Boston Library Consortium Member Libraries

Transcription

1

2

3 Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries

4

5 IHr ) working paper department of economics A Theory of Poitica Transitions Daron Acemogu James A. Robinson massachusetts institute of technoogy 50 memoria drive Cambridge mass

6

7 WORKING PAPER DEPARTMENT OF ECONOMICS A Theory of Poitica Transitions Daron Acemogu James A. Robinson No October 1999 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 50 MEMORIAL DRIVE CAMBRIDGE MASS

8

9 A Theory of Poitica Transitions Daron Acemogu^ James A. Robinson* Juy Abstract We deveop a theory of poitica transitions inspired in part by the experiences of Western Europe and Latin America. Nondemocratic societies are controed by a rich eite. The initiay disenfranchised poor can contest power by threatening socia unrest or revoution and this may force the eite to democratize. Democracy may not consoidate because it is more redistributive than a nondemocratic regime and this gives the eite an incentive to mount a coup. Because inequaity makes democracy more costy for the eite highy unequa societies are ess ikey to consoidate democracy and may end up osciating between regimes or in a nondemocratic repressive regime. An unequa society is ikey to experience fisca voatiity but the reationship between inequaity and redistribution is nonmonotonic; societies with intermediate eves of inequaity consoidate democracy and redistribute more than both very equa and very unequa countries. We aso show that asset redistribution such as educationa and and reform may be used to consoidate both democratic and nondemocratic regimes. Keywords: democracy dictatorship inequaity poitica instabiity redistribution. JEL Cassification: D72 D P16. *We woud ike to thank Abhijit Banerjee Francois Bourguignon Ruth Coier Steve Durauf Jeff Mariano Tommasi Jaume Ventura and seminar Frieden Michae Kremer Dani Rodrik Ken Sokooff participants at MIT Wisconsin NYU Western Ontario Tououse NBER Summer Institute Berkeey and Yae Poitica Science Departments the conference on "Asset Inequaity and Poverty" at the Ministry of Land Reform in Brasiia and LACEA 98 at the Universidad di Teia for usefu suggestions. t Massachusetts Institute of Technoogy Department of Economics E Cambridge MA 02319; e-mai: daronomit. edu ^University of Caifornia Berkeey Department of Poitica Science and Hoover Institute Stanford University; e-mai: jarobins@usc.edu. 1

10 "MASSACHUSETTS INSTITUTE OF TECHNOLOGY

11 1 Introduction Most Northern European countries extended the franchise during the ate 19th and eary 20th centuries and succeeded in consoidating mass democracy. For exampe in Britain foowing the first tentative reforms of 1832 voting rights were significanty extended in 1867 and in They were further expanded in 1919 when universa mae suffrage was introduced and in 1928 when a women were aowed to vote. There were no reversas in this process of democratization. Athough many ess deveoped countries most notaby those in Latin America aso became democratic during the ate 19th and eary 20th century they quicky reverted to nondemocratic regimes. 1 The recent history of many Latin American countries is therefore marred by osciations in and out of democracy. In Argentina for exampe universa mae suffrage became effective in But it was soon overthrown by a coup in Democracy was re-instated in 1946 but fe to a coup in 1955 re-created again in 1973 subverted again in 1976 and finay re-instaed in Why has mass democracy been durabe in many Northern European countries and why has it been so hard to consoidate this set of poitica institutions in ess deveoped countries such as those in Latin America? This paper provides a framework for anayzing this question. We emphasize that in democratic societies the poor impose higher taxes on the rich than in nondemocratic societies. This makes the poor pro-democratic whie simutaneousy giving the rich an incentive to oppose democracy. 2 In nondemocratic societies the poor are excuded from poitica power but shocks for exampe wars or depressions may create a revoutionary threat (or at east costy socia unrest). 3 The rich (eite) wi try to prevent revoution before the mass democratization of the 19th century Britain had eections with a very restricted franchise whie in Argentina nondemocratic regimes have often been miitary dictatorships. We do not distinguish between these different types of non-democratic regimes. We aso define any significant move towards mass democracy as "democratization". 2 For exampe Rodrik (1999) shows that democracies tend to have higher wages and higher abor share. In the context of Latin America there are many exampes of miitary coups specificay aimed at reducing redistribution (see Skidmore 1967 Smith 1978 Stepan 1978 and Waerstein 1980). For exampe in Argentina the abor share which increased rapidy during the democratic regime of fe during the foowing dictatorship (e.g. Di Teia and Dornbusch 1989 p. 6). There were simiar increases in inequaity in Brazi after 1964 when the eft-wing President Gouart was deposed and in Chie the coup of 1973 was aimed in part at stopping and reversing the increasingy redistributive poicies of Aende. 3 Haggard and Kaufman (1995) document the importance of shocks and crises in regime changes. In Acemogu and Robinson (1997) we provide evidence on the importance of wars and depressions in triggering socia unrest from 19th century Europe.

12 by making concessions to the poor for exampe in the form of income redistribution. Because the threat of revoution is often ony transitory current redistribution does not guarantee future redistribution. If this temporary redistribution is insufficient to prevent a revoution as it wi be in a very unequa society the eite wi be forced to make a credibe commitment to future income redistribution. This is what extending voting rights achieves by changing the identity of the future median voter. Democracies are not necessariy permanent because the eite may have an opportunity to mount a coup. The poor woud ike to commit to ow eves of future taxation to prevent this. But since such commitments may not be credibe the eite may prefer to retake power. They are more ikey to do so when due to high taxes democracy is reativey costy for them. Taxes wi be high in turn when inequaity is high. So inequaity destabiizes democracy by making coups more attractive for the rich eite. As a resut a highy unequa society is ikey to fuctuate in and out of democracy. 4 More generay poitica instabiity is more ikey when the stakes economic gains from controing poitica power are greater a view that can be traced to James Madison (1788). In consoidated democracies such as the OECD economies the threat of coups is not important so taxes are determined by the usua trade-off for the median voter between transfers and deadweight osses. There is itte or no variabiity in the amount of redistribution. In contrast in highy unequa economies fisca poicy is more voatie because as a society fuctuates between different regimes the amount of fisca redistribution changes. 5 Even if an unequa society prevents regime changes the threat of coups wi often imit taxes. More generay whie greater inequaity in a consoidated democracy increases redistribution (e.g. Metzer and Richards 1981) an unequa society is ess ikey to be in the more redistributive fuy-consoidated democratic regime. 6 4 The connection between inequaity and poitica transitions may account for why many Latin American countries such as Argentina Brazi Peru and Uruguay have had difficuty in consoidating democracy whie the more equa European countries and Costa Rica have had more stabe democracies. It is aso in ine with the correation between inequaity and poitica instabiity documented by Muer and Seigson (1987) and Aesina and Perotti (1996). This accords with some of the patterns observed in the data. For exampe Gavin and Perotti (1997) show that fisca poicy in Latin America is much more variabe than in Europe. They report that the standard deviation of the growth of transfers in Latin America between 1970 and 1995 was 22.4% as compared to 4.4% for industriaized countries. 6 This may expain the ack of a we-defined monotonic reationship between inequaity and redistribution in the data (e.g. Perotti 1996). Our mode is aso consistent with the consensus view that "popuist" poicies are a resut of high eves of inequaity and suggests an expanation for why these poicies are often reversed radicay by regime changes (e.g. Sachs 1990): popuist poicies arise as a method of radica redistribution when the poor majority reaize that democracy is not consoidated. Countries that consoidate democracy such as Costa Rica and Coombia are therefore ess ikey to foow popuist poicies. For exampe Kaufman and Staings (1991 p. 27) write "estabished democracies (Venezuea

