Institutional Constraints and The Inefficiency in Public Investments

Transcription

1 Institutional Constraints and The Inefficiency in Public Investments Leyla D. Karakas March 14, 017 Abstract This aer studies limits on executive authority by identifying a dynamic channel through which they interact with olicy-making. The model features two agents, one of which becomes the incumbent in each of the two eriods and unilaterally decides on investments to a common-interest ublic good, targeted ork sending and the future level of executive constraints. While loose constraints ermit an incumbent to enact his desired olicies in case of re-election, the same alies to his oonent under the oosite scenario. However, this trade-off changes with the level of the ublic good, because the shared investment interests dominate the conflict over ork sending less and less as the efficiency gains from investing get exhausted. The results indicate that more efficient states of the ublic good are associated with tighter executive constraints in equilibrium. Moreover, equilibrium ublic good rovision is inefficient desite the fact that agents share the same references over ublic good consumtion and this inefficiency increases as the incumbent s re-election chances deteriorate. These findings shed light on why incumbents facing similar electoral environments make different institutional decisions. Keywords : Executive constraints, Pork-barrel sending, Public goods. JEL Classification : D7, D78, H4. Deartment of Economics, Maxwell School of Citizenshi and Public Affairs, Syracuse University, Syracuse, NY lkarakas@maxwell.syr.edu. This aer is based on my dissertation at Johns Hokins University. I would like to thank my advisor Hulya Eraslan, Ali Khan, and the seminar articiants at Johns Hokins University, the 13 th International Meeting of the Association for Public Economic Theory, the 013 Asian Meetings of the Econometric Society, the 015 Annual Conference of the Canadian Economic Association and the 85 th Annual Conference of the Southern Economic Association for their valuable comments. I am also grateful to an anonymous referee whose suggestions greatly imroved the aer. 1

2 1 Introduction Political institutions that imose strong constraints on executive authority are one of the defining features of democracies. Examles of such institutions include legislative rotocol that allows oosition arties to articiate in decision-making, constitutional rotections for minority rights and budgetary rules that limit the imlementation of certain olicies. This aer focuses on how governments decide on future limits to the actions of the executive in the resence of electoral uncertainty. Why some governments constrain their future decision-making ower while others don t is an imortant question to study as such limits determine the degree to which citizens are rotected from the executive s otential abuses of ower, such as exroriation or excessive rent-seeking. 1 While the ast century witnessed the gradual adotion of institutions that constrain executive authority, there still exist countries whose strength of executive constraints fluctuate. In fact, the analyses in Besley and Persson (011) indicate that only a small fraction of the countries that gained indeendence after World War II have managed to successfully integrate ersistently strong executive constraints into their olitical systems. To exlain these historical facts, the existing literature theoretically links the strength of a country s executive constraints to the cometitiveness of its elections by arguing that leaders who are likely to remain in ower romote weaker constraints on their robable future rule, while those who are likely to leave office enact strong constraints in order to rotect themselves from their oonents olicies. 3 However, these studies do not address the question of how strong executive constraints survive electorally owerful incumbents in democratic regimes. 4 Secifically, why do electorally owerful 1 For examle, see Persson, Roland and Tabellini (1997) for a model of how a stronger system of checks and balances decreases the rents the government is able to extract. See Besley and Persson (011), Chater 7 for an in-deth analysis and discussion of these trends, as well as details on the measurement of executive constraints, the samle of countries and the time eriod of analysis. 3 For examle, see Besley, Persson and Reynal-Querol (016). 4 Besley, Persson and Reynal-Querol (016) only rovides evidence that olitical turnover affects the choice of institutions in regimes with weak executive constraints.

3 incumbents resond to their olitical advantage by oting for weak executive constraints only in regimes that are already characterized by such constraints? The goal of this aer is to reconcile this discreancy between the literature s theoretical rediction and the observation that electorally owerful incumbents in strong executive constraint-regimes do not behave as redicted. In order to do so, it offers a country s state of ublic good develoment as an additional determinant of its executive constraints. The model features two eriods and two agents who care about their rivate consumtion and a ublic good. One of the agents exogenously becomes the incumbent each eriod according to a fixed robability. Once realized, the first-eriod incumbent makes the following decisions: i) investment to a ublic good that equally benefits both agents (e.g. environmental sending); ii) rivate transfers of the consumtion good to each agent (e.g. ork sending); and iii) the level of executive constraint for the next eriod that will limit the future incumbent s choices in all three olicy dimensions simultaneously. The executive constraint catures the limit on the extremeness of an incumbent s olicies and is oerationalized by a uniform sending ca on each olicy. The first-eriod incumbent makes olicy decisions under the given executive constraint and chooses the one that will constrain the second-eriod incumbent. Since there exists no executive constraint decision in the final eriod, the second-eriod incumbent only decides on investment and rivate transfers. The executive constraint and the ublic good constitute the two dynamic linkages between the eriods. Corroborating the existing literature, the results indicate that ossessing greater electoral ower leads an incumbent to choose a weaker executive constraint in equilibrium, and vice-versa. However, this result only holds for given states of the ublic good. Secifically, I find that for any given distribution of electoral ower, the executive constraints get tightened in equilibrium as the ublic good aroaches an efficiency benchmark. Together, these two results suggest that different states of the ublic good exlain why incumbents facing similar re-election rosects might make different exec- 3

