Gatekeeping Christophe Crombez, Tim Groseclose and Keith Krehbiel University of Leuven, University of California Los Angeles, and Stanford University September 28, 2005 We are grateful to David Primo for research assistance and to John Huber, Nolan McCarty, and Adam Meirowitz for helpful discussions. Abstract Collective choice bodies throughout the world use a diverse array of codified rules that determine who may exercise procedural rights, and in what order. This paper analyzes several two-stage decision-making models, focusing on one in which the first-moving actor has a unique, unilateral, procedural right to enforce the status quo, i.e., to exercise gatekeeping. Normative analysis using Pareto-dominance criteria reveals that the institution of gatekeeping is inferior to another institutional arrangement within this framework—namely, one in which the same actor is given a traditional veto instead of a gatekeeping right. The analytical results raise an empirical puzzle: When and why would self-organizing collective choice bodies adopt gatekeeping institutions? A qualitative survey of governmental institutions suggests that—contrary to an entrenched modeling norm within political science—empirical instances of codified gatekeeping rights are rare or nonexistent. Gatekeeping Christophe Crombez, Tim Groseclose and Keith Krehbiel A large portion of the procedural complexity in collective choice bodies throughout the world can be characterized in simple game-theoretic models. All that is required as a first-order approximation is to specify analytically features such as unique proposal or veto rights, rules governing amendments, and the order in which decision makers act. Empirical counterparts for these abstractions are plentiful. Examples of individuals or groups who possess unique procedural rights include standing committees in legislatures, the Commission in the European Union (EU), prime ministers in parliamentary governments, and the chief executive in presidential systems. Examples of their special rights include, respectively, bill referrals as stipulated by jurisdictional arrangements, proposal rights as designated by international treaties, rights to call elections as provided for by national constitutions, and the right to veto legislation after its passage as specified in the many state and national constitutions. Rules governing amendments may, likewise, give opportunities for some—but not all—decision-makers to propose modifications to the measure before the body. Finally, rules that determine the order in which players engage in collective choice have force within governments at both intra-branch and inter-branch levels. For example, many legislatures have rules or precedents that require that, upon its introduction, legislation is immediately referred to a standing committee, only to be considered later, if at all, by the full body. Similarly, some separation of powers systems require that the legislative branch initiates policy, but that the executive branch (e.g., the president) acts next and must sign or veto the proposed legislation. Parliamentary systems may also grant proposal rights to extra-legislative bodies or players, such as cabinet ministers. This study presents comparative institutional analyses for a class of simple two-stage decision-making arrangements. Within this class, the focal institutional 1 arrangement is that of gatekeeping. A gatekeeping institution is a collective choice process in which the first-stage player has a procedural right to implement unilaterally an exogenous status quo policy, in which case the second-stage player is denied the opportunity to participate in collective choice. This basic model has been used to study a wide range of political behavior since Denzau and Mackay (1983) first formalized it. The study has four parts: a discussion of the concept of gatekeeping and associated claims in the literature, focusing on the institutions of the U.S. and the EU; a formal analysis of gatekeeping alongside a closely-related model of a traditional veto; a descriptive survey of collective choice bodies that use multi-stage decisionmaking procedures; and a summary. Three main findings parallel the first three sections. First, models of gatekeeping are among the most common in political science, particularly in studies that involve legislatures. Second, models of gatekeeping have remarkably undesirable consequences, compared with similar procedural arrangements. Third, instances of gatekeeping institutions are rare or nonexistent within a sizable and diverse set of legislatures. Definitions and Perceptions The modal view of gatekeeping in the U.S. congressional literature is clear and concise. Committees have it, use it, and benefit from it. Similarly, allegations of Commission gatekeeping in the EU are rampant. Less clear, less concise, but equally important is the question: What is gatekeeping? Clarification rests on a distinction between rights and power. In a multistage collective-choice process, an early-acting individual or group of individuals is said to possess a gatekeeping right if the governing procedures of the body allow the individual or group not to act on specific proposals, and if the certain consequence of such inaction is that an exogenously determined status quo policy remains in effect. A gatekeeping right is therefore a feature of codified rules 2 in the empirical domain and a feature of the game form in the theoretical domain. In contrast, a body or group is said to have gatekeeping power if it has a gatekeeping right and the right produces an outcome that the gatekeeper prefers to the outcome that would have resulted if it did not have a gatekeeping right. Gatekeeping power is therefore a characteristic of a political outcome in a specific empirical domain, and a characteristic of equilibrium play of a game in which a gatekeeping right is postulated to exist. Clearly, then, a gatekeeping right is a necessary condition for gatekeeping power. The literature on gatekeeping typically does not make this distinction, preferring to use the term “gatekeeping power” as an umbrella concept for of our definitions. As the following review illustrates, however, failure to differentiate between gatekeeping as an exogenous codified procedural right and gatekeeping as an action leading to desired outcomes not only confuses matters but also tends to result in overstated claims about ostensible gatekeeping rights and their outcome consequences. The prevailing view on the gatekeeping by U.S. congressional committees can be traced back to 1885, when Woodrow Wilson penned his famous quote: “As a rule, a bill committed is a bill doomed. When it goes from the clerk’s desk to a committee-room it crosses a parliamentary bridge of sighs to dim dungeons of silence whence it will never return” (1956 [1885], 63). A few decades later, Lindsay Rogers reinforced the interpretations and sentiments of Wilson with reference to a more specific example: “In the House of Representatives, even though a majority would wish, say, to change a tariff schedule, it could not do so without the consent of a majority of the Ways and Means Committee. That body has absolute control ” (1926, 131). Rogers acknowledges a procedural change that occurred since Wilson— the House’s discharge procedure. But, in the same sentence, he ridicules it: “There have been, to be sure, rules for the ‘discharge of committees’ but it would have been more accurate to title them for the ‘non-discharge of committees’” (1926, 131). 