Political Careers Concerns and Political Parties Claudio Parés Bengoechea Departamento de Economía Universidad de Concepción Serie Documentos de Trabajo EconUdeC 0-010
Political Careers Concerns and Political Parties Claudio Parés Bengoechea Universidad de Concepción / Toulouse School of Economics (GREMAQ) cpares@udec.cl Abstract Political parties increase the information citizens have about their members. In fact, when an incumbent president reveals his ability he affects his comrades in two ways: first, he adds one more observation to the prior distribution, lowering uncertainty, and he creates a reputation effect if his performance signals a level of ability different of the expected. Hence, the existence of political parties increases effort of its members by decreasing uncertainty but lowers it by diminishing the marginal effect of politician s effort on their probability of being elected. JEL codes: D7, D73, D8. 1 Introduction Political economics often uses personnel economics to understand incentives problems of politicians. On one hand, politicians become agents that need to be selected, monitored once in office, and compete by signalling ability or ideology to extract rents from office or provide public goods. On the other hand, people become the principal and elections provide incentives: incumbent politicians may be reelected, promoted to higher offices or ousted by citizen s ballots. Although elections work as the main mechanism to correct moral hazard and adverse selection of politicians, reelection of the president is often limited to a number of periods and lower level politicians compete to climb up the hierarchy. In fact, in most elections, ministers, senators, governors and majors run for president. The political career is determined by competition among other politicians and the signals they can send to the public and the main signal available is the membership to a political party. In fact, political parties help reducing uncertainty of citizens by sending information about abilities of its members. Jones & Hudson (1998) claim that the lowest cost signal politicians can send to Preliminary manuscript. I thank my advisor Gilles Saint-Paul for precious guideness. I also thank Phillipe De Donder for very useful comments and Hugo Salgado for precious help. 1
voters is their membership on a political party. The signal provides information on policy position [and] information about personality traits of the candidate as the party candidate will have been screened in a selection process for attributes that party organisers regard as important. They conclude that membership of political parties is one of the signals that voters use when attempting to estimate the quality of political representatives. Selection and internal structure of political parties define personality traits and quality of its politicians. The size of the party, the amount of active politicians or potential candidates, the level of internal dissent, the existence of primaries, etc. differ from one party to the other and send different signals about the quality of its members. For example, Caillaud & Tirole (00) argue that candidates need to convince rank-and-file of their party before running for elections and the quality of this filter depends on the mechanism chosen by the party; Murase (004) postulate that high quality politicians handicap themselves by choosing an extreme party to signal their quality. If internal processes of political parties are not fully known by citizens, every time a member of the party reveals his ability, he decreases the uncertainty about the ability of his comrades. In fact, voters have one more observation to discover the distribution of politicians abilities belonging to that particular party. In this way, political parties decrease uncertainty about the ability of the candidates and increase effort made by incumbent politicians that have its membership. Moreover, the same signal changes the reputation of the whole party, either positive or negatively. This reputation effect would decrease the effort of a politician. In fact, a politician that receives a bequest from a comrade knows that his effort has a lower effect on the possibility of being elected. If this bequest is positive, he needs to exert less effort to prove better than an independent challenger. If he inherits a negative signal, he quits the race before running. In any case, the politician exert a lower effort because reputation worsen incentives. Therefore, the existence and strength of political parties involve a trade off: they decrease uncertainty and enhance reputation effects. The first effect increases effort of partisan politicians while the second decreases it. First, uncertainty about abilities, i.e. higher variance of the prior beliefs about the distribution of abilities among politicians, decreases the effort exerted by politicians. In fact, their effort has to signal that they are more able than the challenger and more uncertainty decreases the probability of being pivotal in this competition. Persson & Tabellini (000) claim that equilibrium rents are lower the narrower is the range of uncertainty and the larger is the value of winning the elections (p. 85). Alesina & Tabellini (005) explain the difference of the political model with the personnel economics model of Holmström (1999): if the reward is proportional as a wage or peers recognition of skills uncertainty increases effort, while if the reward is binary reelection or not uncertainty decreases effort. Hence, the decrease in uncertainty created by the membership to a political party increases effort because when all politicians are perceived as alike there is harder competition to prove better than a competi-