13 Our basic anaysis suggests that high eves of inequaity induce frequent regime changes. When in addition to making concessions the eite can use a repression strategy this concusion needs to be modified. Very high eves of inequaity which make democracy costy for the eite may encourage them to use repression to prevent democratization. Therefore ow eves of inequaity are necessary for a democracy to consoidate and high inequaity is ikey to ead to some form of poitica instabiity (either regime changes or socia unrest suppressed by repression). Since the incentives to engage in or avoid fisca redistribution which are generated by underying asset inequaity are a key factor in shaping regime dynamics there is an important roe for asset redistribution e.g. educationa and and reforms. Educationa reforms that increase the earnings capacity of the poor and and reforms that achieve a more egaitarian distribution of assets may consoidate democracy because they are difficut to reverse reduce the preferred amount of fisca redistribution by the median voter and make democracy ess costy for the eite. 7 There is a danger in radica reforms however as their anticipation may make a coup more ikey as in Guatemaa in 1954 Brazi in 1964 and Chie in Athough the reasons for the changes in regimes are numerous confict between different socia groups appears to be important in practice. In Acemogu and Robinson (1997) we presented evidence suggesting that in Britain France Germany and Sweden democratization was in arge part a response to the threat of revoution and socia unrest. In Latin America many instances of democratization appear to have been driven by the same factors. 8 Our paper is reated to the anayses of the poitica economy of redistribution (e.g. Metzer and Richards 1981 Persson and Tabeini 1994 and Aesina and Rodrik 1994 Benabou 1999) and to modes of socia confict (e.g. Roemer 1985 Grossman 1991 Torne and Veasco 1992 Ades and Verdier 1995 and Benhabib and Rustichini 1995). There is a arge poitica science iterature on democratization starting with the work of Coombia and Costa Rica in our study) were aso associated with orthodox macro poicies... It was the transitiona democracies (Peru Argentina and Brazi) that foowed popuist poicies." 7 This may hep to expain why democracy was consoidated in Costa Rica after 1948 and Venezuea after 1958 two countries that instituted and and educationa reforms foowing democratization. 8 Coier (1998) has recenty argued that socia pressure from the masses was the driving force behind the most recent democratizations in Peru Uruguay and Brazi. Simiary in Argentina both the 1912 reforms which institutionaized universa mae suffrage and the withdrawa of the miitary in 1973 which reinstated Peron appear to have been a response to increasing socia unrest (see Rock 1987 Ch. 8 and circa p. 188). In Venezuea the democratizations in 1945 and 1958 both happened in the context of a widespread uprising (see e.g. Levine 1989 p. 256 and Kob 1974 p. 175). The pattern aso appears to be quite simiar in Guatemaa and E Savador (see for exampe Paige 1997).

14 Lipset (1959) and Moore (1966) which emphasizes the structura determinants of democracy (such as income eve and cass composition). More recent work has focused on the strategic interaction between regimes and their opponents and on poitica rather than economic factors (e.g. Rustow 1970 O'Donne and Schmitter 1986 Przeworski 1991 Linz and Stepan 1996). Therborn (1977) and Rueschemeyer Stephens and Stephens (1992) share our emphasis on the actions of the disenfranchised poor as the key to understanding democratization and in our previous work Acemogu and Robinson (1997) we aso emphasized democratization as a commitment to future redistribution. The iterature on coups is much ess deveoped and focuses mosty on how purey poitica factors expain the persistence or coapse of democratic poitics (for exampe Dah 1971 and Linz 1978). This contrasts with our focus on socia confict and redistribution (though O'Donne 1973 aso discusses the importance of socia confict in the context of Latin American deveopment strategies). The paper proceeds as foows. In Section 2 we present our basic mode of poitica deveopment and study the determinants of transitions between regimes. In Section 3 we consider a simpe extension of the mode to discuss how the eite may use repression to avoid democratization. In Section 4 we extend the basic mode to consider asset redistribution. In Section 5 we discuss how the possibiity of making transfers groupspecific increases poitica instabiity. Section 6 concudes. 2 The Basic Mode There are two groups of agents: the poor and the rich (the eite). The poitica state can be democratic or nondemocratic. In a democracy the median voter sets the tax rate and because the poor are more numerous the median voter is a poor agent. In a nondemocratic regime taxes are set by the rich. When the poitica system is nondemocratic the poor can attempt a revoution and the eite decides whether to estabish democracy. When the system is democratic the rich can mount a coup. The costs and benefits of coups and revoutions are stochastic capturing the notion that some periods such as those foowing wars or depressions may be more conducive to socia and poitica unrest. This aso enabes us to mode the fact that those in power cannot commit to future tax rates which wi be determined in future poitica equiibria.

15 2.1 The Environment We consider an infinite horizon economy with a continuum 1 of agents. A proportion A of these agents are "poor" whie the remaining 1 A form a rich "eite". Throughout the paper superscript p denotes poor agent and r denotes rich agent (or member of the eite). We wi treat a poor agents as identica and a members of the eite are aso identica. Initiay poitica power is concentrated in the hands of the eite but A > so that if there is fu democracy the median voter is a poor agent. There is a unique consumption good y and a unique asset with tota stock h (which can be thought of as physica or human capita or and). We begin our anaysis of the economy at time t = where each poor agent has capita h p and each member of the eite has h r > h p. These capita stocks are exogenous. To parametrize inequaity et h r = (1 9)h/{\ A) and h p = 9h/\ where A > 6 > 0 so that a ow eve of 6 corresponds to higher inequaity. Without oss of any generaity we aso normaize the aggregate stock of capita h to 1. The fina good is produced from capita and tota output of an agent is y\ = h for i = pr. A agents have identica preferences represented by E t J2'jLoP t+j c+ j for i = pr where /5 is the discount factor and E t is the expectations operator conditiona on a information avaiabe at time t. Post-tax income is given by y\ = (1 r t )h z + T t where Tt > is the tax rate on income and T\ > is the ump-sum transfer that an agent of group i receives from the state. For now we simpify the anaysis by assuming that taxes are inear and transfers cannot be person specific hence T\ = T t. Group-specific transfers wi be discussed in Section 5. We aso assume that it is costy to raise taxes: at tax rate T t there is a deadweight cost of C(r t ) where C is twice continuousy differentiabe with (7(0) = 0 C"(0) = 0 C'(t) > for a r > 0 and C" > 0. If there were no costs of taxation our genera resuts woud not be atered but some of the comparative statics woud not appy when the tax rate is at the corner i.e. r = 1. To avoid keeping track of this case we assume C'{r) > and C"() = oo which ensure an interior tax rate. The government budget constraint impies T t = r t [Xh p + (1 A)/i r ] C (r f ) = r t C(r t ). The A poor agents are initiay excuded from the poitica process but can attempt a revoution in any period t > 1. We assume that if a revoution is attempted and a fraction < 1 of the poor take part it aways succeeds. However in the course of the revoution a proportion 1 [i > of the capita of the economy is destroyed. After a revoution poor agents who have participated in the revoution take contro of the output fi and share it among themseves. Therefore when a poor agents take part in the revoution each

16 receives fi/x in every period whie the rich receive nothing. A ow vaue of fj impies that a revoution is reativey costy. We assume that // is stochastic and changes between two vaues: h h (j. and \i = with Pr(/x t = \i ) = q independent of the previous reaization. We assume that q < 1/2 so that the threat of revoution is reativey rare. In a democracy the eite have no specia voting power (one-person-one- vote) but they can attempt a coup. After a coup the poitica situation reverts back to the initia status quo with the eite controing poitica power. We ignore the free rider probem among the eite and assume that the cost of a coup for each member of the eite is <p where <p G {ip h <p 1 } <p h > (p > and Px(<p t = <p ) = s. Assume that s < 1/2 so that the coup threat ike the threat of revoution is reativey rare. Ony in state if 1 wi the rich want to mount a coup so we et (p h > +oo. If a coup is mounted then \x = fj 1 at first so that there is no revoution immediatey. Simiary if democratization occurs then democracy starts with the coup cost at <p h impying that a democracy has at east some window of opportunity before a coup can occur. Finay in each nondemocratic period the eite have to decide whether or not to extend the franchise. If it is extended then the economy becomes a democracy and the median voter a poor agent sets the tax rate. The timing of events within a period can be summarized as foows. 1. the state // or <p is reveaed. 2. the poor set the tax rate r if we are in a democracy and the rich set r otherwise. 3. in a nondemocratic regime the rich decide whether or not to extend the franchise. In a democracy they decide whether to mount a coup. If they extend the franchise or a coup takes pace the party that comes to power decides whether to keep the tax t set at stage 2 or set a new tax rate 4. the poor decide whether or not to initiate a revoution. If there is a revoution they share the remaining output of the economy. If there is no revoution the tax rate decided at 2 or 3 gets impemented. 5. incomes are reaized and consumption takes pace.

17 2.2 Anaysis The tax rate r m which maximizes the indirect utiity of a poor agent in the absence of a coup threat satisfies r m = argmax ((1 - r)hp + [t-c (t)]> where the term in square brackets is the ump-sum transfer T. The first-order condition C\r m ) =^ (1) of this probem gives where we used the fact that h p = 9/X. (1) impies that r m is uniquey defined and decreasing in 0. When inequaity is higher the maximum tax rate is aso higher and when 6 = A so that h r = hp we have r m = 0. Hence in the case of compete equaity the median voter sets a zero tax rate and there is no redistribution. A poor agents have the same preferences and when it comes to whether or not to participate in a revoution there is no "free-rider probem" because if an agent does not take part in the revoution he can be excuded from the resuting redistribution. So we can treat a poor agents as one payer. Aso a members of the eite have identica preferences and we can aso treat them as one payer in a repeated game. This economy can therefore be represented as a repeated game between the eite and the poor. We wi characterize the pure strategy Markov Perfect Equiibria of this game in which strategies ony depend on the current state of the word. The state is one of (tp D) ((p h D) (/j E) or (fi h E) where E denotes eite in power (nondemocratic regime) and D denotes democracy. The actions of the eite consist of a decision to extend the franchise 7 = 1 in state E and a tax rate re when 7 = (i.e. when franchise is not extended). Ceary if 7 = 0 the state remains at (-E) and if 7 = 1 it switches to (. D). If the state is {.D) the eite choose = 1 i.e. they mount a coup or ( = 0 no coup. Foowing a coup the eite aso make a taxation decision. The actions of the poor consist of a decision to initiate a revoution p when the state is (.E) and a tax rate r d when the state is (.D). In state {-E) the actions of the poor are conditioned on the current actions of the eite who move before the poor according to the timing of events above whie in state (.D) the actions of the eite are conditioned on the actions of the poor. Then a pure strategy Markov Perfect equiibrium is a strategy combination denoted {a p a r } such that 9 There coud be a coordination probem whereby a poor agents expect others not to take part in a revoution so do not take part themseves. However since taking part in a revoution imposes no additiona costs irrespective of whether it succeeds or not it is a weaky dominant strategy and we therefore ignore this coordination probem.

