Democracy and the Threat of Kingmaking 1

Transcription

1 Democracy an the Threat of Kingmaking 1 Thorsten Janus University of Wyoming Abstract This paper suggests a way to achieve emocratic reforms at no cost espite resistance from incumbent power holers. The key iea is to rewar a takeover by a challenger tomorrow if the incumbent oes not emocratize toay. I also iscuss emocracy movements as challengers, creit constraints, creibility, renegotiation, collusion, reform blocking, emocracy reversals, classical kingmaking, emocracy-from-above, an contemporary Afghanistan an Iraq in relation to the moel. Finally, I briefly iscuss how the plan coul work in practice. 1 Paper Presente at the North American Meetings of the Econometric Society, Boston MA, June4-7, I am grateful to seminar participants at the University of Wyoming an to Jake Kenall, Lauren Malone an Klaas vant t Vel. 1

2 Introuction Weber famously efines a state as a human community that (successfully) claims the monopoly of the legitimate use of physical force within a given territory. (Weber 1919/1972, emphasis in original). Short of large coorination problems within the state (Gurr, Jaggers an Moore 1990, Svensson 2000, Dixit 2007) or the use of external physical force, therefore, presumably states cannot be mae to pursue goals not in their interest, such as human rights an evelopment in many non-emocracies (Olson 1993, Bates 2008, Bueno e Mesquita et al. 2005, Acemoglu et al 2008). This paper examines what it means for the feasibility an cost of welfare improving reforms that they must be compatible with the incentives of incumbent leaers. Specifically, I explore how cheaply emocratic reforms can be bought by a benevolent external party with no access to the use of physical force, although it may eman compensation for its services from nonelite citizens later on. The central result is that, if the thir party can commit to its actions, emocracy can be achieve at zero cost for either citizens or the thir party, that is, the incumbent elite surreners power without compensation or struggle. A few observations motivate the paper. First, in a simple moel with an incumbent power holer an unfree citizens, the loss of the latter shoul excee the gain of the former because the surplus extraction process is costly (e.g. citizens have poor work incentives or invest in low-return expropriation-safe assets). Secon, if this is true, some conition require for the Coase theorem must not be present or citizens coul compensate the ruler for emocratizing. Among the problems may be coorination, free-riing, or commitment inability among citizens, while financial constraints or the incumbent s inability to commit can prevent ex-ante compensation schemes. Thir, however, it seems that with a Coasean surplus available an external party might serve as a financial intermeiary. For example, a collection of international ai organizations or evelope country governments may have better coorination an commitment ability than the citizens of a non-emocracy because of fewer agents involve, lower communication costs, more concern with reputation, or greater ability to esign institutions which tie their hans in future ealings with the incumbent. Moreover, since citizens can in principle compensate 2

3 the incumbent, they can also compensate the intermeiary for compensating the incumbent. Thus, the external party nee not be benevolent an a Pareto-improving move to the efficient frontier shoul be feasible. While to an extent, such Coasean emocracy contracts are alreay being use, since many ai onors promise to rewar reforms, these carrots are typically ineffective. First, conitional ai schemes may be non-creible (Easterly 2002) because political economy reasons (such as lobbying, colonial ties, geopolitics, or investment an export goals) an poverty targeting are ultimately more important for ai ecisions than broa reforms such as emocracy promotion. Also, it may be unacceptable for external actors to let a emocratizing ictator, for example, walk away with millions of ollars given concerns for punishing her abuse of power, eterring new ictatorships, an helping poor citizens. Creibility coul also suffer if incumbency rents are simply very large, as may be true in resource rich nations such as Angola, Cha, The Republic of Congo, or Saui Arabia. Sticks in the form of military invasion threats or even mere trae sanctions (which require multilateral cooperation to be effective) are often politically infeasible an therefore non-creible threat points. Secon, the carrots currently use to rewar reforms may be too isperse or small: political elites can typically sustain goo living stanars espite trae sanctions an if the conition for more ai is that incumbents cannot steal the ai it is unclear that they benefit. Promises of entry to prestigious international organizations or growth may bring large aggregate, but again isperse, benefits. More targete sanctions, like travel restrictions or overseas asset freezes, may be more effective but seem unable to outweigh the present value of incumbency rents an at most promote limite an temporary reforms in practice. In this paper, I propose another approach: rather than rewar the current incumbent if she emocratizes, suppose that a rewar is offere to an agent who eposes the current incumbent an then emocratizes. If the rewar is large enough, this ensures that the agent, henceforth calle the challenger, will pose a creible takeover threat. In turn, the 3

