Voting Rules in a Changing Environment Antoine Loepery Wioletta Dziuda September 2014 DRAFT Abstract We identify a novel distortive e¤ect of supermajoritarian voting rules such as quali…ed majority, …libuster or unanimity. Under such rules, di¤erent voters become pivotal under di¤erent status quos. If policies are continuing in nature and a changing environment creates the need for renegotiation, legislators distort their voting behavior in favor of alternatives that make them pivotal in the future. As a result, players disagree more often, and the policy is less responsive to the environment. We show that the voting distortions are bigger, the greater the supermajoritarian requirement. The distortions generated by supermajoritarian requirements are larger when the environment is volatile. In the last decades, multiple U.S. states have passed constitutional amendmends that require a quali…ed majority to increase taxes. Arguably, these amendments are designed to reduce spending and tax levels. Our results imply that such amendments can severely limit the states’ ability to adjust taxation throughout the business cycle. We also show that they can even fail to achieve their primary goal of keeping taxes low. JEL Classi…cation Numbers: C73, D72, D78 Keywords: Dynamic voting, supermajoritarian voting rules, quali…ed majority, …libuster, checks and balances, continuing policies, polarization, policy inertia. y Kellogg School of Management, Northwestern University. Email: [email protected] Universidad Carlos III de Madrid. Email: [email protected] 1 1 Introduction The …libuster has come lately under heavy criticism. Commentators and pundits have argued that with a highly polarized Congress, supermajority requirements lead to policy inertia and legislative gridlock. Such inertia prevents Congress from reacting in a timely fashion to a volatile environment.1 Even though supermajority requirements frustrate proponents of reforms, it is not clear whether their costs outweigh their bene…ts. They prevent e¢ cient reforms, but they also prevent ine¢ cient ones, and the total e¤ect is ambiguous. Economists and political scientists have predominantly found that supermajority requirements and veto power have desirable properties. They reduce the likelihood of irrational collective decisions (Condorcet 1785), can mitigate dynamic commitment problems (Gradstein, 1999; Messner and Polborn, 2004), and most importantly, prevent policy makers from implementing laws that bene…t only a bare majority, and force them instead to internalize the interests of a greater share of the population (Buchanan and Tullock, 1962; Riker, 1962). In this paper, we provide a novel strategic rationale for the suboptimality of supermajority requirements and veto power. We consider policies that are continuing in nature but require periodic revisions in response to exogenous shocks. In such dynamic settings, supermajoritarian rules implicitly tie the future bargaining power to the current policy, creating incentives to distort voting behavior. We show that these distortions make legislators appear more polarized than they are, and hence lead to excessive policy inertia. In other words, it is not polarization that renders …libuster ine¢ cient; it is the …libuster that leads to a polarized behavior. To see the intuition for this rationale, consider legislators that are ordered on the ideological spectrum. When voting, each legislator takes into account not only how adequate the policy is to the current environment, but also how it a¤ects the identity of the pivotal player in the next periods. Under simple majority, the same median legislator is pivotal independent of the current status quo. Hence, when voting, the median legislator considers only the policy’s adequacy to the current environment, because she expects to be pivotal in the next periods irrespective of the outcome of 1 A …libuster is a parliamentary procedure where one or more members can delay or entirely prevent a vote on a given proposal by extending the debate. In the current political environment, the …libuster e¤ectively means that most major legislation requires a 60% vote to pass. 2 the vote today. Under a supermajoritarian rule, the identity of the pivotal player depends on the policy change under consideration: If the status quo is to the left of the proposed policy change, the pivotal player is more leftist than the median legislator; if the status quo is to the right of the proposed policy change, the pivotal player is more rightist than the median legislator. Therefore, in order to increase the probability of being pivotal tomorrow, relatively leftist (rightist) legislators prefer leftist (rightist) status quos. These incentives induce legislators to vote in a polarized way, and hence lead to more disagreement and status quo inertia. We show that the resulting polarization is bigger, the greater the supermajoritarian requirement is. Hence, increasing the required supermajority leads to more political inaction not only because policy reforms require the approval of a greater number of legislators, but also because all legislators behave as if they were more polarized. Understanding the e¤ects of quali…ed majorities on policymaking is of great importance as many decision bodies operate under such voting rules, and consider or have recently implemented changes to these rules. For example, the EU requires unanimity or quali…ed majority for decision making in the European Council, and generally requires an agreement of the Council and the Parliament. Since the Single European Act, the scope of the voting procedure for unanimity has become more restricted. The Lisbon Treaty further increases the number of areas where quali…ed majority voting in the Council will apply. Our results provide yet another rationale for such changes. In the last decades, several U.S. states have passed constitutional amendments that require a quali…ed majority to increase taxes.2 These amendments were designed to reduce spending and tax levels.3 Between 1996 and 1999, the US Congress has voted every year on an amendment to require two-thirds supermajority in order to increase taxes.4 Our model implies that such amendments, by leading to excessive polarization, can severely limit the states’ability to adjust taxation throughout the business cycle. More surprisingly, we show that they can even fail to achieve their primary goal of keeping taxes low: In our dynamic setting, the policy change that requires higher majority may actually become more prevalent. We show that the distortions created by the supermajoritarian requirements are larger when the 2 See Table 1 in Section 6. If the preferences of the society coincide with the median voter, one should not expect the society to introduce measures to curb taxes. Gradstein (1998), Messer and Polborn (2004) argue that quali…ed majority can be adopted by a median voter as a commitment device not to raise taxes in the future. 4 These attempts failed. 3 3 environment is more volatile. Intuitively, in a volatile environment, the current policy is likely to require a revision next period; hence, securing pivotality for the next period becomes more important than implementing an adequate policy given the current circumstances. This result provides a new rationale for endowing the governments with power to introduce state of emergency during which the usual legislative procedures are suspended and the decision power becomes concentrated. This is also in line with the rules of the EU. The EU requires unanimity for membership of the Union, harmonization in the …eld of social security and social protection, the common foreign and security policy, the granting of new rights to European citizens, anti-discrimination measures, and certain institutional issues such as the electoral system and composition of the Parliament and committees. Arguably, the stance of the member states on these issues is unlikely to change rapidly. On the other hand, the EU requires smaller quali…ed majority for other regulatory areas that are arguably more a¤ected by recurring shocks, such as immigration, crime prevention, transport, structural and cohesion funds, the EU budget. In purely democratic systems, the decision process is based on the simple majority rule. However, in modern democracies, bills and policy reforms have to pass several additional institutional hurdles to be enacted. These checks and balances can take many forms: presidential vetoes, concentrated power to set agenda, committee approval, bicameralism, the possibility of public initiatives, or judicial review by a constitutional court. Their original intent is to limit the concentration of power by increasing the set of individuals whose approval is necessary to enact policies. Our analysis highlights a detrimental, unintended consequences of checks and balances: by changing the identity of future pivotal players, they induce decision makers to behave in a more polarized way, and thus prevent them from implementing Pareto improving reforms. In Section 5.4 we discuss institutional arrangements that can mitigate the polarizing nature of checks and balances. The existing literature on voting rules considers mostly static environment, and hence identi…es their static properties. In contrast, we focus on a dynamic environment in which decision makers’ preferences are a¤ected by shocks. This is arguably relevant, as the preferences of the legislators are stable in few policy areas. For example, legislators’ preferences over …scal policies re‡ect heterogeneous ideologies and constituencies, but they also depend on the business cycle, changes in the country’s credit rating, demographic trends, or the vagaries of public opinion. Similarly, citizens’and legislators’preferences over civil liberties and privacies are a¤ected by shocks such as 4 terrorist attacks and national security threats. The paper is organized as follows. Section 2 discusses the related literature. Section 3 describes the basic model. Section 4 characterizes the general properties of all equilibria. Section 5 shows how these properties vary as we vary voting rules and volatility of the environment. Section 5.4 concludes. All proofs are in the appendix. 