3 Once established, this interpretive bandwagon has proven to be irresistible to researchers in several fields. The following excerpts form a small but representative sample of descriptive research and formal theories about gatekeeping in the U.S. Congress. “The most important functions of standing committees are screening and drafting. The screening function, also known as gatekeeping, is the power to say no. A majority of a committee’s members must support a bill before the committee will send it to the floor... [C]ommittees have life-or-death control over legislative proposals... (Johnson, Miller, Aldrich, Rohde, and Ostrom, 1994, 450, 451). “[C]ommittees, notably in the House, often exert considerable gatekeeping and agenda-setting powers. (Kollman 1997). “Under the new order...A committee’s members, operating without accountability to a majority within either the House or the Senate, could stop any legislation that fell within the committee’s jurisdiction, no matter how widespread the support for the legislation in the Congress or the country” (Dodd and Schott 1979, 75). “[G]overnment by committee vests a tremendous amount of power in the committees and subcommittees of Congressespecially in their leaders... Committee members can bury a bill by not reporting it to the full House or Senate” (Janda, Berry, and Goldman, 1995, 380). “Undoubtedly, the most important tool possessed by committees is their gatekeeping power . (Maltzman 1997, 65) Clearly, these claims are strong. Furthermore, when a relatively few researchers provide qualifications, the qualifications are tepid by comparison. For example, in a work addressing the multiple ways in which legislation can be killed en route to enactment, Herzberg acknowledges that gatekeeping “power” is not absolute. But then she writes, “Given the difficulty of the discharge process, it is not surprising that it is rarely used to prevent blocks.” (1986, 215). Kingdon, likewise, acknowledges that House’s discharge procedure, but he downplays its significance on grounds that it is “difficult to obtain” (1989, 141). Keefe and Ogul, similarly, downplay the significance of discharge mechanisms. “The parliamentary weapons 4 that House members may call upon in attempting to bring obdurate committees to heel are not impressive... [The discharge procedure] is a small-caliber ‘gun behind the door’ whose presence may spur the search for a compromise solution which can be brought to the floor” (Keefe and Ogul, 1993, 237-238). Finally, as attention shifts from descriptive to theoretical research, qualifications are abandoned and gatekeeping in its pure strong form returns. Denzau and Mackay (1983) were the historical counterparts to Woodrow Wilson within the formal modeling literature. Weingast and Moran (1983) used the Denzau-Mackay model to study the FTC and regulatory processes, while Weingast and Marshall’s (1987) influential article on legislative organization elevates gatekeeping to the status of a “property right.” Others followed suit. According to Shepsle and Weingast (1984, 217-218), “One extremely important but subtle rule underpinning committee influence is its power to veto proposals within its jurisdiction: any proposal that fails to make a committee majority better off is simply kept from coming to the floor for a vote.” McCubbins, Noll, and Weingast (1994) agree:“In each chamber of Congress, at least one subcommittee and one full committee have gatekeeping rights in that a bill normally will not be considered by the entire legislative body until it has been approved in committee.”1 Portrayals of gatekeeping are remarkably similar across the Atlantic, where claims that attribute the European Commission’s influence to gatekeeping are abundant. While discussing the Single European Act, a reform treaty signed in 1986, Fitzmaurice (1988, 398), for example, states: “The Commission is given a key ‘gatekeeping’ function, which could become politically uncomfortable.” Lenaerts (1991, 22) discusses the separation of powers in the EU, and claims that “apparently no 1 See also Shipan (1998) on regulatory influence, Ferejohn and Shipan (1990) on congressional-judicial relations, Huxtable (1994) and Dion and Huber (1995) on the House Rules Committee, Segal (1997) on courts, Snyder (1992a, 1992b) on biases in roll call data, and Aldrich (1994) and Cox and McCubbins (1993) on the majority party. 5 other Community institution nor a member state can force the Commission to take a legislative initiative when the Commission thinks such an initiative not to be in the interest of the Community.” Franchino (1999, 9) summarizes the Commission’s role as follows: “[The Commission] is a hybrid body in classical constitutional terms. It carries out traditional administrative functions, frequently shared with national administrations, but also has to provide executive leadership and legislative gatekeeping.” Lenaerts and Van Nuffel (1999, 435 and 439) state “The right to propose legislation means, in the first place, that the Commission can decide whether or not the Community should act. . . Neither the Council nor a member state can compel the Commission to submit a proposal.” Paralleling research on the U.S. Congress, formal theorists also defend their formalization of Commission gatekeeping with reference to perceived stylized facts. Steunenberg (1994, 647) claims that the consultation procedure grants the Commission the “exclusive right of initiative” and suggests that only the Commission “decide[s] whether or not it will initiate a legislative process.” Moser (1996, 836) concurs: “[T]he Commission does have broad agenda control and is not bound by the suggestions of the [Parliament].”2 Although scholars of the EU are more likely than not to argue that the Commission has and benefits from a gatekeeping right, the consensus is not as overwhelming among Europeanists as Americanists. Westlake (1995, 336), for example, recognizes that “many of the Commission proposals are made in response to external factors, from Council through member state requests through to international obligations.” Others claim the Commission’s power has declined. Edwards and Spence (1994, 8) state: “[A]lthough the Commission retains the sole right to initiate legislation, in a number of respects it has become a more formal responsibility rather than the source of power and authority in setting the Community agenda as originally conceived.” Moravcsik (2002, 612) agrees that the Commission has lost power, but considers 2 See also Steunenberg et al. (1999, 344, 352) and Nurmi and Meskanen (1999). 