The Political Economy of Private Legislatures

t Sterling Professor of Law, Yale Law School; Professor, Yale School of Management. : Dean and Lewis F. Powell, Jr. Professor of Law, University of Virginia. Professor Schwartz was an Associate Reporter for the American Law Institute's Enterprise Liability Project. Dean Scott is a member of the Permanent Editorial Board Study Group considering revisions to UCC Article 9 and has been a member of the UCC Article 2 Drafting Committee. The views he expresses here are his own and do not reflect the views of the Study Group. Dean Scott is a member of the American Law Institute. Professor Schwartz had been a member from 1985 until this year. This Article benefited from workshops held at the Iowa College of Business, the University of Virginia Legal Studies Workshop, and the Yale School of Management. We also are thankful for comments from Kenneth Abraham, Bruce Ackerman, Barry Adler, Ian Ayres, Richard Craswell, Hanoch Dagan, Andrew Daughety, Gillian Hadfield, Henry Hansmann, Geoffrey Hazard, Lewis KornhauserJohn Langbein, Saul Levmore, Edward McCaffery, Jeffrey O'Connell, Uriel Proccaccia, 'Jennifer Reinganum, Peter Schuck, Steven Schwarcz, Matthew Spitzer, and Paul Stephan.


INTRODUCTION
A large amount of American law originates with private lawmaking groups. This Article considers two of these groups, the American Law Institute (the "ALI") and the National Conference of Commissioners on Uniform State Laws ("NCCUSL"). The ALI is a self-perpetuating organization of lawyers,judges, and academics. Its primary function is to promulgate restatements of law. These restatements are sets of rules, organized by subject matter, the content of which is partly a function of the case law but also is a function of the ALI's collective view respecting which legal rules are normatively desirable for courts to apply. Restatement rules do not have binding force but are advisory to courts. Inclusion of a rule in a restatement, however, is widely thought to increase the likelihood that courts will follow it.
The National Conference of Commissioners on Uniform State Laws is an organization of "commissioners," each of whom is appointed by the governor of her state. These commissioners are also lawyers, judges, and academics. NCCUSL creates statutescalled "uniform laws"-that it recommends to state legislatures. The states often reject NCCUSL recommendations, but when they do accept them, they commonly enact the NCCUSL statutes as written. The ALI and NCCUSL jointly created America's longest and most influential commercial statute, the Uniform Commercial Code (the -UCC-).
The ALI and NCCUSL currently are engaged in a major revision of the UCC. This process has already resulted in the revisions of UCC Articles 3 and 4 (as well as the promulgation of Article 4A), the recommended repeal of Article 6, and the addition of Article 2A. Completed revisions to Articles 5 and 8 are imminent. The two revisions that remain are especially important events in commercial law: Article 2, regulating sales, has never been revised, and Article 9, regulating secured lending, was last rethought in 1972.
The ALI and NCCUSL have created large portions of American contract and commercial law and have made major contributions in other areas as well, such as tort and property law. Despite these groups' significance, they have never been seriously studied. Rather, uniform laws and restatements have been evaluated as if they were produced by rule-generating "black boxes." Lawyers know that these boxes can produce bad laws as well as good ones, so serious critical attention is devoted to ALI and NCCUSL products. In contrast, the relation between the institutional structures of these organizations and the rules they adopt has been entirely neglected. This lack of attention apparently is because ALI and NCCUSL members are thought to be disinterested legal experts who pursue only the public good: the task is not to study this ideal but rather to extend it to other areas.
The legal community's inattention to the consequences of ALI and NCCUSL procedures, however, is unjustified. Positive political theory teaches us that the form and substance of a law are significantly endogenous to the law-creating institution. Put more simply, a legislature's output is a function both of the preferences of the legislators, whether selfish or altruistic, and of the institutional structure in which the legislators perform. Thus, ALI and NCCUSL outputs also should be endogenous to their organizational forms.
This Article uses the tools of "structure-induced equilibrium" theory' to study the ALI and NCCUSL. These tools were developed to study typical legislatures, but their use is apt here because the ALI and NCCUSL actually do function as private legislatures. The analyst doing structure-induced equilibrium theory identifies the utility functions that participants in the legislative process maximize, specifies the institutional structures that transform participant preferences into legislative outcomes, and then shows what outcomes these preferences and structures will produce.' When this method is applied to an institution that functions as the ALI and NCCUSL do, it reveals that the institution (a) has a strong status quo bias that induces it to reject significant reform; (b) frequently produces highly abstract rules that delegate substantial discretion to courts; and (c) produces clear, bright-line rules that confine judicial discretion commonly when and because dominant interest groups influence the process. These bright-line rules ordinarily advance the interest group's agenda. 8 This Article is primarily positive; its goal is to understand how large private law-making groups such as the ALI work. However, the results we develop do have several normative implications. First, when nationally uniform regulation of a subject is desirable, desirable, the ALI president appoints a "study group" to prepare a report. This action is taken on the recommendation of the director, who consults with NCCUSL and the ALI council; the latter body must approve study group membership. A study group commonly has an academic reporter and other academics as well as practitioners, but the division of authority differs from restatement practice: a UCC study group as a whole, rather than the academic reporters, has the final say as to a draft report's contents. According to the ALI Director, Geoffrey Hazard, the principal criteria governing the composition of a study group are expertise in the subject, credibility with the ALI council, and an appropriate mix of academics and practitioners.
A study group's report is sent to the ALI and NCCUSL, where it proceeds on parallel tracks. The ALI commonly approves study group reports, but approval does not exhaust the ALI's function. Thereafter, the ALI is consulted about the composition of NCCUSL drafting committees and must approve proposed changes to the Code. NCCUSL, in turn, is responsible for putting reports in the form of statutes and for recommending these drafts to legislatures. According to Hazard, study group reports are very influential in determining the content of proposed UCC legislation: "That is the whole idea" of doing reports.