2 Related literature Studying the properties of various voting rules has been a subject of Public Choice for a long time. The literature has found that various rules may be optimal depending on the considered environment as well as the assumed optimality criterion. Quali…ed majorities seem to dominate, however, in a broad class of problems. Since Condorcet (1785), the social choice literature has argued that majority rule may note aggregate heterogeneous preferences in a rational way (Plott 1967). Caplin and Nalebu¤ (1991) show that for a broad class of preferences distribution, a quali…ed majority can solve the Condorcet paradox (see Enelow 1997 for a survey of that literature). Voting rules have been also analyzed with respect to their information aggregation ability (see Grofman and Owen, 1986, for a summary). Austen-Smith and Banks (1996) show that strategic voters may not vote sincerely. As a consequence, Feddersen and Pesendorfer (1998) show that unanimity rule can have perverse incentives. Coughlan (2000) argues, however, that unanimity can be optimal if communication among jurors is allowed and jurors have similar preferences. Another strand of literature focused on the properties of the implemented policy changes. Wicksel (1896) has advocated that since governments are formed to bene…t the citizens, unanimity should be used to assure that all changes are Pareto improving. Similarly, Buchanan and Tullock (1962) single out unanimity as the only rule that guarantees Pareto improvements. Rae (1975) points out that under unanimity the societies may be stuck with an ine¢ cient policy as long as such policy bene…ts one player. Using a similar argument, McGann (2004) argues that simple majority o¤ers the best protection to minorities as it makes it the easiest to overturn minority-harming policies. Guttman (1998) points out that under unanimity, projects that give large bene…ts to all but a few will be vetoed (see also Rae 1969), which is socially detrimental. In response, Tullock (1998) argues 5 that in the presence of transfers, all socially bene…cial changes will be undertaken even under unanimity. Arrow (1998) counters that this result holds only under the assumption that distributive policies can violate the generality of treatment of taxpayers. However, Buchanan (1998) argues that rent-seeking costs would again most likely tilt the verdict in favor of more inclusive decision rules, as such rules minimize the resources wasted on assuring one’s position in the winning coalition. Supermajoritarian requirements have been advocated also on the basis of protecting minority from "the tyranny of majority" (a view associated with James Madison and Hayek 1960). When distributive policies are considered, majority rules create incentives to particularize the bene…ts and collectivize the costs, and the extent of that is higher the lower the required majority. As a result, ine¢ cient distributive policies can be adopted (Buchanan and Tullock 1962, and Riker 1962). Ferejohn, Fiorina, and McKelvey (1987) and Baron (1991) …rst formalized this prediction in models of legislative bargaining. Aghion, Alesina and Trebbi (2004) analyze a model in which higher supermajority means that any reforms: bene…cial as well as harmful are more easily passed. They show that a higher quali…ed majority is better when the probability of expropriation and the cost of it are higher, the bene…t from reforms is small, the distribution of bene…ts is less polarized, and players are more risk-averse. Aghion and Bolton (2003) show that when optimal contracts are possible, unanimity is the best rule. When they are not and when transfers are ine¢ cient, smaller majority may be optimal. The reason for this is that a smaller number of loser-voters need to be compensated in exchange for their vote, which restricts the deadweight loss from transfers. May (1952) shows that the majority rule is the only pairwise decisive and responsive voting rule that satis…es anonymity (all voters treated equally) and neutrality (all alternatives treated equally). With any supermajority rule, the status quo is privileged, and since the status quo is favorable to some voters, these voters are privileged by a supermajority (Rae 1975). The literature has long recognized that under higher supermajority requirements fewer changes can be adopted (see Tsabellis 2002) as any change requires an approval of a larger set of voters. That does not mean, however, that supermajority is suboptimal; the higher the supermajority is, the more e¢ cient the adopted policies are, and under all voting rules Pareto e¢ cient changes are implemented. In contrast, we show that policy inertia of the supermajoritarian rules is exacerbated by players’ dynamic concerns, and such voting rules will result in Pareto e¢ cient policies not being adopted. Persico (2004) and Hastard (2005) compare voting rules in environments in which voters make 6 costly investments in the pre-voting stage (in information and preparedness for reforms respectively). Hastard (2005) shows that higher supermajority requirements lead to more underinvestment; and hence, fewer changes being undertaken. The rationale for this is that higher majority requirement makes each party’s approval more likely to be needed for the change. Hence, parties that have invested heavily will be willing and forced to compensate those who did not for the vote in favor of the change. This decreases the incentives to invest. Persico (2004) shows that a large quali…ed majority can be optimal only if the information available to each committee member is su¢ ciently accurate. When individual information is noisy, however, such rules discourage the committee members’from acquiring information: When information is noisy, it is unlikely that all remaining members of the committee vote exactly in the same way; and hence, that the vote of the voter in question is pivotal. Our paper analyzes a di¤erent environment than the aforementioned literature. In our paper, the preference domain is restricted so that all considered voting rules are decisive. We consider common interest policies with no transfers; hence, the question of tyranny of majority is mute. All information is public, and there are no investments. Instead, we focus on a dynamic environment, where an implemented change may need to be revised later in response to an exogenous shock. There is a strand of literature which recognizes that a quali…ed majority can serve as a commitment tool. In Gradstein (1999), by adopting a quali…ed majority, the society de facto delegates the decision about taxes to voters that prefer lower taxes than the median voter does, herby committing itself to smaller future tax increases. Messner and Polborn (2004) assume that older voters prefer lower taxes than the younger ones. In such a setting, a median voter may prefer to adopt a quali…ed majority as the decision rule for the future, as such a voting rule assures that she is pivotal when she becomes older. In contrast to that, in our paper legislators would like to decrease the quali…ed majority. Moreover, contrary to the aforementioned literature, in our model, the voting rule a¤ects not only who is pivotal, but also how voters vote. In fact, the reason why simple majority rule is better is entirely due to this second e¤ect. Our paper builds upon our previous work. In Dziuda and Loeper (2014), we analyze a twoplayer model with unanimity. We show that players become polarized, interpret this polarization as partisanship, and we focus on the magnitude of the resulting status quo inertia and its robustness. In this paper we focus solely on the relationship between voting rules and voting distortions. Hence, 7 we work with a multi-player model. We show that the e¤ect identi…ed in our previous work does not extend to quasi-dictatorial voting rules. We also tie the magnitude of the distortions generated by non quasi-dictatorial rules to the volatility of the environment. Another related paper is Eraslan and Merlo (2002), who analyze a divide the dollar game in which the size of the size of the surplus to be allocated varies stochastically over time. They show that majorities lower than unanimity may lead to ine¢ cient agreements: a current winning coalition may prefer to divide the current surplus even though waiting for a bigger surplus would be bene…cial because they fear that they lose their place in the winning coalition next period. Our paper is also related to Roberts (1999), which studies a dynamic model of club formation. Even if they prefer a larger club, the current club members may vote against enlarging it out of fear that they will lose the decision power over subsequent enlargements (See also Acemoglu, Egorov, and Sonin, 2014). In our paper, the legislators veto Pareto improving policies our of fear that they lose the decision power over subsequent changes. 