18 a p and a r are best-responses to each other for a possibe states. We wi characterize the Markov Perfect Equiibria by writing the appropriate Beman equations. In a democracy the median voter is a poor agent (since A > 1/2). Let V ((ph D) be the vaue of an agent of type i = pr when there is democracy and when the cost of mounting a coup is h (p. Simiary et V % (ip ) be the vaue of agent i when the cost is ip (in which case there may be a switch to a nondemocratic regime as a resut of a coup). When the state is (ip h D) there are no constraints on the median voter so he wi choose the tax rate r m. The returns to poor and rich agents are: VV D) = h { + 8\e) +P[(1- s)vv D) + svv)]. (2) where i = pr and 6 (9) is the net amount of taxation that a person of type i receives when the tax rate is r m. Hence 8 {6) = T m r m h % and from the assumption that the budget is baanced T m = T m -C(T m ). Note that 6 r (9) < < 6 P (6) and ^ = ^ > whie J} ' = ^ < by the Enveope Theorem. Intuitivey higher inequaity raises the tax rate on the rich whie simutaneousy increasing net transfers to the poor. V(</?') is the vaue to agent i in state (ip D). In this state the poor may reduce the tax rate beow r m say to a eve r d in an attempt to prevent a coup yieding vaues where i V(^ D r d ) =ti + A {9) +p[(- s)v (ip h D) + sv^ip 1 D r d )]. (3) = pr and A (9) =T d - r d h\ Ceary A*(0) < 8 P {6) and A r (8) > 8 r (9) because r d < T m. After observing the tax rate r d the eite decide whether to mount a coup so VV) = max (C [VV E) - if 1 } + (1 - C) VV D rd )) (4) ytf) = CV(^ ) + (i-c)v(^aa 1 where reca that ^ = 1 impies a coup. A coup wi take the society to the state (p where the eite are in power and there is no threat of revoution thus they set r = 0. E) Hence for % = p or r we have: VV E) = ti + (3 [(1 - q)tv E) + SfVV E)\ (5) A coup in state (ip D) ( = 1 is therefore optima for the eite if V r (p E)~V r (^Dr d )><p (6) This is the coup constraint: a coup occurs if the gain to the rich of capturing poitica power and reducing taxation is greater than the cost of the coup <p. 8

19 We can first determine a critica vaue of the cost of coup (p(6qs) such that as ong as ip > (p(9qs) a coup is never beneficia for the rich even if the poor continue to tax at the rate r = r m in state (<p D). This critica vaue is given by (p(9qs) = V r {fj r (<p 1 E)-V D r m ) where V{(ft D r m = h 0) ) is the return to the rich of aways "+-f remaining in democracy with a tax rate r = r m. Imposing (5) 10 we obtain this critica vaue as: When <p > <p(6 q s) the coup threat does not pay a roe and democracy is fuy consoi- The tax rate r m is aways determined by the usua trade-off for the median voter dated.... -ps r (e)(i-q) ^ Q^ S)= -P(-q) V r (fjh E) = V r (tp h D) in (7) baancing transfers against the deadweight osses of taxation. Observe that dg < 0 which impies that a ess unequa society is more ikey to achieve a fuy consoidated democracy. This is intuitive since a greater eve of inequaity makes democracy ess attractive for the rich. We can next determine the vaue of the cost of coup 7p(6qs) such that if ip > </?(# q s) the poor can stop a coup by setting a ow enough tax rate in the state (D <p (or conversey when ip < Tp(9qs) even a poicy of setting r d = does not stop a coup). Since the owest tax rate that the poor can set is r = 0 p(8qs) is given by V r (ti E)-V r (<p DT d = 0)=p(6qs). Combining (2) and (3) and setting V r (ip ) = V r (<p Dr d ) we can cacuate the vaue of aways remaining in democracy for the rich. Prom this we define V r (ip DT d = 0) = + \Zb as the maximum vaue the median voter can crediby commit to give to a rich agent under democracy. Now soving (2) and (5) with V r (fi h E) = V r (cp h D) and V r {<p ) = V r {fi E) - <p\ we obtain:. -(38 r (6)(-s-q) ) Expression (8) is intuitive. Reca that S r (6) < and 1 s q > 0 so 7p(6qs) is decreasing in q and s. If q is high then eite contro foowing a coup wi be short ived because a revoutionary threat wi reoccur quicky. This reduces the expected benefits from a coup. Simiary if s is high the coup constraint binds reguary and because in this state the rich pay reativey ow taxes democracy is ess costy to them. Aso ceary Tp(9qs) < (p(dqs). 10 Note that V r ((i h E) = V r {f h ) D) appies aong the equiibrium path. Since the economy starts with the eite in power we can ony be in a democracy because the franchise was extended. Therefore if a coup takes pace aong the equiibrium path democracy wi occur again in the future when fi = (i h.

20 More important for the focus of the paper is that qq < 0: higher inequaity increases the threshod Tp(Q q s) and makes a coup more ikey because in an unequa society the rich ose more under democracy. If ip > p(6qs) then democracy is semi-consoidated: the poor can avoid a coup by reducing the tax rate beow r m in state (</?' D) and setting r = r d such that V r (fi E) ip = V r ((f D r d ). Athough the society aways remains democratic the threat of a coup is sti important and infuences taxes: the tax rate r d is ess than r m woud have set in the absence of this threat. which the poor This tax rate r d is given by the equation (-p(-s-q))a r {6) + (3{-s-q)8 r {6) + <p {-p(-q)) = 0 where A (9) =T d -r d ti. Impicit differentiation shows that r d is increasing in 6: reduces the tax rate that is required to prevent a coup. 11 dr d jdq > 0 so higher inequaity If ip < fp(0 q s) even a strategy of setting r = by the poor wi not prevent a coup. In this case the society wi revert back to a nondemocratic regime when ip = <p. The poor woud ike to prevent such an outcome and if they coud they woud promise ower tax rates in the future. However such promises are not credibe because future tax rates are determined in future poitica equiibria and once the threat of coup disappears the tax rate wi rise back to r m. Forward-ooking eites reaizing this prefer a coup even though this is a costy outcome for the society. h Now consider the state (fx E). If the poor did not attempt a revoution in this state the eite woud stay in power forever and set r = 0 so the poor woud receive 1 utiity equa to j^s. In contrast with a revoution in state fi = y! they woud obtain V P (R) = x mq\ the per-period return from revoution for the infinite future discounted h to the present. Notice that ony the vaue of fj at the time of the revoution matters hence the per-period return is constant over time (and this aso impies that in the state H = fj = 0 a revoution wi never occur). We now assume Assumption 1: //* > 6 which ensures that when // = (J 1 the revoution threat is binding. In case of a revoution the rich ose everything i.e. V r (R) = 0. They wi therefore attempt to prevent it at a costs. They can do this in two different ways. First they can extend the franchise 7=1 giving the poor their return under democracy V p (cp D). 11 Notice that this equation can be written as (1-0(1-8- q)) [^^ - C(rd )\ + [^f - ^1] 0(1 - s - q) + <p ( - (3(1 - q)) = 0 where r is the maximum tax rate defined by (1). Differentiating and noting that d6 p (9)/d9 < and dr m /d8 < 0 we obtain dr d /d9 > 0. 10