4 utive constraint decisions. The basic trade-off in choosing an executive constraint gives rise to the former result in the above aragrah: The ush for a weaker constraint in order to enjoy olicies closer to the incumbent s ideal in case of his re-election conflicts with the motivation to rotect himself from his oonent s ork sending in the oosite scenario. However, while the distribution of electoral ower determines which of these two incentives dominates in equilibrium, its role is not constant, leading to the latter result: The agents shared investment interests dominate the conflict over rivate transfers at too low or too high states of the ublic good, because there exist large mutually beneficial gains to be realized. At these states, the ability to invest (or disinvest) takes recedence over the ability to make rivate transfers so that both agents discount their olitical disagreement as a result and re-election uncertainty becomes a relatively less imortant determinant of executive constraints. However, as these gains get exhausted, the agents no longer need to tolerate each other s ork sending for the sake of being able to imlement the much-needed investments. Consequently, the incumbent assigns less weight on the future incumbent s ability to invest and more weight on otential rivate transfers when deciding on the next eriod s executive constraint. Comaring ublic good rovision under electoral uncertainty with the lanner s solution indicates that the equilibrium is inefficient and that this inefficiency increases as the first-eriod incumbent s re-election chances deteriorate. The first source of this inefficiency result is the fact that the incumbent cares only about his own utility and therefore does not internalize the ositive externality his investment choice creates. The second source of the inefficiency is more novel and arises from the structure of an executive constraint assumed in the model. By imosing a uniform sending bound on all olicy choices, an executive constraint may duly restrict investments as well as rivate transfers even though the agents do not fundamentally disagree on its otimal level. Consequently, investments get sacrificed to the conflict over rivate transfers even at 4

5 highly inefficient states of the ublic good. These results imly that electoral uncertainty can exlain differences in executive constraints only across countries with similar levels of ublic good develoment. Accordingly, they oint to a country s state of ublic good develoment as an exlanation for why strong executive constraints survive owerful incumbents in democracies: The incentive to kee executive constraints weak in order to allow the future incumbent, regardless of his identity, to invest in the ublic good is absent in societies that have already exhausted such common gains. This aer thus attemts to fill the ga between the existing literature s theoretical rediction and the emirical observations by exloiting the additional dynamics introduced by the ublic good state variable. The rest of the aer is organized as follows: The following section discusses the literature this aer builds uon. Section 3 introduces the model and Section 4 rovides an efficiency benchmark for ublic good rovision by resenting the dictatorshi and the lanner s solutions. Section 5 contains the main results of the aer that characterize the equilibrium under electoral uncertainty. The efficiency of this equilibrium is evaluated in Section 6. Section 7 concludes. Related Literature This aer contributes to a large literature that studies the determinants of institutional constraints on olicy-making. For examle, Lagunoff (001) analyzes minority rights and finds that societies with higher olitical uncertainty exhibit more constraints on majority rule. Aghion, Alesina and Trebbi (004) characterizes the otimal size of suermajorities that are mandated to ass legislation. Acemoglu, Robinson and Torvik (013) shows that voters may elect oliticians who weaken checks and balances as a way to decrease corrution. 5 Another branch of this literature that includes Diermeier, Prato 5 Other aers that study institutions of accountability and executive constraints include Maskin and Tirole (004), Ticchi and Vindigni (010), Wolitzky (013) and Robinson and Torvik (016). 5

6 and Vlaicu (015) and Eguia and Shesle (015) focuses on endogenously determined rocedural rules that constrain decision-making in organizations. Among these studies, Besley, Persson and Reynal-Querol (016) is closely related as the authors theoretically establish the relationshi between electoral uncertainty and executive constraints discussed in the Introduction. It also finds emirical evidence for the negative effect of a resilient leader on adoting strong constraints. However, no equivalent evidence is resented for transitioning from strong to weak executive constraints. Thus, while the authors confirm the imortance of olitical turnover on a government s choice of executive constraints, the fact that the theory fails to hold in democratic regimes rovides a motivation for this aer. 6 The underlying framework of this aer builds on Battaglini and Coate (007), where legislators allocate the revenues obtained by distortionary taxation between a commoninterest ublic good and transfers to their districts through a bargaining rotocol. The authors find that investment will be efficient for all states of the ublic good below a certain threshold. However, as the ublic good accumulates, legislators start to underinvest and allocate ork to their districts. This aer can be interreted as an extension of their main idea to an environment in which the limits to olicy-making are also endogenous. The other main modeling differences are the unilateral decision-making and the lack of individual economic choices in this aer, which allow the analysis to focus on the government s institutional decision. There also exists a large literature that focuses on imlicit constraints to olicymaking that originate from the government s non-institutional decisions. For examle, Alesina and Tabellini (1990) rovides a model in which the current government ties the hands of its successor by borrowing. Similarly, Azzimonti (015) studies growth in an environment in which today s investment affects future governments sending. Besley and Persson (009) analyzes how governments build u fiscal and legal caacity that 6 Karakas (016) is motivated by the same uzzle, but offers exlicit costs of reforming executive constraints as an exlanation for it. 6