3 The model A set of legislators N = f1; :::; n; :::; N g are in a relationship that lasts for in…nitely many periods. In each period, the legislators must decide which of the two alternatives L and R to implement. We will interpret L as a liberal policy such as a high tax rate or high government spending, and R as its conservative counterpart. The utility of legislator n 2 N from implementing alternative R in some period t 2 N is n (t) ; and the utility from alternative L is normalized to 0; without loss of generality. Hence, if n (t) is positive (negative), legislator n prefers the conservative alternative R (the liberal alternative L) in period t: We refer to ( 1 (t) ; :::; N n (t) as legislator k’s current preference in period t, and to (t) = (t)) as the state of nature in period t: We assume that in each period, (t) = (t 1) with probability p; and with probability 1 p; (t) is drawn from a distribution P ( ) : Parameter p measures the persistence of the environment: the higher p; the larger the persistence, or equivalently, the smaller the volatility. We assume that has a …nite expectation. 8 Assumption 1 For all t 2 N; 1 (t) legislators n; m 2 N , m (t) n (t) and < :: < N (t) with probability one, and for any two distinct are of opposite sign with positive probability. Assumption 1 has a natural interpretation in political economy: legislators can be unambiguously ranked on the ideological spectrum. Speci…cally, legislators with lower indices are more leftist than legislators with higher indices. This means that whenever a more leftist legislator disagrees with a more rightist legislator, then the former prefers L and the latter prefers R: Note, however, that this assumption imposes no restriction on the preference distribution of a single legislator nor on the severity of the con‡ict of interest between legislators: all legislators might prefer the same policy arbitrarily often. In each period t; a status quo q (t) 2 fL; Rg is in place. First, (t) is realized and observed by all legislators.5 Then legislators choose which alternative to implement in that period using some voting rule ; which we describe later. The implemented policy determines the legislators’payo¤s in this period and becomes the status quo for the next period q (t + 1). Each legislator maximizes her expected discounted sum of payo¤s, and game by > 0 is the discount factor. We denote the above ( ). Voting rules A voting rule is characterized by a pair of collections of winning coalitions ( L; R) ; which determine the outcome as follows: if the status quo is L (R) in a given period, then it is replaced by R (L) if and only if the set of legislators who vote for R (L) in this period is an element of ( R ). L We impose the following conditions on the voting rule. De…nition 1 A pair of collections of winning coalition if for all q 2 fL; Rg, q =( L; R) 2N 2N is a voting rule satis…es the following: (i) Monotonicity: if C 2 q (ii) Nonemptyness: N 2 q; and C (iii) Properness: for q 0 6= q, if C 2 C 0 , then C 0 2 q, then N n C 2 = q, q0 : Conditions (i) to (iii) are standard in the voting literature (see, e.g., Austen-Smith and Banks 2000). Monotonicity ensures that if a coalition can impose R; then so can a larger coalition. 5 own This assumtion is without loss of generality: in equilibrium, each legislator conditions her action only on her k ; hence, whether she observes other legislators’preferences is inconsequential. 9 Nonemptyness ensures that the voting rule is Paretian. The properness condition means that if a coalition can change the status quo, then the legislators outside this coalition cannot reverse this change. These conditions characterize a large class of voting rules such as majoritarian voting rules, but also other nonanonymous, and nonneutral, voting rules. An example of a nonanonymous voting rule is the combination of a simple majority and a veto legislator v 2 N : R = L = C N : jCj > N and v 2 C : 2 Such nonanonymous voting rules are the de-facto rules for many democracies in which the legislative body uses simple majority but is subject to a veto of the president, or the agenda control of committees. An example of a nonneutral voting rule is the simple majority rule when q = R and a quali…ed majority rule with threshold M > R = C N 2 when q = L: N : jCj > N 2 and L = fC N : jCj > M g : (1) Such rules are used by many US states (see Section 6) to legislate taxation: while simple majority is needed to lower taxes, a quali…ed majority is required to raise them. Equilibrium We look for Markov perfect equilibria in stage-undominated strategies (henceforth, equilibria) as de…ned in Baron and Kalai (1993). A Markov strategy for legislator k in ( ) maps in each period t the current state (t) and the current status quo q (t) into a probability distribution over votes. Stage undomination requires that in each period, each legislator votes for the alternative that gives her the greater continuation payo¤. It rules out pathological equilibria such as all legislators always voting for the status quo.6 Without loss of generality, we assume that when indi¤erent, a legislator votes for R: Comments Two comments on the modeling assumptions are in order. First, the model used in this paper 6 Stage-undominated Markov perfect equilibrium is the dominant concept in the dynamic voting literature. 10 builds on the two-player model analyzed in Dziuda and Loeper (2014). We extend it to more than two players to analyze the impact of the voting rule on the equilibrium outcome, which is the main focus of this paper. We simplify it in other dimensions. Speci…cally, the policy space consist of two alternatives, and we assume a speci…c form of correlation of players’ preferences over time. The latter assumption is only for expositional simplicity. The former assumption implies that the policy making process within each period depends on the voting rule but not on the allocation of other procedural rights, e.g., amendments rights or proposal rights. Therefore, this simpli…cation allows us to isolate the impact of the voting rule, independently of the details of the bargaining procedure.7 Second, our model applies only to continuing policies: policies that do not have an expiration date and remain in e¤ect until a new agreement is reached. Many legislative decisions have this feature. For example, about two-thirds of the U.S. federal budget— called mandatory spending— continues year after year by default. Outside of the …scal sphere, many ideologically charged issues such as taxes, immigration, …nancial regulation, minimum wage, civil liberties, and national security are typically regulated by permanent legislation. 4 Equilibrium 4.1 Pivotal voters and the dispersion of power Since most of the literature on voting is cast in an inherently static setting, we will use such a setting as a benchmark. Consider hence the game in which legislators play a single period of ( ). In the unique undominated equilibrium of this game, legislators vote according to their current preferences. That is, each legislator n votes for R whenever n 0 and for L if n < 0; irrespective of the status quo. Under such behavior, Assumption 1 implies that if a legislator n votes for R, then so do all more rightist legislators k n: Therefore, by monotonicity of ; if the status quo is L; policy R is implemented whenever there exists n 2 N such that n 0 and fn; :::; N g 2 n < 0; and L: Conversely, status quo L stays in place if there exists n 2 N such that 7 One can use the technics developed in Dziuda and Loeper (2014) to extend the analysis to a one dimensional space of alternatives under some reasonable conditions on the bargaining procedure. However, such an extension would come at a signi…cant cost in terms of technical and expositional complexity, and the main results would be qualitatively unchanged. 