Organized Interests, Legislators, and Bureaucratic Structure Stuart V. Jordan and Stéphane Lavertu Preliminary, Incomplete, Possibly not even Spellchecked. Please don t cite or circulate. Abstract Most studies of statutory constraints on bureaucratic structure and process focus on legislators specifically on how legislators use structure and process to insure that policy outcomes conform to their interests. This paper develops a quite different (although not incompatible) approach. We treat structural choice as primarily driven by competition between organized interest groups. We build a formal model in which two competing interests first lobby legislators over the terms of legislation, and then compete to influence bureaucratic choice under conditions determined by that legislation. The model suggests that organized interests induced preferences over bureaucratic structure can be at odds with those of legislators with interest groups preferring legislation that imposes tight structural constraints, while legislators prefer legislation that allows for greater discretion.
Introduction Conflicts over the content of legislation do not end when the legislation at issue passes. Legislation must be implemented, and implementation inevitably requires choices that are themselves substantively important and malleable through lobbying, litigation, and legislative oversight. Even the most specific legislation for instance, a revenue bill that sets a precise tax rate on a list of specifically-named assets leaves high-stakes choices undetermined: Which returns will revenue officials choose to audit? How many staff will be allocated to auditing returns that may be liable to this particular tax? On what terms will the government settle cases of delinquent payment? How aggressively will the government prosecute alleged fraud? Because of this, a number of authors e.g. Terry Moe, McCubbins, Noll & Weingast, McCubbins & Schwartz have developed a view of legislation that emphasizes the significance to politics and policy of statutory constraints on bureaucratic structure and process. Constraints on the both the substance and process of executive decisions, they point out, can "stack the deck" in favor of one interest relative to others. Moreover, legislators can use structural design to economize on the costs of overseeing implementation decisions relevant to their policy or electoral interests. Thus, in this view, struggles among interest groups and lawmakers over legislation are just as much about statutory restrictions on bureaucratic structure and process as they are about substantive restrictions on policy choices. With a handful of important exceptions we discuss below (McNollGast commitment story, Moe, DeFig), studies in this vein have focused on explaining structural choices in terms of legislators interests in policy outcomes. For the most part, these explanations point out that legislators want bureaucrats to adopt policies that achieve particular outcomes, but are limited by a relative lack of expertise, information, and time. Structural and procedural restrictions are one among a range of tools (including statutory restrictions on the substance of policy and ex-post oversight) legislators use to solve this problem. This paper develops a quite different (although not incompatible) approach. We treat structural choice as primarily driven by competition between organized interest groups. Interest groups are an essential part of the mechanisms through which procedural provisions shape the substance of bureaucrats choices. Think, for instance, of provisions that single out certain bureaucratic deci- 2
sions as subject to de-novo judicial review. To the extent that such provisions alter bureaucrats incentives, they do so by giving organized interests an additional tool (i.e. litigation) they can use to control bureaucrats. The actual substantive effects of such provisions, then, depend largely on which interest groups use that tool and how they use it. Organized interests preferences over procedure, then, are likely to be intense, and, possibly, at odds with those of legislators. Our specific goal in this paper is to explore the logic of these induced preferences. We build a formal model in which two competing interests first lobby legislators over the terms of legislation, and then compete to influence bureaucratic choice under conditions determined by that legislation. This provides a depiction of interests groups structural preferences and the realization of those preferences in structural choice that is tightly integrated with the role they play in the operation of structural constraints. We use the model specifically to generate intuitions about how legislators structural preferences might differ from those of organized interests. In the model, organized interest find lobbying whether of legislators or bureaucrats costly. All else equal they prefer to avoid making campaign contributions, spending time and effort to cajole policymakers, and spending money to hire lawyers and professional lobbyists. Thus they prefer bureaucratic structures that enable them to get what they want with a minimum of lobbying effort. Legislators, on the hand, like to be lobbied. Lobbying brings with it material perks (catered lunches, golf trips, rides in jets), campaign contributions, and the psychic rewards of being treated as someone with power and authority. Thus when contemplate structural alternatives, they consider the streams of lobbying rents they will accrue under alternative structural regimes, and all else equal prefer structures that direct as much lobbying as possible to the legislature. The major substantive implication of these premises regards induced preferences over what McCubbins Noll and Weingast call ex-ante vs. ex-post means of legislative control of the bureaucracy. We show that statutes that stack the deck ex-ante in favor of one interest relative to another substantially curtail the stakes lobbies have in legislators use of their ex-post oversight powers. This is true even when those statues impose no formal limitations legislative oversight. On the other hand, an open-ended statute i.e. one that does not build an automatic bias into the bureaucratic process has the effect of intensifying each lobby s effort to get legislators to intervene in the bureaucracy on its behalf. Legislators, then, obtain more lobbying rents by when 3