tor. In this internal competition, the marginal effect of effort on the probability of being elected is higher the more certain are prior beliefs about politician s abilities. Second, reputation decreases the effort of politicians. On one hand, if reputation of the party is positive, a politician who belongs to that party does not need to exert too much effort to prove better than the benchmark. In other words, he can rest on his party s oars and win elections with less effort than a politician without membership. On the other hand, if reputation of the party is negative, a politician who belongs to it exerts less effort because his effort has a little effect on his perceived ability. Therefore, is he can do little to change his probability to be elected president, he throws in the towel and exert less effort than a politician without membership. Accountability adds a third effect. In fact, different offices have different impact on the utility of citizens. The evaluation citizens do about the ability of politicians occupying these offices depends on the accountability of the office itself. Offices differ in bureaucracy, media coverage or simply because they provide different public goods. Hence, the job (and the evaluation) of a senator is different from the job of a governor or a major, but all the three may want to run for president. Therefore, the effort exerted by a politician change from one office to the other. The more accountable is an office, the more effort is exerted by the politician occupying it. This paper presents a career concerns model in which neither politicians nor voters know the ability of politicians in office. Agents exert effort to provide public goods and citizens infer their ability from their performance. High ability politicians are elected to go up the political hierarchy and low ability politicians retire. The existence of political parties introduce a trade off in the incentives of politicians that allows to study political institutions: uncertainty vs. reputation. Politicians have to compete in elections against an outsider or another politician. In the first case, the politician has to prove better than an unknown challenger that is believed as good as the average politician. In the second case, the challenger is also under evaluation and he may prove better (or worse) than expected. This uncertainty about the opponent decreases effort of politicians because the relevant signal is no longer their ability but the difference with their opponent. Following the reasoning about uncertainty above, this limited competition decreases the effort exerted by politicians. Finally, this model allows to analyse a cost that has not been taken into account in political economics literature about decentralization (to the best of my knowledge). Centralization implies that politicians in all the offices of government belong to the same political party. Hence, if parties are informative about the ability of its members, having a bigger sample make the prior beliefs about candidates much more precise that several parties in power under decentralization. In fact, under decentralization politicians may belong to different parties and the evaluation of each politician would be lees precise, this divided government may lose information revealed by strong political parties in a centralized system. Therefore, to the usual accountability effect that increases effort made by 3
politicians when institutions are decentralized, I add the effects of political parties, namely the uncertainty and the reputation effect. Although reputation effect works in the same direction of accountability, the uncertainty effect may counter the balance and justify that in a country with strong political parties, centralization leads to better outcomes. Set up Assume a country with a two-level hierarchy: a president is the head of the national level and a governor is the head of the regional level. The president provides national public goods Y using his ability η P and some effort E. The governor that could also be a major, senator or another politician occupying an office provides public good y using also his ability η and his effort e. Y = η P +E (1) y = η +e () Politicians can not be reelected and they can only go up the hierarchy. President has to retire after his mandate and governors can become president with probabilitypafterelections,otherwiseheretires. 