11 fn + 1; :::; N g 2 = L: Hence, if legislator n is such that fn; :::; N g 2 L and fn + 1; :::; N g 2 = L, this legislator is pivotal when q = L in the sense that the outcome of the vote coincides with her preferences. Condition (i) and (ii) in De…nition 1 guarantee that such a legislator exists and is unique. This leads to the following de…nition. De…nition 2 For a given voting rule ; the pivotal legislator under status quo L is denoted by l ( ) and is characterized by: fl; :::; N g 2 L and fl + 1; :::; N g 2 = L: The pivotal legislator under status quo R is denoted by r ( ) and characterized by f1; :::; rg 2 R and f1; :::; r 1g 2 = R: For instance, under the unanimity rule, r = N and l = 1, while under the simple majority rule, when N is odd, r = l = then f1; :::; l 1g 62 N +1 2 . R: Note that condition (iii) in De…nition 1 implies that if fl; :::; N g 2 Together with condition (i) ; this implies that l L; r : the legislator pivotal under status quo L is more leftist than the legislator pivotal under status quo R:8 When is such that l = r; we shall say that is quasi-dictatorial. In that case, the equilibrium outcome of the static game is the same as it would be if the unique pivotal voter were the dictator. Quasi-dictatorial rules include rules in which one player is a dictator, but also simple majority when the number of voting players is odd. Similarly, majority voting with the median voter as the agenda setter is quasi-dictatorial. On the other hand, systems in which the legislature operates under simple majority but is subject to the presidential veto are not quasi-dictatorial. The following partial order on the voting rules will be useful in the subsequent analysis. De…nition 3 Let under if L 0 L and 0 and R be two voting rules. The dispersion of power is greater under 0 than 0 : R In words, the dispersion of power increases when changing the status quo requires the approval of a greater set of legislators. When restricted to majoritarian rules, a voting rule has greater 8 Our de…nition of pivotal legislators follows Krehbiel (1999). Note that l and r are also the lowest and the highest quasi-median voter of Acemoglu, Egorov and Sonin (2014). 12 dispersion of power if it requires a higher threshold to approve the change. In particular, the dispersion of power under the unanimity rule is greater than under the simple majority rule, as under the latter, a bare majority is needed to change the status quo, while under the former, an agreement of all legislators is required. It is easy to see that when the dispersion of power under 0 is greater than under , then the pivotal voters are more extreme under l( Note that when 0 ) l( ) r( ) r( 0 0 : ): (2) is quasi-dictatorial, the dispersion of power under is minimal because l ( ) = r ( ), so from (2), any further decrease in the dispersion of power must leave the unique pivotal voter, and thus the equilibrium, unchanged. 4.2 Equilibrium characterization Let us turn now to the dynamic game of ( ) : The following proposition characterizes the equilibria ( ): Proposition 1 Suppose p < 1: In all equilibria of ( ), legislators use cuto¤ strategies: there exists c 2 RN such that for all n 2 N ; legislator n 2 N votes for R if If n is quasi-dictatorial, then c = (0; :::; 0) is the unique equilibrium. If cn and for L if n < cn . is not quasi-dictatorial, in any equilibrium c; (i) c1 > :: > cN ; (ii) cl( ) > 0 > cr( ) : Proposition 1 states that each legislator votes for R if and only if her current payo¤ from R is higher than a certain threshold. The threshold is 0— and hence, there is no distortion in the voting behavior as compared to the static game— if and only if the voting rule is quasi-dictatorial. For any other voting rule, each legislator n votes as if there was a single period, and her current preferences were given by n cn instead of n: The term n captures legislator n’s preferences over the alternative implemented in a given period, while the term cn re‡ects her preferences over the next status quo, given the continuation play prescribed by the equilibrium. Therefore, one can view the term n cn as legislator n0 s intertemporal preferences. 13 When there is some dispersion of power, part (i) states that legislators’intertemporal preferences are more polarized than their current preferences in the sense that the di¤erence between the intertemporal preferences of any two legislators is always bigger than the di¤erence between their current preferences: For all n > m; ( n cn ) ( cm ) = ( m m) n + (cm cn ) > n m: Therefore, the dispersion of power exacerbates the initial polarization of the legislators. They strategically vote as if their ideological positions where more spread out than they are. To see how this polarizing e¤ect of the dispersion of power a¤ects the equilibrium outcome, note …rst that from part (i), c satis…es Assumption 1. Therefore, the same reasoning as for the static game shows that the statically pivotal legislators l ( ) and r ( ) are also pivotal in the dynamic game ( ) when the status quo is L and R; respectively. Part (ii) further states that the intertemporal preferences of these legislators are biased in favor of the status quo under which they are pivotal. More formally, in any period t in which the status quo is L and l( ) (t) 2 0; cl( ) ; legislator l ( ) prefers to stay at the status quo L even though policy R would give her a greater payo¤ in that period. Similarly, in any period t in which the status quo is R and r( ) (t) 2 cr( ) ; 0 ; legislator r ( ) prefers to stay at the status quo R even though policy L would give her a greater payo¤ in that period. Hence, the dispersion of power creates excessive status quo inertia. The magnitude of the strategic polarization and the excessive status quo inertia is captured by the magnitude of the voting thresholds cl( ) and cr( ) : The intuition for Proposition 1 is as follows. Legislator l is pivotal as long as the status quo stays at L: Therefore, to keep control of the policy in the next period, she is willing to sacri…ce her current payo¤ and cling onto L even if L gives her a lower payo¤ than R. How much of her current payo¤ is she willing to sacri…ce to remain pivotal in the next period? Being pivotal matters only if it leads to a di¤erent outcome than not being pivotal, and this happens only in the states in which l ( ) and r ( ) disagree. Consequently, as one can see from the proof of Proposition 1, the voting threshold of legislator n satis…es (1 cn = 1 p) p Z (cn f 2RN : l <cl and 14 r cr g n ) dP ( ): From Assumption 1, distinct pivotal legislators disagree with positive probability in every period, which explains why the voting distortions are nonzero whenever the voting rule is not quasidictatorial. We end this section on a technical note. Polarization in this model feeds on itself: the more r distorts her behavior, the more likely it is that status quo R stays in place, so the more important it is for l to defend status quo L; and thus the greater the incentives for l to behave in a polarized way. As a result, there may be multiple equilibria, which makes it hard to derive comparative statics. The proposition below, however, states that there exists an equilibrium in which the voting distortions are the smallest. This result allows us to perform comparative statics on the least polarized equilibrium in the subsequent sections. Proposition 2 There exists an equilibrium c 2 RN such that for any other equilibrium c0 2 RN ; c0r cr and cl c0l : In the results below, for any voting rule we call it simply the equilibrium. 5 ; c ( ) refers to the least polarized equilibrium, and 9 The determinants of polarization 5.1 Dispersion of power The following proposition shows that polarization, and hence status quo inertia, increase with the dispersion of power. 0 Proposition 3 If the dispersion of power is greater under cr( 0 ) If n 0 cr( 0 ) ( ) cr( )( ) 0 cl( )( ) than under , then cl( 0 ) ( ) cl( 0 ) 0 : (3) is not quasi-dictatorial, if (l ( ) ; r ( )) 6= (l ( 0 ) ; r ( 0 )) ; and if for each n; the distribution of has full support,10 then all of the above inequalities are strict. 9 There exists also a most polarized equilibrium c such that for any other equilibrium c0 ; cr c0r and c0l comparative statics for this equilibrium would be the same as for the least polarized one. 10 More precisely, what is needed is that for all a < b; the probability that n 2 (a; b) is strictly positive. 