7 imrove their ability to rovide ublic goods in the future. Bowen, Chen and Eraslan (014) solves a dynamic bargaining model to study how different budgetary rules affect ublic good rovision. 7 By allowing the government to make both a olicy and an institutional choice, this aer aims to also contribute to this literature by introducing a channel through which these two decisions interact in equilibrium. 3 The Model The economy consists of two agents, A and B, who live for two eriods and care about a rivate consumtion good y and a ublic good g. In each eriod t, the economy roduces an exogenous surlus equal to K t units of the consumtion good, which is shared equally between the two agents. The ublic good g is a stock variable and can be roduced from the consumtion good y using a linear roduction technology g = y 1, where is a ositive constant. Agent i s stage ayoff is given by u(y i, g) = y i + λg α (1) for i = A, B, where λ is a constant that reresents the relative imortance of the ublic good to the rivate consumtion good and α (0, 1). There is no discounting. In each eriod, an incumbent is realized according to a fixed robability q i for i = A, B, where q A + q B = 1. Uon selection, the first-eriod incumbent unilaterally makes both a olicy choice under the given executive constraint and an institutional choice on the future executive constraint, whereas the second-eriod incumbent only chooses olicy. The executive constraint and the ublic good constitute the dynamic linkages between the eriods. The rest of this section describes these two tyes of choices: 7 Other examles of models in which today s olicy choices constrain tomorrow s decision-makers by altering the status-quo include Battaglini, Nunnari and Palfrey (01), Nunnari and Zaal (014), Cunha and Ornelas (015), Agranov, Frechette, Palfrey and Vesa (016) and Bowen, Chen, Eraslan and Zaal (017). 7

8 Policy Choice: In each eriod t, the incumbent decides on investment I t to the stock of the ublic good g t and rivate transfers of the consumtion good y A,t and y B,t resectively to agents A and B. These olicies are financed through a total lum-sum tax τ t shared equally between the agents. Equation (1) imlies that while the agents agree on the otimal level of the ublic good, they disagree over rivate transfers of the consumtion good. The ublic good may reresent defense or environmental rojects that both agents enjoy equally. On the other hand, the transfers are targeted; while both agents are required to ay their equal share of taxes for a given transfer of the consumtion good to agent i, only agent i receives utility from it. There exist logistical costs of making rivate transfers such that a unit transfer costs the economy b > 1 units of the rivate good. The initial stock of the ublic good is exogenously given by g 0 0. The stock of the ublic good dereciates each eriod at the rate δ (0, 1) and its evolution is described by g t = (1 δ)g t 1 + I t. Investments to the ublic good are reversible so that I t (1 δ)g t 1. Given the roduction technology g = y 1, the arameter reresents the rice of investment in terms of the consumtion good. I assume that a unit of disinvestment also costs units of the consumtion good and can be consumed as a rivate good. 8 The government must balance its budget in both eriods such that by A,t + by B,t + I t τ t () for t = 1,. Since the lum-sum taxes on the agents endowments of the rivate good are required to be uniform, agent i s eriod-t utility after the eriod-t incumbent makes 8 Since the ublic good can be laundered into the consumtion good by disinvesting and taxation is non-distortionary, a sufficiently low rice could make it rofitable to invest just to consume these investments in the next eriod. In order to rule out this ossibility, I assume that > 1 so that it is strictly not otimal to invest beyond the ublic good state where marginal benefit of a unit of investment equals its marginal cost. 8

9 his decisions becomes K t + y i,t + λg α t τ t. (3) Since an incumbent would never leave tax revenues unsent, () always holds with equality. Thus, substituting the government budget constraint into (3) yields K t + y i,t + λg α t 1 (by A,t + by B,t + I t ). (4) In order to simlify the model, I assume b < 3 so that the choice of τ t need not be searately considered. This is because when b (1, 3 ), an incumbent always sets τ t = K t. Institutional Choice: To simlify the resentation of an executive constraint, first note that incumbent i does not receive utility from the rivate consumtion of the other agent so that y j,t = 0 for j i and t = 1,. Therefore, his olicy choice can be described by the air (I t, y i,t ) R. An executive constraint imoses the same uer bound (but no lower bound) on an incumbent s sending on each olicy. Secifically, letting ȳ t and Īt resectively denote the maximum sending ermitted on self-transfers and investments (or disinvestments), I assume that ȳ t = Īt l t for all t. Taking the institutionally feasible set in R as given, the first eriod incumbent i chooses his sending on (I 1, y i,1 ) from this set and decides on the institutionally feasible set that will constrain the second-eriod incumbent s olicy choice. The assumed structure for an executive constraint imlies that this choice is equivalent to choosing a square with its center at the origin in R and whose corners reresent the most extreme olicies ermitted for either incumbent. Letting yi for i = A, B and I g 1 denote the agents ideal exenditures in the second eriod for a given level of the ublic good g 1 and the axes measure sending in terms of the rivate consumtion good, Figure 1 illustrates two institutionally feasible sets: While the outer square ermits the imlementation of both 9