21 Second they can choose to maintain poitica power 7 = 0 but redistribute through taxation. In this case the rich impose a tax rate r e where and give the poor return V p h E r e (fj ) ^(/A T e ) = ti + rf{9) +0 \qv\ii h E^) + (1 - g)vv )] (9) and rf(6) = T e r e h. So the poor receive net income (1 r e )h p from their own earnings and transfer T e r e C (r e ) giving them income of hp + rf(9). If next period we are k sti in state fi = fx then redistribution continues. But if in the next period the economy switches to jj = fj redistribution stops. This captures the notion that the eite cannot commit to future redistribution uness the future aso poses an effective revoution threat. Aso note that r e < r m that is the eite wi not tax themseves at a rate higher than r m since this is the rate that maximizes redistribution to a poor agent. If this tax rate is not sufficient to stop a revoution then no tax rate r e G [0 1] wi do so. With either democratization or redistribution by the eite the poor may sti prefer a revoution. Thus given the actions 7 and r e of the eite the vaue to the poor in the state (n h E) is V p ( f j h E)=m^(v p (R)^V p (ip h D) + (- 1 )V p ( f ih ET e )). Combining (5) and (9) we cacuate the maximum utiity that can be given to the poor without extending the franchise: vy^-o.^ 1 -^ -*(«> 1 \j where we used the fact that to give maximum utiity to the poor the eite woud set r = r m hence rf (9) = 6 (9). The poor compare (10) to V P {R). This defines a critica vaue of fji h (10 ) Ji(eq) = \(-(3(-q))6 p (6) + (11) such that V p (]J(6q)ET - r m ) = V P (R). For 1 > fi h > J(9q) a revoution is so attractive for the poor in state fj h that even the maximum amount of redistribution by the rich cannot stop it. Democratization is therefore the ony option eft to the eite. Notice aso that ^^ = 1 + A( - (3(1 - q))^1 > 0 so high inequaity reduces the revoution threshod because the poor are worse off in a nondemocratic regime. (/i / For fih < ~j2(9 q) democratization can be avoided by redistributing to the poor in state E). In this case the tax rate that the eite have to set in order to avoid revoution is 11

22 r = r e such that V p h ((j ET e ) = V P h (R). which is increasing in fj and decreasing in 9 (i.e. increasing in the eve of inequaity). 12 Finay we restrict attention to the area of the parameter space where democratization prevents a revoution that is V p (ip h D) > V P (R). Since democracy is not necessariy an absorbing state the vaue function V p h ((p D) takes into account the future possibiity of coups. The vaue to the poor of a semi-consoidated democracy is higher than that of a democracy subject to coups so it suffices to ensure that the vaue to the poor of an unconsoidated democracy is greater than V r (R). Soving (2) and (5) simutaneousy with V p (n h E) = V p (ip h D) and V p (ip ) = V p {fx E) we obtain this sufficient condition as: 13 Assumption 2: (1 _ (3 {1 _ s _ q))9 + A ( - 0(1 - p q ))8 (9) h ~ 1-/3(1---?) " "' "" / Now we can estabish the foowing resut (proof in the text): Proposition 1: Suppose Assumptions 1 and 2 hod and the society starts in a nondemocratic regime. Then: 1. If fi h < Ji(6 q) then the society remains nondemocratic. When fj = /i t = 0 and there is no redistribution. When // h = fi r r e where V p (n h Er e = ) V P (R). A higher eve of /jh or ower 9 (higher inequaity) increase both the amount of redistribution and the variabiity of fisca poicy. 2. If fjh > j(9 q) and <p > (p(9 q s) then we are in a fuy consoidated democracy. The society switches to democracy the first time /tx = // and h remains democratic thereafter and taxes are aways given by r = r m. 3. If h fi > ~p>{9q) and </?' < (p(9qs) but <p >!p(9qs) then we are in a semiconsoidated democracy. The society switches to democracy the first time [i = fj h and remains democratic thereafter. When <p = <p h tp = tp\ T = r d < r m such that V r (/i E) - ip = V T {up 1 D r = r"\ When r d ). A ower ip reduces the amount of redistribution and increases the variabiity of fisca poicy. A ower eve of 9 (higher inequaity) increases the variabiity of fisca poicy but may increase or decrease the overa eve of taxation. T - (\-e)(i-0(i- q )) + \-e so de - (x-e)-xc(t'-)- since T < L < i-/3(i-</)' ana y A )/ A C'(j m ) > C'(r e ) (by the convexity of C) we have % < Assumption 2 wi hod when democracy is sufficienty redistributive. This eads to an interesting trade-off: a highy redistributive democracy eads to poitica instabiity but if the potentia for redistribution is too imited democratization does not prevent revoution. 12

23 4. If fi h > /!(# q) and ip < Tp(9 q s) then we are in an unconsoidated democracy. The society continuousy switches regimes. In a nondemocratic regime when fi = n the eite set r = 0 and when fj h fj democracy when ip = <p h they extend voting rights. In a r = r m and when ip = ip there is a coup. A ower eve of 9 (higher inequaity) increases both the eve of redistribution and the variabiity of fisca poicy. The variabiity of fisca poicy is greater than in the other three cases and there is ess redistribution than in cases 2 and 3 and more than in case 1. In the first type of equiibrium where fih < Ji(9q) a revoution is sufficienty costy that given the amount of inequaity and the vaue of q u the eite can avoid it by redistributing. Therefore in state /i' the eite set r 0 whie in state jj 1 they redistribute by setting the tax rate r e which is just enough to stop a revoution. In this equiibrium there is never democratization and the amount of redistribution is reativey imited. More inequaity nonetheess increases the eve of redistribution in this regime because the rich are forced to choose higher taxes to prevent a revoution in the state (n h E). Now consider the case with y! 1 > ](6q). When the economy transits into state h /j the rich can no onger maintain their poitica power via redistribution and must extend the franchise. cp There are three types of equiibria depending on the vaue of <p. If > <p(9 q s) democracy once created is fuy consoidated. When the state first moves from fj to n h the eite are forced to extend the franchise. After this the poor aways set t = r m. In this type of society the amount of redistribution is at its highest eve there is very itte or no fisca voatiity and the threat of a coup pays no roe once the society becomes democratic. countries. We interpret this case as simiar to the situation in most OECD It is more ikey to arise when 9 is high that is when the society is fairy equa. The second possibiity is that <p < <p(9qs) but cp > p(9qs). In this case democracy is not fuy consoidated; if the poor were to set a tax rate r m in the state (<p D) a coup woud occur. However the poor can avoid a coup by setting a ower tax t = T d in state (ip D) which is just sufficient to dissuade the eite from mounting a coup. Athough the society aways remains democratic it is in some sense "under the shadow of a coup" as the threat of a coup keeps overa redistribution beow the eve 14 With q ow the revoution threat is transitory thus the poor reaize that they wi ony receive transfers for a short whie and then transfers wi cease. We might think of this as a non-credibe promise of future redistribution by the eite. Unconvinced by this promise the masses woud attempt a revoution. The revoution is ony prevented by franchise extension. 13

24 of a fuy consoidated democracy. 15 In fact we show beow that higher inequaity may reduce redistribution in a semi-consoidated democracy. The fina type of equiibrium invoves n h > ~p(8q) and ip < Tp(8qs). In this case democracy is unstabe: when the state moves to ip a coup is reativey attractive for the eite and cannot be hated by reducing taxes. fuctuate between democracy and eite contro. the eite in power and they set r = 0. franchise after which the poor set r = r m. As a resut the economy wi stochasticay More specificay the economy starts with h Whenever the state moves to /j. they extend the But as soon as the state goes from (<p h D) to (ip D) they mount a coup regain poitica power and set r = 0. The variabiity of fisca poicy is therefore highest in this equiibrium and the amount of redistribution is ess than in cases 2 and 3 but more than in case 1. Higher inequaity increases redistribution in this regime because it increases the tax rate when there is democracy whie there is never any redistribution during dictatorships. Notice that in this case when the poor are in power they set the maximum tax rate fuy anticipating that redistribution wi eventuay come to an end as a resut of a coup. This resut may hep to expain the existence of highy redistributive but reativey short-ived popuist regimes of Latin America (see for exampe Kaufman and Staings 1991 and Sachs 1990 on popuism). There are four concusions to be drawn from this anaysis. The first inks inequaity to regime changes. A decrease in 8 reduces J(8 q) and increases Tp(8 q s) and <p(8 q s). Therefore societies with more initia inequaity are more ikey to switch between democracy and dictatorship and ess ikey to have a fuy consoidated democracy. So our resuts are in ine with the empirica findings of a positive association between inequaity and poitica instabiity (e.g. Muer and Seigson 1987 and Aesina and Perotti 1996). The second concusion pertains to the ink between inequaity and redistribution. To see this fix the cost of coup ip and define 8 > 8 L such that <p = H <p(9 qs) and L tp = p(8 qs). Moreover suppose that fjh > ~p(8 L q). When 8 > 8 H <p > <p(8qs) so inequaity is sufficienty ow that democracy is fuy consoidated. Now consider an increase in inequaity (a reduction in 8). This wi increase redistribution at first as in the standard modes of voting over redistribution (e.g. A^p < 0. However as 8 fas beow 8 H Metzer and Richards 1981) since democracy is no onger fuy consoidated but semi-consoidated i.e. if 1 (p(8qs)<p(dqs)). In this case the poor are forced to reduce taxes from T m to r d in the state (</?' )) so overa redistribution fas. In fact 15 Wantchekon (1998) argues this has been the case in E Savador and Weyand (1996) argues that the redistribution in democratic Brazi is restrained by the threat of miitary intervention. This resut is aso reated to Eman and Wantchekon (1998) and Wantchekon (1998) who anayze how the threat of confict initiated by the oser of a democratic eection affects the voting outcomes. 14