B. NCCUSL
The Conference creates uniform laws that it proposes to state legislatures. It also creates model acts when it wishes to influence the direction of law reform but believes that an area is too unsettled to support the adoption of a uniform law. The Conference is composed of "commissioners" who are appointed for three-year terms by the governors of their states. Commissioners are lawyers, judges, and academics. 8 Appointment as a commissioner is regarded as nonpolitical, and commissioners commonly are reappointed. A state can have as many commissioners as it likes, but voting at national meetings is done by state on important issues, including whether to approve a proposed uniform law. Each state has one vote, which reflects the choice of a majority of the state's 'The NCCUSL constitution requires commissioners to be members of the bar. 400 (1994) [hereinafter HANDBOOK] (citing NCCUSL constitution, § 2.4). There currently are more than 300 commissioners. See id. at 1-11.

602
UNIVERSITY OF PENNSYLVANIA LAW REVIEW [Vol. 143:595 commissioners. The Conference meets annually for a week or less to review the laws and model acts that its committees create. 9 An academic reporter for a recent project described NCCUSL as "a state-sponsored and -funded nonpolitical organization dedicated to promoting improvement and uniformity in state law."" 0 When the Conference decides that a uniform law is advisable, it appoints a drafting committee." 1 According to prominent commissioners, members of drafting committees "are selected to achieve a breadth of experience and perspective." 12 Academics commonly are reporters for these committees, but "real lawyers" are in charge;" 8 that is, the lawyer commissioners and reporters have the final say concerning the content of any proposal that is forwarded to NCCUSL commissioners for discussion and adoption. 14 Despite this real-world emphasis, NCCUSL has difficulty getting its products enacted. Of more than two hundred proposed uniform acts, 107 have been adopted in fewer than ten states; seventy-seven have been enacted in less than five.' 5 The Conference has had its greatest success in the commercial area. Of the twenty-two acts adopted in more than forty states, nine have been commercial.' 6 At least four others, such as the Uniform Act to Secure the Attendance of Witnesses from Without a State in Criminal Proceedings and the Uniform Durable Power of Attorney Act, have dealt uncontroversially with the legal system itself. 17 The Conference also has been moderately successful in the probate and trust field.'"

C. The ALl and NCCUSL Approach
The ALI and NCCUSL take a similar intellectual approach. Both groups purportedly prefer to deal with technical issues that legal expertise can resolve, not matters whose resolution requires controversial value choices or would be aided by social science or philosophical skills. 1 9 Consequently, both groups perceive themselves as doing technical (in the old-fashioned phrase "scientific") work, not policy analysis, and both prefer nationally uniform solutions. 20 " The ALI's approach is best articulated in the report that was the basis of its founding and that remains influential today. That report recited: The fact that a man is a lawyer does not make him an expert on the tariff •.. or enable him to speak with authority on hundreds of other questions of existing or proposed law debated ... in legislative assemblies. It is therefore ... vital to the permanent usefulness of the proposed organization that we regard its activities as restricted to improvements in existinglaw ... in relation to those subjects of which the legal profession has expert knowledge.
It is the province of... legislative bodies.., to express the political, economic and social policies of the nation .... The proposed organization should concern itself with such matters as... the details of private law.... It should not promote or obstruct political, social or economic changes .... ... Changes in the law which are, or which would, if proposed, become a matter of general public concern and discussion should not be considered, much less set forth, in any restatement of the law such as we have in mind. Changes which.., will carry out more efficiently ends generally accepted as desirable are within the province of the restatement to suggest. Report of the Committee on the Establishment of a Permanent Organization for the Improvement of the Law Proposing an American Law Institute, 1 A.L.I. PROC. 1, 4-5, 15 (1923) [hereinafter Report Proposing the ALI].
NCCUSL also claims to adopt this approach. Knowledgeable observers commonly say that NCCUSL takes a conservative approach to law reform, avoiding "novel" areas. Its current guidelines for creating uniform laws recite that the commissioners should avoid subjects that are "entirely novel" and "controversial because of disparities in social, economic or political philosophies among the various states." HANDBOOK, supra note 8, at 440. 2 1 See Report Proposing the ALl, supra note 19, at 35 (stating that the proposed organization [the ALI] will do "constructive scientific juristic work"); id. at 65 (concluding that "[tjhe work of the proposed Institute is analogous.., in respect to the character of the labor involved.., to those institutions founded to investigate the cause of disease"). The successful motion to found the ALI, made on the basis of the Report, recited that the ALI's object "shall be to promote the clarification and simplification of the law and its better adaption to social needs .. and to encourage and carry on scholarly and scientific legal work." William D. Lewis These positions can be summarized in an "official" ALI and NCCUSL view: public-spirited lawyers volunteer to resolve important legal issues in technically correct and politically uncontroversial ways; their results are embodied in restatements and proposed uniform laws that courts should follow or state legislatures should enact as written. 2 This preference prevailed for the UCC; most states did adopt it largely as it came from the ALI and NCCUSL. 2 2 Despite the clarity of the intellectual approach propounded by the ALI and NCCUSL, however, the evidence as exemplified in the UCC itself and other laws and restatements shows that both groups often fail to act in accordance with their animating ethos. These institutions do venture into areas where values conflict and traditional legal expertise is insufficient to generate effective solutions to the problems at hand.