4 value of the incumbency rent rops to zero an a small rewar ensures emocratization by the incumbent. This scheme may aress at least some of the problems with the existing carrot-an-stick proceures. First, it oes not unuly rewar the incumbent an therefore oes not set a ba preceent for paying off hostage takers or create a perceive unfair outcome which coul lea to political opposition ex-post. Also, by outsourcing the regime change an reforms to the challenger, the countries posting the rewar o not put the lives of their own soliers at stake in military conflict, which coul harm the creibility of the takeover threat. Their possibly limite knowlege of how to achieve the reforms is also less problematic. Secon, unlike trae an investment sanctions for example, the payment irectly targets the agent who controls reform implementation, making it cost-efficient. The spirit of the plan is consistent with Krain (2005), who fins that among ongoing government sponsore genocies or politicies between 1995 an 1997 only interventions irectly targeting the perpetrator or aiing the target raise the probability of stopping or slowing the process, while impartial interventions ha no effect (Stanton 2004, Valentino 2004). One remaining issue may be that the value of office holing coul be large, which coul persuae the challenger to become a new ictator after eposing the initial incumbent. However, partly the external party only pays the rewar after emocratization is achieve an partly, as I iscuss below, the mechanism can be applie repeately, so that the original challenger becomes threatene by a new challenger. Also, while the present value of infinite incumbency is large in some autocratic nations, precisely for that reason incumbency is har to sustain: leaers have a high probability of being forcefully remove from office even without the external intervention. 2 The paper procees as follows. Section 2 reviews some relate literature. Section 3 first presents the basic moel an the central result of the paper that emocracy can be 2 I thank Anton Kim an Markus x for this point. 4

5 achieve at zero cost espite the loss to the incumbent. Then the rest of Section 3 consiers emocracy movements as challengers, creit constraints, creibility, renegotiation, collusion, reform blocking, an emocracy reversals in the moel an argues that the results are generally robust to these complications. Section 4 relates the moel to classical kingmaking, emocracy-from-above, contemporary Afghanistan an Iraq, an consiers how to implement the scheme in practice. 2. Relate Literature The paper relates to the large literature on the political economy of reforms, which explains why efficient reforms may either be elaye (Rorik 1999, Lau, Qian an Rolan 2000, Alesina an Drazen 1991, Haggar an Kaufman 1995, Sturzenegger an Tommasi 1998) or fail to materialize at all (Acemoglu an Robinson 2006b, Parente an Prescott 2000, Chauhry an Garner 2007, Fernanez an Rorik 1991). This paper follows the (often implicit) assumption in that literature that the winners an the losers, which here are the citizens an the government, cannot write a Cosean contract ue to commitment or asymmetric information problems. 3 I eviate from this literature, however, by aing an external party who can in fact contract with the government an, more importantly, with a contener for government control which I call the challenger. In particular, the external party can commit to paying for emocracy once achieve. I argue below that moerate egrees of information asymmetry woul not change the result an justify the external party s commitment ability in section 3.8. The paper also relates to the political competition an economic outcomes literature (Persson an Tabellini 2002, Roemer 2001), particularly in the absence of full emocracy (Acemoglu an Robinson 2006a, Paro I Miquel 2007, Grossman 1991). Closely relate 3 Microfounations of incomplete contracting are iscusse in Hart an Moore (2003) an Acemoglu (2003). Azam an Mesnar (2003) show how commitment or asymmetric information problems can make a Coasean contract to avoi a civil war, resembling a challenger takeover in this paper, infeasible. Here, however, it is precisely the commitment of the external party not to renegotiate with the incumbent if it fails to emocratize that ensures zero-cost emocracy (see the section on renegotiation below) 5

6 here is Mulligan an Tsui (2006, henceforth MT), who use a monopolistic competition analogy to show how the threat of entry (implying takeover, not uopoly) to the political market can iscipline an otherwise abusive incumbent. This paper shares the MT result that a more severe entry threat promotes iscipline but iffers since MT o not consier Coasean contracts an external meiation between citizens an incumbents. Thus, their moel can explain goo policies by autocrats but not permanent uncompensate emocratic transitions. On the other han, however, the contingent payments scheme here works like an external commitment to subsiize the entry of a pre-selecte challenger if the incumbent fails to behave in their moel, so it is similar to a special policy intervention. In fact, in their policy section, they briefly iscuss the problem that challengers may be willing to challenge if an only if they themselves can steal political rents without punishment by external agents, that is, the ol rulers can only be remove if the new rulers are just as ba. My scheme resolves that problem, however, by rewaring the new ruler for goo behavior, which here implies emocratizing. That rewar is compatible with a rise in the welfare of citizens even shoul they be require to foot the bill ultimately, because the transition to emocracy brings them an even larger gross welfare increase. MT s moel may be appropriate if one consiers emocracy reversible an a matter of egree - a change in parameters in the same basic game while the moel here may be may be appropriate when emocracy is an irreversible transition to a new institutional equilibrium (Przeworski 1991, Weingast 1997). However, I show below that a emocratic reversal option makes no ifference for the results of the paper. 3. A Democratization Moel I moel citizens, the initial incumbent, a possible contener for political control calle the challenger, an an external party in a sequential move game. All agents are risk neutral an citizens have zero payoff from the non-emocratic status quo an a welfare gain of W > 100 from emocracy, while the incumbent I earns a rent of 100 in the status quo. The rent is less than citizens loss because the rent extraction process is inefficient 6