6 the Parliament as the beneficiary: “For over a decade, the [Parliament] has been progressively usurping the role of the Commission as the primary agenda-setter visà-vis the Council in the EU legislative process.” Likewise, a few formal theorists deny that the Commission has gatekeeping rights altogether. Tsebelis (1994, 131) concludes: “In any case, all three institutional actors can in fact place items on the legislative agenda.” Crombez (1996, 204) writes: “Only the Commission can initiate the [consultation] procedure, but it is required to make a proposal if the Council or the Parliament requests one. Thus the Commission has monopoly proposal power, but it does not have gatekeeping power.” Schulz and König (2000, 655) echo his remark: “While the Commission has the formal authority to propose legislation, the Council or the [Parliament] may request that the Commission submit a proposal (Art. [208] EC and Art. [192] EC). That is, the Commission has proposal power but no gatekeeping power.” In light of the excerpts on the gatekeeping rights of U.S. congressional committees and the EU Commission—and the fact that they are neither atypical nor exhaustive—it seems potentially fruitful not only to revisit and inspect the fundamental properties of the gatekeeping model, but also to undertake a comparative analysis that might provide some insights into conditions under which we might expect, or expect not, to observe various institutional arrangements. Comparative Institutional Analysis Suppose that two actors, L and C, must choose a policy x within the convex policy space X ∈ Rn . One can interpret these actors as a (one-member) legislature and a (one-member) committee. However, there are several other interpretations for the two actors, including two branches of government in a law-making process. For instance, L could be interpreted as Congress and C could be interpreted as the president. Let uL (x) and uC (x) be the utility functions of L and C. We assume that these functions are strictly quasi- concave, that is, that indifference curves 7 are strictly convex. (Geometrically, utility functions are single-peaked with no flat spots.) We also assume that the functions are maximized in the interior of the policy space. Let l and c be the points at which the functions are maximized. Define these as the ideal points of L and C. So that the analysis is non-trivial, assume l = c. Finally, we assume that if either actor is indifferent between two policies, the policy chosen is the one that benefits the other actor more. We consider two, simple, two-stage games. In the gatekeeping game, play begins with C who chooses whether to propose a bill or not. If C chooses no bill, the game ends, and the resulting policy is the status quo, q ∈ X, which is given exogenously and known to both players. If C decides to propose a policy, then L is allowed to adopt any policy in the second stage.3 In the traditional veto game, players move in the opposite order. First, L chooses any bill b ∈ X. Second, C chooses whether to accept the bill as-is (without amendment) or to veto (kill) the bill. If C accepts, then the final policy is b. Otherwise, the final policy is q.4 In both games the equilibrium concept is subgame-perfect Nash. To compare these games it is useful to formalize criteria for evaluating institutions. Definition. A game weakly-Pareto-dominates-at-q another game if for a given status quo q, both players weakly prefer all equilibria of the first game over all equlibria of the second game. A game Pareto-dominates-at-q another game if for a given status quo q, both players weakly prefer all equilibria of the first game over all equlibria of the second game and at least one player strictly prefers all equilibria 3 Although we stipulate that C’s action space as a dichotomy, other versions of this game (e.g. Denzau and Mackay 1983; Gilligan and Krehbiel 1987) assume that the action space is a continuum. This distinction is not outcome consequential. 4 Shepsle and Weingast (1987) develop a somewhat more elaborate version of this model which came to be known as an ex post veto. Although the empirical veracity of this institution is debatable, all that is important for our analysis is that institutions such as gatekeeping, the ex post or traditional veto, closed rules, etc., can in principle exist. This minimal requirement is not controversial in light of Article 1 Section 5 of the Constitution which states that “each House may determine the rules of its proceedings.” 8 of the first game over all equilibria of the second game. Finally, A game Pareto dominates another game, if for all q, that game weakly-Pareto-dominates-at-q the other game, and for at least one q the game Pareto-dominates-at-q the other game. In at least two senses our definition of Pareto dominance is stronger than the usual definition. First, ours compares a set of equilibria with another set, not just two particular equilibria. By our definition, if one game Pareto dominates another, then for at least one q the worst equilibrium of the first game must be better than the best equilibrium of the second. Second, our definition compares outcomes over a set of different status quos rather than just one status quo. Therefore, it is stronger, for instance, than a definition that require only that one game dominate another in terms of expected values. Example 1: One-dimensional policies Suppose the policy space is one-dimensional, i.e. X ⊆ ℜ. Suppose L and C’s preference for a policy depends upon the distance of the policy from their ideal point (or any positive monotonic transformation of this distance). Without loss of generality, assume c < l. We graph the outcomes for each game in Figure 1. First, note that for each possible status quo both games have a unique equilibrium policy. The figure considers four intervals for the status quo: (I) to the left of 2c − l, (II) between 2c − l and c, (III) between c and l, and (IV) to the right of l. If the status quo is in interval I, III, or IV, then both games produce the same policy outcome. The important difference between gatekeeping and veto institutions arises when status quo policies lie somewhat exterior to the Pareto interval on C’s side. These points are defined precisely by interval II. When the status quo lies in this interval, behavior in the two models is quite different. In the gatekeeping model, C exercises gatekeeping because he prefers the status quo to L’s ideal point, the policy that would result if C allows the game to proceed. The behavior of C can be summarized as self-interest trumping social optimality. Although the status quo is outside the set of Pareto policies and is, thus, not socially optimal, the status quo is nevertheless 9 better for C than is L’s ideal point. In contrast, when the status quo is in interval II the traditional veto game does produce a Pareto-improving outcome. Here L acts as a Romer-Rosenthal (1978) agenda setter. She chooses the policy closest to her ideal point subject to the constraint that C will not veto it. The constraint is binding in equilibrium. That is, C ends up being indifferent between the status quo and the policy that L proposes. Meanwhile, L strictly prefers the policy to the status quo. As Figure 1 shows, L is indifferent between the institutions if the status quo lies in intervals I, III, and IV, but she has a strong preference for the veto institution in interval II. To the extent that the organization in question is self-governing—that is, has the capacity to choose its own rules—this normative observation has potentially powerful predictive implications. The difficulty for the two players to obtain a Pareto-improving outcomes under gatekeeping is due to a commitment problem. By definition, when outcomes are not Pareto optimal, alternative outcomes exist that make at least one player strictly better off and the other player not worse off. Such outcomes cannot be obtained, however, when the main body acts last. Any hypothetical bargaining agreement in the contract curve, (c, l), cannot be sustained as an equilibrium because, under the open rule, L has a second-stage incentive and the institutional capability to renege on the agreement. The gatekeeping institution, therefore, seems to be deeply flawed from the perspective of social efficiency, and the generality of this previously unknown observation is focal point below. Example 2: Two-dimensional policies The qualitative observations extracted from Example 1 also hold when the policy space is two-dimensional. As in the previous example, suppose that C and L have Euclidean preferences. Consider the indifference curve of C that contains l. We illustrate this in Figure 2. There are three relevant regions in which the status quo 10 can lie. (1) If the