How Does Electoral Competition Affect Bureaucratic Performance

Bureaucrats and Politicians: How Does Electoral Competition Affect Bureaucratic Performance? Abstract This paper examines two competing mechanisms by which electoral competition affects bureaucrats’ performance: re-election concerns and dynamic incentives. Incumbent politicians in competitive constituencies have higher re-election concerns. This gives them a greater incentive to monitor their bureaucrats, resulting in faster project execution. On the other hand, politicians in low competition constituencies have a high probability of being re-elected and therefore typically have longer tenures. This gives them the ability to provide better incentives to bureaucrats using long term contracts. To isolate these mechanisms, I construct a unique dataset from India by matching details of bureaucrats’ work histories with individual local public good projects under the Member of Parliament Local Area Development scheme. The main results show that bureaucrats approve projects faster when the incumbent politician’s probability of winning is higher. Moreover, bureaucrats perform better when they are up for promotion and when they know that the politician is likely to be in office at the time of promotion. These findings suggest that access to dynamic incentives is the dominant mechanism through which political competition affects bureaucrats’ performance. 1 Introduction Political competition is said to yield benefits to the citizens just as competition in economic markets yields benefits to consumers. 1 There is a large theoretical literature and an increasing number of empirical studies which show that lack of political competition may lead to adverse policy outcomes compared to competitive constituencies.2 The underlying mechanism that drives these results is that due to re-election concerns, the voters gain influence in disciplining the politicians only when political competition is high. While these studies mostly examine policy choices made by politicians, the implementation of these policies by bureaucrats also matters for governance.3 In the above context, the question this paper addresses is: how does political competition affect the bureaucrats’ performance? If re-election concerns are also an important determinant of project implementation, then we should find that bureaucrats perform worse in less competitive constituencies.4 Contrary to this prediction, using data from India, this study finds that bureaucrats perform better when the incumbent politician’s probability of winning is higher. In this paper, we introduce another important mechanism that rationalizes these results. This is related to the ability of the politicians to provide incentives to the bureaucrats. In low competition constituencies, the incumbents typically have a high probability of returning to office in the next term (assuming no term limits). This provides them with an additional way of incentivizing bureaucrats that their counterparts in competitive constituencies do not have: promising future rewards. The use of dynamic incentives enables the incumbents in low competition areas to implement higher effort levels and hence we get a negative relationship between electoral competition and bureaucrat’s performance. 1 See Bardhan and Yang (2004) and Bardhan and Mookherjee (2010) for a discussion on this. Besley et. al (2010) show that lack of political competition may lead to policies that hinder economic growth. Nath (2014a) shows that in absence of political competition, local elites exert disproportionate influence on the allocation of spending on local public goods. Brown and Hunter (1999); Lake and Baum (2001) and Hecock (2006) show that competition increases the level of spending. 3 The bureaucrats we have in mind are career civil servants who enter the bureaucracy through a meritorious entrance exam. Since they are non-elected government officials, the only way political competition can affect their performance is through the influence exerted by politicians. 4 The effect of re-election concerns on bureaucrats’ performance has been studied by Roggers (2014). Using data from Nigeria he shows that politicians in high competition constituencies are more likely to delegate public good projects to autonomous bureaucrats rather than governmental agencies. The autonomous agencies, on an average, perform better and hence he finds a positive relationship between competition and bureaucratic performance. 