they refrain from the imposition of ex-ante controls. Lobbies, on the other hand, do better under statutes with strong procedural controls, because such controls reduce the intensity of competition over bureaucratic choice, and thus dissipation of rents through lobbying. The next section of the paper discusses related literature. We then describe the model and describe its equilibrium. We conclude by interpreting the model s implications for induced preferences over bureaucratic structure. 4
2 Related Literature Coming soon... 5
Statutory Contest Oversight Contest Implementation Contest. Lobbies choose statutory efforts (e s,, e s,2 ). 2. Legislator chooses statutory bias s {,, 2}. 3. Lobbies choose oversight efforts (e v,, e v,2 ). 4. Legislator chooses oversight bias v {,, 2}. 5. Lobbies choose implementation efforts (e m,, e m,2 ). 6. Bureaucrat chooses policy x {, 2}. Display : Moves and Timing 3 Model The political environment we try to depict is complex: We wish to capture lobbying contests over legislation in which all participants the lobbyists and the legislators are motivated largely by their expectations of how future lobbying contests will unfold under alternative statutes. Further, we wish to capture two distinct means through which legislators influence bureaucratic choice, and thus through which they earn lobbying rents writing statutes, on the one hand, and using ex-post oversight to influence choices authorized by existing statutes, on the other. Thus, in fact, there are three separate venues in which lobbying can take place in this environment: lobbies can lobby bureaucrats directly (say, by participating in rulemaking or litigation), they can lobby legislators to apply informal pressure to bureaucrats (for instance, to make budgetary threats), or they can lobby legislators over the language of authorizing statutes. We capture this structure through an idealized depiction in which lobbying over bureaucratic structure and bureaucrats policy choices takes place in three sequential contests between two lobbies lobby and lobby 2 over the decisions of a single legislator and a single bureaucrat. (See Display.) First there is a statutory contest in which the lobbies compete to influence the legislator s choice of statutorily-specified bureaucratic structure. Second, there is an oversight contest in which the lobbies compete to influence the legislator s use of his powers of ex-post control of bureaucratic choice. Finally, there is an implementation contest in which the lobbies compete to influence directly the bureaucrat s choice of actual policy. The three contests in the model jointly combine to determine a single policy decision x {, 2}. This decision is a simple distributional choice between the two lobbies effectively the allocation of a single indivisible good between the two where x = indicates that the good goes to lobby, 6
Lobby Lobby 2 Legislator Bureaucrat { es, if s = c s {s q} + e s,2 if s = 2 {x = } (e s, + e v, + e m, ) {x = 2} (e s,2 + e v,2 + e m,2 ) } { ev, if v = c v {v } + e v,2 if v = 2 e m, σ {s = 2} ν {s = 2} if x = e m,2 σ {s = } ν {s = } if x = 2 } Display 2: Payoffs e s,i 0 e v,i 0 e m,i 0 s {,, 2} q {,, 2} c s > 0 σ > 0 v {,, 2} c v > 0 ν > 0 Lobby i s effort in the statutory stage Lobby i s effort in the oversight stage Lobby i s effort in the implementation stage The direction of statutory bias chosen at the statutory stage The status quo direction of statutory bias The cost to the legislator of changing the statutory status quo i.e. setting s q The magnitude of bias generated by statutory deck stacking The direction of pressure applied to the bureaucrat through oversight. The legislator s cost of applying oversight pressure. The magnitude of the bias generated by oversight activities. Display 3: Notation and x = 2 indicates that the good goes to lobby 2. Each lobby cares only about the effort it expends in lobbying and about the policy ultimately implemented by the bureaucrat. (See Display 2.) Since there are three contests, there are three distinct kinds of lobbying effort: e s,i stands for lobby i s level of effort in the statutory contest, e v,i stands for lobby i s effort in the oversight contest, and e m,i stands for lobby i s effort in the implementation contests. In each contest, each lobby bears the cost of its effort regardless of the outcome of the contests. This means in particular that lobbyists in this model cannot bind themselves to contracts in which they reward or punish a policymaker conditional on the policymaker s decision. Instead, each contest is an all-pay contest. Specifically, lobby s payoff from policy decision x if it expends efforts 7