25 in a semi-consoidated democracy the reationship between inequaity and taxation is ambiguous. The average tax rate is r a = (1 s)t m + sr d. r m is increasing in inequaity whie r d is decreasing. If the cost of taxation C(t) is highy convex then the second effect dominates and the average tax rate fas as inequaity rises though it is sti more ikey that taxation increases when inequaity increases. 10 Intuitivey higher inequaity makes a coup more attractive for the eite so to prevent the coup the poor have to reduce the tax rate substantiay in the state </?' eading to ower redistributive taxation on average. As inequaity increases further we have 9 < 6 L so tp < Tp{9 q s) and democracy is now unconsoidated with ower overa redistribution than both in fuy and semi-consoidated democracies. Therefore there is a nonmonotonic reationship between inequaity and redistribution with societies at intermediate eves of inequaity redistributing more than both very equa and very unequa societies. 17 The third impication of our anaysis is reated to fisca voatiity. The reationship between fisca voatiity and inequaity is ikey to be increasing. Within each regime higher inequaity eads to more variabiity. Moreover higher inequaity makes Case 4 which has the highest amount of fisca variabiity more ikey. This may expain why fisca poicy has been much more voatie in Latin America than in the OECD (Gavin and Perotti 1997). The fina impication of our anaysis is that the costs of redistribution wi aso have an impact on the equiibrium poitica system. Suppose that the cost of taxation becomes ess convex so that C(T m ) is unchanged but C'(T m ) decreases. Since deadweight osses from taxation are now ower the median voter wi choose a higher eve of taxation. However as r m increases so wi S r (9) so democracy becomes more costy to the eite and hence ess ikey to be consoidated. This impies that in societies where taxation creates ess economic distortions for exampe in societies where a arge fraction of the GDP is generated from natura resources democracies may be harder to consoidate. 1G A sufficient condition for the average tax rate to fa as inequaity rises is r n (^m\ ^ (-«-/?(-«-g))[(-a)c'(t")-(0-a)1 \ T I > As[(-j3(-s- 9 ))T«'+/3(-«- 9 )ri'] ' 17 Previous expanations of why high eves of inequaity may not generate high eves of redistribution have emphasized different ideas. First the rich may contro more votes (Benabou 1999 Rodriguez 1998). Second the poor may view themseves as upwardy mobie and not vote for high redistribution (Wright 1996 Benabou and Ok 1998). Third if redistribution fufis an insurance function and if insurance is a norma good increasing inequaity may reduce the demand for insurance and hence redistribution (Moene and Waerstein 1998). the constituency for redistribution (Roemer 1998). Finay there may mutipe dimensions on which citizens vote spitting 15

26 3 Repression So far we have assumed that the rich can mount a coup in a democracy but we have not aowed them to use repression when the society is nondemocratic in order to prevent socia unrest. Such repression is observed in many cases for exampe in Indonesia in 1965 and in E Savador in In this section we wi show that the possibiity of repression wi quaify some of our resuts. The main resuts of our anaysis so far are: 1. democracies are more ikey to consoidate in more equa societies. 2. frequent regime changes are more ikey in more unequa societies. In this section we show that in very unequa societies the eite may have so much to ose from democratization as to prefer a repression strategy to suppress revoution and prevent democratization. As a resut the reationship between inequaity and regime changes is potentiay nonmonotonic as we but democratic consoidation continues to be more ikey in a more equa society. 18 To anayze these issues in the simpest possibe way suppose that the eite are in power and we have fih > J(8q) and ip > ip(9qs) so that if the society democratizes it wi remain so forever and the rich wi obtain V r ((p D r m ) = *_i '. Assume aso that the rich can hire an army with the soe purpose of suppressing revoutionary threats at per period cost M for each member of the eite and that this strategy competey avoids the threat of revoution. 19 It is cear that with this strategy the rich wi have a return V r (Rep) = [Zff - ft foows immediatey that the rich wi find it beneficia to use repression if M < -8 r (9) (12) This condition wi be satisfied if inequaity is sufficienty high i.e. if 9 < M 9 where qm _ p (-M). Next define 9 R such that fih = ~[i{9 R q); so the eite cannot prevent a. This impies that when 6 R < M 6 revoution with redistribution if 9 < 9 R there wi be no democratization: a eve of inequaity that is arge enough to make democratization necessary wi aso make miitary repression desirabe. More interestingy if 9 R > M 9 a society with 9 G (8 9 ) wi democratize because socia unrest cannot be prevented by redistribution and miitary repression is too costy. In contrast if 9 < M 9 then 18 Poitica instabiity is more ikey in more unequa societies aso as ong as socia unrest suppressed by repression is counted as "poitica instabiity". 19 For exampe the army may be financed by taxation in which case M is the tax paid by each member of the eite for this purpose. 16

27 inequaity is so high that the eite are wiing to pay for miitary repression in order to prevent democratization. 20 Finay if 9 > 9 R then there wi be no democratization either this time because the eite can prevent socia unrest by redistribution. It is sti true that among the countries that democratize those with greater inequaity are ess ikey to consoidate democracy and wi therefore osciate between democracy M and dictatorship. It is ony those with 9 G (9 9 R ) and ip > 7p(9 q s) that transit to and consoidate democracy. The second condition requires inequaity to be ow. Therefore our main resut that ow eves of inequaity are conducive to the consoidation of democracy continues to hod in this mode with repression. Poitica instabiity either in the form of frequent regime changes or repression is aso more ikey when inequaity is high. 4 Asset Redistribution We now extend the mode of Section 2 to aow for asset redistribution as we as fisca redistribution. To simpify the anaysis we ignore the possibiity of repression and focus on the configurations of interest rather than giving a compete characterization of a the possibe cases. Let us define 9q to be the initia distribution of assets and use 9 for the distribution after redistribution has been undertaken. Unike fisca redistribution if asset inequaity is reduced it is permanent 21 and cannot be reversed but it aso has permanent costs. We mode the costs by assuming that asset redistribution reduces the tota stock of assets in the economy. If h is the initia stock then the post-redistribution stock is H(9) where if is a concave twice continuousy differentiabe decreasing function i.e. H'(.) < and H"(.) < 0 so asset redistribution reduces tota resources. Aso obviousy h = H(9 ) = Asset Redistribution under Democracy Reca that 9 L is defined by (p =!p(9 L qs) where 8 L > 9 and assume <p < 7p(6 qs) so that without any asset redistribution the economy woud osciate between regimes. Hence for democracy to be (semi-) consoidated inequaity needs to be reduced i.e. 9 needs to be raised to 9 L. Aso assume that 9 R > 9 where jjh = R ~p(9 q). We first assume 20 Notice that the formuation with a fixed cost of miitary repression is not essentia for this resut. Even if the degree of repression is variabe it wi be the societies with greater inequaity where the eite have more to ose from democratization that use more repression. 21 For our genera resuts asset redistribution does not need to be permanent it ony needs to be harder to reverse than fisca redistribution. In practice it may be easier to reverse asset distributions than democracy which is utimatey an empirica question. 17

28 that 9 R is very high so we ignore it in this subsection. Let V p h ((p D9) be the present discounted vaue of a poor agent in a democracy when the cost of a coup is <p h there is democracy and the distribution of assets is given by 9. V p (ip h D 9) = 9H(9)/X + 8 P {9) + [(1 - s)v p {ip h D 9) + sw p (9)] (13) where 8 P {9) is defined as in the previous section and h p = 9H(9)/X. The continuation vaue W p (9) after the coup cost transits to state ip depends on 9 reative to 9 L. If 9 < 9 L then the poor wi be unabe to stop a coup by redistributing and therefore: W?(6) = 9H{9)/\ + p[(- q)w?{0) + qv p (<p\ D 9)} (14) On the other hand if 9 > 9 L coups can be stopped so: W$(6) = 9H{9)/\ + A p (9) +p[(- s)v p {ip h D 9) + sw$(9)\ (15) where A p (9) = r d ( - h p ) - C{r d ) is defined as in the previous section (and A p (0) = 8 P {9) if</>' ><p(9qs)). Soving (13) and (14) simutaneousy (i.e. imposing W p (9) = Wf(9)) we obtain that which is ejm/x + V g(g) ' ^ = TTs DsS'(e) (1-/3K1-W--)) the vaue function that appies when coups occur aong the equiibrium path. (16) When 9 > 9 L we function when coups are avoided as: 22 sove (13) and (15) (i.e. W p {9) = W$(0)) to obtain the vaue vm. «wa ± i^m (i -y) where X S ^ [-2^ +.^k^^g^pm^ ] >.. vm + x > vm (17) To determine equiibrium asset redistribution et 9' argmaxf 1 p (^) bearing in mind that we might be at a corner soution with 6' = 6>o Aso et 9" = aigmaxv (9). Then: where no asset redistribution is chosen. 1. If 9" > 9 L the masses wi redistribute assets up to 9". Intuitivey in this case the eve of redistribution that the poor prefer ignoring the coup constraint aso prevents coups. This case is iustrated in Figure 1. "Observe that \6 P (6) + (1 - X)8 r (e) = -C(r m ) and AA?(0) + (1 - \)A r (6) = -C(t"). Combining these with the equation that determines r d in footnote 11 we obtain A p (8) = ^'^i-p\-j- q ))^ ~ ^p- + {1 %T-7- q )) [~ gi T 1 ~ 5P ( e )} Substituting in (17) we obtain V%(6) = V?{6) + X. 18