2 2 Our measure of bureaucratic performance is the time taken to approve local public goods projects. The details of projects are obtained from Members of Parliament Local Area Development (MPLAD) scheme. Each member of parliament (MP) gets a fixed sum of money to spend on infrastructure projects within his constituency. The politician has full control over the type of project, the cost and the location. These projects, however, have to be approved by the bureaucrats in the administrative district where they are to be constructed. We use work histories of bureaucrats to match them with individual projects under the MPLAD scheme. This allows direct identification of the bureaucrats who approved the projects and how long they took to approve them. To identify the causal effect of political competition on bureaucratic performance, we take advantage of a previously unexplored information shock that resulted in an effective term limit for some incumbents. The shock we exploit is the announcement of changes in reservation status of some of the constituencies as result of re-districting of all electoral boundaries in India.5 When these boundaries were redefined, the population shares of minority groups changed as well, resulting in changes in the reservation status accordingly. 6 The announcement of the changes was made in December 2007 and they were to be enforced in the 2009 elections. The incumbents who were affected by this change knew that their probability of winning in 2009 was zero. The politicians in the control group were not affected by the news and therefore their probability of winning does not change. We compare the bureaucratic performance in the two groups before and after this shock. The differences-in-difference (DID) strategy gives us a causal effect of change in political competition on the bureaucratic performance. Our results show that as the probability of winning goes to zero due to the information shock, the average sanctioning time increases by 13%. Now, probability of winning going to zero means that the incumbent’s re-electoral motives become weaker. Moreover, since the bureaucrats know that the politician is not going to get re-elected, any promise of future rewards made by the politicians are not credible. Since both the reelection concerns and access to dynamic contracts shut down with the exogenous shock, this result by itself does not help us isolate the two channels. In order to identify which of the mechanisms drives the effect of electoral competition on bureaucratic performance, we compare the sanctioning times of bureaucrats in constituencies 5 Reservation of a constituency for a minority group means that only those candidates who belong to the minority group can contest elections. 6 The electoral boundaries did not change throughout the period of the study (1999-2009). The announcement was made in the middle of the 2004-2009 legislative term. The new boundaries came into force in the 2009 elections, at the end of the period. 3 that are a party stronghold with those that are not. The variable stronghold is a dummy which takes value 1 if the same party won all four elections prior to the period of study. This is our measure of competition - incumbents in strongholds have a high probability of winning again and hence face very little electoral pressure. The results of this empirical model tell us what happens when we move from probability of winning close to half (non-strongholds) towards probability of winning close to one (strongholds). Our results show that in constituencies with party strongholds, projects are sanctioned 11% faster. This means that bureaucrats perform better in constituencies where probability of winning for politicians is higher. Taken together with the results from the natural experiment, our findings are consistent with the dynamic contracts mechanism: the politicians in low competition constituencies have longer tenures and therefore have access to dynamic contracts that provide better incentives to bureaucrats. This in turn improves bureaucratic performance. However, if probability of winning goes to zero, the promise of future rewards are no longer credible and hence bureaucrats take longer to approve projects. One of the empirical concerns we need to address is that of selection: politicians in strongholds may be able to get better performing bureaucrats and that may drive the results rather than incentives/ability of politicians to monitor the bureaucrats. In order to control for selection, we take advantage of the fact that the administrative and electoral boundaries do not perfectly overlap in India. A single administrative district may have two or three electoral constituencies that overlap with it. Since the bureaucrat sits in an administrative district and the politician is the elected representative of the electoral constituencies, we have situations where one bureaucrat may deal with two or even three politicians. Comparing the performance of a single bureaucrat with multiple politicians helps us deal with the selection problem. It may be argued that better performance of bureaucrats in strongholds may have nothing to do with incentives, but instead be due to rent-seeking motives. Politicians and bureaucrats may collude to push certain types of projects faster in order to gain rents. In order to alleviate these concerns, we provide two pieces of evidence. Firstly, we show that there is no systematic bias in types of projects that are sanctioned faster. All types of projects - roads, water, health, and irrigation - have