15 cl : The Proposition 3 states that if a voting rule 0 is replaced by a voting rule with a greater dispersion of power, the polarization of the pivotal players increases for two reasons. First, from (2), more extreme legislators become pivotal. From Proposition 1 part (i) and (ii), we know that if we …x the voting rule , legislators that are more extreme than l ( ) and r ( ) are more polarized. This e¤ect explains the four inner inequalities in (3). Second, the two outer inequalities in (3) further say that the polarization of a given legislator increases as the dispersion of power increases. The intuition for the second e¤ect is as follows. As argued in Section 4.2, legislators distort their behavior in order to remain pivotal in future decisions. Pivotality, however, has value only in states in which the pivotal legislators disagree. Since the pivotal legislators under ideologically extreme than the pivotal legislators under 0 are more ; they are more likely to disagree. Hence, 0: the incentive to distort behavior is higher under Proposition 3 has important implications for policy inertia. From Proposition 1, the status quo stays in place independent of its identity if l( ) (t) cl( )( ) 0< r( ) (t) cr( )( ). So as the dispersion of power increases, status quo inertia increases not only because a reform requires the approval of policy makers that are farther apart on the ideological spectrum— that is, r( ) (t) l( ) (t) and move farther apart— but also because policy makers on either side of the political spectrum are less likely to agree— that is, cl( )( ) and cr( ) also move farther apart. Few pundits and academics dispute that the political polarization in Congress have grown dramatically in the recent decades (see Table 2). At the same time, a …libuster— once an infrequently used tool reserved for the most important legislation— became one of the central features of American politics (see Table 3). Political scientists have come forward with various explanations for the increase in polarization (polarized electorate, gerrymandering, primary elections, economic inequality, money in politics, media, majority-party agenda control, and party pressures) and argued that an increased polarization increases the use of the …libuster, at the same time making the consequences of the …libuster more severe (Barber and McCarty 2013, Koger 2010). Proposition 1 suggests that there may also exist a reverse relationship. Since it has became easier to …libuster— just the threat of the …libuster is su¢ cient to block a bill— legislators understand that reversing policy changes in the future became harder. As a result, they distort their behavior. As a consequence, they appear more polarized. 16 Bafumi and Herron (2010) indeed showed that legislators seem to be taking positions that are even more extreme than their constituency. This observation is consistent with a theory that there was no change in the ideological position of the legislators, but their behavior became more polarized in response to the easier use of the …libuster. The following example demonstrates numerically the e¤ect of increasing dispersion of power. 5.1.1 Example: Increasing the supermajoritarian requirement Suppose p = 0; that is, preferences are redrawn every period. For each n 2 N ; let " (t) ; where " (t) n N (0; 1). The parameter n n (t) = is interpreted as the ideological position of the legislator. That is, the ideological di¤erences between legislators are constant across periods, and their preferences are a¤ected by the common shock. The shock " can be interpreted for example as the bene…ts from public spending: when they are high, legislators become more favorably inclined to increase taxes (implement L), and when they are low, they are more inclined to lower them (implement R). The bene…t from public spending ‡uctuates in response to the state of the economy, wars, natural disasters, investors’sentiment, etc. Suppose that the median legislator’s position is distributed around the median legislator. Let m = 0; and legislators are symmetrically = 0:9: In this setting, increasing the quali…ed majority means that the ideological positions of the pivotal legislators become more polarized r = l $ . Panel A of Figure 1 below demonstrates the voting distortions of the pivotal player l in the least polarized equilibrium, cl ; as increases (the distortions for player r are symmetric). Note that as increases, the polarization of the voting behavior, as measured by becomes 10 times larger than the polarization of the current preferences 17 r cl r l : l cr ; Panel B plots the probability that the current policy (be it L or R) stays in place. The dashed curve depicts this probability for the static game, and the solid curve depicts it for the dynamic game ( ) : As the supermajoritarian requirement increases, the status quo stays in place more often even in the static game. However, it stays in place much more often in the dynamic game. For r above 0:45; the probability of status quo inertia is almost 1: Hence, we have complete gridlock in which the initial status quo is replaced only with negligible probability. 5.2 Nonneutral voting rules Some democratic institutions use di¤erent voting rules depending on the policy change under consideration. For example, in 16 U.S. states, a bill that increases taxes requires the approval of a quali…ed majority in each house (plus the governor’s signature), while tax decreases can be approved using simple majority. Similarly, in the U.S. budget process the Byrd Rule requires a …llibusterproof majority to pass bills that raise the de…cit and a simple majority to pass bills that lower it.11 Arguably, the explicit goal of these voting arrangements was to limit the size of the public sector and the growth of the public debt. The analysis below shows, however, that they may fail to achieve their primary objectives. If we interpret policies L and R as high and low taxes, respectively, then introducing the nonneutral voting rules described above can be modelled as follows: a simple majority rule replaced by the voting rule quo L— i.e., 0 L = L— 0 is de…ned in (1): It entails the same dispersion of power under status but a greater dispersion of power under status quo R— i.e., Hence, the median player m is always pivotal under ; but under 0; 0 R R: she is only pivotal when taxes are high (q = L) : When taxes are low (q = R) ; the pivotal legislator r is more rightist than the median legislator. Clearly, if we …x the legislators’voting strategy, increasing the hurdle to raise taxes has a direct e¤ect of increasing the likelihood that status quo R stays in place. However, this reallocation of voting power has a strategic e¤ect as well, as it a¤ects legislators’equilibrium behavior. Proposition 11 The U.S. federal budget process is governed by the Congressional Budget Act of 1974, which prohibits the use of the …libuster against budget resolutions. This act was amended in 1985 (and later in 1990) by the Byrd Rule to allow the use of a …libuster against any provision that increases the de…cit beyond the years covered by the reconciliation measure. 18 3 implies that after the nonneutral rule is introduced, the median legislator will distort her behavior in favor of higher taxes L: In other words, a …scally conservative voting rule makes the median legislator less willing to reduce taxes, because she realizes that the the bias in the voting rule makes it more di¢ cult to increase them in the future. Hence, whether the bias in the voting rule makes policy R more likely on the equilibrium path depends on the relative strength of the direct and strategic e¤ect. As the example below illustrates, the strategic e¤ect might dominate. That is, a …scally conservative voting rule might in fact generate higher taxes. 5.2.1 Example: Increasing the supermajority for tax increases The distribution of payo¤s is like in Section 5.1.1. Under , the median legislator m is always pivotal, and from Proposition 1, legislators vote according to their current preferences. Therefore, R replaces L when m increase is equal to 1 Under 0, " (t) 0 and L replaces R when m ( and that they decrease is ( m ); " (t) < 0. The probability that taxes m ): the median legislator m is still pivotal when taxes are high (q = L) ; so l = m: However, when taxes are low (q = R), a more rightist legislator r > m is pivotal. First, note that if legislators continued voting the same way as they do under ; then 0 would indeed increase the probability that R is implemented in a given period. High taxes L would be still replaced by low taxes R whenever the median voter preferred lower taxes, but low taxes would be replaced by higher one less often: with probability 1 ( r) < 1 ( From Proposition 1 we know that in the dynamic game voting behavior. Legislator n will vote for L if will replace L) when m " (t) n " (t) m ): ( 0 ) ; however, legislators distort their cn : In this case, taxes will decrease (R cm ; and taxes will increase (L will replace R) when r " (t) < cr : Proposition 1 implies further that cr < 0 < cm : Hence, the probability that taxes increase is 1 ( r cr ); which is lower than under simple majority 1 ( m ): However, the probability