29 2. If 9" < 9 L and V[{9') > Vf(0 L ) then the poor wi redistribute to 9\ and coups wi occur aong equiibrium path. 3. Otherwise the eve of redistribution wi be 6 L. This case iustrated in Figure 2 is probaby the most interesting one for our purposes as it iustrates that the poor may choose a high eve of redistribution in order to prevent coups. e L e" Figure 1: An increase in (p increases the gap between Vf(9) and V${9) (i.e. raises X in (17)) making coups ess attractive for the eite. This reduces the concessions that the poor have to make to defend democracy. Therefore it makes them more wiing to choose a higher eve of asset redistribution to consoidate democracy. Aso an obvious comparative static pertains to the eve of inequaity: if 9$ > 9 L there wi not be a motive to redistribute assets in order to consoidate democracy. So we may expect asset redistribution to emerge as a method of consoidating democracy especiay in reativey unequa democratic regimes that are expected to be threatened in the future. Overa therefore the main impication of this anaysis is that asset redistribution can hep to consoidate democracy. Whenever the choice of the poor is 9 L or greater coups no onger occur aong the equiibrium path because asset redistribution has changed the eve of inequaity permanenty and makes coups ess attractive for the eite. 19

30 9' 9" 9 L Figure 2: In practice asset redistribution appears to have payed such a roe in a number of instances. In Acemogu and Robinson (1997) we argued that educationa expansion in 19th century Britain and France was in part a resut of democratization and Engerman Marisca and Sokooff (1998) argue the same for Latin America. In Britain and France these and other poicies reduced inequaity and there were no significant reversas in the process of democratization. In Costa Rica the educationa and and reforms that reduced both earnings and and inequaity after the democratization in 1948 appear to have heped with the consoidation of democracy (see Yashar 1997 for this argument and Vias 1995 for some numbers). 23 The situation in Venezuea after the return to democracy in 1958 which ed to a and reform redistributing 19.3 percent of agricutura and 24 aso provides some support to this view (see Powe 1971). 4.2 Anticipated Asset Redistribution and Poitica Instabiity We have so far iustrated the roe of asset redistribution in consoidating democracy. Since asset redistribution is permanent and costy to the eite the anticipation of such 23 Chaker (1995) writes that "the most remarkabe egaitarian measure in Costa Rica occurred in the 1960's and 1970's when the concentration in income distribution was reduced. outcome rather than a cause of democratic poitics." 24 See Tabe 10.2 in Cardoso and Hewege (1992). Interestingy this was an 20

31 redistribution may aso make a coup more attractive. We wi now anayze this using a simpe extension of our mode. Suppose that ip > ~ip(9oqs) (i.e. 9 > 9 L ) so that democracy is (semi-)consoidated without asset redistribution. Consider the first period of democracy in the state with <p h. The poor may want to redistribute assets this time not to consoidate democracy but to increase their incomes. However we assume that there is a one period deay between the egisation and the impementation of asset redistribution. For exampe and reforms invove administrative deays. If during this period the state switches from (ip h D) to ((p D) then the rich may mount a coup to avoid asset redistribution before it is impemented. To simpify the anaysis we assume that during the process of asset redistribution (i.e. in both the period of egisation and impementation) there wi be no fisca redistribution. To anayze this case first write the vaue to the rich after asset redistribution changes inequaity to 9 as V T (D) ~ 1 ' - = (1 r 9)H{9)/{\ - A) + P( - s)6 (9) + j3sar(9) 1-/3 (whether the state is (<p D) or h ((p D) does not matter as there is no current fisca redistribution). h Suppose now that asset redistribution is egisated in the state ((p D) then the vaue to the rich is VV.D) = (1 ~^Hx {do) + p [(1 - s)w(d) + sm^{v r (D)V r (f/e) - <p 1 }] This vaue function takes into account that if during the period of administrative deay the state switches to <p (/*' ) is the rich may mount a coup. The vaue to the rich in the state VV E) = (- 9 Q )H(9 )/(1 - A) + (3 [(1 - q)v T (fi E) + qv r & h D) Soving these equations the critica vaue tp at which V r (D) = V r (fi E) <p is + 0q) U r (6) + P( - s)6 r {6) + psa r (6). -(1 V = ~ -P(-q) where k t (6) = (1-6)H(6)/(1 - A) - (1 - )H(6 )/(1 - A) < is the oss for the rich as a resut of asset redistribution. In the case where <p < ip the eite woud mount a coup foowing asset redistribution egisation if the cost of the coup were to change to ip. Define k p (6) = 0H(d)/\ - 9 H(9 )/\ simiary then we can notice that kp(9) + - /5(1 s)8 p (9) + PsA p (9) is the discounted average per-period transfer to the poor hence k p (9) + 21

32 (3(1 - s)s p (6) + (3sA p (9) > k p (9) + (3(1 - s)8 p (9) + (3sA p (9) for any 9 since 9 is chosen optimay. Since taxation and asset redistribution are costy this in turn impies that K r (d) +/3{1 - r s)6 (B)+PsA r (d) < K r r (8 ) + f3( - s)6 (9 )+(3sA r (9 ) < (38 T (9 )(1 -s-q). Hence <p > Tp as defined by (8). So there exist vaues of ip (7p ip) such that without asset redistribution democracy is consoidated but if during the period of administrative deay after asset redistribution the cost of a coup switches to <p the eite wi attempt a coup. If s the probabiity that the cost of the coup switches to tp is ow enough the masses may prefer to enact asset redistribution despite its potentiay destabiizing effects. Therefore the main concusion of this subsection is that asset redistribution which is generay in the interest of the poor and often usefu in consoidating democracy may create a temporary period of instabiity for a democracy. A number of coups in Latin America appear to have been motivated by a desire to prevent radica and reform. For exampe in Brazi a centra aim of the coup in 1964 was to prevent the attempt by the eft-wing President Gouart to by-pass the veto of the Congress and use other means to push through agrarian reform (see for exampe Skidmore 1967 and Waerstein 1980). Simiary most schoars argue that the agrarian reform after 1952 in Guatemaa was the main motivation for the coup of 1954 (Handy 1984 Trudeau 1993) and that the increasing radicaization of Aende's poicies especiay on and reform precipitated the coup of 1973 in Chie. 25 The same is true in Venezuea where the 1948 and reform aw was immediatey repeaed by the incoming miitary government (see Dorner 1992 p. 47 and Powe 1971). 4.3 Asset Redistribution in Nondemocratic Regimes The rich may aso wish to undertake asset redistribution in order to stop a revoution or democratization. Therefore our anaysis in this subsection wi revea another strategy for the eite when faced with the threat a revoution in addition to democratization and repression. We continue to assume n h > ~p(9 q) 26 but suppose that ip > <p(9 qs) so that democracy if created wi be fuy consoidated. We aso assume that redistribution away from the poor is not possibe i.e. 9 >9q. 2o See Vaenzuea (1989). Brown (1989 p. 236) writes "a second generation of [agrarian] reform was ceary gaining momentum in the atter part of Aende's administration - a fact that was not ost on counterreform eements that utimatey supported Pinochet's coup d'etat." Of the and originay expropriated by Aende's government 43% was returned to previous owners or excuded from the reform by other means see Jarvis (1989 p. 249). 26 Hence > 0o so that the eve of inequaity that prevents democratization is ess than the initia eve. 22