significantly less delays in strongholds. Secondly, we explicitly look at the tenures of the bureaucrats to construct measures of how far they are from promotions. We find that bureaucrats perform better in the year that they are up for promotion. Moreover, they perform worse when they know that the politician is not likely to be in office at the time of promotion. These results suggest that bureaucrats are indeed responding to incentives. The main contributions of this paper can be summarized as follows: firstly, we examine 4 a hitherto unexplored mechanism that connects electoral competition to bureaucratic performance: access to dynamic contracts. Secondly, we develop an empirical methodology to identify the causal effect of electoral competition on bureaucratic performance and to isolate the mechanism that drives the relationship. Finally, we construct a rich and disaggregated dataset that allows us to [a] directly observe the behavior of bureaucrats and [b] control for selection and unobserved abilities that typically pose a threat to identification of causal mechanisms. The rest of the paper is organized as follows. Section 2 provides details of the institutional background: the bureaucrats and the politicians as well as the way in which they interact in our setting. In section 3, we outline how the alternative mechanisms work through the lens of a canonical efficiency wage model that is adapted to match our setting. We provide details of the data and our empirical identification strategy in sections 4 and 5 respectively. While section 6 provides empirical results, section 7 concludes. 2 Institutional Background 2.1 The Bureaucrats The bureaucrats I study in this paper belong to the Indian Administrative Service (IAS). These IAS officers are federal government employees recruited through a nationwide competitive examination conducted by the independent Union Public Service Commission. A few IAS officers within each state may be recruited through the State Civil Services. The latter ones are posted only within the state and typically do not hold positions in the ministries or departments in New Delhi. 7 Once the direct recruits are chosen, they all undergo training together. Thereafter, they are assigned to one of the states in a quasi-random manner and this assigned state is known as their Cadre. They train under their assigned superiors for about four to five years in different districts. They are subsequently assigned to a district as the head administrator. This post is known as collector or district magistrate 8 . In terms of their career path, the bureaucrats can serve as collectors roughly between years 5-16 of their careers. In their 9th year, they are promoted to the junior administrative grade. There is no screening: everyone is promoted. 9 This is basically a salary hike. It is in the 13th and the 16th years that the promotions are 7 The officers that enter the IAS through the UPSC exam are referred to as ‘Direct Recruits’ and the ones that come from the State Civil Services are called ‘Promotees’. 8 The Promotees typically become IAS officers pretty late in their careers. 9 See http://persmin.gov.in/DOPT/EmployeesCorner/Acts_Rules/IASPromotionGuideLines.pdf 5 based on screening of the bureaucrats. As collectors, the bureaucrats are responsible for law and order, collection of land revenue and various taxes, land acquisition and land assessment, crisis administration, and serve as the development officer. The main role of the collector relevant for this study is that he is Ex-officio Chairman of District Rural Development Authority Agency which carries out the various developmental activities. Any development project that has to be executed in the district must be approved by the collector. The IAS officers are civil servants and, as per the directives of the Constitution of India, they cannot be hired or fired by the politicians. The bureaucrats are assigned to various posts in each state by the corresponding Department of Personnel and Training. The executive order of each assignment is signed by the top bureaucrat of each state known as the Chief Secretary. There is some evidence, however, that politicians may influence the assignments of the bureaucrats. Iyer and Mani (2012) show that when the leader of the party in power in a state changes, the probability of reassignments of bureaucrats goes up. Hence, politicians may use the threat of reassigning the bureaucrats to different posts as a control mechanism. Moreover, the politicians can potentially influence bureaucrats’ job assignments once they are up for the promotions described above. Later, this paper explores how political competition affects the ability of the politicians to use these promotions as an incentive mechanism. 