that taxes decrease is also lower: it is equal to the probability that the median voter votes for R; which is ( m cm ) < ( m ): Hence, one can see that compared to ; the probability of tax increases goes down, but the probability of tax decreases goes down as well. So when starting from a low tax level, the constitutional amendment will keep taxes low for longer. However, once they increase, it will be harder to lower them when the need arises. If the second e¤ect is strong enough, the average tax rate in 19 the long run can go down. Figure 2 below depicts the invariant probability of L (probability that in a distant future the policy is L) as we vary r: The dashed line is the probability of L under simple majority, the dotted curve demonstrates how this probability would change under 0 The solid curve demonstrates what happens in equilibrium under and Panel B for m = if players played a static game. 0: Panel A is drawn for m =0 0:5. When the median player is on average indi¤erent between low and high taxes (i.e., she is indi¤erent when " (t) = 0), then increasing the majoritarian requirement to raise taxes indeed makes high taxes less prevalent. In fact, it makes them less prevalent than what a static model would imply. However, when the median player on average prefers high taxes, then increasing the majoritarian requirement may result in higher average taxes in the long run. The empirical evidence on the e¤ect of supermajoritarian rules on taxes is relatively scant and relatively inconclusive. For the period 1980-2008, the average tax level in the states in which tax increases require a supermajority approval is nearly identical as in the states without such supermajority requirements. The average state with no supermajority requirement taxes as a share of personal income have been between 9.7 percent and 10.9 percent , while in the seven states with strict supermajority requirements they have been between 9.7 percent to 10.8 percent (Leachman, Johnson, Grundman, 2012, and Jordan and Ho¤man, 2009).12 On the other hand, using …xed e¤ects models, Knight (2000) and Besley and Case (2003) …nd that supermajoritarian requirements reduce taxes by about $50 per capita. 12 These calculations are based on Arizona, California, Delaware, Mississippi, Nevada, Louisiana, and Oregon. Knight (2000) reports that in 1995, among the continental states, states with supermajoritarian requirements had identical average e¤ective tax rates of 7.13% as states without such requirements. 20 5.3 Volatility The following proposition shows that the polarizing e¤ect of the dispersion of power is greater in more volatile environments. Proposition 4 Let be a voting rule that is not quasi-dictatorial. Let p; p0 2 R be such that p0 < p < 1; and let c and c0 be the equilibrium thresholds with persistence p and p0 respectively. 0 Then c0r < cr < 0 < cl < c0l : When the environment is static, that is, when p = 1; then c = 0: The intuition for Proposition 4 is the following. As the volatility of the environment increases— that is, as p decreases— the policy implemented today is more likely to require a revision tomorrow; hence, remaining pivotal tomorrow becomes more important relative to implementing the right policy today. In the extreme case when p = 1; preferences never change; hence, a policy adopted today will never be revised. As a result, all players vote according to their current preferences. This result implies that the inertial e¤ect of the dispersion of power is greater in more volatile environments. This suggest that simple majority rule should be used in policy domains which require frequent adjustments, such as …scal or monetary policies, and requiring a quali…ed majority is less costly for policies over which preferences are more stable, such as constitutional laws. 5.4 Discussion and Conclusions When binary decisions are taken via quali…ed majority rule, two sorts of mistakes can be made. First, no quali…ed majority wants to move away from the status quo though changing the status quo would be socially optimal (type 1 mistakes). Second, a quali…ed majority wants to move away from the status quo, but it would be socially optimal to stay at it (type 2 mistake). By increasing the quali…ed majority requirement in a static setting, we increase the likelihood of type 1 mistakes, but decrease the likelihood of type 2 mistakes. Whether this is welfare improving depends on the expected cost of each type of mistake (see Gutmann 1998). Our results show that in dynamic settings and volatile environments, increasing the quali…ed majority may increase the likelihood 21 of type 1 mistake quite substantially. Hence, the optimal quali…ed majority is lower than the one implied by the static game. Propositions 3 and 3 have important consequences for constitutional design. As we discussed in Section 1, in most modern democracies there exists a system of check and balances. Admittedly, these checks and balances are not designed to smooth the decision process. Rather, their role is to limit agency costs and abuses of power by any government branch. Our model shows, however, that when checks and balances are introduced in a decision process, they tend to make legislators more polarized, which can greatly exacerbate their inherently inertial e¤ect. Hence, a system of checks and balances should be complemented with solutions that mitigate polarization. One option is to allow or require a more frequent use of sunsets, especially during volatile times. Sunset provisions are clauses attached to a legislation that determine its expiration date. They break the dynamic linkage between today’s decision and tomorrow’s pivotality; hence, mitigate the voting distortions identi…ed in this paper. Legal scholars and politicians, starting from Thomas Je¤erson, have long encouraged the use of sunsets.13 The modern conception of sunset provisions was spelled out in Lowi (1969). The main rationale was to ensure that obsolete policies and agencies do not last solely on the basis of legislative inertia. Our paper provides an explanation for the source of this inertia. Over the past three decades, Congress has increasingly employed sunset provisions to tax legislation. In 2001, Congress enacted the largest tax cut in twenty years, the Economic Growth and Tax Relief Reconciliation Act (EGTRRA), and two years later, it passed the third largest tax cut in history, the Jobs and Growth Tax Relief Reconciliation Act (JGTRRA). Both acts included sunset provisions. Sunset provisions have been recently used also to legislate gun ownership (1994 Anti Crime Act which banned the manufacture, sale, and possession of certain type of assault weapons) and civil liberties (2001 USA PATRIOT Act). The more frequent use of sunset provisions for major legislations coincided with higher polarization and gridlock in the US Congress. If this polarization re‡ected a true change in ideological positions of the legislators, the sunsets should not facilitate a bill passage more than they did when polarization were small. If on the other hand, the increased 13 “[T]he power of repeal is not an equivalent [to mandatory expiration]. It might indeed be if [...] the will of the majority could always be obtained fairly and without impediment. But this is true of no form. [...] Various checks are opposed to every legislative Proposition [...] and other impediments arise so as to prove to every practical man that a law of limited duration is much more manageable than one which needs a repeal.” (See Woods 2009, p. 93). 22 polarization of the US Congress is an endogenous phenomenon driven by the recent changes in voting rules and in norms governing the use of the …libuster, then sunsets should become more attractive. In this paper, we abstracted from the ability of legislators to use sunsets, as we intended to focus solely on the distorting e¤ect of di¤erent voting rules. However, a formal analysis of the interaction of voting rules and the frequency (and impact) of sunset provisions is an interesting avenue for future research. 6 Tables Table 1. Source: National Conference of State Legislators. 23 Table 2. Source: Barber and McCarty (2013) Table 3. Source: US Senate. 