7 (e.g. tax istortions, public monopolies, price an quantity controls, corrupt bureaucracies, expropriation risk). The external party E maximizes the welfare of citizens minus any payments to the incumbent or the challenger unless otherwise state. Thus, whenever E has a non-negative payoff, citizens can compensate it for its services once emocracy is achieve an coul alternatively use a competitive financial intermeiary to achieve emocracy. The challenger R has value c of holing power an cost c of taking power. The challenger is entirely self-intereste an I assumew > c> v, where if the secon inequality i not hol the challenger shoul have taken power alreay. 4 c may reflect transfers provie by the incumbent precisely to prevent a takeover or the challenger s cost of assembling a winning coalition (Mulligan an Tsui 2006). For the reasons iscusse above, citizens cannot engage in a Coasean bargain with the incumbent without assistance from E. The payoffs are known to all. 5 Section 3.11 shows how, if emocracy were reversible, the challenger coul be prevente from reversing emocracy once introuce by the incumbent. The timing is as follows. The country is initially a non-emocracy. Then, first, the I R external party E offers non-negative contingent payments { p, p } to the incumbent if it 4 An interesting extension may be if entry leas to a power uopoly rather than replacement. This is realistic for faile or weak state environments such as Afghanistan, Colombia, Russia in the 1990s, or Somalia an other parts of Sub-Saharan Africa. Power centers may then compete for loyalty an prouce consumption goos an conflict, mixing ieas from Tiebout s sorting moel an the conflict moels of Hirshleifer (1995, 2001) an others. 5 Uncertainty about the incumbent s (challenger s) value of being in office woul not change the results since all players are risk-neutral. If only the incumbent (challenger) knows her value of being in office an the external party oes not, this makes no ifference because in equilibrium the incumbent is pai zero an the challenger is pai her takeover cost c for emocratizing, regarless of their office holing values. If the external party oes not know c (even after asking citizens), suppose that the external party knows c to lie in [, c] c with c < W. Then the external party can ensure the same qualitative outcome by promising rewar c to a emocratizing challenger, although the latter earns an information rent. However, plausibly the takeover cost can be reasonably approximate an shoul be relatively small compare to citizens gross benefitw. A similar argument applies if the challenger is creit-constraine an must be pai up to c up-front an v to emocratize once in office (see section 3.4.). 7

8 emocratizes an to the challenger if, after taking power, it emocratizes. E can commit to these payments (see section 3.8.). Secon, after E presents its plan, the incumbent ecies to emocratize or not. If it emocratizes, the game ens. Thir, the challenger R ecies to take power or not. If it oes not take power, the game ens. Fourth, if the challenger has taken power it ecies to emocratize or not an the game ens in either case. Once emocracy is introuce it is irreversible an the government earns no rent. The challenger is the only source of political risk for the incumbent, although an exogenous risk of being epose 1 µ simply lowers the incumbent an challenger office rents to 100µ an v µ, respectively, which lowers the rewars the external party has to offer. The game tree below summarizes events, where E, I, an R are the external party, the incumbent an the challenger; is emocratize, n o not emocratize, t is take power an nt is o not take power. I look for subgame perfect equilibriums. exit Stage 1 E: { P I, P R } Stage 2 I n Stage 3 Stage 4 nt R t R n E: 0 I: 0 R: v-c E: 0 I: 100 R: 0 E: W- P I I: P I R: 0 E: 0 I: 100 R: 0 E: W- P R I: 0 R: P R -c 3.1. Unchallenge Incumbent The situation without the challenger is a benchmark. Now the external party negotiates irectly with the incumbent. This really requires a separate game but is easily capture in R the game tree above by assuming that E is force to set p = 0. Then, since c> v, the I challenger has no incentive to take power an E sets p = 100, the incumbent emocratizes an the payoffs for E an I are W 100 an 100. However, beyon the financial cost, this act of simply buying off the incumbent coul be perceive as unfair 8

9 ex-post, which woul eroe its creibility ex-ante. Also, this metho coul set a ba preceent encouraging the rise of new ictatorships Challenge Incumbent With the challenger involve in the game, backwar inuction shows that in the unique I subgame perfect equilibrium E sets p = 0 an p R = c. Now, if the incumbent oes not emocratize in stage 2, the challenger will take power an emocratize in stages 3 an 4. The challenger prefers to emocratize in stage 4 since c> v an to take power in stage 3 since p R c= c c 0. Consequently, the incumbent can o no better than to emocratize in the stage 2. This gives the central result of the paper: emocracy can be achieve at zero cost. In principle, E can also conition the payment to the challenger R p on the incumbent being punishe after R takes power. This will raise the incumbent s incentive to emocratize in stage A Democracy Movement as the Challenger This payments scheme may be akin to supporting a emocracy movement, which is a common policy avice. In the scheme, however, first, the challenger is only pai if the incumbent oes not emocratize. Secon, the rewar c turns a self-intereste challenger into a e-facto emocracy movement, so in principle emocracy forces nee not preexist in the country. However, emocracy movements o have a special property as challengers, which I now iscuss. Suppose that a emocracy movement can be moele as altruistic with large benefits from regime change βw, β (0,1]. 6 Then its failure to take office before toay implies that c > βw, where c is the movement s takeover cost otherwise, the movement woul have taken power alreay. This alone suggests that using emocracy movements are a costly way to provie emocracy (although no payment is mae in equilibrium). However, notice also that the emocracy movement is R actually willing to challenge for a rewar p = c β W < c : the movement partly 6 β is the movement s share of the aggregate benefit of emocracy an can reflect the size of the movement in the population, its intensity of preference for emocracy, an members egree of altruism. β large s plausible for a broa base movement an a moerate population size if emocracy is a public goo 9