(e s,, e v,, e m, ) is {x = } (e s, e v, e m, ), and lobby 2 s payoff from policy decision x when it expends efforts (e s,2, e v,2, e m,2 ) is {x = 2} (e s,2 e v,2 e m,2 ), where the indicator function { } equals if the expression in its argument is true and equals 0 otherwise. The remaining features of the model the moves in each contest and the payoffs of the bureaucrat and legislator are easiest to understand by describing each contest separately. We start with the final implementation contest, and move backwards in time to the oversight contest, and statutory contest. 3. The Implementation Contest The implementation contest takes place in two stages. First, the lobbies choose their implementation efforts e m, and e m,2. Second, the bureaucrat chooses the final policy x {, 2}. (See Display.) In each contest in this model, lobbying is best thought of as subsidizing the targeted officeholder s implementation of the choice that the lobby seeks. In the implementation contests specifically, the bureaucrat s payoff from adopting policy i is linearly increasing in lobby i s effort. For instance, one might think of e m,i (lobby i s implementation effort) as effort devoted to the supply of technical information, white papers, regulatory language, legal briefs, public relations campaigns, or even in-kind administrative services that make it easier for the bureaucrat to adopt and implement the policy favored by lobby i. In addition to the subsidies he receives from the lobbies, the bureaucrat s payoff from his policy choice depend on the terms of the statute authorizing his decision (determined in the statutory contest), and on pressures created by legislative oversight (determined in the oversight contest). We model both of these methods of bureaucratic control as imposing costs that the bureaucrat must pay to implement one policy or the other. For instance, think of a statute that stacks the deck From here forward we use male pronouns to refer to the bureaucrat, female pronouns to refer to the legislator, and call each lobby it. 8
by granting rights to seek judicial review of decisions that impose penalties on regulated parties, but not on decisions to refrain from imposing penalties. One can think of this as favoring regulated parties by raising the potential costs (litigation, or the steps necessary to avoid it) of imposing a penalty relative to refraining from doing so. Formally (see Display 3), we use s {,, 2} to indicate the direction of any deck-stacking in the authorizing statute. s = i indicates that the statute favors lobby i, and s = indicates that the statute favors neither lobby relative to the other. We use v {,, 2} in the same way to indicate which lobby is favored by the legislator s oversight activities. We capture differences in the magnitude of the effects of statutory language and oversight using two parameters, σ and ν. If the statute favors, say, lobby, then, gross of the effects of lobbying efforts and legislative oversight, the bureaucrat must pay a cost σ > 0 in order to allocate the good to lobby 2 instead of lobby. If oversight activities favor, say, lobby 2, then, gross of the effects of lobbying efforts and statutory language, the bureaucrat must pay cost ν in order to allocate the good to lobby. Conditional on the statutory bias s, the oversight bias ν, and the implementation efforts of the lobbies e m,, e m,2, the bureaucrat s payoff from adopting policy x = is e m, σ {s = 2} ν {v = 2} ; And the bureaucrat s payoff from adopting policy x = 2 is e m,2 σ {s = } ν {v = }. 3.2 Oversight Contest The oversight contest immediately precedes the implementation contest. (See Display ) It determines the direction v {,, 2} of incentives or pressure the legislator applies to the bureaucrat through oversight hearings, budgetary threats, and other non-statutory means of control. Like the oversight contest, it proceeds in two stages. First, the lobbies simultaneously choose oversight efforts (e v,, e v,2 ). Second, the legislator chooses the direction of pressure v. As in the implementation contest, the legislator s decisions are subsidized by the lobbies her marginal benefit of setting v = i is increasing in lobby i s oversight effort. Unlike the implementation 9
contest, however, the legislator has three possible decisions to pressure the bureaucrat in favor of lobby, in favor of lobby 2, or to apply no pressure whatsoever. The legislator finds the first two options inherently costly. Effective oversight i.e. oversight that truly affects bureaucratic incentives requires the legislator to expend valuable resources. If nothing else, she must take the time to communicate her demands to the bureaucrats in question. She may also need to organize and attend oversight hearings, or organize a coalition of other legislators to create a credible threat of budgetary or statutory changes. We therefore model the legislator s payoffs from the oversight contest as e v, if v = ; c v {v } + e v,2 if v = 2; where c v > 0 is the legislator s marginal cost of pressuring the bureaucrat, relative to her marginal benefit of a lobby s subsidy. 3.3 Statutory Contest The statutory contests has the same basic structure as the oversight and implementation contests: First the lobbies choose levels of statutory effort (e s,, e s,2 ), then the legislator chooses the direction of deck-stacking s {,, 2}. Each lobby s effort, moreover, subsidizes deck stacking in it s favor making it more attractive to the legislator. Finally, the legislator finds some actions more costly than others. We assume specifically that there is an existing status quo statute, that itself has a directional bias q {,, 2}. We assume that the legislator finds altering this status quo, no matter the direction of its bias, inherently costly. It requires, if nothing else, space on the crowded legislative agenda that could be used for other bills. Formally, given a deck-stacking decision s and statutory efforts (e s,, e s,2 ) by the lobbies, the legislator s payoff from the statutory contest is: e s, if v = ; c s {s q} + e s,2 if v = 2; Finally, we assume that the legislator s total payoff (i.e. from the combination of statutory and oversight contests) is the sum of his payoffs from the two contests. 0