33 The vaue function for the rich when the state is /r and inequaity is equa to 9 is: FV E 9) = (I- 9)H(e)/( - A) + p [(1 - q)v r (fj E 6) + qw r (9)\ (18) where h r = (1 9)H{9)/{\ A) and W r (0) is their return after the state switches to /A If # < # K then democratization woud take pace in the state (/x ft E). The assumption ip > <p(9q q s) aso ensures that <p > (p(9 q s) for any 9 > 9 so democracy once created wi aways be fuy consoidated. Our above anaysis then impies that the return to - \ the rich under a consoidated democracy is W[{9) = 1 e ) H \ e )/0-- x ) +s \e ) _ Hence using W r (9) = W{{9) in (18) we find that when 9 < 9 R V{(n E9) = ^% + (i-m-%-)) is the vaue function of the eite when they cannot prevent democratization. the eite may wish to undertake asset redistribution in order to reduce 6 r (9) In this case depending on whether asset or fisca redistribution is more costy to them. In what foows we assume that asset redistribution is sufficienty costy that the eites wi not do this so r 27 axgmaxv 1 {fi E9) = 9. In contrast if 9 > 9 R when the revoution threat arises. democratization can be avoided by redistributing to the poor The continuation vaue of the eite in state (fj h E) is: WZ(9) = (1-9)H{9)/{\ - A) + rf{6) + {3 [(1 - q)v T ^\ E 9) + qw[{9)] (19) W)=M«1^ (20) Soving (18) and (19) with W r {9) = Wff) we obtain Whether the eite wi choose asset redistribution is determined simpy by comparing V{(fi E6 ) and V{((j 1 E9 r ). If V{(fj!E9 ) < V{(n E9 R ) then the eite prefer to prevent democratization and wi choose the minimum redistribution sufficient to prevent democratization that is 9 = 9 R. Otherwise they wi choose not to redistribute so 9 = 9q. The main impication is that the eite may choose to redistribute assets in order to prevent democratization. 28 Two cases appear to fit this impication nicey. In a 1949 reform South Korea redistributed 50% of the agricutura and in Korea. Haggard (1990 p. 55) argues that the reforms were aimed at defusing rura insurrections and counteracting the destabiizing spiovers from and reform in North Korea. Taiwanese and reforms "Formay -H(e ()+(\-e a )H'(e n ) t oi many ggqgoj < (i_/3)(i_a) + (i-/3)(i-/3(i-<?)) - Q a 28 This resut is reated to previous anayses of and reform such as Grossman (1993) and Horowitz (1993) which argue and reform can prevent revoution though these papers do not compare asset and fisca redistribution. Aso in contrast to these papers asset redistribution in our economy may prevent not ony a revoution but aso democratization. 23

34 of that redistributed 24.6% of the and (Ho 1978 p. 163) aso appear to have been an attempt to defuse rura protest (see Amsden 1985). In the words of Ch'en Ch'eng the governor of Taiwan at the time of the reforms "...the situation on the Chinese mainand was becoming critica and the viages on the isand were showing marked signs of unrest and instabiity. It was feared that the Communists might take advantage of the rapidy deteriorating situation" (quoted in Haggard 1990 p. 82). Interestingy unti very recenty both South Korea and Taiwan remained reativey nondemocratic especiay compared to other countries with simiar per capita income eves. It can aso be observed that asset redistribution as a method of preventing democratization is more ikey to emerge when 6 r (9) the transfers away from the rich in a democracy are arger. This resut may hep expain why asset redistribution emerged in South Korea and Taiwan where the threat of communism made socia unrest very costy to the eite but not in the Phiippines where the threat was ess serious. Finay athough we have modeed assets redistributed by the eite as equay divided among a the poor a more appeaing interpretation may be a strategy of co-opting whereby the eite redistribute assets seectivey to groups among the poor who are important for the threat of revoution and who can be persuaded to switch sides with such transfers. This is equivaent to our formuation but this interpretation may fit the exampe of Mexico in the 1930s better where sma groups of peasants that supported the ruing party were given and. 5 Targeted Transfers and Poitica Instabiity Our anaysis so far has assumed that transfers are not targeted to a particuar group. This assumption is ceary unreaistic in many settings. In practice transfers are often means-tested and taxes are non-inear. Perhaps more importanty groups with diverse poitica interests are often geographicay or ethnicay segmented making transfers to ony one group quite easy. It is straightforward to extend our anaysis to aow for this possibiity. In particuar suppose that when group i is in power it can aocate a fraction a 1 of a tax revenue to itsef (where a p > A and a r > 1 A are parameters taken as given). This changes the equations in Section 2 in an obvious manner and aso raises the possibiity that even the rich may choose a positive eve of taxation when in power in order to tax the poor and transfer some of the resources to themseves. The key equations from Section 2 are modified as foows. The poor wi now set a 24

35 tax rate T m given by a p - 6 when they are not constrained by the threat of a coup and the rich wi set a tax rate A r nd satisfying when they are not constrained by the threat of revoution (and if a T < 1 6 r nd = 0). When the state is (<p h D) the returns to poor and rich agents continue to be given by equation (2) in Section 2 with the ony difference that now 6 P (8) = ^-T m r m h p and S r (9) = ^Tm - T m h r. In the state (<p D) the poor may once again change their redistributive strategy in order to prevent a coup. With targeted transfers however they can do better than setting r d = 0: they can tax a income at the rate r nd and target a fraction a r of the redistribution to the rich (exacty as the rich wi do when they have power). Therefore the maximum transfer that the poor can make to the rich is x r {@) = j^t nd T nd ht > (if x r (@) were negative r = (fi woud be better). The vaue functions in the nondemocratic regime wi aso change because in the state E) the rich wi now tax at the rate r nd and redistribute a fraction a r of the revenues to themseves. Hence: V\v\ E) = h + X {e) + [(1 - g)vv E) + qv\n\ E)\ (21) where x r (#) = ^T nd - r nd h r is as defined above and x p {6) = ^T w - r nd K?. The coup constraint is sti given by equation (6) and the critica vaue for the cost of a coup is now found to be >(*«') = 1-/3(1-) W Expression (22) differs from (8) because it has the additiona term x r > in the numerator and aso because with targeted transfers the cost of democracy to the rich 8 r is higher. As a resut!p(0qs) is greater than 7p(9qs). More generay both an increase in a p and in a r raise ~fi(0 q s) making coups more ikey because they raise the stakes (a p raises 8 T and a r raises x r )- With targeted transfers controing the power to tax is more vauabe; the rich are forced to make greater net transfers to the poor in a democracy and are abe to receive net transfers in a nondemocratic regime. Therefore the rich are wiing to pay a greater cost to regain power. This resut aso puts one of our basic resuts from Section 2 into perspective; greater inequaity increases poitica 25

36 instabiity because it increases the stakes in the poitica game giving greater incentives to each group to fight to regain power (a point aso emphasized by Madison 1788). h A simiar argument impies that the critica vaue for fi ~jt(9q) is now ower (i.e. ess than J(6q) as given by (11) in Section 2). So when the government can direct transfers poitica power has more economic vaue and osciations between different regimes are more ikey. This impies that poitics wi be more unstabe in countries such as many in Africa where ethnic or geographic segregation of different groups makes targeted transfers possibe. This extension aso emphasizes that our mode can be appied to poitica confict between any two groups rather than ony between the rich and the poor. For exampe the two parties can be thought of as rura vs. urban interests in confict over transfer of resources to urban areas or two ethnic groups with the one in power transferring resources from the other group and the group out of power contesting poitica power. 6 Concusions In this paper we have deveoped a simpe theory of poitica transitions. Our theory emphasizes the roe of the threat of revoution and socia unrest in eading to democratization and the desire of the rich eite to imit redistribution in causing switches to nondemocratic regimes. Inequaity emerges as a crucia determinant of poitica instabiity as it encourages the rich to contest power in democracies and aso often encourages socia unrest in nondemocratic societies. Therefore inequaity is generay conducive to frequent regime changes. However in very unequa nondemocratic societies the rich may choose a repressive strategy in order to prevent democratization and revoution. In a cases democracy is more ikey to be consoidated if the eve of inequaity in the society is imited and inequaity is ikey to ead to poitica instabiity either in the form of frequent regime changes or repression of socia unrest. Inequaity is aso ikey to ead to fisca voatiity as the redistributive regime changes with the poitica system of the society. Nevertheess inequaity does not necessariy ead to more redistribution. Unequa societies switch between regimes and in nondemocratic regimes there is no redistribution. Our theory suggests that asset redistributions may be used to stabiize both democratic and nondemocratic regimes but the anticipation of a radica asset redistribution such as a and reform may destabiize an otherwise consoidated democracy because the eite may mount a coup specificay to avoid the reforms. 26