10 internalizes the payoff of the external party. Thus, if c β W < c, paying a emocracy movement is more cost effective than paying an orinary self-intereste challenger. 7 If the external party E represents evelope country governments, it may also fin it easier to convince its omestic voters to rewar the emocracy movement A Creit Constraine Challenger Suppose that the challenger R has zero initial wealth an woul have to receive a transfer of c up-front to make the takeover threat creible. This transfer must occur between the time the incumbent ecies to emocratize or not an the challenger ecies on a takeover. Now, if E offers no aitional payment once the challenger is in power, the challenger will trae off v for holing on to power for zero if it emocratizes an will stay in power. Then E s contingent payment scheme must inclue a pre-emocratization payment of p R pre = c an a emocracy-contingent payment p R = v, making the total cost of emocracy (as before, out of equilibrium) c+ v. More generally, if the challenger has initial wealth a c, ( c a) must be pai upfront an max{ v, a} after the take-over, implying total cost ( c a) + max{ v, a}. 8 Thus, for a< v the total cost eclines in a an if a v it is constant at c. E shoul then prefer a wealthy challenger. Altogether, the I optimal scheme is = max{ c a,0}, p = 0, p R = max{ v, c max{ c a,0}}. 9 With p R pre 7 A complication may be if the emocracy movement prefers to ecline the offer an hope that the external party will fin someboy else to emocratize the country, giving the movement a net gain of c. I abstract from this complication, but at worst it means the emocracy movement woul have to be pai c > c βw. The moel also technically rules out such free-riing since E makes a take-it-or-leave-it offer. 8 E minimizes R R R R p pre + p subject to p pre c a p v R R, an p p c pre + (an non-negative payments). E is best off making the takeover just feasible an backloaing the rewar for emocracy, that is, withholing as much as possible of the payment until R has fulfille all its uties. A takeover can be ensure by p R pre R R = c a, so by the last constraint c a+ p c p a, which combine with the secon constraint gives the state result. The external party may further improve the outcome by traing off the higher takeover probability if she raises the rewar for the greater cost. For simplicity, I maintain the perfect information assumption, however. 9 This follows by replacing c a p R pre = with max{ c a,0} p R pre = in the previous footnote. 10

11 ifferent takeover costs among potential challengers, E shoul look for the challenger i so that max{ c a,0} + max{ v, c max{ c a,0}} is lowest. i i i i i i 3.5. A Creit Constraine Democracy Movement If E sponsors a emocracy movement with zero wealth, then the movement must be pai c up front. However, with a emocracy movement there is no agency problem once it hols office sinceβ W > If E can make the emocracy movement repay the ifference between its actual an reservation payoffs c ( c β W ) = βw once it has taken office then the creit constraint is irrelevant. Even if the movement cannot be force to repay, however, presumably a emocracy movement is part of the people an the external party is altruistic, so giving a surplus to the emocracy movement may be acceptable. 11 Thus, at low wealth levels, even a very high cost emocracy movement may be preferable because, unlike an orinary challenger, it presents no agency problem once 10 One coul imagine even a emocracy movement has gains to retain office after taking power. If that rewar vˆ satisfies office rent β W < vˆ vˆ βw, then an agency problem arises even with a emocracy movement: it has net. However, this largely correspons to the emocracy movement acting like the selfintereste challenger. 11 Formally, suppose that the movement is part of the people. Suppose first that E can enforce repayment of βw from a emocracy movement let M with initial wealth a. E minimizes the takeover cost p p M pre M pre p M + subject to the creit, incentive compatibility (IC), an iniviual rationality (IR) constraints M M M c a, p βw an p + p c βw. E can o no better than set = c a pre p M pre an satisfy the IR constraint with equality by choosing p M a βw =. This implies that IC hols. E s payoff is the payoff to the emocracy movement plus the payoff to remaining citizens minus the cost of hiring the emocracy movement ( βw c + ( c a) + ( a βw )) + (1 β ) W ( c a) ( a βw ) = W c M suppose that E cannot enforcement repayment. Then IC becomes 0 p. If a βw. Alternatively, > the same solution as before applies. If not, then E sets p M pre M = c a, p = 0 an the movement earns a surplus β W c + c a) = βw a ( rather than its zero surplus before. However, the altruistic external W c + ( c a) + (1 β ) W ( c a) = W c β. party still earns ( ) 11

12 in office. Formally, this requires c < ( c a) + max{ v, a}. However, I now suggest ways for the external party to avoi the agency rent of a self-intereste challenger Avoiing the Challenger s Agency Rent Rather than pay the rent ( c a) + v c= v a to a self-intereste challenger with a< v, another option may be open to the external party: once in power, a challenger who refuses to emocratize for compensation a (making its total profit ( c a) + a c= 0 ) may in principle be treate like the incumbent was. That is, the contingent payments scheme can be applie as p I 2 = 0, 2 p R 2 = c with I2 an R2 enoting the new incumbent an the new challenger an 2 c the new challenger s cost. The secon challenger can be bought for only c 2 because she is threatene herself by a thir challenger if she refuses to emocratize for compensation a 2 once in office an so forth. This of course requires an infinite sequence of challengers, but at the n' th challenger the cycle can start over. Thus the agency problem inuce by a creit constraint may be resolve by a scheme I p = 0 p Ri pre i i = c a, Ri p = a i where Ri p pre an Ri p are the pre- an post payments mae to the i' th challenger Avoiing the Challenger s Agency Rent with Collusion One may conjecture that with n small the challengers coul collue (or collue with the incumbent, but I iscuss this case below), so that effectively there is only one challenger an it seems that the external party must pay ( c a) + max{ v, a} after all, where the variables now escribe the challengers as a group. The challengers coul then earn a rent up to v. However, first, with sufficient rewars i n { c } i= 1 even a country which appears to have few potential challengers at first sight may prouce a large number n (e.g. ethnic, religious, 12