3.4 Summary and Discussion of the Model Displays, 2 and 3 summarize the timing, payoffs, and notation of the model. There are five key features that are important to note. First, the model depicts the bureaucrat as having direct control over the actual policy itself he choose the actual policy x {, 2} in the final implementation contest. The legislator s powers on the other hand specifically, setting the direction of the bias of statutory language s {,, 2}, and the direction v {,, 2} of any informal pressure through oversight (merely) impose costs the bureaucrat must incur in order to implement one decision or the other. Second, we capture the magnitude of the costs the legislator can impose through statutory language and oversight through two parameters σ and ν. If the legislator writes a statute that favors, say, lobby (sets s = ), then, gross of the effects of oversight and lobbying, the bureaucrat must pay a cost σ in order to allocate the good to group 2. Similarly, if the legislator takes an oversight action that pressure the bureaucrat if favor of lobby (sets v = ), then gross of the effects of statutory language and lobbying, the bureaucrat must pay a cost ν in order to allocate the good to lobby 2. These parameters allow us to model the effects of variation in two features of lawmaking and law-implementation institutions the legislator s overall ability to control bureaucrat choice (σ + ν), and the relative effectiveness of ex-ante (σ) vs. ex-post (ν) methods of control. The third key feature of the model is the legislator s ability to refrain from biasing bureaucratic choice either via statute or oversight. When the legislator sets s =, she authorizes bureaucratic choice without biasing it in one direction or the other. Similarly, the legislator can refrain from oversight (set v = ), and let bureaucratic choice play out under whatever biases are imposed by existing legislation. Fourth, the supply of legislation and oversight in this model is costly. First, there is always a legislative status quo in place (q {,, 2}), and changing that status quo (setting s q) requires the legislator (perhaps with the assistance of the lobbies) to expend resources. Second, oversight that effectively pressures the bureaucrat in one direction or the other (i.e. setting v = or v = 2 instead of v = ) is always costly. We model these costs with two parameters c s (the cost of changing the statutory status quo) and c v (the cost of pressuring the bureaucrat through oversight). As with the parameters we use to capture the effectiveness of legislative controls, the cost parameters allow us
to explore two concepts: c s + c v captures the costliness to legislators of controlling the bureaucracy overall, while the relative sizes of c s and c v capture the relative costs of changing the status quo and pressuring the bureaucracy through oversight. In combination with the effectiveness parameters σ and ν, moreover, these cost parameters capture the levels of efficiency of lawmaking ( σ c s ) and oversight ( ν c v ) institutions. Finally, there is an important feature of legislative and bureaucratic choice that most formal models have and this one does not: parameters meant to depict legislators and bureaucrats inherent preferences over policy. There are two ways to interpret the absence of these parameters here. One is to think of this model as describing a mechanism through which policymaker s revealed preferences are formed. From this point of view, the model provides a simplified depiction of revealed preference in which the effects of factors other than lobbying by organized interests are set aside. Alternatively, one can think of this model as depicting the implications of a particular hypothesis about legislators and bureaucrats preferences specifically that they seek to minimize the resources they must expend to supply the actions lobbyists demand, and maximize the rents they can collect by virtue of their ability to supply these goods. 2
4 Equilibrium This game has a unique subgame perfect equilibrium. We describe it starting from the implementation contests, and working backwards from there. Figures and 2 summarize events in the full equilibrium in the cases where q = and q = (the case of q = 2 is the same as q = with the labels reversed). Throughout, we restrict attention to portions of the parameter space that satisfy three conditions. First, we restrict attention to parameter values at which lobbying occurs along the path of play in every contest in the sense that at least one group exerts positive effort. This requires first of all that the biases created through statutory language and legislative oversight are not so large that a group is unwilling to exert enough effort to overcome them in the bureaucracy. Specifically, σ +ν <. It also requires that each lobby is willing to lobby enough to compensate the legislator for changing the statutory status quo, and to pressure the bureaucrat through oversight. Specifically, c s < σ, c v < ν. 2 Second, we restrict attention to the case in which statutory language is more effective than expost oversight as a tool for controlling bureaucratic choice. Formally, the cost σ that deck-stacking can impose on bureaucrats for certain decisions is strictly larger than the cost ν that oversight can impose. The alternative to this assumption e.g. that ex-post oversight is more effective than exante legislative language is certainly interesting, and might shed light on a number of real-world circumstances. However, our preference is to provide a more in-depth analysis of a single case than a series of necessarily brief treatments of multiple cases. In any event, results for all arrangements of parameters are available upon request from the authors. Third, we show below that in equilibrium, the legislator gets more rents at the oversight stage when the statue does not stack the deck in favor of either group than she does when the statute stacks the deck. This last restriction concerns the size of this difference in rents relative to the cost the legislator must bear to change the status quo. Specifically, we restrict attention to values of c s, 2 In fact, when σ > ν, these conditions are sufficient but not necessary. When σ > ν, at least one lobby exerts positive effort with positive probability in each contest in equilibrium, and all the propositions we state in this section hold, if and only if all the following are true: σ + ν <, c v < ν and c s < σ + ( ) ( ) 3 ν cv c 2 v ν ν + cv 2 ν + 3