10 y y A y I g 1 y B Figure 1: Two reresentations of executive constraints in R. agents ideal olicies, the smaller square is more restrictive. 9 The motivation for treating an institution as a constraint on ayoff-relevant olicies is to emhasize its instrumentality: agents only have intrinsic olicy references and care about executive constraints insofar as it enables or inhibits their imlementation. Similar to a budget constraint, the underlying rationale for the resent structure is that an incumbent should not be able to fine-tune his institutional choice. Choosing a singleton set whose only member is the incumbent s ideal olicy is one extreme examle of such fine-tuning and should be ruled out. Instead, if an incumbent exands the set of ermitted olicies, he does so for both otential future incumbents to enjoy. However, an executive constraint is also different from a budget constraint in the sense that decreasing sending on one olicy at either corner of the institutionally feasible set does not free u resources for the other. The requirement that sending cas must be uniform across olicies limits the overall extremeness of a olicy ackage by tying together the agents common interest and conflict. 9 In this figure, the first two quadrants reresent a olicy sace for agent A while the third and the fourth quadrants are for agent B. Although the agents ideal olicies do not lie in the same sace, it is convenient to exress them as here by exloiting the fact that agents only make self-transfers. 10

11 4 The Benchmark Solutions Studying executive constraints as an endogenous limit on olicy-making is only nontrivial under the ossibility of government turnover; a dictator who will forever remain in office or a lanner who cares about the society s aggregate utility would never willingly tie their own hands. Both would choose an executive constraint that would not be binding on their olicy choices in the next eriod. Nonetheless, their otimal olicies create benchmarks for evaluating ublic good rovision under electoral uncertainty. For simlicity, the following roosition assumes that the first-eriod executive constraint is binding and the initial stock of the ublic good g 0 is sufficiently low that disinvestment is never otimal Proosition 1. The stocks of the ublic good rovided resectively by the dictator D and the lanner P in the final eriod are given by g D g P = min = min { (bλα { (bλα ) 1 1 α, (1 δ) g 0 + (1 δ)ī1 + K ) 1 1 α, (1 δ) g 0 + (1 δ)ī1 + K } } ; (5). (6) Notice that the two solutions are equivalent when the latter term in (5) is less than the former, which reresents the case in which the dictator and the lanner allocate their entire budgets to the ublic good in the second eriod and therefore do not make any rivate transfers. This haens when the marginal benefit of an additional unit of investment exceeds the constant marginal cost by a sufficiently large margin that this imbalance erseveres even at the better state of the ublic good attained by exhausting the budget. For examle, a sufficiently low initial stock, a restrictive first-eriod exec- 10 All the roofs are in the Aendix. 11 Battaglini, Nunnari and Palfrey (01) analyzes a similar lanner s roblem with an infinite horizon in which the rivate transfer cost arameter b is set equal to 1. 11

12 utive constraint or a high dereciation rate may drive such high investment needs. In contrast, the levels of the ublic good given in the first terms of (5) and (6) equate the marginal cost of an additional unit of investment with its marginal benefit resectively for the dictator and the lanner. It is when the investment costs required to bring the ublic good to these levels are within budget that the dictator and the lanner make rivate transfers of the consumtion good. While the dictator makes only self-transfers, any allocation of the remaining resources after investment sending between the two agents is otimal for the lanner. 1 A comarison of (5) and (6) indicates that ublic good rovision is higher under a lanner for all states of the ublic good at which the dictator is making rivate transfers. Intuitively, this is because the lanner maximizes the aggregate utility of the two agents, whereas the dictator cares only about his own utility. 13 Accordingly, the dictator does not factor into his investment decision the ositive externality it creates. Since each agent enjoys the rivate consumtion good linearly, the lanner s solution maximizes the society s welfare. However, the dictatorshi solution also rovides a useful benchmark as it reresents an intermediate case between this otimum and the olitical equilibrium. 5 Equilibrium under Electoral Uncertainty This section solves for the Subgame Perfect Nash equilibrium under electoral uncertainty. The given incumbent at t = 1 chooses a olicy vector (I 1, y A,1, y B,1 ) and the executive constraint l, taking g 0 and l 1 as given. At the beginning of t =, the new incumbent is realized according to the fixed robability q i for i = A, B. Now taking g 1 = (1 δ)g 0 +I 1 and l as given, he chooses (I, y A,, y B, ). The main results will characterize how the 1 The same logic alies if g 1 is assumed to be too high, for examle due to a combination of a high g 0, a low Ī1 and a low δ, such that the marginal cost of an additional unit of investment exceeds its marginal benefit at the state (1 δ)g 1 and the otimal action is to disinvest. 13 This can alternatively be exressed by asserting that the dictator is identical to a lanner who assigns all the weight to his own agent-tye when maximizing the society s aggregate utility. 1