13 class or region base groups, military factions), incluing factions within the current government. As long as one or more challengers cannot be persuae to join the coalition the competitive race between challengers remains. Though it may be harer to take power from a coalition of many challengers, this shoul in principle only raise c i for the challengers outsie the coalition an emocracy is obtainable at no cost as long as the scheme remains creible. Secon, suppose that all challengers o collue. To sustain this coalition, the original challenger must pay each of the others a fraction of v an it is willing to pay up to v to avoi a takeover. However, even if it istributes the full v to the other challengers, the external party can in principle simply top the offer. For example, if a group of size from a pool of symmetric challengers can make or break the office of the original challenger, then the external party can commit to paying the take-over cost plus a premium of v / nˆ to each among a pre-selecte group of nˆ challengers to take power from the original challenger if it fails to emocratize. As long as this is creible, zero cost emocracy can still be obtaine. The challengers can no longer commit not to betray each other. n< ˆ n 3.8. Ensuring Creibility While the external party pays nothing for emocracy in equilibrium, the out-ofequilibrium threat to pay c to the challenger (or, in the worst case with maximal agency rent, up to c+ v ) must be creible. Otherwise, the equilibrium unravels. Several factors coul enhance creibility, however. First, ieally the payment only covers the takeover cost, but even in the worse case where neither the option to use a emocracy movement nor the options to limit the challenger s agency rent iscusse above can be use - the rent will still not be pai to an incumbent with a checkere history but to a new political player with a better history, limiting perceive injustice an setting no preceent of payoff off a ictator. The payment funs coul also be manage by an inepenent international institution. 13

14 Secon, if E can buil a reputation for making goo on its promise payments, it can make similar schemes creible in other countries, so it has a natural incentive to follow through. Thir, buying emocracy of valuew for a cost of no more than c< W after the takeover is ex-post efficient an is consistent with citizen preferences. If citizens repay the external party, then the latter also makes a non-negative profit. Fourth, as long as an ex-post payment of v is creible, E can make the ex-ante payment max{ c a, c v} to any challenger before the incumbent ecies to emocratize. This applies the well known result that a sunk cost can make a threat creible because the resiual cost of carrying out the threat is smaller. Fifth, the payment for emocracy is narrowly targete to the challenger who now hols the key to emocracy. As iscusse earlier, this may be more effective than, for example, a promise of more future trae, ai, foreign investment or invitations to join prestigious international organizations: such rewars may bring large aggregate benefits, but the irect benefit to the power holer may be limite (Krain 2005, Stanton 2004, Valentino 2004). Sixth, as a political outsier the challenger may lack popular legitimacy an increasingly so if it fails to emocratize an be unable to manage the complex system of rent extraction, patronage an governance evelope by the previous incumbent (Rajan an Zingales 2003). This shoul lower the value of v an eventually the maximum cost of the takeover ( c a) + max{ v, a} for a wealth-constraine challenger, while it leaves the cost at c for challengers unconstraine by wealth Renegotiation Offer by the Incumbent Suppose that after E proposes the plan, but before the incumbent emocratizes in the first perio or E makes any payments to the challenger, the incumbent offers to emocratize in exchange for 0 < x< c, where c is the cost of getting the challenger to emocratize instea. Thus, the incumbent seeks to exploit his cost avantage ( 0 < c ) in supplying emocracy. The incumbent may have some leverage because if the bargain fails E gets W c rather thanw, paying an extra c for emocracy. Using the Nash proceure, the bargaining surpluses earne by E an I, enote x an y, respectively, solve 14

15 max( x ( W c))( y 0) subject to x, y x + y= W, which leas to y= c / 2. Thus, E saves half of the cost of refusing the bargain an may be tempte to bargain. If we further assume that the challenger, once in power, will renegotiate to get a payment higher than c, the loss for E if the first perio bargain fails is even greater an the incumbent woul earn even more. However, just as E may wish to establish a reputation for paying I R its promise rewars for emocratizing { p, p }, it may want to establish a reputation for not negotiating with incumbents, who are after all similar to hostage-takers in emaning payment in exchange for the political freeom of citizens. Accepting a bargain which rewars the incumbent may also be seen as unjust or prevente by retribution motives, as iscusse earlier Collusion between Incumbent an Challenger Suppose that the incumbent can ecie not to emocratize in the first perio, but before the challenger ecies on taking power or not the incumbent offers her a share of the incumbency rent of 100. As with the case of colluing challengers, however, first, the incumbent may not be able to collue with a whole competitive stream of challengers an, secon, whatever offer the incumbent makes to the challenger the external party can top it: even if the incumbent offers the challenger the full 100 to avoi the takeover, the external party can promise to pay the takeover cost plus 100 as well, making the takeover threat creible. 12 In other wors, the challenger cannot commit to conuct the takeover shoul the incumbent fail to emocratize. As an example of the limitations facing incumbent trying to pay off challengers, in 2005, Zimbabwe s total military spening was 4% of GDP. This ignores transfers mae to military agents outsie the military buget, such as lan ownership transfers, but on the other han less than the whole military buget is a pure rent. Since furthermore the military is a powerful player in Zimbabwe s politics by most account, I will assume that the military buget is about half of Mugabe s office rent. In that case, the most Mugabe 12 For example, Sunni militants recent turn against al-qaea in Iraq may reflect a belief that, financially, the US is the best long run bet for an ally. 15