1 Presidentextractprivatebenefit B and bear a cost C(E) from holding office with C(0) = 0 and C,C > 0, while governors extract private benefit b and bear a cost c(e) from the effort they exert in office, with c(0) = 0 and c,c > 0. Utility for the president, u P, and for the governor, u G, are u P = B C(E) u G = b c(e)+p u P (3) Citizens observe provision of both public goods and extract utility U(Y, y). Indirectly, more effort and more ability from politicians give citizens more utility. Taking the career concern model on Persson and Tabellini (000, ch. 4), timing is as follows: 1. Nature puts random politicians in office. The ability of the president is revealed, while the ability of the governor remains unknown. Citizens know that ability follows a normal distribution with unknown parameters but a prior mean η 0 and a prior variance σ if there is no additional information.. Politicians decide on their efforts. Lower level politicians have no information about their own ability when deciding on their effort. 1 If the governor could stay in power he should be evaluated in some way, however this evaluation would require him to prove better than the outside option, i.e. a politician with ability η 0, but this is, in fact, the condition to be elected president as it will be defined below. 4
3. Public goods Y and y are determined by (1) and (). Citizens observe Y and y but neither η nor e. 4. Elections are held. President retires and the governor runs for president. Citizens infer his ability from production of public good y and elect him if he is perceived as a good politician. If he is not elected, a random politician with expected ability η 0 is elected president. 5. The incumbent politician or outsider seat on president s office and a new politician is randomly put in the other office. Efforts are set, and public goods are determined. The decision of the president is simple. In stage, he has no incentive to exert any effort because if he provides a positive effort he would decrease his utilitywithnocompensation. Hence, hisoptimalchoiceistoexertefforte = 0. In the final stage 5 the national public good production Y would come only from the ability of the president η P, therefore citizens vote for the most able politician to occupy that office. In stage 4, citizens have two choices: the governor and a random outsider. Therefore, the governor is elected president next period if he is perceived as more able than the outsider. The inferred ability of the governor at the moment of elections is η = y e (4) where e is the solution to the politician s optimization problem in stage that is to be derived. In stage, i.e. when deciding his effort, η is a random variable. Therefore e can not depend on its value because it is unknown at the moment of making the decision. Hence, the probability that the governor is elected president is the probability that voters infer that his ability exceeds the outsider s, i.e. p = Pr[ η η 0 ] Combining () and (4), η η 0 is equivalent to η η 0 e+e DefineF( )asthecumulativedistributionfunctionofη andf( )asitsdensity function. p = 1 F (η 0 e+e ) Using the first order condition that maximizes (3) and the fact that e = e in equilibrium, c (e ) = f (η 0 ) B 5
.1 Uncertainty and Reputation Effect Political parties are ex ante identical. However, the revealed ability of the president adds one more observation to the beliefs about the ability his comrades. A new observation of a distribution with unknown mean and variance has two effects: it reduces the variance and it may increase or decrease the mean of the posterior beliefs (see DeGroot (1970) p. ) Therefore, if the distribution of η is normal, when the politician belongs to the party of the president, the distribution of his ability is also normal with mean η 0 +β and variance δ σ, with β > 0 if and only if η P > η 0 and δ < 1. If the politician belongs to a different political party, there is no new information and η N ( η 0,σ ). Partisanship of politicians define the distribution f( ) and, hence, the effort e. Career concerns oblige active politicians to exert effort until the point they expect to be perceived just more able than the competitor. Define the effort of and independent politician as e I and the effort of a partisan politician as e P. c ( e I) = c ( e P) = B πσ { } B β πδ σ exp δ σ The first effect of political parties is the uncertainty effect. The new information about a member of the party (the president) decreases the variance of the perceived distribution of abilities of partisan politicians. This decrease of the variance makes the election prospects of the politician more sensitive to his effort, i.e. effort is more effective to move up marginally from the average quality. Hence, the governor has stronger incentives to exert effort. The second effect is the reputation effect due to the difference between the prior mean of the distribution and the actual performance of the member of the party (the president). A president who performs better than expected send a positive signal to the public about the ability of his comrades. In the same way, a politician who performs worse than expected send a negative signal. Any signal (positive or negative) implies that the effort of the governor who belongs to the same party than the president has a lower marginal effect on his election prospects. If the signal is non-zero, the effort exerted by the politician is less effective than in a situation of independency. If the president reveals better than the average, the candidate needs less effort to prove better than the average. If the president reveals worse than the average, the governor is expected to be worse than the average and his effort has a lower effect on the probability of being elected president. In both cases the reputation effect, either positive or negative, decreases the effort exerted by the governor. After a negative signal revealed by the president, a governor member of that party may want to leave the party and run as independent. However, quitting is a dominant strategy for any politician belonging to that party when it is too late because the signal has already been sent. 6