13 equilibrium executive constraint deends on the distribution of electoral ower and the state of the ublic good. The ayoff-relevant states of the world in a given eriod are the executive constraint described by the uniform sending ca and the ublic good. A ure olicy strategy for incumbent i in eriods t = 1, consists of a air of rivate transfer rules ρ i j,t : R + R + for j = A, B such that ρ i j,t (l t, g t 1 ) lt b that γ i t(l t, g t 1 ) (1 δ)g t 1 lt and an investment rule γ i t : R + R + such for a given air (l t, g t 1 ). The equilibrium rivate transfer rules ρ i j,t (l t, g t 1 ) = y j,t for j = A, B yield the amount of the consumtion good incumbent i allocates to agent j in eriod t = 1, for any given l t and g t 1. The equilibrium investment rule γ i t(l t, g t 1 ) = g t yields the level of the ublic good incumbent i determines for eriod t = 1, through his investment choice I t. In addition, a ure institutional strategy for incumbent i in eriod 1 is an executive constraint rule θ i : R + R +, where θ i (l 1, g 0 ) = l yields the uniform sending ca that constrains the olicy choices in t = for any given l 1 and g 0. Definition 1. The strategy rofile σ i = (ρ i A,1, ρi B,1, ρi A,, ρi B,, γi 1, γi, θi ) for incumbent i {A, B} is a Subgame Perfect Nash equilibrium if and only if (1) Given l and g 1, incumbent i s olicy strategies ρ i j, (l, g 1 ) for j = A, B and γ i (l, g 1 ) solve max y i, + λg α by i, + by j, + I y i,,y j,,i (7) subject to 0 by i, l ; 0 by j, l ; I (1 δ)g 1 ; I l ; g = (1 δ)g 1 + I and by i, + by j, + I K, where j i and i {A, B}. () Taking as given the otimal olicy strategies ρ i j, (l, g 1 ) for j = A, B and γ i (l, g 1 ) that incumbent i {A, B} would follow in t =, incumbent i s institutional strategy θ i (l 1, g 0 ) = l and olicy strategies ρ i j,1 (l 1, g 0 ) for j = A, B and γ i 1 (l 1, g 0 ) solve 13

14 for any given l 1 and g 0 max y i,1 + λg1 α by i,1 + by j,1 + I 1 y i,1,y j,1,i 1,l (8) + [ q k ρ k i,(l, g 1 ) + λ(γ k (l, g 1 )) α bρk i, (l, g 1 ) + bρ k j, (l, g 1 ) + γ k(l ], g 1 ) (1 δ)g 1 k=i,j subject to 0 by i,1 l 1 ; 0 by j,1 l 1 ; I 1 (1 δ)g 0 ; I 1 l 1 ; g 1 = (1 δ)g 0 + I 1 ; l 0; and by i,1 + by j,1 + I 1 K 1, where j i and i {A, B}. Condition (1) requires the otimality of the second-eriod incumbent s olicy choices subject to the economy s budget constraint and the executive constraint determined by the revious incumbent. Condition () imoses otimality on the first-eriod incumbent s olicy and institutional choices. Anticiating the otimal olicy strategies that would be followed by either ossible incumbent as functions of the executive constraint and ublic good states that he would ass on to the second eriod, the first-eriod incumbent maximizes his dynamic utility subject to the given constraints. In order to simlify the exosition of the analysis, I assume that the initial stock of the ublic good g 0 is sufficiently low that disinvestment does not occur in equilibrium. This is a realistic assumtion that does not alter the main results. Assumtion 1. The initial stock of the ublic good g 0 is such that λαg α 1 0 > b. Characterizing an incumbent s equilibrium institutional strategy first requires understanding the second-eriod olicies that would be followed by either agent. Similar to a dictator, the second-eriod incumbent i {A, B} only makes self-transfers so that ρ i j, (l, g 1 ) = 0 is true in equilibrium for any state air (l, g 1 ) and j i. Since incumbent i faces both a budget constraint and an executive constraint, there are three ossibilities for his otimal olicy choices in t = deending on the given l and K : First, if the executive constraint l is sufficiently weak that K < l, then it is the 14

15 budget constraint that binds incumbent i s investment and self-transfer choices. In this case, equilibrium olicies are identical to a dictator s in the second eriod. Second, if the economy s endowment of the consumtion good K is sufficiently large that the budget constraint would not bind before the executive constraint, then incumbent i s olicy choice deends on whether l is in fact binding on investment. If it is, then incumbent i chooses ρ i i, (l, g 1 ) = l b and γ i (l, g 1 ) = (1 δ)g 1 + l, sending the maximum amount ermitted by the executive constraint on each olicy. If it is not binding, then only the self-transfers are constrained. This requires K l. Finally, the executive constraint may bind the investment choice while the budget constraint binds the self-transfer such that γ i (l, g 1 ) = (1 δ)g 1 + l and ρ i i, (l, g 1 ) = K l b. This requires l K < l. 14 It is imortant to note that the agents would make the same investment choice in the second eriod, because they enjoy the ublic good equally and second-eriod investment only confers consumtion benefits. On the other hand, while they agree on the amount of self-transfers, each would target this amount to himself. Since the consumtion good enters the agents utilities linearly, the institutional choice can be analyzed based on whether the the budget constraint would still bind rivate transfers if the executive constraint allows for the common ideal investment sending. The following roosition describes an incumbent s equilibrium institutional strategy based on whether it restricts investments: Proosition. For all l 1 and g 0 such that I γ i 1 (l 1,g 0 ) K, incumbent i s equilibrium institutional strategy θ i (l 1, g 0 ) restricts future investments as well as rivate transfers if and only if b q i > 1. For all other l 1 and g 0, the condition b q i > 1 is necessary, but not sufficient, for θ i (l 1, g 0 ) to restrict future investments. Proosition characterizes an incumbent s otimal institutional choice by focusing 14 Note that a fourth ossibility in which the budget constraint binds investment while the executive constraint binds the self-transfer does not exist, because this would imly I = K < l and y i, = 0. Since l > 0, y i, is determined by the budget, not the executive, constraint. 15