status quo is on or outside this indifference curve, then both games produce the same policy outcome, l. (2) If the status quo is inside the indifference curve but not on the contract curve, then the two games produce different policies. In the gatekeeping game the resulting policy is the status quo. (3) If the status quo is on the contract curve between l and c (the line segment with these points as endpoints), then both games result in the same policy outcome, the status quo. (If C were to propose a bill, L would amend it to l. Because C prefers the status quo to l, C will instead exercise its gatekeeping right.) In the traditional veto game, L wants to choose a bill that maximizes her utility subject to the constraint that C weakly prefers the bill over the status quo. When the status quo is in this region, the solution can be found by drawing the indifference curve of C that includes the status quo point. The solution to L’s maximization problem is the point where this indifference curve intersects the contract curve. Define this point h(q). See Figure 2 for an illustration. C weakly prefers this outcome to q, the outcome of the gatekeeping game. Meanwhile, L strictly prefers this outcome to q. Consequently, the traditional veto game Pareto dominates the gatekeeping game. [Figure 2] Both examples demonstrate the inferiority of gatekeeping as an institution. Accordingly, they suggest that actual law-making bodies should be reluctant to choose this institution. One aspect of Example 2 suggests that law-making bodies should be even more reluctant to choose this institution when the policy space is many-dimensional. To see this, consider the indifference curve of C that passes through the ideal point of L. Call the points inside this curve the non-extreme points. It might be reasonable to believe that it is rare for a status quo to lie outside this region in an actual law-making body. Within the non-extreme region some status quos—namely, those that lie on the contract curve between C and L— cause the two games to produce identical outcomes. In the one-dimensional case these points compose half the measure of the non-extreme region. (The contract 11 curve is the points between c and l. This curve is half the length of the set of nonextreme points—the points between 2c − l and l.) However, in the two-dimensional case these points are a zero-measure set. (Note that the contract curve is a onedimensional set, while the non-extreme points are a two-dimensional set.) Further, the same relation will be true when the policy space has even more dimensions than two. Thus, while in the one-dimensional case there is a significant chance that the two institutions produce the same outcomes, in the many-dimensional case, there almost zero chance that they produce the same outcome. Generalizations Examples 1 and 2 generalize to a setting in which preferences are not necessarily Euclidean and the number of policy dimensions is arbitrarily large. Proposition 1. The traditional veto game Pareto-dominates the gatekeeping game. Proof. See Appendix A. Proposition 1 and the models on which it is based may elicit several ancillary questions or concerns that we address prior to the empirical application. Exaggeration of gatekeeping rights. Probably the most common concern is that the gatekeeping model overstates the procedural rigidity of its real-world counterparts. In the U.S. Congress, for example, several mechanisms are probable antidotes to gatekeeping excesses: e.g., the Senate’s practice of attaching nongermane riders to bills, the House’s discharge procedure, and the House Rules Committee’s ability to propose special rules to expedite consideration of otherwise blocked legislation (Oleszek 2004, 143-4). As an empirical matter, for reasons that will highlighted in the next section, we are entirely sympathetic to the claim that the gatekeeping model ignores important procedural realities. As a theoretical matter, however, it is important to begin by confronting the literature on its own terms, so Proposition 1 conforms to the industry-standard strong-form gatekeeping model which ignores these complexities. That being said and done, it is now instructive to ask 12 whether the key Pareto-dominance insight offered by Proposition 1 holds under more realistic—which is to say, weaker—forms of gatekeeping in which, say, player L may discharge a gatekeeper at some positive cost.5 Proposition 2, stated formally and proved in Appendix B, shows that it does. Specifically, a weak form of the gatekeeping institution is Pareto dominated by a similarly weak form of the traditional-veto institution. In short, it is not the absoluteness of the rights in the formalization that drives the result. Rather, the result reflects a deeper and more pervasive limitation in institutions of gatekeeping as compared with more traditional veto institutions. Bicameralism. Another component of institutional complexity that might be expected to temper the Pareto-dominance results is bicameralism. It is straightforward to show that, at best, two chambers with gatekeepers will generate the same Pareto-dominance result and, at worst, the addition of a gatekeeper in the added chamber will exacerbate the inferiority of the gatekeeping institution relative to the traditional veto. Suppose a committee in the U.S. House refuses to report a bill to the floor. Even if the Senate were to pass a companion bill, this would neither compel the House committee to act nor permit the House to act. Instead, under the assumptions of the standard model, the House committee would continue to withhold the bill. Worse yet, bicameralism can cause gatekeeping to be even less preferred to a traditional veto than under unicameralism. In the unicameral setting, half the players (L) strictly prefer the veto game, while the other half (C) are indifferent between the veto game and the gatekeeping game. However, in general, in a bicameral setting three-fourths of the players (L1 , L2 ), and one Ci would strictly prefer the veto game, while only one-fourth (the gatekeeping Cj ) would be indifferent.6 5 The standard model can be thought of as a special case in which the cost of discharge is infinite. 6 Example: Let the House and Senate (say, median voters) have ideal points of zero. Let the House committee have an ideal point of 2, and let the Senate committee have an ideal point of 3. Suppose the status quo is 5. In the bicameral gatekeeping game the Senate committee will exercise its gatekeeping right. Meanwhile, in 13 Weak institutional preferences. As illustrated in figure 1, C’s preference for the for veto institution is weak. In three of the four intervals in which the status quo may lie, outcomes are identical in the veto and gatekeeping models, and, in the fourth (interval II), C is indifferent between playing the veto game or gatekeeping game while only L strictly prefers the veto game. Analytically, this is not particularly bothersome, indifference in equilibrium is common in multistage games with complete information and can typically be broken via ǫ arguments. More substantively (and stepping slightly outside the model), there are two additional reasons that C (as well as L) will strictly prefer the veto game to the gatekeeping game. The first reason involves uncertainty. Suppose that L does not know Cs ideal point. Then it can be shown that, for all status quos, C weakly prefers the veto game to the gatekeeping game, and, for some status quos, C strictly prefers the veto game.7 The second reason involves experimental evidence from the wellknown ultimatum game in which one subject A proposes a division of externally provided money between subject A and subject B. The proposal by A—like L’s bill in the veto game—is a take-it-or-leave-it offer. If B accepts the offer, then the two players divide the money as A specified. However, if B rejects the offer, then neither subject receives any money. The pure bargaining model predicts that the the veto game the House and Senate (say, acting as a conference committee) would optimally propose a bill of 1. Neither committee would veto b = 1, so this would be the final policy. Note that the House and Senate plus the House committee strictly prefer the veto game, while the Senate committee is indifferent between the two games. Interestingly and intuitively, each chamber has especially strong preferences that its rival chamber not give gatekeeping rights to its committee. 