16 coul pay the military to avoi a coup is 8% of GDP per year. This is a high offer to beat for the external party, but harly impossible, since external official ai to Zimbabwe in 2005 was 11% of GDP. If a takeover coul lea to an improve political an economic situation in currently chaotic Zimbabwe, then foreign spening on a takeover is more cost effective than spening on ai. Likewise, in Mauritania with a emocratically electe presient topple in a coup in August 2008, the 2005 military buget-to-ai ratio was 33.6% (close to Zimbabwe s 36.4%), for Guinea-Bissau with a faile coup attempt in November %, an for Sub-Saharan Africa as a whole 33.3% (Worl Development Inicators 2005 an author s calculations). 13 Therefore, even oubling transfers to the military woul require a limite rise in external funs. In fact, the recent coup efforts in Mauritania an Guinea-Bissau suggest that the military may alreay be earning close to the full office rent in those places (otherwise, rulers coul have averte the coup attempts by paying the military more). This may also be true in Zimbabwe currently:..troops are not being issue with firearms as senior officers are no longer sure who they can trust Low-ranking soliers have riote in the streets of Harare on several recent occasions after being unable to obtain cash an Officially the Zimbabwe National Army is 40,000-strong, but informe estimates put its real numbers at a maximum of 30,000, with its ranks thinning ramatically ue to esertion by soliers an officers up to the level of captain (Telegraph, Dec. 8, 2008) o Challengers or Preemptive Takeover, Suppose that there simply are no challengers which can be promise a rewar to create a creible takeover threat or that a single existing challenger can choose not only to take power an emocratize but also to block emocratization by the incumbent. For example, in some countries a monolithic military may have a firm grip on power an not be susceptible to internal ivisions or the military can easily conuct a coup if a civilian puppet leaer is not obeient. Then, since the military is not reware in equilibrium uner the scheme it may prefer an easy preemptive coup before the incumbent 13 Unfortunately, no ata for military expeniture for Iraq an Afghanistan, which I iscuss below, are available after

17 emocratizes. Either way, the external party must negotiate with an internally unifie an unchallenge office holer. In this case, policy options may inee be limite. However, first an perhaps least attractive, simply paying the price to buy out the incumbent may be worth paying. There must then be guarantees against reform reversal. This can partly be helpe by backloaing the payment to the incumbent, but tying its hans to commit to non-reversal is also in the incumbent s interest to earn the external party s trust. A military regime, for example, may commit to non-reversal of emocracy by restructuring, emobilizing, integrating iverse ethnic groups an proviing eucation an economic growth to raise the opportunity cost of time for emobilize soliers an generally promote popular support for emocracy an lock in the requisite institutions. Secon, an perhaps more attractive, the external party coul provie its own challenger by eploying military forces, offering a rewar to a neighboring country to invae an accomplish a takeover, or by hiring a private military company with a reputation concern preventing it from colluing with the incumbent. South African base Executive Outcomes playe a key role in suppressing insurgencies in Angola an Sierra Leone in the 1990s (Howe 1998) an there was no switch of loyalties espite the moerate payment an the iamon wealth the UNITA or RUF rebel groups might have offere. Alternatively, if the problem is a preemptive takeover by a challenger to block a emocratizing incumbent, the task of the externally provie force may be to block the challenger from blocking emocratization (as Executive Outcomes may have one in the two examples since the governments of Angola an Sierra Leone both were emocratically electe) Democracy Reversals I assume above that once emocracy is extene it is irreversible (or very costly to reverse). This may not always be the case. In particular, suppose that in equilibrium the incumbent were to emocratize in the first perio so that the challenger never gets to be in power, but that the challenger can stage a coup against the new emocratic 17

18 government. However, the starting point of the moel is precisely an unemocratic incumbent, so as long as incumbents can be force to emocratize without compensation there is no point in staging the coup in the first place. Aitionally, if the takeover cost is still c> v or simply above v when starting from emocracy instea of autocracy, then there is no coup incentive. 4. Relate Sources of Regime Change 4.1. Classical Kingmaking The payments scheme may resemble classical kingmaking : by structuring incentives to persuae the current incumbent I to emocratize an the challenger R to take power if I refuses, the external party E is essentially implementing a new regime in the country. Historically, however, kingmaking has a mixe recor. For example, controlling for reverse causality, Iyer (2007) shows that Inian states with irect rather than inirect British rule in the colonial era have worse public goos outcomes at present an Easterly et al. (2008) show that eveloping country rulers installe or proppe up by either superpower in a given five year perio uring the Col War were associate with substantially ecrease emocracy in the following five year perio. 14 Domestic militaries have also backe or installe unemocratic rulers historically, for example in Algeria, Brazil, an Chile (Werenfels 2004, Gees 1999) an Acemoglu et al (2000) an Dell (2008) ientify long run etrimental effects of what the former label extractive institutions set up by colonizers. However, in all of these cases the external actor, unlike above, probably i not aim to improve omestic welfare but to further its own economic or political interests (Bueno e Mesquita an Downs 2006), so supporting a ruler with congruent preferences an incentives i not mean a benevolent ruler. In fact, given the external party s objectives some of these kingmaking plans may have been successful. 14 Bueno e Mesquita an Downs (2006) summarize the effects-of-interventions literature. Acemoglu an Robinson 2006b present a theory relevant to kingmaking. For recent non-acaemic work on kingmaking in the Mile East, see Brysac an Meyer (2008), on Central Asia Hopkirk (1992) an Brysac an Meyer (2006), an on Equatorial Guinea Roberts (2006). An amazon.com search on the CIA alone yiels ozens of books etailing such interventions. 18