2.2 The Politicians India has a federal parliamentary system of democracy. The parliament is the supreme legislative body. There are two houses in the parliament - the lower house is called the Lok Sabha (House of the People) and members of this house are directly elected by the citizens. The upper house is called the Rajya Sabha (Council of States) and the members of this house are elected by the state legislative assemblies. This analysis will focus on the members of parliament (MPs) in the Lok Sabha. There are 543 Lok Sabha constituencies. In accordance with the Constitution, elections are held every five years where candidates are selected through universal suffrage. India has a plurality system where the candidate with the highest vote share wins (also called "first-past-the-post"). There is a multi-party system and candidates are allowed to contest independently as well. There are no term limits for politicians in India. 6 2.3 The MPLAD Scheme The local public goods projects that this paper studies are the ones provided under the MPLAD (Member of Parliament Local Area Development) Scheme in India. Under this scheme, each MP is given a fixed budget of Rs. 20 million (0.5 million USD) per year to spend anywhere within his constituency. The money can only be spent on asset building projects. This means that politicians cannot hire employees, give grants and loans, or purchase inventory or stock with this money. Moreover, the guidelines say that acquisition of land, building assets for individual benefits and building religious structures is not permissible. In short, most of the permissible works are construction-based and are for infrastructure development within the constituency. One of the particularly relevant features of these projects is that they are highly visible. According to the guidelines, the MP who funded the project must visit the work site and unveil a plaque detailing the project when it is complete. The MP’s name is written on the plaque along with how much money was spent and how long it took for the project to be completed. The fact that the public knows these details gives the politicians an incentive to ensure that projects are completed in a timely manner. Since the projects are executed by the bureaucrats, this provides us a very nice set up to see whether political competition affects the performance of bureaucrats in executing these highly visible projects. 2.4 The Bureaucrat-Politician Interaction MPLAD is a unique scheme where we observe the decision of each politician separately and can also observe the performance of bureaucrats. When the money from the fund is allocated to the politicians, they send a recommendation letter to the collector. This recommendation letter details the following: [1] the type of projects the politician wants (roads, drinking water, education, etc) [2] the cost of each project and [3] the location of each project. The total cost of various projects recommended by the MP has to be within the fixed budget of Rs. 20 million. Hence, the politician has complete control of what to choose, where to build, and how much to spend on each project. Once the bureaucrat receives the recommendations for the project, the project goes through a sanctioning process. The collector chooses the implementing agency and sends the proposal to its chief engineer. The engineers inspect the site, prepare a technical and feasibility report, and send this report back to the bureaucrat. The collector reviews the structural and financial report and then approves the project if everything is sound. The time between receiving the recommendation and approving the project will be henceforth referred to as the time taken 7 to sanction the project. The collector has the full authority over the sanctioning of the project and the politician has no say in it 10 . Over the period of this study, the official guidelines suggest that the number of days taken to sanction should not exceed 45 days. Once project is sanctioned, the engineers and lower level bureaucrats execute the project with the collector having full control over the implementation process. The bottomline is that the members of parliament depend completely on the bureaucrat to carry out the projects that will ultimately bear the politician’s name. This provides incentives to the politician to monitor the bureaucrat as his image is at stake. 2.5 Administrative and Electoral Boundaries Electoral boundaries in India are intended to create constituencies such that each politician represents the same number of citizens in the parliament. The Delimitation Commission of India is responsible for drawing the boundaries based on population figures from the census. Before our study period, the Delimitation Committee was set up in 1952, 1963, and 1973. In a 1976 constitutional amendment, the government suspended delimitation until after the 2001 census so that states’ family planning programs would not affect their political representation in the Lok Sabha. The report of the 2001 census came in 2003 and the new boundaries were applicable only in 2009. Hence, during the entire period of this study, 1999-2009, the electoral boundaries remained the same. Figure 1 depicts the electoral boundaries. As we can see, the size of the constituencies are not