16 on two searate regions of the ublic good determined by the states l 1 and g 0. First, suose l 1 is sufficiently weak and g 0 is sufficiently large that the agents common ideal investment exenditure in the second eriod, I γ1 i (l, is small relative to the size of the 1,g 0 ) economy in that eriod. In this region, there exists sufficient room for rivate transfers in the budget that an uer bound on exenditures equal to the common ideal investment sending would constrain transfers first before the budget constraint. The roosition states that the equilibrium executive constraint also restricts investments in order to further restrict rivate transfers if and only if b is sufficiently large and q i is sufficiently low that b q i > 1 is satisfied. This is because when the cost of making transfers is high and the chances of himself enjoying these transfers is low, the incumbent s incentive to rotect himself against his oonent s otential ork sending by restricting transfers is strong enough to warrant sacrificing some common-cause investments. Note that while q i < 1 is necessary for this condition to hold, it is not sufficient - as the arameter b decreases, the level of q i below which the incumbent finds it otimal to restrict investments also decreases. On the other hand, when b q i 1, the equilibrium executive constraint ermits the agents ideal investments. The analysis indicates that when this is the case, the otimal l equals I γ1 i (l 1,g 0 ) for q i q j < b 1 and satisfies l K I γ i 1 (l 1,g 0 ) otherwise. Intuitively, given that investment won t be sacrificed to his rotection motive in equilibrium, the incumbent chooses the tightest constraint that still allows for the ideal investments if his re-election robability is sufficiently small and the cost of transfers is sufficiently high, and lets the budget constitute the sole constraint on transfers otherwise. For all other states of l 1 and g 0 for which the economy s second-eriod investment needs are relatively large, there exist less room in the budget for transfers and thus lower risk from allowing for the ideal investment sending. As stated in Proosition, if the incumbent chooses an executive constraint in equilibrium that restricts investments in this region, then the condition b q i > 1 must be satisfied. However, the converse is not 16

17 necessarily true. To see this, note that the budget constraint is always the one binding transfers in this region and any l [ K, I γ1 i (l ) would simly imose a bound on in- 1,g 0 ) vestments without simultaneously restricting transfers. This can clearly not be otimal for any arameter values. An incumbent with a sufficiently low q i that satisfies b q i > 1 may still choose to leave investments unrestricted since the budget offers him rotection. If the incumbent chooses not to restrict investments in equilibrium, then any executive constraint that ermits the ideal investment exenditures would be otimal. 15 An imlication of Proosition is that there can only be one agent whose equilibrium institutional strategy would restrict investments. This is because if θ i (l 1, g 0 ) restricts investments in equilibrium, it must be true that q i < 1 and q j > 1 for j i. Since q j > 1 imlies b q j < 1, which is a sufficient condition for θ j (l 1, g 0 ) to allow for the agents common ideal investment sending in equilibrium, it follows that the agents would have oosite institutional strategies with regards to restricting investments. However, it is ossible in equilibrium for each agent to refer leaving investment sending unrestricted. These results imly the following relationshi between the equilibrium executive constraints and electoral uncertainty: Corollary 1. Incumbent i s equilibrium institutional strategy θ i (l 1, g 0 ) is non-decreasing in his re-election robability q i for any given (l 1, g 0 ) and i {A, B}. The roof of Corollary 1 shows that the equilibrium rule θ i (l 1, g 0 ) is monotonically increasing in q i for all l 1 and g 0 for which the agents shared ideal investment sending is less than the economy s endowment in the second eriod and is constant for all others. Since the budget constraint binds investments in the latter case, it is uninteresting as the 15 Note that an incumbent would never restrict investments in equilibrium if the model allowed for b = 1. Suose l just allows for the agents common ideal investment sending. Since lum-sum taxes are uniform, a unit decrease in l imlies one unit of tax savings for the incumbent when l also binds transfers and an exected utility gain from extra transfers equal to q i b when it doesn t. His exected loss equals q i when l binds transfers lus investment equal to 1. Since λαgα 1 = here, the utility loss b from investment equals 1. When b = 1, the incumbent comares a loss of 1 + qi with a gain of 1 when b l binds transfers and a loss of 1 with a gain of q i when it does not: the marginal cost of restricting investments always exceeds its marginal benefit. 17

18 ideal self-transfers equal zero. However, the former case demonstrates an incumbent s institutional choice trade-off between the incentives to imlement his referred olicies in case of his re-election and to avoid artially financing his oonent s self-transfers in the oosite scenario. When his re-election robability equals zero, the incumbent is motivated solely by his investment interests and rotecting himself from his oonent. As his electoral rosects imrove, the incentive to make self-transfers starts laying a role in equilibrium, resulting in gradually weaker executive constraints and thus lower investment sacrifices. Once his re-election robability reaches a threshold, the incumbent no longer finds it otimal to restrict investments. 16 In this range of higher re-election robabilities for which the incumbent does not restrict investments, the equilibrium executive constraint is either at one of two extremes deending on the incumbent s reelection robability or is constant, as described below Proosition. Corollary 1 reroduces within the context of the resent model the main result of the literature on the relationshi between olitical turnover and the strength of executive constraints. However, it is imortant to recognize that the state of the ublic good is held constant for this result. The discussion following Proosition exlained how the otimal executive constraint deends on the incumbent s future ideal investment and rivate transfer olicies, which in turn deend on the current state of the ublic good and the economy s endowment. The following corollary formalizes this relationshi between the state of the ublic good and the equilibrium executive constraints: Corollary. For all arameter values such that incumbent i s equilibrium institutional strategy satisfies θ i (l 1, g 0 ) < I γ i 1 (l 1,g 0 ) for any given (l 1, g 0 ), it is non-increasing as the ublic good g 0 aroaches the dictatorial benchmark state ĝ ( bλα ) 1 1 α for i {A, B}. Focusing on arameter values for which incumbent i restricts investments as well as 16 This threshold is given by q i = b 1 for the first case discussed below Proosition in which I K γ1 i (l 1,g 0 ). For the second case, it is given by some robability q i for which the equation that defines the otimal l < I equals K γ1 i (l 1,g 0 ). 18