19 With slight moifications, the moel coul precisely escribe self-intereste kingmaking: instea of promising c to a challenger who takes power an emocratizes, the external actor coul promise c if the challenger takes over an subsequently hans the reins to the external actor whenever I refuses to hans over the reins immeiately. Using the scheme to achieve such malevolent takeovers, however, may not reverse its normative benefit, because one can simply treat the former external actor as the current incumbent an a truly welfare maximizing external actor, E2, can then apply the scheme. In other wors, E2 can offer rewar p R2 R2 R2 2 = c to a new challenger R2 with cost c. This plan is E welfare improving as long asw > v, where actor. 15 E v is the value of power for the external 4.2. Democracy from Above Like kingmaking, achieving emocracy from above, an especially when backe by external actors, tens to be problematic (Herman an Broahea 1984, O Donnell et al. 1986, Whitehea 1991). Recently in Iraq an Afghanistan emocracy builing has worke partly at best. Moreover, even countries who emocratize with a substantial omestic impetus for emocratization an without being force to emocratize by military conflict of the threat thereof, may have problems. This has been true in Latin America, Africa an the Former Soviet Union uring the 1980s an 1990s, where at most electoral emocracy but harly eep emocracy exists. Often, leaers opt for an appearance of emocracy to win international or omestic legitimacy an then restrict civil liberties, intimiate opponents or change constitutional term limits. Recent examples are Colombia, Kenya, Russia, Rwana, Venezuela, an Zimbabwe. However, in all of these examples leaers may in fact have vali (if self-intereste) reasons to block eep emocracy because eep emocracy woul threaten their office rents. In contrast, however, in the emocratization scheme of this paper, both the incumbent an the challenger have incentives to emocratize: the incumbent faces a creible threat of 15 In an extreme case, if E W v - for example, if the country is a very important pawn in a game between superpowers an its value epens on it being a non-emocracy - then for E to hol power is efficient (at least in the partial equilibrium which ignores the welfare effects for the other superpower an the rest of the worl), although it is etrimental to citizens - who can in principle but may not in practice be compensate. 19

20 expulsion, so she has no office rent to protect, an the challenger either has no office rent as well ue to a secon takeover threat or she is compensate for emocratizing. Amittely, getting ri of the opposition is a way for the incumbent to remove the challengers before they can take power (Mulligan an Tsui 2006). To avoi such actions by the incumbent, the external party coul a either a premium for going out peacefully or a penalty for an incumbent who is epose after resisting. The former approach has been recently taken by the Mo Ibrahim founation 16 which rewars African leaers $500,000 for ten years an $200,000 per year thereafter million ollars per year for their remaining lifetime if they follow proper political proceures uring their career. More precisely, as the founation website explains The Prize is aware on an annual basis to a former executive Hea of State or Government in sub-saharan Africa. Eligible caniates will have taken office through proper elections an left having serve the constitutional term stipulate when taking office. ( Also, even if the incumbent still prefers to repress the opposition, first, she cannot repress external official or private military forces. Secon, if the incumbent equips the military to repress the opposition toay he has no guarantee that the military will not be tempte to epose him tomorrows. As iscusse in section 3.10, there are limits to how much the incumbent can pay the military to remain loyal an the external party can potentially top that offer. This being sai, however, there may ultimately still be incumbents for whom the scheme leas to engame behavior, which can be extremely estructive. The engame problem shoul therefore receive very careful attention (Mulligan an Tsui 2006). A secon problem, on which the moel is mostly silent, however, is that even willing incumbents may be unable to emocratize because the prerequisites for eep emocracy cannot be provie top-own or overnight (Finlay 2007, Mill 1984). However, the incumbent or challenger oes at least have incentives to facilitate this process an they are free to seek avice an elegate the task to others, incluing popular reform movements (Diamon 1992). The relevant comparison also is not the status quo versus a perfect emocracy but the status quo versus the best achievable emocracy an certainly 16 See 20