c v and ν at which c s > ( 3 ν c ) v 2 ( ν 2 + c v ν ). Under this condition, the rent the legislator captures under a statute that does not favor either group is smaller than the costs the legislator must bear to change the status quo. Thus, absent lobbying by the groups, the legislator always has a net bias in favor of the status quo. The merit of this case (as opposed to the one with the opposite inequality) is that it induces an equilibrium that is somewhat simpler to describe. Substantively, however, there is no important difference between the two cases. Finally, we should note that not every feature of equilibrium we characterize is limited to parameter values that meet all three of these restrictions. In each of the following propositions, therefore, we state exactly which conditions are necessary for the proposition to hold. 4. In the Implementation Contest For any given path of play, let b i be the total bias in favor of group i resulting from the statutory bias s and oversight bias v chosen along that path. Specifically, b i = σ {s = i} + ν {s = i}. Given b and b 2, the bureaucrat must pay a cost of b 2 to adopt policy x = and b to adopt a policy of x = 2. Thus, the bureaucrat s net bias in favor of (say) x = is b b 2. For instance, if the statute is biased in favor of group while oversight is biased in favor of group 2, then b = σ, b 2 = ν, and, gross of the effects of lobbying efforts, the bureaucrat has a net bias in favor of group of b b 2 = σ ν. Proposition 4. (Equilibrium in the Implementation Contest). Suppose σ + ν <. In the unique subgame perfect equilibrium, conditional on the net biases b and b 2, play in the implementation contest is as follows: If b b 2, then: (i) Lobby plays a mixed strategy over its implementation effort e m, in which it exerts zero effort with probability b b 2, and its effort is uniformly distribution on (0, [b b 2 ]) with 4
probability [b b 2 ]. (ii) Lobby 2 plays a mixed strategy over its implementation effort e m,2 in which it exerts zero effort with probability b b 2 and its effort is uniformly distributed on (b b 2, ) with probability [b b 2 ]. (iii) If e m,2 e m, > b b 2, the bureaucrat sets x = 2 with probability, and if e m,2 e m < b b 2, the bureaucrat sets x =. e m,2 e m, = b b 2 occurs with probability 0 along the path of play. Lobby s expected utility, (gross of statutory and oversight effort costs, conditional on b, b 2 and evaluated just before implementation efforts are chosen) is b b 2. Lobby 2 s expected utility is 0. Lobby wins i.e. x = with probability 2 + 2 (b + b 2 ) 2. If instead b < b 2 everything is as above with the labels and 2 switched. There are three things about this portion of the equilibrium worth highlighting. First, in this portion of the parameter space, statutory language and ex-post oversight can combine to make the bureaucrat s decision a foregone conclusion almost certain to be made in favor of one of the two lobbies. Specifically, if s = v = i for one lobby i, then as σ + ν approaches, the probability that lobby wins approaches, and the expected efforts of the lobbies approach zero. Thus, although the bureaucrat is in principal able to choose a policy that contravenes the one favored by statute in this model, for σ + ν close enough to, the probability that the bureaucrat makes a choice different from the one favored by the legislator can be made arbitrarily small. Second, if neither lobby is advantaged in the net, both lobbies get expected utility 0 from the implementation contest. One can think of this as resulting from the fact that when there is no bias in favor of one lobby or the other, the competition between the two lobbies is perfect, so that each is driven to expend (in expectation) the full value of its potential gains from the policy choice. Third, when there is a net bias in favor of one lobby, the favored lobby s expected utility is exactly equal to the size of that bias. Thus, the advantaged lobby s utility varies with the magnitude of its advantage. On the other hand, the disadvantaged lobby gets an expected payoff of 0 regardless of the magnitude of the bias against it. Thus, the disadvantaged lobby has an incentive to lobby the legislator to change the terms of the implementation contest only if the legislator has the ability to 5