19 rivate transfers in equilibrium, the analysis behind Corollary indicates that the equilibrium executive constraint gets tighter as the initial ublic good aroaches a state at which the marginal benefit of a unit of investment equals its marginal cost if γ1 i(l 1, g 0 ) is increasing in g 0, and is constant otherwise. Intuitively, the agents shared investment needs are ressing enough at low states of the ublic good that investing would yield large efficiency gains. These are the states at which disagreement between the agents over rivate transfers is more muted so that the incumbent has an incentive to kee the executive constraint relatively weak in order to enable the much-needed investments, regardless of how low his re-election robability might be. However, as the ublic good moves away from these states toward ĝ, the common cause over investing starts to disaear. Accordingly, the equilibrium executive constraints get tighter as the agents no longer have a reason to tolerate each other s ork sending for the sake of enabling investments. At the dictatorial benchmark state, all the mutually beneficial gains from investing have been exhausted and rivate transfers take center stage in the incumbent s institutional choice. This is observed if γ1 i(l 1, g 0 ) = ĝ so that the initial state g 0 ceases to lay this role in determining the equilibrium executive constraint. Notice that the above result is not true for arameter values at which the incumbent s equilibrium institutional strategy leaves the agents ideal investment sending unrestricted. This is due to the fact that once the ideal investments are ermitted and there still exists room in the budget for making self-transfers, the conflict over rivate transfers becomes the sole determinant of executive constraints in equilibrium. The state of the ublic good affects the equilibrium here only by determining the conditions the otimal executive constraint must satisfy. 17 Intuitively, choosing an executive constraint that will move the state of the ublic good to the dictatorial benchmark allows disagreement between the agents to become the dominant force and institutional choice 17 Secifically, recall that l = I γ i 1 (l 1,g 0 ) for sufficiently low qi and satisfies l K I γ i 1 (l 1,g 0 ) for sufficiently high q i if I γ i 1 (l 1,g 0 ) K and satisfies l I γ i 1 (l 1,g 0 ) if I γ i 1 (l 1,g 0 ) ( K, K ). 19

20 to become unfettered from the agents common investment interests. Corollary offers a qualification to the main result established in the literature and re-iterated in Corollary 1: While executive constraints resond to changes in electoral uncertainty, this resonse deends on the economy s level of ublic good develoment. However, before resenting how these results hel exlain the emirical uzzle on why electorally owerful incumbents may resond differently to a change in their re-election rosects, it is necessary to first comlete the analysis by describing an imortant equilibrium feature of an incumbent s first-eriod investment strategy. Proosition 3. The equilibrium investment strategies that would be followed in t = 1 are such that γ A 1 (l 1, g 0 ) γ B 1 (l 1, g 0 ) for any given (l 1, g 0 ) and q A q B if and only if the same constraint does not bind both strategies and I γ i 1 (l 1,g 0 ) < K. Proosition 3 states that desite sharing the same fundamental references over the ublic good, agents with different electoral rosects would make different first-eriod investment choices in equilibrium in the absence of a mutually binding constraint. That investment in the first eriod is not solely motivated by a shared interest in ublic good consumtion is driven by two searate incentives: maniulating the future incumbent s olicy choices through the executive constraint and through the budget. To demonstrate the former tye of maniulation, suose l 1 and g 0 are such that θ i (l 1, g 0 ) would constrain both y and I choices in equilibrium, which requires q i < 1. For this case, Corollary shows that the equilibrium executive constraint gets tighter as the state of the ublic good imroves. In order to exloit this relationshi and take advantage of the tighter executive constraint and thus lower self-transfers a higher state of the ublic good would bring about, the first-eriod incumbent, who must be sufficiently weak electorally, invests more in the first eriod than the amount a simle interest in ublic good consumtion would rescribe. Alternatively, suose l 1 and g 0 are such that it is the budget constraint that would bind y in equilibrium. Since this is ossible only if 0

21 investment is unrestricted by l, the otimal y must be increasing in g 1. In the absence of a binding constraint on investment, the otential of making more self-transfers constitutes the future benefit of investing today. With his entire endowment taxed in either incumbency scenario due to the non-binding executive constraint, an incumbent with a ositive robability of re-election would always invest more in order to maniulate the future allocation of the budget, with this incentive growing stronger with his re-election robability. The results in this section yield insights into why incumbents facing similar re-election rosects may behave differently. Comaring two societies with similar states of ublic good develoment but with different re-election rosects for their incumbents, Corollary 1 indicates that the equilibrium features a stronger executive constraint in the country with the greater rosect for incumbent turnover. However, Corollary comlements this story by introducing variations in ublic good states as an additional driver of incumbent behavior: Now comaring two societies that are characterized by a similar distribution of electoral ower but are at different stages of ublic good develoment, the equilibrium executive constraint is tighter in the higher ublic good society as long as the executive constraint is relevant for investment choices. This is because rivate consumtion lays a relatively more imortant role in equilibrium in societies that have already exhausted the common gains from investing in the ublic good. Because the agents enjoy the ublic good equally, the only source of dynamic asymmetry in the model is electoral ower, the extent of which determines the divergence between the agents otimal institutional and investment choices. The following section focuses on the imlications of this divergence for ublic good rovision by comaring it to the dictatorshi and lanner benchmarks established in Section 4. 1