21 all Western emocracies has a long an turbulent evelopment perio. That alone harly iscreits reforms. Lastly, countries which were initially emocratic but saw usurpation of power by leaers or military coups may be easier to reform since the country has a recent experience with emocracy. Those cases may also make for more legitimacy an easier jugment calls on whether to apply the scheme at all. For example, in 1994 the US Clinton aministration threatene to invae Haiti (in Operation Uphol Democracy) unless presient-by-coup General Ceras returne power to the emocratically electe Presient Aristie whom he ha espose in That sovereignty infringement by the US was not perceive as illegitimate by US voters an Ceras i inee cee power to Aristie. Finally, the expectation that a contingent payments scheme will be offere in the future may encourage challengers an emocracy movements to grow in the present an the external party can then pick the most cost-effective challenger to inclue in the rewar scheme. Therefore, even a country which appears to have no reainess for emocracy may improve its emocracy basis in response to the scheme Iraq an Afghanistan toay The most visible emocracy-abroa projects in the worl toay may be the US-le attempts to emocratize Iraq an Afghanistan. Although the final outcomes cannot be juge yet an a full analysis is beyon the scope of this paper, the prospects compare to the costs so far are consiere iscouraging by many. For example, critics argue that while both countries now have emocratically electe governments, the principles of that emocracy are not generally accepte. Aitionally, local political, social, economic, or even military control of parts of each country is in the hans of warlors an traitional or religious leaers, bureaucrats succumb to corruption, an government capacities are weak. Many fear that once foreign powers exit the security situation will eteriorate as social groups vie for control (Blattman an Miguel 2008). Thus, emocracy builing in Iraq an Afghanistan may suggest infeasibility of the kin of emocratization scheme propose in this paper. However, Iraq an Afghanistan iffer from the scheme propose above in important ways. First, rather than post a rewar for government takeover followe by 21

22 emocratization, external actors intervene irectly to overthrow the incumbent Ba ath an Taliban regimes. Therefore, they ha to pay the takeover cost which is not pai in equilibrium in the moel, an while the overthrow itself was cheaply obtaine in both cases, the military invasions an inuce state failure an arme conflict will have large long term costs. Ha the incumbent regimes been persuae to emocratize voluntarily an responsibly uner a scheme similar to here, then violence, state failure an instability might have been avoie. With incumbent cooperation, since each incumbent ha military control of its country at the time, it is unclear that insurgencies uring the transition woul have been feasible an popular support for such insurgencies might have been limite ue to the peaceful emocratic en goal coorinating expectations (Rolan an Verier 2003). Aitionally, even if the incumbents i have to be overthrown amittely, the US i issue threats without effect - a sponsore challenger may have been a more efficient means. 17 Inee, perhaps the overthrow threat was not taken seriously precisely because the two regimes knew how costly invasions woul be for the US. The Taliban at least ha many enemies, limite control of the population, an the power of the state was weak, suggesting potential challenger i exist in Afghanistan. Iraq was more tightly controlle an fining a creible challenger may not have been easy. However, only the Sunni minority benefite from the Saam regime while Kurs an Shiites were represse. Moreover, far from all Sunnis benefite an only staye loyal perhaps because a Shiite or Kurish controlle government woul be worse (Paro I Miquel 2007). That concern shoul be less uner a creible emocratization scheme. Secon, it cause problems in Iraq an Afghanistan that the external party became the new incumbent. First, the ifferent culture, history of non-benevolent kingmaking, 17 Military contractors such as Blackwater have been wiely use by the US government in the War on Terror, although for routine an well efine tasks rather than government overthrows, emocratization, or other long run goals. While presumably such contractors alone cannot ensure both a government takeover an viable emocratization, they may be hire by the challenger to support it militarily. As mentione, South Africa base Executive Outcomes probably playe a crucial role for the emocratically supporte governments of Sierra Leone an Angola when critically threatene by insurgencies in the 1990s (Howe 1998). 22

23 intimiating power, an well known economic or strategic interests of the Westerns powers in both Iraq an Afghanistan create suspicion (Carothers 2004). Secon, the external parties were ill informe about the right way to reform each country (Easterly 2002, 2006). Thir, the massive funs they brought in le to corruption, rent-seeking, Dutch Disease-like an agency problems. Fourth, the fact that one ay they will be gone leas to uncertainty an possibly anticipation of future group conflict (Englebert an Ron 2004), although amittely the Western powers o plan a graual withrawal. 18 These problems shoul be less severe (if still present) with a omestic challenger: after eposing the incumbent, the challenger has only a single an transparent goal of emocratization, for which it is only reware once the goal is achieve; it shoul have better local knowlege; the inflow of foreign capital uring the transition time woul be less; an its military an political presence in the country is not expecte to isappear before the goal is transition is complete How Woul This Work in Practice? A vali objection to the moel may be the following: In practice, it is unclear that rulers will always accept a eal even with a creible takeover threat. For example, they may appreciate prestige, be paranoi an mistrusting of foreign agents trying to propose the eal (especially one involving a threat of overthrow if they ecline), or see themselves as national heroes. They may also have their hans tie by next-in-comman cares, military groups, or the clan, regional or ethnic groups they belong to Although the US an others have prestige at stake, they face no takeover threat if they fail to achieve emocracy an if ultimately a stable an frienly emocracy cannot be achieve, then they may prefer stable an frienly over emocracy. For example, the popularly supporte emise of the Shah in Iran harly helpe the US, nor i the fall of right wing ictatorships in Latin America which brought to power a generation of leftist populists. Other examples of emocracy not helping Western interests may be Russia an the Palestinian Authority. US emocracy assistance per country eeme eligible for emocracy assistance in a recent stuy was $2.07 million over an $3.66 million (9% of total US ai) in 2003 (Finkel et al. 2006). These numbers are substantial but smaller than the emocracy focuse rhetoric woul suggest. 19 Zimbabwe is a recent example, see Military Leaers Making the Decisions in Zimbabwe, Washington Post Foreign Service, April 16, 2008, p. A12. 23