reverse the direction of the net bias. Marginal changes that reduce the magnitude of the net bias without reversing its direction are worthless to the disadvantaged lobby. In the case we study here where statutory language has a larger marginal effect than ex-post oversight on bureaucratic choice, this has a critical implications for the oversight contest. 4.2 In the Oversight Contest To get a sense of what occurs in the oversight contest, first consider a path of play in which s = so that the statute chosen in the statutory contest stacks the deck in favor of lobby. If the legislator decides not to exert the effort required to pressure the bureaucrat through oversight (i.e. sets v = ), lobby goes into the implementation contest with a net advantage (and thus expected payoff) of σ. Through ex-post oversight, the legislator can increase or decrease this advantage by ν. However, in the case in which σ > ν (statutory language is more powerful than ex-post oversight), oversight can change the magnitude of the net bias, but not its direction. Therefore, lobby benefits from oversight in its favor, but lobby 2 can not get enough benefit from oversight to justify the cost of lobbying for it. Thus, in equilibrium, along any path in which s =, lobby 2 exerts zero oversight effort in equilibrium. Lobby, on the other hand, is willing to exert exactly ν lobbying effort to get the legislator to pressure the bureaucrat in its favor. In the case where c v < ν, this is more than enough to get the legislator to do so. However, since lobby faces no competition from lobby 2 in the oversight venue, it needs to pay the legislator no more than her exact cost of supplying pressure through oversight. Thus, in equilibrium, lobby 2 exerts zero effort, lobby exerts exactly c v effort, and the legislator pressures the bureaucrat to favor of lobby with certainty. Proposition 4.2 (Equilibrium in the Oversight Contest with a Biased Statute). Suppose that c v < ν < σ and ν + σ <. In equilibrium, along any path of play in which s =, events in the oversight contest are as follows: (i) Lobby sets e v, = c v. (ii) Lobby 2 sets e v,2 = 0. (iii) If e v, = c v, the legislator set v =. Lobby s expected payoff from the oversight contests (gross of statutory effort costs, conditional 6
on s = and evaluated just before oversight efforts are chosen) is σ + ν c v. Lobby 2 s expected payoff from the oversight contest is 0. The legislator s payoff from the oversight contest (gross of statutory lobbying efforts and any costs of changing the status quo legislation, conditional on s =, and evaluated just before oversight efforts are chosen) is 0. On the other hand, if s =, then the legislator does have the ability to determine the net direction of the bias at the implementation stage. Each lobby is in particular willing to exert as much as ν effort to get the legislator to pressure the bureaucrat in its favor. Thus the lobbies compete, and as a result the legislator earns rent specifically gets a subsidy larger than his costs of providing oversight c v. The following proposition describes the details. Proposition 4.3 (Equilibrium in the Oversight Contest with a Neutral Statute). Suppose that c v < ν < σ and ν + σ <. In equilibrium, along any path of play in which s =, events in the oversight contest are as follows: (i) Each lobby i {, 2} plays the same mixed strategy. In that strategy e i,v = 0 with probability c vν, and e v,i is uniformly distributed on (c v, ν) with probability cv ν. (ii) If both lobbies exert zero effort, the legislator sets v = (thus applying no pressure to the bureaucrat). If e m, > max {c v, e m,2 } the legislator sets v =. If e m,2 > max {c v, e m, }, the legislator sets v = 2. Conditional on the event that at least one lobby exerts positive effort, max {e m,, e m,2 } > max c v, e m,, e m,2 with probability. Each lobby s expected payoff (gross of statutory effort costs, conditional on s =, and evaluated just before oversight efforts are chosen) is 0. The legislator s expected payoff (gross of statutory lobbying efforts and any costs of changing the status status quo, conditional on s =, and evaluated just before oversight efforts are chosen) is ( 3 ν c v ν ) 2 ( 2 + c v ν ) Propositions 4.2 and 4.3 together generate one of the model s most important substantive implications: A statute that biases the bureaucratic process in favor of one lobby relative to another results in lobbying over legislative oversight of the bureaucracy that is relatively lop-sided with 7
little or no effort by the lobby disadvantaged by the statute, and consistent participation by the lobby favored by the statute. On the other hand, lobbying over legislative oversight of the bureaucracy is more intense and more competitive under an authorizing statute that does not create an advantage for one side or the other. Because of this, the legislator earns more lobbying rents from her oversight activities under an un-biased statute than she does from a biased statute. Further, although the lobbies prefer different biased statutes, the sum of their expected utilities from a biased statute (σ + ν c v ) is strictly larger than the sum of their expected utilities from a biased statute (0). Thus there is a fundamental conflict between the statutory design that is (jointly) best for the lobbies, and the one that is best for the legislator. 3 4.3 In the Statutory Contest The statutory contest determines how this conflict is resolved. Recall that we restrict attention to the case in which the marginal gain in rents the legislator earns by passing a neutral statute is smaller than the legislator s marginal cost of changing the legislative status quo. Thus, when the status quo statue is biased in favor of one of the lobbies, an unbiased statute is never adopted in equilibrium neither lobby has an incentive to pay for that change to the status quo, and the legislator is unwilling to bear the costs without subsidies from a lobby. In the statutory contest, then, one lobby defends the status quo, and the other tries to get it replaced with a statute that is biased in the opposite direction. The former lobby enjoys an advantage that is proportional to the cost the legislator must bear to write and pass any new legislation. Proposition 4.4 (Equilibrium in the Statutory Contest when the Status Quo Legislation is Biased). Suppose that c v < ν < σ, σ + ν <, c s < σ, and c s > ( 3 ν c ) v 2 ( ν 2 + c v ν ). If q = (the status quo legislation stacks the deck in favor of lobby ), events in equilibrium in the statutory contests are as follows: 3 In the case in which σ < ν (when deck-stacking is less effective than ex-post oversight), similar results hold: When the deck is stacked, lobbying in the oversight contest is lop-sided, with both sides participating, but the side favored by the legislation exerting greater effort that the un-favored party; And lobbying regarding oversight is more intense and generates more rents for the legislator when the statute is neutral than when it favors either lobby. Thus the same sort of conflict between the lobbies and the legislator arises. 8