Proposition 1. Political parties increase incentives of their members compared to independent politicians if and only if β < δ σ ln 1 δ (5) Proof. Governors that belong to the same political party than the president would exert more effort than an independent one if e P > e I c ( e P) > c ( e I) { } B β πδ σ exp δ σ > B πσ which, after some manipulation leads to the result. On one hand, this result implies that political parties can only be effective if the reputation effect is low. The stronger the reputation signal the president send about his party, the bigger the reputation effect and the decrease in incentives for his comrades. Uncertainty, on the other hand, has a limited effect. Even if its effect is always positive, it is limited. If the signal of parties is weak (high δ) society wins by increasing the power of the signal (lowering δ). However, if the signal of parties is strong, society prefers to decrease the power of the signal. In fact, the effect reaches a maximum at δ = e 1/ 0.6065. In Figure 1, the marginal cost of effort of politicians is determined when the Gauss curve crosses the η 0 line. In all the panels, the black solid line represents the prior distribution of and independent politician and the blue dashed line, the distribution of a politician who belong to the party of the president. The reputation effect moves the prior distribution to the right (if positive) and decreases the marginal effect of effort in the probability of being recognized more able than the benchmark η 0 and decreases effort made by politicians as shown in Panel (a) of the figure. Moreover, the uncertainty effect due to the decrease in the variance of the prior distribution increases effort made by politicians as in Panel (b) of the Figure 1. Finally, Panels (c) and (d) show two examples of a situation in which both effects are present and one dominates over the other: in (c) the reputation effect dominates and decreases effort, and in (d) the uncertainty effect dominates and increases effort.. Accountability Effect Governors of different regions, ministers, senators or representatives have different tasks and are evaluated differently by citizens. Hence, the career path of Proof by maximization of the RHS of condition (5). 7
Figure 1: Reputation and Uncertainty efforts. a politician is different if he occupies office A or B, not because of his ability but because the nature of the office itself. In fact, different offices have different impact on the utility of citizens. Bureaucracy, media coverage or simply different nature of the public good provided by each office make the evaluation of politicians ability different from one office to the other. Assume two different public goods y A and y B that are provided from efforts e A and e B and the ability of the politician η. y k = η +e k +ε k (6) where ε k N(0,σk ), k = A,B, is a random shock specific to each public good production that is realized in stage 3. Notice that σa < σ B means that the politician is more visible in office A than in office B. The accountability effect decreases the variance of the perceived ability of politicians. In fact, perceived ability becomes η k = y k e k ε k (7) where η k is the perceived ability of the politician in office k = A,B. Hence, the probability that politician i is elected president is with k l. p k = Pr[ η k η l ] (8) 8
Following the methodology of section, define G k ( ) as the cumulative distribution function of η +ε k and g k ( ) as its density function. Therefore, c (e k) = g k (η 0 ) B (9) To illustrate the accountability effect, take the effort exerted by an independent politician in each office. Independency of the politician means that the variance of his perceived ability in office k is σ + σk. Therefore, efforts are defined by c (e k) = B π(σ +σ k ) (10) Without lose of generality, assume that office A is more accountable than office B, i.e. σa < σ B. This make the perceived ability of the politician to be more precise in office A than in office B. This increased accountability that comes from the administrative nature of the office occupied by the politician increases his effort, e A > e B..3 Summary Summarizing the effects, political career concerns are affected by three effects: uncertainty, reputation and accountability. If a politician belongs to a political party, his performance give better signals about his quality than an independent politician. This decrease in uncertainty increases his effort. However, if previous information about his comrades is far from the mean, the reputation effect makes him exert less effort, either because he can rest on party s oars or because he