37 References Acemogu Daron and James. A. Robinson (1997) "Why Did the West Extend the Franchise? Democracy Inequaity and Growth in Historica Perspective" CEPR Discussion Paper No Ades Aberto and Thierry Verdier (1996) "The Rise and Fa of Eites: A Theory of Economic Deveopment and Socia Poarization in Rent-Seeking Societies" CEPR Discussion Paper No Aesina Aberto and Roberto Perotti (1996) "Income Distribution Poitica Instabiity and Investment" European Economic Review Aesina Aberto and Dani Rodrik (1994) "Distributive Poitics and Economic Growth" Quartery Journa of Economics Amsden Aice H. (1985) "The State and Taiwan's Economic Deveopment" in Peter B. Evans Dietrich Rueschemeyer and Theda Skocpo eds. Bringing the State Back In Princeton University Press Princeton NJ. Benabou Roand (1999) "Unequa Societies" NBER Working Paper forthcoming American Economic Review. Benabou Roand and Efe A. Ok (1998) "Socia Mobiity and the Demand for Redistribution: The POUM Hypothesis" NBER Working Paper #6795. Benhabib Jess and Ado Rustichini (1996) "Socia Confict and Growth" Journa of Economic Growth Brown Marion (1989) "Radica Reformism in Chie: " in Wiiam C. Thiesenhusen ed. Searching for Agrarian Reform in Latin America Unwin Hyman Boston. Cardoso Eiana and Ann Hewege (1992) Latin America's Economy: Diversity Trends and Conficts The MIT Press Cambridge MA. Chaker Cynthia H. (1995) "Eections and Democracy in Costa Rica" in Mitche A. Seigson and John A. Booth eds. Eections and Democracy in Centra America Revisited University of North Caroina Press Chape NC. Cohen Youssef (1994) Radicas Reformers and Reactionaries University of Chicago Press Chicago IL. Coier Ruth B. (1998) Paths Towards Democracy: The Working Cass and Eites in Western Europe and South America forthcoming book Cambridge University Press. 27

38 Dah Robert A. (1971) Poyarchy Yae University Press New Haven. Dorner Peter (1992) Latin American Land Reforms in Theory and Practice University of Wisconsin Press Madison WL Di Teia Guido and Rudiger Dornbusch (1989) "Introduction: The Poitica Economy of Argentina " in Guido Di Teia and Rudiger Dornbusch eds. Poitica Economy of Argentina University of Pittsburgh Press Pittsburgh PA. Eman Matthew and Leonard Wantchekon (1998) "Eectora Competition under the Threat of Poitica Unrest" Unpubished Paper Department of Poitica Science Yae University. Engerman Staney L. Eisa Marisca and Kenneth L. Sokooff (1997) "Schooing Suffrage and the Persistence of Inequaity in the Americas " Unpubished Paper Department of Economics UCLA. Gavin Michae and Roberto Perotti The (1997) "Fisca Poicy in Latin America" NBER Macroeconomics Annua edited by Benjamin Bernanke and Juio Rotemberg MIT Press Cambridge. Grossman Hersche I. (1991) "A Genera Equiibrium Theory of Insurrections" American Economic Review Grossman Hersche I. (1993) "Production Appropriation and Land Reform" American Economic Review Haggard Stephan (1990) Pathways from the Periphery: The Poitics of Growth in the Newy Industriaizing Countries Corne University Press Ithaca NY. Haggard Stephan and Robert R. Kaufman (1995) The Poitica Economy of Democratic Transitions Princeton University Press Princeton NJ. Handy James (1984) Gift of the Devi: A History of Guatemaa Dumont Press Toronto. Ho Samue (1978) Economic Deveopment of Taiwan Yae University Press New Haven CT. Horowitz Andrew W. (1993) "Time Paths of Land Reform: A Theoretica Mode of Reform Dynamics" American Economic Review Jarvis Lowe S. (1989) "The Unraveing of Chie's Agrarian Reform " in Wiiam C. Thiesenhusen ed. Searching for Agrarian Reform in Latin America Unwin Hyman Boston. Kaufman Robert R. and Barbara Staings (1972) "The Poitica Economy of Latin American Popuism" in Rudiger Dornbusch and Sebastian Edwards eds. The 28

39 Macroeconomics of Popuism in Latin America University of Chicago Press Chicago IL. Kob Gen L. (1974) Democracy and Dictatorship in Venezuea Shoestring Press Hamden CT. Levine Danie H. (1989) "Venezuea: The Nature Sources and Prospects of Democracy" in L. Diamond J.J. Linz and S.M. Lipset eds. Democracy in Deveoping Areas: Latin America Lynne Reimer Bouder Co. Linz Juan J. (1978) Crisis Breakdown and Reequiibration Johns Hopkins University Press Batimore MD. Linz Juan J. and Afred Stepan (1996) Probems of Democratic Transition and Consoidation Johns Hopkins University Press Batimore MD. Lipset Seymour M. (1959) "Some Socia Prerequisites for Democracy: Economic Deveopment and Poitica Legitimacy" American Poitica Science Review 53. Madison James (1788) The Federaist: A Coection of Essays Written in Favor of the New Constitution New York. Metzer Aan H. and Scott F. Richard (1981) "A Rationa Theory of the Size of Government" Journa of Poitica Economy Moene Kar O. and Michae Waerstein (1997) "Inequaity and Redistribution" Unpubished Paper Department of Poitica Science Northwestern University. Moore Barrington (1966) The Socia Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern Word Beacon Press Boston MA. Muer Edward N. and Mitche A. Seigson (1987) "Inequaity and Insurrections" American Poitica Science Review O'Donne Guiermo (1973) Modernization and Bureaucratic Authoritarianism: Studies in South American Poitics University of Caifornia Institute for Internationa Studies Berkeey CA. O'Donne Guiermo and Phiip C. Schmitter (1986) Transitions from Authoritarian Rue: Tentative Concusions about Uncertain Democracies Johns Hopkins University Press Batimore MD. Perotti Roberto (1996) "Growth Income Distribution and Democracy" Journa of Economic Growth Paige Jeffrey M. (1997) Coffee and Power: Revoution and the Rise of Democracy in Centra America Harvard University Press Cambridge MA. Powe John D. (1971) Poitica Mobiization of the University Press Cambridge MA. Venezuean Peasant Harvard 29

40 Persson Torsten and Guido Tabeini (1994) "Is Inequaity Harmfu for Growth? Theory and Evidence" American Economic Review Prezworski Adam (1991) Democracy and the Market Cambridge University Press. Rock David (1987) Argentina : From Spanish Coonization to Afonsin University of Caifornia Press Berkeey CA. Roemer John E. (1985) "Rationaizing Revoutionary Ideoogy: A Tae of Lenin and the Tsar" Econometrica Roemer John E. (1998) "Why the Poor do not Expropriate the Rich: An Od Argument in New Garb" Journa of Pubic Economics Rodriguez Francisco (1998) "Inequaity Redistribution and Rent-Seeking" Unpubished Paper Department of Economics University of Maryand at Coege Park. Rodrik Dani (1999) "Democracies Pay Higher Wages" forthcoming in the Quartery Journa of Economics. Rueschemeyer Dietrich Eveyn H. Stephens and John D. Stephens (1992) Capitaist Deveopment and Democracy University of Chicago Press Chicago IL. Rustow Dankwart C. (1970) "Transitions to Democracy: Toward a Dynamic Mode" Comparative Poitics Sachs Jeffrey D. (1990) "Socia Confict and Popuist Poicies in Latin America" in R. Brunetta and C. DeH'Aringa eds. Labor Reations and Economic Performance NYU Press New York. Skidmore Thomas E. (1967) Poitics in Brazi ' An Experiment in Democracy Oxford University Press New York NY. Smith Peter H. (1978) "The Breakdown of Democracy in Argentina " in Juan J. Linz and Afred Stepan eds. The Breakdown of Democratic Regimes: Latin America The Johns Hopkins University Press Batimore MD. Stepan Afred (1978) "Poitica Leadership and Regime Breakdown: Brazi" in Juan J. Linz and Afred Stepan eds. The Breakdown of Democratic Regimes: Latin America The Johns Hopkins University Press Batimore MD. Therborn Goran (1977) "The Rue of Capita and the Rise of Democracy" New Left Review Torne Aaron and Andres Veasco (1992) "The Tragedy of the Commons and Economic Growth: Why Does Capita Fow from Poor to Rich Countries?" Journa of Poitica Economy Trudeau Robert (1993) Guatemaan Poitics: The Popuar Strugge for Democracy Lynne Reimer Bouder CO. 30

41 Vaenzuea Arturo (1989) "Chie: Origins Consoidation and Breakdown of a Democratic Regime" in L. Diamond J.J. Linz and S.M. Lipset eds. Democracy in Deveoping Countries: Latin America Lynne Reimer Bouder CO. Vias Caros M. (1995) Between Earthquakes and Vocanos: Market State and the Revoutions in Centra America Monthy Review Press New York NY. Waerstein Michae (1980) "The Coapse of Democracy in Brazi: Its Economic Determinants" Latin American Research Review Wantchekon Leonard (1998) "Strategic Voting in Conditions of Poitica Instabiity: The 1994 Eections in E Savador" Unpubished Paper Department of Poitica Science Yae University. Weyand Kurt G. (1996) Democracy without Equity: Faiures of Reform in Brazi University of Pittsburgh Press Pittsburgh PA. Wright Randa D. (1996) "Taxes Redistribution and Growth" Journa of Pubic Economics Yashar Deborah J. (1997) Demanding Democracy: Reform and Reaction in Costa Rica and Guatemaa 1870's-1950's Stanford University Press Stanford CA. 31

42

43

44

45

46 inn Date Due Lib-26-67

47

48