22 6 The Efficiency Imlications of Equilibrium The revious section analyzed a feedback loo between the states of the ublic good and the executive constraint: The state of the ublic good affects the otimal executive constraint by determining the intensity of the agents conflict over self-transfers. In turn, the incumbent chooses the final state of the ublic good under this constraint. Therefore, in addition to the budget constraint, an institutional structure that restricts the overall extremeness of olicies instead of imosing searate sending limits on each imlies that the investment decision cannot be uncouled in equilibrium from an incumbent s reference for making self-transfers. Based on the insights from the revious sections, the following roosition evaluates the equilibrium under electoral uncertainty: Proosition 4. The equilibrium rovision of the ublic good under electoral uncertainty is inefficient. In order to highlight the role each of the model s main assumtions lay on this inefficiency result, Proosition 4 can be analyzed in two arts: First, note that the dictator s ublic good rovision is lower than that of the lanner, excet for when the budget constraint binds both their second-eriod investment choices. As discussed in Section 4, this is due to the fact that the dictator only cares about himself while the lanner maximizes the agents joint utility. The second source of the inefficiency is the role re-election uncertainty lays on an incumbent s institutional choice as the inherent conflict over rivate transfers may lead the incumbent to also restrict future investments for further rotection. Universal taxation for financing agent-secific consumtion constitutes the underlying driver of this effect. Accordingly, equilibrium under electoral uncertainty would be equivalent to a dictatorshi in the absence of distributive olitics. Since the choice of an executive constraint in equilibrium that also restricts investments is the main driver of the wedge between the dictator s and the incumbent s investment rules, Corollary 1 imlies that the efficiency of equilibrium imroves as the

23 first-eriod incumbent s re-election robability increases and hence his rotection motive loses force. Similarly, any arameter changes that affect the equilibrium investment rule γ1 i(l 1, g 0 ) while maintaining the validity of the condition in Corollary that requires the equilibrium institutional rule to restrict investments can be exected to imact efficiency. For examle, under this condition, a higher rice of investment and a lower cost of self-transfers b lead to a weaker executive constraint and therefore less restrictions on investments in equilibrium as the resulting lower ublic good state g 1 makes common cause more ronounced between the agents. Likewise, a higher value of λ, for instance due to a natural disaster, would lead to a tighter executive constraint that would restrict investments more in equilibrium as it leads to a greater state of g 1 and romotes the agents conflict over self-transfers. 7 Concluding Remarks This aer analyzed how a government facing re-election uncertainty determines the olitical institutions that limit executive ower by focusing on its trade-off between roductive investments and ork sending. An incumbent s institutional choice is shaed by his oosing incentives to invest in a common-cause ublic good while restricting his oonent s otential future self-transfers. The intensity of the disagreement between the agents over rivate transfers and the distribution of electoral ower together determine which of these effects dominates in equilibrium. The main results indicate that common-cause investments get sacrificed to the agents conflict over rivate transfers when the cost of making transfers is sufficiently high and the incumbent s re-election robability is sufficiently low. For all states of the ublic good at which the agents shared ideal investment sending for the second eriod is not constrained by the budget, the equilibrium executive constraint gets monotonically weaker as the incumbent s re-election robability increases. However, this relationshi 3

24 holds only for a given state of the ublic good, which affects equilibrium by determining whether an incumbent s investment or self-transfer interests are dominant. Holding electoral ower constant, more investment gets sacrificed in equilibrium relative to the economy s needs to restrictive executive constraints as the ublic good aroaches the dictatorial benchmark state. Together, these results imly that ublic good rovision in inefficient in equilibrium even though the agents share the same references over its consumtion. The finding that both electoral uncertainty and the state of the ublic good determine the equilibrium executive constraint sheds light on why incumbents facing similar re-election rosects may nonetheless make different institutional choices. The aer makes a number of modeling choices for tractability that could be modified. First, the re-election robabilities are exogenously fixed so that an incumbent only considers the olicy imlications of his institutional choice. Introducing voter references into this framework would yield richer dynamics, whether these references are defined only over olicies or the voters are endowed with institutional references. Second, the model focuses on one among various ossible structures for an institutionally feasible set. While the focus here is on institutions that limit the extremeness of olicy choices, different structures could be more aroriate for settings in which incumbents have the ability to imose searate limits on olicies. Finally, the model does not allow agents to make economic decisions such as labor and considers only lum-sum taxation of the economy s exogenous surlus. While inherently restrictive due to ruling out distortionary taxation, this modeling choice allows the discussion on the efficiency of ublic good rovision to focus exclusively on the distortions created by an endogenous executive constraint. 4