uniform; some are large and some are small. The smaller constituencies are in areas where the density of population is higher. This is especially true for Uttar Pradesh in the north and West Bengal in the west. In contrast to electoral boundaries, the administrative boundaries are drawn based on land area. This is done so that each district collector is responsible for land revenue, law and order and development works for the same area of land. Unlike electoral boundaries, district 10 The collector has full control of the technical feasibility checks. In some of the districts we visited, we found that collectors required the chief engineers of all the implementing agencies to provide a weekly report of the works in progress. In other districts, this practice was not enforced. If the bureaucrat does not monitor the engineers, the reports are not sent to the collector on time. Bureaucratic delays therefore come from two sources: [a] bureaucrat not monitoring the feasibility checks and [b] upon receiving the report, bureaucrats taking a long time to reviewing and approving it. 8 boundaries do not change over time. There are cases where district boundaries are redrawn and this happens when the states split into two. In 1998, three of the largest states in India split and due to this districts also split into smaller sizes. However, this happened before our period of study and does not affect our analysis. Since the electoral boundaries and administrative boundaries are drawn according to different dimensions, these boundaries do not perfectly coincide. One electoral constituency may overlap between two administrative districts and vice versa. Figure 2 shows these overlaps across India. If we zoom in, the overlaps of boundaries are clearer, as shown in figure 3. The highlighted portion illustrates a possible district-constituency overlap situation where one bureaucrat works with three politicians at any given point in time. We exploit this feature to control for selection and unobserved ability of bureaucrats. We provide more details of this once when we discuss the empirical identification strategy. We also exploit the the fact one politician may deal with multiple bureaucrats at a point in time. This is analyzed in section 6.6. 3 Theoretical Framework As described earlier, there are two ways in which electoral competition can affect the politician’s incentives to induce better bureaucratic performance. In order to study how these mechanisms work, we adapt a canonical efficiency wage model with endogenous monitoring to make two modifications: [a] let the wages be fixed exogenously and [b] consider a two period dynamic contract with one-sided commitment where the principle commits to promises of future payoffs. The model is useful for two main reasons: [1] it furthers our understanding of how the potential mechanisms work and [2] the comparative statics derived in the model provide us with testable predictions which we can use to identify which mechanism is at work in the data. 3.1 The Environment Consider a politician (‘he’) who has a public good project in each period t that has to be executed by a bureaucrat (‘she’). The bureaucrat has to exert an effort to implement the project. Let et ∈ [0, 1] denote the effort level chosen by the bureaucrat in period t. The output takes values πt ∈ {0, 1}. Putting in higher effort increases the probability of successfully implementing the project such that P rob(πt = 1) = et . Putting in effort is costly and to obtain an interior solution, the cost is assumed to be convex: c0 (e) > 0 and c00 (e) > 0. For simplicity, we assume that c(et ) = e2t . 9 There are two features of the institutional setup that limit politician’s ability to directly control the bureaucrat. Firstly, the wages of the bureaucrats are fixed and determined exogenously. Politicians cannot affect these wages. Secondly, politicians cannot hire or fire the bureaucrats. 11 In lieu of these direct methods of control, politicians have indirect ways in which they can control the bureaucrat. The politicians can give a letter of recognition in each period if the bureaucrat puts in effort. Let’s assume that this gives a utility r to the bureaucrats. Secondly, and more importantly, the politicians can influence the re-assignment of bureaucrats to other administrative districts (Iyer and Mani (2012)). If bureaucrats perform well, politicians can reward them by providing them more lucrative assignments in the future. Let us now see how these two instrument of control are used by politicians to provide incentives to the bureaucrat for putting in higher effort. At t = 0, the politician announces that the bureaucrats will be monitored with probability qt ∈ [0, 1] in period t ∈ [1, 2]. If the bureaucrat is monitored and is caught shirking, she does not get a letter of recognition in the current period. Moreover, the politician can further punish her in the future by assigning her to an undesirable post (hence giving a low continuation value V 0 ). The principal would like to