatory?," Unpublished manuscript, is Harvard University. 56. Robinson, James A. and Ragnar Torvik (2005) "White Elephants," Journal of Public Economics, 89, 197-210. 57. Economy Rodrik, Dani (1995) "The Political Rogoff (eds.), Handbook of Trade Policy," in G. of International Economics, vol. 3, Grossman and K. Amsterdam; North-Holland. 58. Roett, Riordan (1999) Brazil: Politics in a Patrimonial Society, Westport, CT: Praeger. 59. Rothstein, Bo and Eric M. Uslaner (2005) "All for All: Equality, Corruption, and Social Trust," World Politics, 58(1), 41-72. 60. Saint-Paul, Gilles (1996) tions," 61. Economic "Exploring the Political Economy of Labour Market Institu- Policy, 11(23), 265-315. Schiavo-Campo, Salvatore, Giulio de Tommaso and Amitabh Mukherjee (1997) "An International Statistical Survey of Government Employment and Wages," The World Bank, Policy Research Working Paper 1806. Andrei and Robert 62. Shleifer, W. Vishny (1994) "Politicians and Firms," Quarterly Journal of Economics, 109(4), 995-1025. 63. Tabellini, Guido and Alberto Alesina (1990) "Voting on the Budget Economic Review, 64. it Rules? State Appara- Davide and Andrea Vindigni (2005) "Endogenous Constitutions," University Pompeu Fabra, Working Paper 896. 66. Tilly, Charles (1990) Coercion, Capital 67. Do When and State Power under Feudalism, Capitalism and Socialism, London; Verso. 65. Ticchi, MA; American 80(1), 37-49. Therborn, Goran (1980) What Does the Ruling Class tuses Deficit," and European States, AD 990-1990, Cambridge Blackwell. Van Rijckeghem, Caroline and Rate of Temptation: Beatrice Do Wages Weder (2001) "Bureaucratic Corruption and the in the Civil Service Affect Corruption, Much?," Journal of Development Economics, 65(2), 307-331. 55 and by How 68. Weyland, Kurt (1996) "Obstacles to Social Reform in Brazil's New Democracy," Com- parative Politics, 29(1), 1-22. 69. Wilson, John D. (1990) "Are Efficiency Improvements in Government Transfer Policies Self-Defeating in Political Equilibrium," Economics 70. World Bank (2004) Doing Business in the World: Bank, Oxford University Press. 56 and Politics, 2, 241-258. Understanding Regulation, World e ° a, Q S a 5 o a £P OO q o _; a X CJ 0; O C CJ cn — — O I CO r<-) 9© CO CN CN CN r- so "* d Ch OS CN (1) 00 i/"l -^^2 CO e E cj ca CJ > O ,"3 OX) <L) i r> iff oo T) J-J ca ^ cn g :r~ OS d C "* * c CJ aj 03 W5 00 o e > ,> aua 11 CJ os£ m so d ^ ca B, 1) Q o «5 CU S^ -. °° OO To -a T3 a c * a. CD W s 0> O r- OS d 9.6, —• IT) 00 OS <N Ov — o ; OJ CU S go? °^ CN ° cn £v C a 3 OJ c? 5 B O CD o <*-i ^1- oo — CN CN O -h SO ,- in ,&M OJ s k. cu > o a > CJl n g OX) d cfl •-* 1; CJ CJ -XJ a, u > cj K S 3 u a.-a M E =3 T3 E S3 O 6J3 S3 T3 C CJ Oh <* o CN d S3 1- U s S3 o 3 H k S CQ -n B o OJ ys >h C3 CJ at a CO CJ B O >i CJ SO £ 1-1 CJ Oh -O > o O a o a o C3 ca 2w S-"o a a. o CJ Q "3 o o a, u-i -= o H cj S2 ca cj ^ CJJ^t- D. ca a J3 cs H CJ c o cj Oh Oh m — u CJ) O so T3 ca a, <4H o CJ C > bo C3 a CJ o 1H CJ a, a! i> — Mh QQ £ o a G a o Redistribution o CN and Public Sector Relative Wage: Unconditional Relationship • Hungary D. Q O • Netherlands • Italy • Israel • Belgium • Bulgaria # p ortuga co CD | • Botswana •fiBfeaSd • Greece CD • RomanBGermarTy • Latvia • South Africa • Estonia CD Nicaragua Q. X W — - anad h C %Tu>; f^ritius • Pakistan • Venezuela •Philippines 0) > outnKorea o .Ecuafl<£ o CNJ i -2-1012 1 1 1 Wage to Average Government coef = -4.9601277, (robust) se = 1.0127675, Figure The • Honduras • BolWShile* ' t •Thailand India 1 Per Capita GDP = -4.9 between average government and central government expenditures as percentage of GDP for the sample of countries with an average polity index for the period 1991-98 greater than or equal to 5 (see text for details). wage 1. figure reports the fitted values of the unconditional relationship GDP to per capita Redistribution o CM • 0. Q and Public Sector Relative Wage: Conditional Relationship Hungary • Netherlands • §l<te ftepub(Wlsfi5ael •Belgium „ «n i- j • Pojppdanct Botswana •Nicaragua • Sweden. •Estonia . Qri ffirtugal • Fiji i • Romania , CD CD • South 3 Afric • Honduras • Greece CD Q. X »JUruguay' _ •T Turkey M ^_ • Germany LU lb E CD > O « pm^fe Mallyfi&lg,; mpi^r^M^ea • ' Ecuador • Thailand ifcCanada • Colornfe^-gentina • USA O o CM -2-1012 — 1 1 1 1 I Average Government coef = -3.6604963, (robust) se = 1.0318145, Figure The t Wage to Per Capita GDP = -3.55 between average government and government expenditures as percentage of GDP for the sample of countries with an average polity index for the period 1991-98 greater than or equal to 5. The control variables are the log of per capita real GDP and the average polity index (see text for deta ils) wage 2. figure reports the fitted values of the conditional relationship to per capita GDP . 60(H '33