7 Example: Consider a status quo in interval II. In the certainty case, L always proposes a bill that leaves C indifferent between it and the status quo and that C accepts. In the uncertainty case, either of two additional cases occur with nonzero probability. (i) L offers a bill that C strictly prefers to the status quo and that C accepts; or (ii) L offers a bill that C likes less than the status quo and that C vetoes. Case (i) makes C strictly better off than the counterpart gatekeeping game, while case (ii) is utility-neutral in this regard, so C strictly prefers the veto game. We can likewise show that, for all q, L’s expected payoff in the veto game strictly exceeds that of the gatekeeping game. 14 A proposes to keep all of the money, or to leave at most a penny for B, however, many experiments find that sharing is much more common (see, for example, Roth 1995). A common interpretation of these findings is that A suspects B might be revengeful in some manner. To the extent that such unmodeled motives are at play in collective choice settings as well as in laboratory bargaining settings, choices by L that affect C’s well-being are likely to be more accommodating than our model’s strict prediction, in which case C will have a strong rather than weak preference for the traditional veto institution over the gatekeeping institution. Falsifiability. A final concern is whether our theoretical results have falsifiable implications. We maintain that they are falsifiable on at least two levels: policy choice and institutional choice. In terms of policy choice, the Pareto-dominance result implies that if a status quo policy lies in interval II of figure 1, then it is out of equilibrium in terms of the larger model of institutional choice and will therefore be unstable. That is, a hypothetical gatekeeper will not be permitted unilaterally to maintain such a status quo, because some alternative institutional arrangement will be in place (or will be adopted) that empowers the legislature to move the policy into interval III (including, possibly, the L’s ideal point). On the other hand, if genuine gatekeeping rights do exist, then status quo points in interval II elicit effective gatekeeping, are theoretically stable, should be empirically identifiable, and would constitute refutation of our main result.8 Falsifiability at the level of institutional choice builds on the same logic. Because gatekeeping institutions are Pareto-dominated by veto institutions, observation of the former when the latter are possible would compel us to reject the theory. Stated more practically and positively, to the extent that it appears that collective choice organizations have 8 We suggest that an appropriate test of this implication should account for the possibility that it is costly to introduce and pass new legislation. Accordingly, occasional findings of status quo points in interval II would be tolerated. However, the distribution of status quo points in this interval should be less dense than a comparable region, say, an identically-sized region on the opposite side of L’s ideal point. For a test with these broad features see Krehbiel (2005). 15 gatekeepers, closer inspection should reveal that apparent gatekeeping is not codified, can be circumvented in codified ways, or both. Empirical Instances of Legislative Gatekeeping What, then, are the central procedural tendencies of collective choice bodies with regard to gatekeeping and related institutions? Our search for gatekeeping spans the globe in three phases: the United States, the European Union, and elsewhere. Gatekeeping in the U.S. Claim 1. In U.S. legislative bodies, gatekeeping rights are not granted to standing committees. The most straightforward case is the U.S. Senate. Due to its absence of a germaneness rule, the Senate has a de facto semi-automatic weapon with which to discharge its committees. Any measure may be brought up as a rider to another measure—the so-called vehicle—through the normal, open, amendment process. Therefore, U.S. Senate committees do not have gatekeeping rights. In the U.S. House of Representatives, the situation is somewhat more intricate procedurally than in the Senate, but, when all is said and done, the conclusion is much the same. Any House member may file a petition to discharge a bill from a committee, as long as the bill had been referred to the committee for at least thirty days. If the discharge petition gathers at least 218 signatures, a motion to discharge is put on the House’s Discharge Calendar. After seven legislative days, the motion becomes privileged business on the second and fourth Mondays of the month (excluding the last six days of a session). Any member who signed the petition may be recognized to offer the discharge motion, which is debated for 20 minutes and then subject to a vote. If a majority votes affirmatively on the discharge motion, it becomes permissible to make a motion for immediate consideration of the bill. 16 Finally, if this motion prevails by a majority, the bill comes before the House under an open rule. These conclusions are not unique to the U.S. national government. According to information provided by the National Council of State Legislatures, seventy state legislative bodies operate under Mason’s Manual of Parliamentary Procedure, which provides for discharge. Additionally, 22 state assemblies have mandatory reporting requirements of standing committees (i.e., automatic discharge), while 41 have explicit rules for discharge similar to that of the U.S. House. Much more often than not, the threshold is simple majority or less. Claim 2. Neither chamber of the U.S. Congress grants gatekeeping rights to the majority party. Cox and McCubbins (2004) have recently formalized some of the arguments of their influential book Legislative Leviathan under the rubric, “procedural cartel theory.” According to this theory, the majority party can prevent the floor from considering bills that it opposes. That is, the majority party and/or its leaders have gatekeeping rights. As we have defined terms, the congressional majority party has neither gatekeeping rights nor (by implication) gatekeeping power. The former requires a rule that stipulates that the concurrence of a majority of the majority party is necessary before the chamber can consider any bill. Clearly, no such rule exists. Nor do Cox and McCubbins claim otherwise. More weakly and sensibly, they argue that the majority party benefits from various non-codified behavioral practices, such as the ability to discipline members on key procedural or organizational choices . In turn, these practices pertaining to partisan agenda control are sufficiently effective that law-making is observationally equivalent (in terms of legislation produced and not produced) to a