make V 0 → −∞ since this threat would be enough to induce the bureaucrat to work and it would mean that he will not have to monitor the bureaucrat. However, such a threat is not credible and hence we assume that V 0 ≥ 0. Also, monitoring is costly. Let cost of monitoring bureaucrat be C(qt ) = qt2 such that Cq > 0, Cqq > 0. The second way to incentivize the bureaucrat is to promise a higher future payoff in the form of more lucrative assignments. This is captured via the future continuation value V which the bureaucrat discounts at rate β(φ) ∈ [0, 1] where φ is the extent of electoral competition in the constituency politician represents. The discount rate depends on the level of competition because if the electoral constituency is competitive, then the probability of winning in the next election is low. In such an event, promises of future rewards are less credible and the discount rate is close to zero. Hence, β 0 (φ) < 0. Effect of electoral competition on optimal effort through beta will henceforth be referred to as "access to dynamic contracts" mechanism. The politician offers a contract (q1 , q2 , e1 , e2 ) to the bureaucrat and the latter chooses e0 to maximize his own payoffs. The bureaucrat’s expected payoff in period zero is therefore given by: 11 This setting is not unique to India. It exists in many other countries, especially the commonwealth countries where there exists some form of civil service. 10 U B = 2 X β t−1 w + (1 − qt )r + qt 1{et = et }r − et2 0 h 0 i t=1 Let R(e, φ) denote the present value of gross payoffs the politician gets when the project is implemented. If the project is not implemented, he gets a zero payoff. Note that the payoff function is concave in effort: Re (e, φ) > 0 and Ree (e, φ) < 0. This ensures a unique solution. The electoral motives are captured in this gross payoff function: when Reφ (e, φ) > 0, it means that the marginal increase payoff when the project is implemented increases as the extent of competition increases. This is exactly what the reelection concern motive would suggest. To √ get a closed form solution, we assume R(e, φ) = R(φ) e. The politician’s expected payoffs are: U P = 3.2 h i t−1 R√e − q 2 . β t t t=1 P2 The Maximization Problem The politician’s problem can be written recursively as: Period 1: maxe1 ,q1 ,V,V 0 √ R(φ) e1 − q12 + β(φ)Q(V ) (1) subject to: IC1 : h i w + r + β(φ)V − e21 ≥ w + (1 − q1 ) r + β(φ)V + q1 β(φ)V 0 (2) w + r + β(φ)V − e21 ≥ 0 (3) PC1 : NN: V ≥ 0; V0 ≥0 (4) Period 2: Q(V ) = maxe2 ,q2 √ R(φ) e2 − q22 (5) subject to: IC2 : w + r − e22 ≥ w + (1 − q2 )r 11 (6) PC2 : w + r − e22 ≥ V (7) ICt represents the incentive compatibility constraint in period t and says that the payoff the bureaucrat gets from putting in effort e0t = et is at least as much as the payoff from shirking and putting in zero effort. The participation constraint (P Ct ) ensures that, in each period, the payoff from putting in effort is at least as much as what the bureaucrat gets when he does not work on executing the politician’s project. In the first best scenario, the effort is observable. There is zero monitoring in both periods and the effort level chosen by the politician is a solution to the above problem without the IC constraints and without q. In this scenario, the participation constraint binds at optimum in each period. Consider the second period: if P C2 is not binding at optimum, then the effort level can be increased without violating the constraint. This creates a contradiction because it increases the politician’s payoffs. Hence, the participation constraint binds in period two and √ we get eF2 B = w + r − V F B . Consider now the first period. If the participation constraint is not binding, the politician can now do two things. He can increase e1 and increase his payoffs in period 1, or reduce V and increase his payoffs in period 2. Hence, at optimum, the participation constraint of period one should also bind. Solving the first period problem gives p us: eF1 B = w + r + βV F B . How do the effort levels compare to the case where the politician offers a static contract each period? In the static contract case, the politician implements the same level of effort: √ es2 = w + r = es1 . Note that eF1 B > es1 and eF2 B < es2 . This shows that by having access to dynamic contracts, the politician can use the promise of a higher future payoff in order to improve performance in the first period. By keeping his promise in the second period, he gives the bureaucrat a higher payoff by reducing the effort level the bureaucrat is asked to implement. Let us now turn to the second-best problem defined by equations (1)-(6). Similar to the first best scenario, e∗1 > e∗2 . Even in the presence of moral hazard, the politician can front-load the incentives to induce the bureaucrat to work harder in the first period of the contract. To curb deviations from the contracted effort level, the politician engages in costly monitoring. The solution to the second-best problem is collected in the following proposition: 12