throws in the towel. Finally, the quality of the information revealed by a particular office about the ability of the politician that occupies it also affects his effort. In fact, a more visible office pushes politicians to exert higher effort, which is given by { c B (e ik ) = π(δ i σ +σk ) exp βi } (δi σ +σk ) (11) where β i = {0,β}, δ i = {δ,1} define if politician i belongs to the party of the president or not and σk measures the (inverse) accountability of office k. 3 Active Political Competition Politicians often compete against colleagues from other offices instead of only outsiders when trying to become president. In this section, citizens choose between two active politicians, i.e. the two candidates occupy public offices at the moment of elections and there is no outsider. 9
As in the previous section, people infer abilities of both politicians in stage 4 and vote for the most able agent. Instead of proving better than a random outsider, politicians have to prove better than each other to be elected president. However, both the ability of politician i and j are unknown. Therefore, the probability that politician i is the best alternative to become president is p i = Pr[ η i η j ] = Pr[ η i η j 0] (1) Ifweassumethatthepriordistributionsofabilitiesarenormalasinsection andthatpoliticiansoccupyofficesaandb,thedifference η i η j followsanormal distribution G( ) with mean β i β j and variance V ( δ i +δ j ρ) σ +σ A + σ A where β i and δ i account for the fact that politicians belong or not to the party of the president and ρ accounts for the membership of both politicians to the same party, which would mean that their abilities are somehow correlated. Using equation (1, and solving the maximization problem of the governors, the effort exerted by politicians is defined by c (e i) = g(0) B (13) where g( ) is the density function of the difference of abilities, η i η j and β i = {0,β}, δ i = {δ,1} and ρ = {0,ρ} depending on the political affiliation of politicians. Finally, equation (13) can be written as { } c (e B i) = exp (β i β j ) πv V (14) There are two differences respect to the benchmark case. First, the increased uncertainty of the outcome of elections due to the fact that there is no fixed outside option decrease the effort made by politicians. In this sense, weaker political competition decreases the effort made by politicians. Second, if both politicians belong to the same political party, the uncertainty effect remains while the reputation effect is cancelled out by the reputation of the competitor. Using a similar model, it is possible to include competition against both active and outsider politicians, however the results would be similar. 4 Decentralization Decentralization has been studied using personal economics emphasizing the trade off between spillovers and accountability. Seabright (1996) underlines that centralization allows benefits from policy coordination but has costs in terms of diminished accountability. Moreover, Bolton & Roland highlight that 10
decentralization has a bureaucratic cost because it duplicates offices while it allows a better fit of regional policies. However, partisan politics seem not to be important in this context. In fact, only O Neill (005) uses the influence of political parties to explain decentralizing reforms. She claims that if a party foresee that it would loose next elections, it would set up reforms to keep some power after loosing presidency. Centralization implies that the president and the governor belong to the same party because regional officers are appointed by the central power. Hence uncertainty and reputation effects are implied in the effort of governors by definition. Moreover, decentralization implies increased accountability of the governors. In fact, decentralization allows for a divided government between the president and the governor: if the president belongs to party A it does not imply that the governor also does, which is the case in a centralized system. The previous section shows that governor would provide more effort under decentralization because of the accountability effect. However, the possibility of divided government in this system may vanish both the reputation and uncertainty effects. Adding up all the three effects, decentralization produces more effort from governors because of the accountability effect and the reputation effect, while it produces less effort from governors because it reduces uncertainty. Define e C and e D as the efforts provided by a governor under centralized and under decentralized systems respectively. For simplicity, assume there is no reputation effect (β = 0) and that the probability of a divided government, i.e. that the president and the governor are elected from the same party, occurs with probability q under decentralization. Therefore, the expected effort under each systems are The difference can be written as E[e C ] = e P C (15) E[e D ] = q e P D +(1 q) e I D (16) E[e C e D ] = q (e P C e P D) +(1 q) ( e P C e I ) D (17) The first term on the RHS is positive and it is the usual accountability effect. However, the second term may be positive if the uncertainty effect is high enough. Moreover, if there is a high probability of a divided government in the sense of having the president and the governor belonging to different political parties, a centralized system may be better than a decentralized one because the decreased uncertainty political parties give to voters. Indeed, the second term of the RHS is (under no reputation effect) ( e P C e I D = B 1 π δ σ +σc ) 1 σ +σd (18) 11