theoretical process in which the majority party has a strict gatekeeping right. Even if this were true, our analysis poses a puzzle for which the present lit17 erature offers no solution. A straightforward application of our propositions show that the procedural cartel model, too, is a Pareto-dominated institution. To see this, simply redefine C as the median of the majority party and let L be the median voter of the entire chamber. The cartel model is now game-theoretically identical to our gatekeeping model. But now consider an alternative institution where the majority party is given a traditional veto instead of a gatekeeping right. Because the proposition applies to this set of conditions, it follows that it is in the interest not only of a majority of the legislature (which inevitably includes many majority party moderates and some minority party moderates) but also (weakly) a majority of the majority party to switch from a gatekeeping institution to a veto institution. In other words, our model provides a unique explanation for the observed non-codified status of putative majority-party gatekeeping. Claim 3. No Constitutionally-specified participant in lawmaking in the U.S. (House, Senate, President) has gatekeeping rights. In the U.S. lawmaking process, Congress writes the bills, and the president has an approximate traditional veto.9 One could imagine an alternative institution in which the Constitution allowed only the president to write bills, but in which Congress could amend these bills without the prospect of a presidential veto. Structurally, such an arrangement is identical to our gatekeeping game where C is the president and L is the Congress. As Proposition 1 demonstrates, this structure is Pareto-dominated by the actual structure of the U.S. Constitution. It is therefore noteworthy and consistent with our analysis that the Founding Fathers did not opt for gatekeeping. More generally, no constitution of any U.S. state opts exclusively for gatekeeping while most of them employ some form of traditional veto.10 9 The veto can be overriden by a two-thirds majority of both houses. Although some constitutions give the executive a gatekeeping right, it is paired with a traditional veto, in which case the latter feature is analytically significant while the former is not. For instance, in Maryland only the governor can propose public works projects. This is tantamount to possessing a gatekeeping right over 10 18 Another possible constitutional structure is to allow the executive to write bills, but to give the legislature gatekeeping rights. That is, for instance, suppose the process begins with Congress, which writes a bill or chooses to exercise its gatekeeping right after which the president can amend the bill (if and only if it exists). This arrangement is identical to our gatekeeping game, subject to the relabeling of Congress as C and the president as L. It follows that this institution, too, is Paretodominated by one in which the president first writes a bill after which Congress has a traditional veto. Although these arrangements seem far-fetched, it is noteworthy that one of them is in fact used. Specifically, the U.S. president is sometimes granted fast-track trade promotion authority to negotiate agreements with foreign countries, subject to congressional approval. Upon receiving such proposed agreements, the Congress must accept or reject them without amendment. In other words, this is a form of the traditional veto model whose existence is consistent with our formal models. If, instead, trade promotion authority gave Congress only a prior gatekeeping right, then we would have considered this a falsification of our theory. Gatekeeping in the EU Claim 4. In the European Union, neither the Commission, nor Council, nor Parliament is granted gatekeeping rights. Upon initial consideration, the procedural status of the Commission of the European Union appears to be an instance of gatekeeping, because the Commission has codified exclusive rights to initiate legislation under both of the EU’s main lawmaking procedures: consultation and co-decision. But while monopoly initiation is a necessary condition for gatekeeping, it is not sufficient if other participants in the larger legislative process can compel the Commission to act. It is the unilateral ability to impose the status quo that provides sufficiency, and it is this ability that such projects. It is easy to show that the additional existence of the veto is sufficient to preclude the non-Pareto outcomes that can occur with gatekeeping alone. 19 the Commission lacks. More specifically, both the Parliament and the Council can play a significant role in determining the legislative agenda. The Parliament can request that the Commission make a proposal (Art. 192 EC). Similarly, a simple majority of the Council can demand that the Commission submit a proposal (Art. 208 EC). Should the Commission fail to respond, the Parliament and the member states can sue the Commission in the European Court of Justice for failing to fulfill its Treaty obligations (Art. 232 EC). Although neither the Parliament nor the Council can force the Commission to make a proposal with specified content, the Commission is obliged to act, whereupon the relevant question becomes whether subsequent actors have the right to amend such an induced proposal. The answer to this question is affirmative, therefore, an apt summary is that the commission has monopoly proposal rights, but it does not have formal gatekeeping rights. This procedural fact has practical consequences, too. Since the late 1980s the EU has informally been operating on the basis of annual legislative programs, assembled via negotiations between representatives of all three institutions—Commission, Council and Parliament—each of which has influence. The Council Presidency, in particular, plays an increasingly important role in this regular process. In short, most EU legislation is not initiated unilaterally by the Commission. Finally, the Commission also has the right to change its proposal during the legislative process (Art. 250(2) EC). This procedural feature approximates a traditional veto that, if genuine gatekeeping were to exist, would nullify the consequences of gatekeeping as shown above(note 9). Gatekeeping elsewhere Claim 5. The appearance of gatekeeping rights in other legislatures is not uncommon, but such appearances are often incomplete and inaccurate. The actual existence of such rights in any major legislative body remains to be demonstrated. 20 France. The French status of legislative gatekeeping is the same as in the U.S. Huber explains: “French committees cannot exercise gatekeeping power and are never granted a closed rule, because the government controls the parliamentary agenda (Article 42 of the Constitution” (1992, 678). In other words, French committees do not possess a gatekeeping right. Germany. Also much like the situation in the United States, Germany has a discharge procedure. Johnson explains: “Paragraph 60(3) [of the rules of the Bundestag] contains a clause enabling the originator of a motion referred to a committee to demand a report to the Bundestag after six months. In principle this is intended to prevent committees from ‘burying matters’ referred to them...” (1979, 121).11 Israel . Upon initial consideration, it seems that the Israeli parliament is a viable candidate for genuine gatekeeping. Hazan writes, “Other than political sanctions. . . there is no formal mechanism by which a committee can be forced to deal with a particular bill.” (1998, 171) However, in a personal communication (8/13/99), Hazan notes that an absolute veto right does not exist, because the government can withdraw a bill from committee at any time and reintroduce it in another committee. Italy. Several sources claim that Italian committees wield extraordinary power, up to the point that they can legislate unilaterally. Doring, for example, confirms that Italian committees can enact legislation without the approval of the plenary” (1995b, 657). Though strong, this claim is not inconsistent with our main argument for two reasons. First, a committee that has direct lawmaking rights as described it is a one-stage dictator—not gatekeeper in a two-stage process as in our model— and its optimal strategy is to act by implementing its ideal point. This is a Pareto optimal outcome (even when embedded in a larger game in which, say, the legislature 11 Johnson is somewhat ambiguous about the overall consequences of Bundestag procedure: “...