7 Appendix Throughout the appendix, we use the following notations: 24 Notation 1 For all p 2 [0; 1] ; l; r 2 N ; and c 2 RN ; (1 Hn (p; l; r; c) = 1 p) p Z (cn f 2RN : l <cl and r n ) dP ( ): (4) cr g We omit p; l; and r in the argument of H when these parameters are clear from the context. The next remark and lemmas derive some important properties of H. Remark 1 Note that (Hl ; Hr ) (c) depends on c only though (cl ; cr ) : Moreover, the …xed points of c ! H (c) are in bijection with the …xed points of (cl ; cr ) ! (Hl ; Hr ) (cl ; cr ) in the following sense: if c is a …xed point of H; then (cl ; cr ) is a …xed points of (Hl ; Hr ), and reciprocally, if (cl ; cr ) is a …xed points of (Hl ; Hr ) ; then there exists a unique c0 that is a …xed point of H and such that (c0l ; c0r ) = (cl ; cr ). cl and c0r Notation 2 De…ne the order ( ; ) as follows: (c0l ; c0r ) ( ; ) (cl ; cr ) if c0l cr : We say that (Hl ; Hr ) (cl ; cr ) is isotone in the order ( ; ) if whenever c0 ( ; ) c then (Hl ; Hr ) (c0l ; c0r ) ( ; ) (Hl ; Hr ) Lemma 1 The mapping (cl ; cr ) ! (Hl ; Hr ) (cl ; cr ) is isotone in the order ( ; ) ; and for n 2 fl; rg Hn (cl ; cr ) is (1 p) Lipshitz continuous in cn : Proof. For all cl ; c0l ; cr ; c0r such that cl Hl (c0l ; c0r ) Hl (cl ; cr ) (1 (1 p) p) = c0r ; c0l and cr Z f l <cl and + Z f r 0 l <cl and cr g r c0l cl dP ( ) c0r gnf l <cl and r cr g c0l l dP ( ) The integrand of the two integrals on the right-hand side of the above equation are positive on their domains of integration, so Hl (c0l ; c0r ) Hl (cl ; cr ) : Therefore, Hl (cl ; cr ) is increasing in (cl ; cr ) in the order ( ; ) : A similar proof shows that Hr (cl ; cr ) is decreasing in (cl ; cr ) in the order ( ; ) ; so (Hl ; Hr ) is isotone in (cl ; cr ) in the order ( ; ) : When cr = c0r ; the above equality implies 25 Hl (c0l ; cr ) Hl (cl ; cr ) (1 (1 p) p) = Z f l <cl and + Z f cl For all l 2 [cl ; c0l ] ; cl l c0l (1 cr g r 0 l <cl and cl dP ( ) cr g r c0l l dP ( ) cl ; so the above equality implies in turn that Hl (c0l ; cr ) Hl (cl ; cr ) (1 (1 p) so Hl (cl ; cr ) is c0l p) c0l cl P l < c0l and r c0l cr cl ; p) Lipshitz continuous in cl : The proof for Hr (cl ; cr ) proceeds in the same way. Lemma 2 There exists a < 1 such that (Hl ; Hr ) [ a; a]2 [ a; a]2 and all the …xed points of (Hl ; Hr ) are in [ a; a]2 : R : Proof. Let k k = maxn2N j n j dP : ( ), and kck = maxn2fl;rg jcn j. From (4), we see that for all (cl ; cr ) 2 R2 and all n 2 fl; rg ; jHn (cl ; cr )j is bounded by (1 p) 1 p point of (Hl ; Hr ), then kck (1 p) 1 p (k k + kck). So if (cl ; cr ) is a …xed (k k + kck), which in turn implies that kck The following lemma shows that the equilibria of the game (1 p) 1 k k a. ( ) can be characterized as the …xed points of the mapping H: Lemma 3 If is an equilibrium, then is a cuto¤ strategy pro…le for some pro…le of cuto¤ s c such that cn is decreasing in n. A pro…le of cuto¤ s c is an equilibrium if and only if it is a …xed point of the mapping c ! H (c) ; and cn = Proof. Let 1 p (1 (1 p) p) Pr (f l < cl and r cr g) Z f l <cl and n dP r ( ): (5) cr g be a pro…le of Markov strategies, and for all q 2 fL; Rg ; let Wn (q) be the continuation value of the game for player n at the beginning of some period t; conditional on the status quo being q; on the state being redrawn at the beginning of the period, and on players playing . Since players play Markov strategy, and since the state is redrawn, Wn (q) does not 26 depend on previous states. Since we can assume without loss of generality that players revise the policy only when the state is redrawn, the di¤erence in continuation value for player n from implementing policy R versus L in a period with state n can be expressed as follows: + (1 p) Wn (R) + p ( 0 + (1 = 1 1 n p n + (1 p) Wn (L) + p (0 + (1 + (1 1 p) (Wn (R) p p) Wn (R) + p (:::)) (6) p) Wn (L) + p (:::)) Wn (L)) : In any period with status quo q (t) 2 fL; Rg and state (t) 2 Rn ; if (6) is nonnegative (negative), then R (L) is the only stage-undominated action for player n: Manipulating (6), this implies that n prescribes stage-undominated actions if and only if it prescribes n to vote for R if and only if n (t) (1 : Wn (R)) = cn : p) (Wn (L) (7) Since a stage-undominated strategy must prescribe stage-undominated actions for all possible realizations of n (t) (even those that may arise with probability 0), this shows that a stage-undominated strategy must be a cuto¤ strategy, and for any strategy pro…le ; the unique stage-undominated cuto¤ is given by the above condition.14 Let D ( ) denote the set of realizations of such that no winning coalition of players vote for the alternative di¤erent than the status quo when the state is : Note that the status quo in period t + 1 a¤ects the policy outcome in t + 1 only when (t + 1) 2 D ( ) : In such states, the status quo stays in place. Hence, Wn (R) Wn (R) Wn (L) = Wn (L) is simply the expectation of (6) over all Z 1 2D( ) 1 p n which delivers Wn (R) Wn (L) = 1 (1 1 + p 14 R p) (Wn (R) p 2D( ) n dP (1 p) R 2 D ( ), so Wn (L)) dP ( ) ; (8) ( ) 2D( ) dP ( ) : Technically, the strategy does not have to specify the behavior for that are not in the support of P ( ) ; so if the support of is not connected, the thresholds may not be uniquely determined. We ignore this issue for simplicity, as the resulting multiplicity does not reverse any of our …ndings and only complicates the exposition. 27 Substituting the above expression for Wn (R) Wn (L) into (7), we obtain that if c is an equilibrium cuto¤ pro…le, cn is decreasing in n: Substituting (7) into (8), we obtain (1 cn = 1 We have shown that if p) p Z (cn n ) dP ( ): (9) 2D( ) is an equilibrium, is a cuto¤ strategy pro…le for some pro…le of cuto¤ c such that cn is decreasing in n. Therefore, voters vote as if there is a single period and their current preferences is given by n cn instead of n; and c satis…es Assumption 1. Therefore, the same reasoning as in Section 4.1 shows that players l and r are pivotal in all periods in which the status quo is L and R; respectively (see De…nition 2). So, up to a zero measure set, D ( ) can be rewritten as a function of the preference realization of the pivotal legislators only: 2 RN : D( ) = l < cl and cr : r (10) Substituting the above expression for D ( ) into (9), we obtain that c must be a …xed point of H; and (5) follow immediately from (4). Conversely, let c be a …xed point of H; and consider the corresponding strategy pro…le : Then (4) must hold, and (4) implies that cn is decreasing in n; and thus that (10) holds. Therefore, Wn (L) must still satisfy (8), and viewed as an equation in (9) also holds. Note that Wn (R) Wn (R) Wn (L), (8) has a unique solution. Manipulating (9), we obtain that the same equation, and is thus equal to Wn (R) is an equilibrium. By Lemma 3, in any equilibrium , players use status-quo- independent cuto¤ strategies, and the pro…le of cuto¤s c is such that c = H (c) and c1 If H (c) = (0; :::; 0) : Suppose now that Hl (c) 2 RN : is quasi-dictatorial, then l = r so 0 2 RN : Hr (c) ; and thus that cl l < cl and r satis…es Wn (L). Therefore, (7) holds, which implies that the cuto¤ strategy cn is stage undominated for player n; so Proof of Proposition 1. 1+ p (1 p) cn l < cl and r cr ::: cN : = ;; and therefore, c = is not quasi-dictatorial. One can easily see from (4) that 0 cr . Since l 6= r; Assumption 1 implies then that > cr has strictly positive probability, and thus that Hl (c) > 0 > Hr (c), which proves property (ii) : 28 Since 2 RN : l < cl and r > cr has strictly positive probability, Assumption 1 implies that the integral on the right-hand side of (5) is strictly increasing in n; which proves property (i) : Proof of Proposition 2. Using Lemma 3 and Remark 1, to prove Proposition 2, it su¢ ces to prove that the …xed point of (Hl ; Hr ) are a complete chain for the order ( ; ) : From Lemma 2, the …xed points of (Hl ; Hr ) lie in [ a; a]2 which is a complete lattice for the order ( ; ) ; and (Hl ; Hr ) [ a; a]2 is included in [ a; a]2 : Lemma 1 together with Tarski’s …xed point theorem imply that the set of …xed points of the restriction of (Hl ; Hr ) on [ a; a]2 (and hence the set of …xed points of (Hl ; Hr ) on R2 ) is a complete lattice in the order ( ; ) : Let (cl ; cr ) and (c0l ; c0r ) be two …xed points of (Hl ; Hr ) ; and suppose w.l.o.g. that c0r c0l < cl ; then since from Lemma 1, Hl is (1 cr : If p) Lipshitz continuous in cl ; we obtain cl = Hl (cl ; cr ) < Hl c0l ; cr + cl c0l Hl c0l ; c0r + cl c0l = cl ; where the last equality follows from Hl (c0l ; c0r ) = c0l : Hence, we have a contradiction. Therefore, c0r cr implies that c0l cl ; which shows that the set of …xed points of (Hl ; Hr ) is completely ordered by ( ; ) ; and is thus a chain. 0 Lemma 4 If the dispersion of power under cr( 0) 0 is greater than under cr( ) ( ) 0 cr( 0 cl( ) ( ) cl( 0) ; then 0 : (11) Proof. Let D = D0 = r( ) r( 0) cr( 0) 0) and 0 29 and l( ) l( < cl( 0) < cl( 0) 0) 0 ; 0 : Assumption 1 together with (2) imply that modulo a zero probability event, D H (l ( 0 ) ; r ( 0 ) ; c ( 0 )) ; and since cl( cl( 0 0) = 0) 0) 0) l( Z p) c p D0 l( Z p) c p D l( Z p) c p D l( (1 1 (1 1 (1 1 = Hl( A similar argument shows that cr( ( 0) 0 ( 0) Hr( 0 l( 0) 0 l( ) ) dP ( ) 0) l( 0) l ( ) ; r ( ) ; cl( ) 2 D0 ; we have that is positive for all 0) dP ( ) 0) 0 dP ( ) ; cr( 0) l ( ) ; r ( ) ; cl( 0) 0) with Lemma 1, this implies that for all (cL ; cR ) 2 R2 such that cL D0 : Since c ( 0 ) = cl( 0 : ( 0 ) ; cr( 0) 0) ( 0 ) : Together ( 0 ) and cR cr( 0) ( 0) ; (Hl ; Hr ) (l ( ) ; r ( ) ; cL ; cR ) ( ; ) (Hl ; Hr ) l ( ) ; r ( ) ; cl( ( ; ) cl( 0) Using the notations of Lemma 2, if C = implies that Hl( ) ; Hr( ) 0 ; cr( a; cl( (l ( ) ; r ( ) ; C) 0) 0) 0 cl( 0) )( ( 0 ) ; cr( ) ; cr( 0) )( ; cr( 0) 0 : ( 0) cr( 0) ( 0 ) ; a , the previous two inequality C: Therefore, (cL ; cR ) ! Hl( ) ; Hr( admits a …xed point in C which is smaller than cl( Since cl( 0 0) 0) ( 0 ) ; cr( 0) ) (l ( ) ; r ( ) ; cL ; cR ) ( 0 ) for the order ( ; ) : ) is the smallest …xed point of this mapping, it must be smaller than ( 0) : Proof of Proposition 3. We …rst show that four inner inequalities in (3): cr( 0) ( ) cr( )( ) 0 cl( )( ) cl( 0) ( ): When is quasi-dictatorial, Proposition 1 implies that these inequalities in (3) hold with equality. When is non dictatorial, Proposition 1 part (i) and (ii) imply that they hold strictly. 