(i) Lobby plays a mixed strategy in which e s, = 0 with probability distributed on (0, σ + ν c v c s ) with probability cs σ+ν c v. (ii) Lobby 2 plays a mixed strategy in which e s,2 = 0 with probability distributed on (c s, σ + ν c v ) with probability cs σ+ν c v c s σ+ν c v, and e s, is uniformly c s σ+ν c v, and e s,2 is uniformly (iii) If e s, > e s,2 c s, the legislator sets s = with probability, and if e s,2 c s > e s,, the legislator sets s = 2 with probability. One of these strict inequalities holds with probability, and thus the legislator sets s = with probability 0. Lobby s expected utility (evaluated just before statutory efforts are chosen) is c s, Lobby 2 s expected utility is 0, and the legislator s expected utility is: ( 3 (σ + ν c v) c s σ + ν c v ) 2 ( 2 + c s σ + ν c v ) If q = 2, events in equilibrium are given as in the description above, with the labels and 2 switched. Now consider what occurs when the status quo favors neither lobby. In this case, the legislator strictly prefers the status quo to any alternative changing the status quo requires her both to the pay the cost of writing and passing new legislation and to forgo the additional rents she earns from oversight under a neutral statute. One can show, however, that in this case where c s < σ and c v < ν, each lobby is willing to exert the effort necessary to induce the legislator to pay both of these costs. To simplify the description of the resulting contest, let R(ν, c v ) ( 3 ν c ) v 2 ( 2 + c ) v ν ν Denote the rent the legislator captures in the oversight contest when the s =. Proposition 4.5 (Equilibrium in the Statutory Contest when the Status Quo Legislation is Neutral). Suppose that c v < ν < σ, σ + ν <, c s < σ, and c s > R(ν, c v ). If q =, then events in equilibrium in the statutory contest are as follows: (i) Each lobby i {, 2} plays a mixed strategy in which e s,i = 0 with probability R(ν,cv)+cs σ+ν c v, and e s,i is uniformly distributed on (R(ν, c v ) + c s, σ + ν c v ) with probability R(ν,cv)+cs σ+ν c v. 9
(ii) If both lobbies exert 0 effort, the legislator sets s =. If e s, > max {R(ν, c v ) + c s, e s,2 }, the legislator sets s =. If e s,2 > max {R(ν, c v ) + c s, e s, }, the legislator sets s = 2. One of these events occurs with probability. Each lobby s expected utility (evaluated just before statutory efforts are chosen) is 0. The legislator s expected utility is R(ν, c v ) + ( 3 (σ + ν c v) R(ν, c ) v) + c 2 ( s 2 + R(ν, c ) v) + c s. σ + ν c v σ + ν c v 20
Figure : Equilibrium when the status quo is biased towards group (q = ). Lobby sets e m, = 0 with probability c s σ+ν c v and otherwise mixes uniformly on (0, σ + ν c v c s ). Lobby 2 sets e m,2 = 0 with probability c s σ+ν c v and otherwise mixes uniformly on (c s, σ + ν c v ). s = with probability 0. ( ) 2 s = with probability 2 + c s 2 σ+ν c ( v R(ν,cv)+c s σ+ν c v ). s = s = s = 2 Lobby sets e v, = ν c v. Lobby 2 sets e v,2 = 0. v = with certainty. Each lobby i sets e v,i = 0 with probability cv ν, and otherwise mixes uniformly on [c v, ν]. v = with probability ( cvν ) 2. Lobby sets e v, = 0, Lobby 2 sets e v,2 = ν c v. v = 2 with certainty. v = v = i for either v = 2 lobby with probability [ ( ) ] c vν 2. Lobby sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [0, (σ + ν)]. Lobby 2 sets e v,2 = 0 with probability σ + ν, and otherwise mixes uniformly on [σ + ν, ]. x = with probability 2 + 2 (σ + ν)2. v = 2 v = v = 2 Lobby sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [σ + ν, ]. Lobby 2 sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [0, (σ + ν)]. x = with probability 2 2 (σ + ν)2. Lobby sets e v, = 0 with probability ν, and otherwise mixes uniformly on [0, ν]. Lobby 2 sets e v, = 0 with probability ν, and otherwise mixes uniformly on [ν, ]. x = with probability 2 + 2 ν2. Each lobby i mixes uniformly on [0, ]. x = with probability 2. Lobby sets e v, = 0 with probability ν, and otherwise mixes uniformly on [ν, ]. Lobby 2 sets e v, = 0 with probability ν, and otherwise mixes uniformly on [0, ν]. x = with probability 2 2 ν2 2
Figure 2: Equilibrium when the status quo is unbiased (q = ). Each lobby i sets e m,i = 0 with probability R(ν,c v)+c s, and otherwise mixes uniformly on ( σ+ν c v R(ν,cv)+c s σ+ν c v, σ + ν c v ). ( ) s = with probability R(ν,cv)+c 2. s σ+ν c v s = i for either lobby with probability 2 [ ( ) ] 2 R(ν,cv)+c s σ+ν c v s = s = s = 2 Lobby sets e v, = ν c v. Lobby 2 sets e v,2 = 0. v = with certainty. Each lobby i sets e v,i = 0 with probability cv ν, and otherwise mixes uniformly on [c v, ν]. v = with probability ( cvν ) 2. Lobby sets e v, = 0, Lobby 2 sets e v,2 = ν c v. v = 2 with certainty. v = v = i for either v = 2 lobby with probability [ ( c vν ) ] 2. Lobby sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [0, (σ + ν)]. Lobby 2 sets e v,2 = 0 with probability σ + ν, and otherwise mixes uniformly on [σ + ν, ]. x = with probability 2 + 2 (σ + ν)2. v = 2 v = v = 2 Lobby sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [σ + ν, ]. Lobby 2 sets e v, = 0 with probability σ + ν, and otherwise mixes uniformly on [0, (σ + ν)]. x = with probability 2 2 (σ + ν)2. Lobby sets e v, = 0 with probability ν, and otherwise mixes uniformly on [0, ν]. Lobby 2 sets e v, = 0 with probability ν, and otherwise mixes uniformly on [ν, ]. x = with probability 2 + 2 ν2. Each lobby i mixes uniformly on [0, ]. x = with probability 2. Lobby sets e v, = 0 with probability ν, and otherwise mixes uniformly on [ν, ]. Lobby 2 sets e v, = 0 with probability ν, and otherwise mixes uniformly on [0, ν]. x = with probability 2 2 ν2 22