[I]n practice it has not been necessary to make use of this provision, and committees have found more refined ways of holding up progress than crude neglect.” (Johnson 1979, 121.) 21 delegates this right to the committee). Second, there are indications elsewhere that such claims about Italian committees are overstated. DiPalma, for example, notes that controversial legislation is rarely passed through this procedure (1977, 194200). Likewise, on the issue of whether committees in the Italian parliament can impose the status quo even over the wishes of the rest of the parliament or the government, D’Onofrio writes, “If, on the contrary, a majority is solidly behind the Government, no committee would be able to hold the bill without acting on it” (1979, 740). United Kingdom. Although there is some evidence that Great Britain’s parliamentary committees are undergoing a process of institutionalization (Hibbing 1988), British committees are typically ad hoc and usually regarded by scholars as weak relative to those in Congress and in other parliamentary governments. We found neither claims nor evidence that they have a gatekeeping right. Japan. An almost literal form of gatekeeping occasionally occurs in the Diet when committee members “block the doors leading into a committee room thereby preventing access to it, or surrounding a chairman’s raised seat and thereby making it impossible for him to convene the meeting” (Baerwald 1979, 347). Several qualifications are important, however. First, such obstruction is typically executed by committee minorities rather than by a committee majority. Second, voting is not the mechanism for obstruction. Third, and most important, these non-codified tactics are not ultimately effective because, consistent with our prediction, codified institutional features provide procedural mechanisms for addressing them. Specifically, the majority has the option of using a Kuoko Saiketsu, or “forced vote.” This option is often exercised on acrimonious legislation. Norway. Norwegian committees do not have gatekeeping authority over legislation. The Storting does not have to send bills to committee. Further, the Presidium—a six-member body that performs tasks similar to those performed in the U.S. by the Speaker, Rules Committee and parliamentarian—can unanimously 22 vote to initiate the equivalent of a discharge and send a bill to another committee if no recommendations are made. Poland. Committees do not have absolute veto rights, according to Olson et al. (1998, 114). These legislatures exhaust those for which we were able to obtain gatekeepingspecific information.12 However, our concluding generalization is tentative, because the N is small and the sample is not random. As a whole, the empirical evidence on gatekeeping exhibits a consistent pattern that is broadly consistent with our Paretodominance propositions. Although there are several instances in which committees appear to be gatekeepers, and there are some instances in which committees appear also to be powerful, there are no instances in which such power is closely related to a strict gatekeeping right as defined here and as modeled elsewhere. Summary When juxtaposed with the preponderance of quoted excerpts about gatekeeping in the literature review, the information about gatekeeping in the previous section seems conspicuously contradictory. How can so many scholars be wrong, and so few be right? Happily, some reconciliation of positions is possible with reference to the conceptual distinctions addressed at the outset. First, recall that many authors cite as evidence of gatekeeping the fact that many bills die in committees. If their implicit definition of gatekeeping is not structural, as is ours, then this is simply an instance of defining gatekeeping as a behavioral regularity rather than as a codified, constraining institution. Then, rather than a contradiction, what exists is an unfortunate but substantively significant difference in usage of terms. Similarly, other authors combine the evidence of dead bills with a belief that committees are generally powerful and come up with the intuitively additive claim 12 For a number of other countries, we consulted various sources but could not establish one way or another whether committees had gatekeeping rights. 23 that committees have “gatekeeping power.” This, too, does not contradict the procedural facts—it merely ignores them while making a broader but imprecise claim about the nature of committees (or parties) in legislatures, and, likewise while making a reckless inference about gatekeeping. So, again, some of the apparently large discrepancies in the literature may be attributed to semantic differences. They are not, however, trivial differences. On the contrary, an improved understanding of the interplay between political institutions and political behavior seems, fundamentally, to rest upon a sharp distinction between institutions and behavior when developing models that effectively predict behavior (initially) and institutional design (eventually). An item that remains on the research agenda, therefore, is: why do we see so much behavior that deceptively appears as if some player(s) have a codified institutional gatekeeping right? The judicious use of formal models—even simple ones—provides a foundation for answering such questions. By demonstrating theoretically that gatekeeping procedures are Pareto-dominated, and by demonstrating empirically that codified gatekeeping is rare-to-nonexistent as predicted, future studies can benefit from a new and healthy skepticism toward old and perhaps ailing institutional explanations. 24 References Aldrich, John H. 1994. A Model of a Legislature with Two Parties and a Committee System. Legislative Studies Quarterly 19: 313-40. Bendiner, Robert. 1964. Obstacle Course on Capitol Hill. Toronto: McGraw-Hill. Calabresi, Guido and A. Douglas Melamed. 1972. Property Rules, Liability Rules, and Inalienability: One View of the Cathedral, Harvard Law Review 85: 10891114. Cox, Gary W., and Mathew D. McCubbins. 1993. Legislative Leviathin. Berkeley: UC Press. Cox, Gary W., and Mathew D. McCubbins. 2004. Setting the Agenda. Typescript: UCSD. Crombez, Christophe. 1996. Legislative Procedures in the European Community. British Journal of Political Science 26: 199-228. Dodd, Lawrence, and Richard Schott. 1979. Congress and the Administrative State. New York: John Wiley. D’Onofrio, Francesco. 1979. Committees in the Italian Parliament. In Committees in Legislatures: A Comparative Analysis, eds. John D. Lees and Malcolm Shaw. Durham, NC: Duke University Press. Denzau, Arthur T., and Robert J. Mackay. 