Therefore, if there is a high probability of divided governments and political parties are very informative about ability of politicians (such that ( 1 δ ) σ > σ C σ D )3, thedecreaseinuncertaintymayrevertthebenefitsofdecentralization without spillovers or bureaucratic costs associated to such a reform. 5 Conclusions Political parties play a key role as informative signal for citizens to elect the best politician in office. This information has different consequences depending on the system and political configuration of countries. First, the ability of a politician who belongs to a known political party is more precise than the ability of an independent politician. In other words, a politician how belongs to the political party in power faces both internal and external competition and, to get a political promotion, he needs to improve his evaluation about performance by increasing his effort more than an independent politician. Second, politicians inherit the reputation of their party. Actually, if the party send some signal about the ability of its members, either positive or negative, it is more difficult to the politician to differentiate. His effort becomes less effective in signalling ability and incentives are lowered. In summary, political parties decrease uncertainty and enhance reputation effects. The first increases incentives and the second, lowers them. Although the uncertainty effect is limited, if this trade-off is added to an institutional framework, as decentralization reforms, a new cost of decentralization arises: it increases political uncertainty and may lower incentives of politicians. In fact, if political parties are informative about politician s abilities, centralization is better in signalling and selecting the best politician in the president s office. References [1] Alesina, Alberto & Stephen E. Spear, 1988, An overlapping generations model of electoral competition. Journal of Public Economics, vol. 37, p. 359-379. [] Aragón, Fernando, 008, Choosing a candidate: party primaries and quality of politicians. Mimeo. [3] Ashworth, Scott & Ethan Bueno de Mesquita, 007, Valence Competition and Platform Divergence. Mimeo. [4] Caillaud, Bernard & Jean Tirole, 1999, Party governance and ideological bias. European Economics Review, 1999, vol. 43. [5] Caillaud, Bernard & Jean Tirole, 00, Parties as Political Intermediaries. The Quarterly Journal of Economics, vol. 117, No. 4, pp. 1453-1489. 3 Notice that, by definition σ C > σ D 1
[6] Castanheira, Micael; Benoit S Y Crutzen & Nicolas Sahuguet, 006, Intraparty Politics and Electoral Competition. Erasmus University Rotterdam, Mimeo. [7] Crutzen, Benoit, 004, Intra-party Discipline, Rent Extraction and Electoral Rules. Mimeo. [8] DeGroot, Morris, 1970, Optimal Statistical Decisions New York, Mc- Graw Hill. [9] Enikopolov, Ruben & Ekaterina Zhuravskaya, 007, Decentralization and political institutions Journal of Public Economics, Vol. 91, pp.61-90. [10] Geys, Benny & Jar Vermeir, 008, Party cues and yardstick voting European Journal of Political Economy, vol. 4, p. 470-477. [11] Harrington, Joseph, 199, The role of party reputation in the formation of policy. Journal of Public Economics, 199, vol 49. [1] Hazan, Reuven, 003, Introduction. The Journal of Legislative Studies, vol. 9. No. 4, pp. 1-11. [13] Jones, Phillip & John Hudson, 1998, The Role of Political Parties: An Analysis Based on Transaction Costs. Public Choice, vol. 94. No. 1, pp. 175-189. [14] Lazear, Edward & Sherwin Rosen, 1981, Rank-Order Tournaments as Optimum Labor Contracts Journal of Political Economics, vol 89, no. 5. [15] Padró-i-Miquel, Gerard & Erik Snowberg, 009, Accountability with Strong Parties. London School of Economics and Political Science, Mimeo. [16] Parés, Claudio, 008, Is it reasonable to allocate power to appointed Regional Authorities? Toulouse School of Economics, Mimeo. [17] Persson, Torsten & Guido Tabellini, 000, Political Economics, Explaining Economic Policy Cambridge, MA, The MIT Press. [18] Poutvaara, Panu & Tuomas Takalo, 007, Candidate Quality. International Tax and Public Finance, vol. 14, pp 7-7. [19] Seabright, Paul, 1996, Accountability and decentralization in government: An incomplete contracts model European Economic Review, Vol. 40, Iss. 1 (Jan), pp. 61-89. [0] Snyder, James & Michael Ting, 00, An informational Rationale for Political Parties. American Journal of Political Science, vol. 46, no.1 pp. 90-110. 13
[1] Ström, Kaare, 1990, A Behavioral Theory of Competitive Political Parties. American Journal of Political Science, Vol. 34, No. (May), pp. 565-598. [] Tadelis, Steven, 00, The Market for Reputations as an Incentive Mechanism. Journal of Political Economy, 00, vol. 110, no. 4. [3] Tirole, Jean, 1996, A Theory of Collective Reputations (with applications to the persistence of corruption and to firm quality). Review of Economic Studies, vol. 63, pp. 1-. 14