5 Discussion We showed above that the legislator s expected payoffs are larger from biased legislation than from neutral legislation, while the opposite holds for each lobby. In this section we explain the intuition behind this result and examine the determinants of the intensity of these structural preferences specifically, the marginal differences in players payoffs from the two types of legislation. Lobby gets 0 expected utility in equilibrium in the event that neutral legislation or legislation biased towards lobby 2 is passed. On the other hand, its payoff from legislation that biases the bureaucratic process in its favor is σ + ν c v. It is unsurprising that this marginal difference is increasing in the magnitude of the effects of statutory provisions on bureaucratic choice (σ). In some sense, σ is the direct effect of the bias created in Lobby s favor by its preferred legislation. The relevance of the magnitude of the effect of oversight (ν) and the cost oversight, however, may be not be immediately obvious. Recall that statutory controls on the bureaucratic process severely curtail the benefits that the dis-advantaged lobby can reap from legislative oversight. Under the parameter values studied here, this effect is especially pronounced: If the statute is biased in favor lobby, the legislator s oversight powers are too weak to improve lobby 2 s expected payoff in the implementation contests. As a result, when the statute is biased in its favor, lobby faces no competition from lobby 2 in lobbying the legislator to provide favorable oversight. As a result it is able to gain the benefit of that oversight (ν) without having to expend the full value of those benefits on lobbying. Specifically, it only needs to compensate the legislator for the cost c v of providing that oversight. That cost is lower than the value of the oversight to lobby, and thus lobby enjoys a profit of ν c v. Stated more briefly, the direct effect of statutory bias on the bureaucratic process itself has a secondary effect of biasing the competition over legislative oversight. The value of that secondary effect is captured by the difference between the value of winning that latter contest (ν) and the cost c v that the lobby must pay to obtain it, given its advantage. The legislator s expected payoff from biased legislation is 0. This is because the lack of competition the advantaged lobby faces for the legislator s oversight services. As a result, the legislator ends up selling those services at cost and thus being no better off than she would be if she were not lobbied at all. On the other hand, her expected payoff (and thus marginal gain in rent) from 23
an unbiased statute is R(ν, c v ) = ( 3 ν c ) v 2 ( ν 2 + c v ν ). This payoff is strictly decreasing and convex in the legislator s cost c v of providing oversight, and strictly increasing and convex in the magnitude ν of the effect of oversight. 4 The key to understanding these effects is to recognize that when the legislation is unbiased, competition for the legislator s oversight services is two-sided and intense. Because unbiased legislation gives no prior advantage to either side in the bureaucracy, the direction of pressure that the legislator applies through oversight can have a substantial effect on the payoffs of either lobby. Specifically, each lobby is willing to pay as much as ν to get the legislator to intervene on its behalf. Because of the stiff competition, each lobby overpays with some probability in equilibrium each plays a mixed strategy in which it bows out of the competition (exerts zero effort) with probability cv ν, and with probability cv ν mixes effort uniformly over (c v, ν). In the latter event, the lobby fully covers the legislator s cost of oversight (c v ), and pays an additional 2 (c v ν) amount of effort in expectation. Roughly speaking, the amount of rent the legislator earns is given by the difference between the amount the lobbies are willing to pay for her services (ν) and her cost of supplying those services (c s ). As these quantities diverge (ν increases or c s decreases) the degree of overpayment increases driving up the legislator s rents. The more general lesson of this paper is that statutes can affect the allocation of rents across political actors and across policy making venues specifically by altering the competitiveness of the lobbying environment. Changes to structure that create an advantage for lobbies on one side of an issue reduce the incentives their opponents have to participate in policymaking. This in turn allows the advantaged lobbies to get what they want with less lobbying effort. The result is that any policymaker that has authority to affect how existing statutes are implemented is lobbied less when 4 Specifically, which is positive since c v < ν. Further, ( ) R(ν, c v) = c2 v, c v ν 2 ν R(ν, cv) = 2 ( 3 cv ν ) [ ( ) ] 2 3 4 + 2 cv ν and 2 ν R(ν, cv) = 2 ( ) 2 cv 2 3 ν 24
those existing statutes stack deck in favor of a particular policy outcome. The result is an induced preference on the part of any such policymaker for statues that create a competitive environment between opposing interests statutes that in particular create no particular advantages for one side relative to another. 25