1983. Gatekeeping and Monopoly Power of Committees: An Analysis of Sincere and Sophisticated Behavior. American Journal of Political Science 27:740-761. Diermeier, Daniel, and Keith Krehbiel. 2003. Institutionalism as a Methodology. Journal of Theoretical Politics 123-44. Dion, Douglas, and John D. Huber. 1996. Procedural Choice and the House Committee on Rules. Journal of Politics 58:25-53. DiPalma, Giuseppe. 1977. Surviving Without Governing. Berkeley: UC Press. Doring, Herbert. 1995. Is Government Control of the Agenda Likely to Keep ’Legislative Inflation’ at Bay?. In Parliaments and Majority Rule in Western Europe, ed. Herbert Doring. Frankfurt: Campus Verlag. Edwards, Geoffrey, and David Spence. 1994. The Commission in Perspective. In The European Commission, eds. Geoffrey Edwards and David Spence. London: Catermill International Ltd. 25 Ferejohn, John, and Charles Shipan. 1990. Congressional Influence on Bureaucracy. Journal of Law, Economics, and Organization 6: 1- 20. Fitzmaurice, John. 1988. An Analysis of the European Community’s Cooperation Procedure. Common Market Law Review 4: 389-400. Franchino, Fabio. 1999. The Determinants of Control of Commission’s Executive Functions. European Integration Online Papers 3: 1-27. Gilligan, Thomas, and Keith Krehbiel. 1987 Collective Decision-Making and Standing Committees: An Informational Rationale for Restrictive Amendment Procedures. Journal of Law, Economics, and Organization 3: 287-335. Hazan, Reuven. 1998. Political Reform and the Committee System in Israel: Structural and Functional Adaptation. Journal of Legislative Studies 4(1): 163-187/ Hibbing, John R. 1988. Legislative Institutionalization with Illustrations from the British House of Commons. American Journal of Political Science, 32(3): 681712. Huber, John D. 1992. Restrictive Legislative Procedures in France and the United States. American Political Science Review 86:675-687. Huxtable, Phillip A. 1994. Incorporating the Rules Committee: An Extension of the Ferejohn/Shipan Model. Journal of Law and Economics 10:160-167. Janda, Kenneth, Jeffery M. Berry, and Jerry Goldman. 1995. The Challenge of Democracy: Government in America. 4th ed. Boston: Houghton Mifflin. Johnson, Nevil. 1979. Committees in the West German Bundestag. In Committees in Legislatures: A Comparative Analysis, eds. John D. Lees and Malcolm Shaw. Durham, NC: Duke University Press. Johnson, Paul E., Gary J. Miller, John H. Aldrich, David W. Rohde, and Charles W. Ostrom, Jr. 1994. American Government: People, Institutions, and Policies. 3rd ed. Boston: Houghton Mifflin. Keefe, William J., and Morris S. Ogul. 1993. The American Legislative Process. 8th ed. Englewood Cliffs, NJ: Prentice- Hall. Kingdon, John W. 1989. Congressmen’s Voting Decisions. 3rd ed. Ann Arbor, MI: University of Michigan Press. Kollman, Ken. 1997. Inviting Friends to Lobby: Interest Groups, Ideological Bias, and Congressional Committees. American Journal of Political Science 41:519544. 26 Krehbiel, Keith. 2005. Pivots. In Barry Weingast and Donald Wittman eds., Oxford Handbook of Political Economy (forthcoming). Lenaerts, Koen. 1991. Some Reflections on the Separation of Powers in the European Community. Common Market Law Review 28: 11-35. Lenaerts, Koen, and Piet Van Nuffel. 1999. Constitutional Law of the European Union. London: Sweet and Maxwell. Maltzman, Forrest. 1997. Competing Principles: Committees, Parties, and the Organization of Congress. Ann Arbor: University of Michigan Press. McCubbins, Mathew, Roger Noll, and Barry Weingast. 1994. Legislative Intent: The Use of Positive Political Theory in Statutory Interpretation. Journal of Law and Contemporary Problems 57:3-37. Moravcsik, Andrew. 2002. In Defence of the ‘Democratic Deficit’: Reassessing Legitimacy in the European Union. Journal of Common Market Studies 40: 603-24. Moser, Peter. 1996. The European Parliament as a Conditional Agenda Setter: What are the Conditions? A Critique of Tsebelis (1994). American Political Science Review 90: 834-8. Nurmi, Hannu, and Tommi Meskanen. 1999. A Priori Measures and the Institutions of the European Union. European Journal of Political Research 35: 161-79. Oleszek, Walter J. 2004. Congressional Procedures and the Policy Process, 6th Edition. Washington, D.C.: CQ Press. Olson, David M., Ania Van Der Meer-Krok-Paszkowska, Maurice D. Simon and Irena Jackiewicz. 1998. Committees in the Post- Communist Polish Sejm: Structure, Activity, and Members. In The New Roles of Parliamentary Committees, eds. Lawrence D. Longley and Roger H. Davidson. London: Frank Cass. Rogers, Lindsay. 1926. The American Senate. New York: F.S. Crofts. Romer, Thomas, and Howard Rosenthal. 1978. Political Resource Allocation, Controlled Agendas, and the Status Quo. Public Choice 33:27-43. Roth, Alvin E. 1995. Bargaining Experiments. Hanbook of Experimental Economics, John H. Kagel and Alvin E. Roth, editors. Princeton, NJ: Princeton University Press. Schulz, Heiner, and Thomas König. Institutional Reform and Decision-Making Efficiency in the European Union. American Journal of Political Science 44: 653-66. 27 Segal, Jeffrey A. 1997. Separation-Of-Powers Games in the Positive Theory of Congress and Courts. American Political Science Review 91:28-44. Shepsle, Kenneth A., and Barry R. Weingast. 1984. When Do Rules of Procedure Matter? Journal of Politics 46:206-221. Shepsle, Kenneth A., and Barry R. Weingast. 1984. The Institutional Foundations of Committee Power. American Political Science Review 81:85-104. Shipan, Charles R. 1998. Political Influence, Regulatory Regimes, and the FDA. University of Iowa: Typescript. Snyder, James M. Jr. 1992b. Gatekeeping or Not, Sample Selection in the Roll Call Agenda Matters. American Journal of Political Science 36:36-39. Snyder, James M. Jr. 1992a. Committee Power, Structure-Induced Equilibria, and Roll Call Votes. American Journal of Political Science 36:1-30. Snyder, James M. 1992. Committee Power, Structure-Induced Equilibria, and Roll Call Votes. American Journal of Political Science 36: 1-30. Steunenberg, Bernard. 1994. Decision Making Under Different Institutional Arrangements: Legislation by the European Community. Journal of Institutional and Theoretical Economics 150: 642-69. Steunenberg, Bernard, Dieter Schmidtchen and Christian Koboldt. 1999. Strategic Power in the European Union: Evaluating the Distribution of Power in Policy Games. Journal of Theoretical Politics 11: 339-366. Tsebelis, George. 1994. The Power of the European Parliament as a Conditional Agenda Setter. American Political Science Review 88: 128-42. Weingast, Barry R., and Mark J. Moran. 1983. Bureaucratic Discretion or Congressional Control? Regulatory Policymaking by the Federal Trade Commission. Journal of Political Economy 91: 765-800. Weingast, Barry R., and William Marshall. 1988. The Industrial Organization of Congress. Journal of Political Economy 96: 132-63. Westlake, Martin. 1995. The Council of the European Union. London: Catermill International Ltd. Wilson, Woodrow. 1956 [1885]. Congressional Government. Baltimore: Johns Hopkins Press. 28 Outcome gatekeeping game l veto game 2c-l c Pareto optimal Outcomes I II 2c-l III c IV l Status quo Interval of Pareto dominance Figure 1. Outcomes and Pareto Dominance in the Gatekeeping and Veto Games Region of status quo Outcome of gatekeeping game Outcome of veto game I l l II q h(q) III q q C and L’s preferences for the gatekeeping (G) and veto (V) games C and L are indifferent between G and V C is indifferent and L strictly prefers V to G C and L are indifferent between G and V l Region I q1 h(q1) h(q2) c Region III P: Set of q2 Pareto optimal outcomes Region II Figure 2. Extension to Two Dimensions