30 We now show the two outer inequalities in (3): cl( 0) ( ) cl( ( 0 ) and cr( 0) 0) ( 0) cr( 0) ( ). Let D = cr( r( ) D0 = r( )( cr( 0) ) and 0 0) l( ) and < cl( 0) l( )( ) ; < cl( 0 0) : From (5), cl( 0 0) cl( 0) = ( ) = 1 p 1 p Z (1 p) (1 p) Pr (D0 ) D0 Z (1 p) (1 p) Pr (D) D l( dP ( ) ; 0) l( dP ( ) : 0) D0 ; and for almost all Assumption 1 together with (11), modulo a zero probability event, D 2 D0 ; l( 0) > 0; so the above equations show that cl( shows that cr( 0) ( 0) cr( 0) 0) ( 0) (12) cl( 0) ( ) : A similar reasoning ( ). It remains to show that these inequalities are strict when (l ( ) ; r ( )) < (l ( 0 ) ; r ( 0 )) : Suppose to …x ideas that r ( ) < r ( 0 ) ; the proof in the case l ( ) > l ( 0 ). By de…nition of D and D0 ; < cr( )( ) and From Proposition 1, cr( )( ) r( ) 1, r( ) < cr( )( ) implies l( cr( 0) 0 and cl( 0) r( If r ( ) < r ( 0 ) and r( < cl( 0) 0) 0) 0 0) ( 0) and l( 0) < cl( 0 0) D0 n D 0, so Assumption 1 implies that with probability ( 0 ) : Therefore, the above inclusion implies that 2 cr( 0) 0 ; cr( )( D0 n D ) (13) is not quasi-dictatorial, from what precedes, we have that cr( 0) 0 cr( 0) ( ) < cr( )( ): The above inequalities and the full support assumption imply that the probability of the event on the left-hand side of (13) is strictly positive. Therefore, P (D0 ) > P (D) : Moreover, cl( Substituting the latter inequalities in (12), we obtain cl( 31 0) ( 0 ) > c l( 0) 0) ( 0 ) > 0. ( ) : A similar reasoning shows that cr( 0) ( 0 ) < cr( 0) ( ). Proof of Proposition 4. Note …rst that when is quasi-dictatorial, Proposition 4 follows trivially from Proposition 1. Suppose now that is not quasi-dictatorial. Using the notations of Proposition 4, Lemma 3 and Remark 1 imply that (c0l ; c0r ) is a …xed point of (cL ; cR ) ! (Hl ; Hr ) (p0 ; cL ; cR ) : Let (cL ; cR ) be such that c0r cR and cL c0l : (14) Applying H to the above inequality, and using Lemma 1 and @Hn @p = (1 ) (1 p)(1 p) Hn ; : we have that c0l = Hl p0 ; c0l ; c0r < Hl p; c0l ; c0r and 0 If we denote C = [0; c0l ] Hr (p; cL ; cR ) Hl (p; cL ; cR ) Hr p; c0l ; c0r < Hr p0 ; c0l ; c0r 0; c0r : [c0r ; 0] ; the above inequalities show that (Hl ; Hr ) (p; C) C: Since C is a complete lattice, Tarski’s theorem implies that (cL ; cR ) ! (Hl ; Hr ) (p; cL ; cR ) admits a …xed point in C: Moreover, the above inequalities imply that for all such …xed point, the inequalities in (14) hold strictly. From Lemma 3 and Remark 1, (cl ; cr ) is one of these …xed points. References [1] Acemoglu, D., G. Egorov, and K. Sonin, 2014. Political Economy in a Changing World. Journal of Political Economy, forthcoming. [2] Aghion. P., A. Alesina, and F. Trebbi, 2004. Endogenous Political Institutions, The Quarterly Journal of Economics, 119, 565–611. [3] Aghion P. and P. Bolton, 2003. Incomplete Social Contracts. Journal of the European Economic Association. 1, 38–67. [4] Arrow K. J., 1998. The External Costs of Voting Rules: A Note on Guttman, Buchanan, and Tullock. European Journal of Political Economy, 14, 219–222. 32 [5] Austen-Smith, D. and J. Banks, 1996. Information Aggregation, Rationality and the Condorcet Jury Theorem. American Political Science Review, 90, 34–45. [6] Bafumi, J. and M. C. Herron, 2010. Leapfrog Representation and Extremism: A Study of American Voters and Their Members in Congress. American Political Science Review 104, 519–542. [7] Barber M. and N. McCarty, 2013. Causes and Consequences of Polarization. In Negotiating Agreement in Politics, eds. J. Mansbridge and C. J. Martin. American Political Science Association. [8] Baron, D., 1991. Majoritarian Incentives, Pork Barrel Programs, and Procedural Control. American Journal of Political Science, 35, 57–90. [9] Besely, T. and A. Case, 2003. Political Institutions and Policy Choices: Evidence from the United States. Journal of Economic Literature,41, 7–73. [10] Buchanan, J. M.,1998. Agreement and E¢ ciency: Response to Guttman. European Journal of Political Economy, 14, 209–213 [11] Buchanan, J. M. and G. Tullock, 1962. The Calculus of Consent. Ann Arbor, MI, University of Michigan Press. [12] Caplin, A., and B. Nalebu¤, 1988. On 64%-majority rule. Econometrica, 56, 787–814. [13] Condorcet, N., 1785. Essai sur l’application de l’analyse a‘la probabilite´ des de´cisions rendues a‘la pluralite´ des voix. Paris: Imprimerie Royale. [14] Coughlan, P. J., 2000. In Defence of Unanimous Jury Verdict: Mistrials, Communication and Strategic Voting. American Political Science Review, 94, 375–393. [15] Dziuda, W. and A. Loeper, 2014. Dynamic Collective Choice with Endogenous Status Quo. Northwestern University, mimeo. [16] Enelow J., 1997. Cycling and Majority rule. In D. Mueller, ed., Perspective on Public Choice: A Handbook, Cambridge: Cambridge University Press, 149–62. 33 [17] Eraslan, H. and A. Merlo, 2002. Majority Rule in a Stochastic Model of Bargaining. Journal of Economic Theory, 103, 31–48. [18] Gradstein, M., 1999. Optimal Taxation and Fiscal Constitution, Journal of Public Economics, 72, 471–485. [19] Feddersen, T. and W. Pesendorfer, 1998. Convicting the Innocent: The Inferiority of Unanimous Jury Verdicts under Strategic Voting. American Political Science Review, 92, 23–35. [20] Ferejohn, J. A., M. P. Fiorina, and R. D. McKelvey, 1987. Sophisticated Voting and Agenda Independence in the Distributive Politics Setting. American Journal of Political Science, 31(1), 169-93. [21] Gradstein, M. 1999. Optimal Taxation and Fiscal Constitution. Journal of Public Economics, 72, 471–485. [22] Grofman, B. N. and G. Owen, 1986. Information Pooling and Group Decision-Making. JAI Press, Greenwich, CT. [23] Guttman, J., 1998. Unanimity and Majority Rule: The Calculus of Consent Reconsidered. European Journal of Political Economy, 14, 189–207. [24] Hastard, B., 2005. Majority Rules and Incentives. Quarterly Journal of Economics, 120, 535– 568. [25] Hayek, F. A., 1960. The Constitution of Liberty. University of Chicago Press. [26] Jordan, M. M., and K. U. Ho¤man, 2009. The Revenue Impact of State Legislative Supermajority Voting Requirements. Midsouth Political Science Review, 2009, 10, 1-15. [27] Knight, B., 2000. Supermajority Voting Requirements for Tax Increases: Evidence from the States. Journal of Public Economics, 76, 41–67. [28] Krehbiel, K. 1999. Pivotal Politics: A Re…nement of Nonmarket Analysis for Voting Institutions, Business and Politics, 1, 63–81. [29] Leachman, M., N. Johnson, and D. Grundman, 2003. Six Reasons Why Supermajority Requirements to Raise Taxes Are a Bad Idea. Report for Center on Budget and Policy Priorities. 34 [30] Lowi, T.J., 1969. The End of Liberalism: Ideology, Policy, and the Crisis of Public Authority. New York, NY, Norton. [31] May, K. O., 1952. A Set of Independent, Necessary and Su¢ cient Conditions for Simple Majority Decision. Econometrica 20, 680–684. [32] McGann, A. J., 2004. The Tyranny of the Supermajority: How Majority Rule Protects Minorities. Journal of Theoretical Politics, 16, 53–77. [33] Messner, M. and M. Polborn, 2004. Voting on Majority Rules. Review of Economics Studies, 71, 115–132. [34] Persico, N., 2003. Committee Design With Endogenous Information. Review of Economic Studies, 70, 1–27. [35] Plott, C. R., 1967. A Notion of Equilibrium and Its Possibility Under Majority Rule. American Economic Review, 57, 787–803. [36] Rae, D. W., 1969. Decision-Rules and Individual Values in Constitutional Choice. The American Political Science Review, 63, 40–56. [37] Rae, D. W., 1975. The Limits of Consensual Decision. The American Political Science Review, 69, 1270–1294. [38] Riker, W. H., 1962. The Theory of Political Coalitions. New Haven: Yale University Press. [39] Roberts, K., 1999. Dynamic Voting in Clubs. Unpublished manuscript. [40] Tullock, G. 1998. Reply to Guttman. European Journal of Political Economy, 14, 215–219. [41] Tsabellis, G. Veto Players, 2002, Princeton: Princeton University Press. [42] Wicksel, K., 1896.A New Principle of Just Taxation. Reprinted in R. Musgrave and A. Peacock, eds., Classics in the Theory of Public Finance, New York: St. Martin’s Press, 1967. [43] Brett F. Woods. 2009. Thomas Je¤erson: Thoughts on War and Revolution. Algora Publishing, New York. 35
© Copyright 2026 Paperzz