D. A Typology of Rules Produced by the ALI and NCCUSL
In order to evaluate the products of private law-making groups such as the ALI and NCCUSL, we begin with a taxonomy of the types of legal rules such groups produce. Legal rules commonly take one of three forms. These forms are conveniently illustrated by considering the task of creating a traffic rule regulating the speed of automobiles. A Model 1 traffic rule would recite that driving more than X miles per hour is prohibited. Such a rule is binary, in [The Commissioners] are an elite group. Most of the Commissioners are prominent lawyers... chosen because they have a more intellectual interest in uniform law' than would a typical legislator....

...
[T]he principal argument that the Commissioners can make on behalf of a uniform law when it is considered by a state legislature is its technical and substantive superiority over a law born in the back room of a state legislature and sired by a lobbying organization. White, supra note 15, at 2096-97. 21 See, e.g., Miller et al., supra note 12, at 1453 ("It is the position of the Conference that uniform laws should ordinarily be adopted 'as written,' except in unusual circumstances where minor changes may be necessary to conform with paramount principles of local policy, law, and conditions."); DonaldJ. Rapson that it purports to distinguish between compliance and noncompliance with a single criterion. The rule also is "bright-line" because it restricts the set of information on which application turns to objective facts-X miles per hour-and thus is relatively easy to apply.
Model 1 rules have the virtue, from the viewpoint of the legislature, of confining the rule applier's discretion (assuming that the rule applier actually enforces rules as written; the discretionary enforcement by police of the fifty-five-mile-per-hour speed limit illustrates the difficulty of constraining discretion even when a rule is clear on its face). The vice of Model 1 rules is their crudeness: such rules are invariably both under-and over-inclusive. Thus, a Model 1 rule can influence behavior only imperfectly." Model 2 rules are written on a higher level of abstraction. For example, a Model 2 rule would require persons to drive in a "reasonable manner." Such abstract statements vest more discretion in the rule applier than Model 1 rules permit. A possible consequence of not articulating the specific criteria for application of the rule in the rule itself is an increased risk of "misapplication" (the rule may be applied in a manner inconsistent with the intent of the rule maker). 2 4 On the other hand, some believe, stating the underlying norm (in this case "reasonable driving") in the rule itself can increase the likelihood that the rule will be enforced in accordance with its animating purpose. 25 A Model 3 rule attempts to find a middle ground between the first two. Such a rule both includes and then purports to illuminate the underlying norm that is set out in a Model 2 rule. As an example: Persons must drive in a manner reasonable under the 23 See ROBERT E. ScoTr & DOUGLAS L. LESLIE, CONTRACT LAW AND THEORY 339-41 (2d ed. 1993) (discussing features of precise or per se rules and context-sensitive or multifactored rules). A potential defendant who predicts that a Model I rule will result in her conduct being subject to sanction may be deterred from socially desirable activity (such as transporting a sick friend to the hospital in time to secure treatment). On the other hand, a potential defendant who incorrectly predicts that the rule will exonerate her conduct may engage in socially undesirable activity (such as driving too fast in a snow storm). For more on the appropriate "fit" of Model 1 rules, see Isaac Ehrlich & Richard A. Posner, An Economic Analysis of Legal Rule-Making, 2 J. LEGAL STUD. 257, 268 (1974) ("Greater specificity of legal obligation generates allocative inefficiency as a result of the necessarily imperfect fit between the coverage of a rule and the conduct sought to be regulated."). 214 See Scorr & LESLIE, supra note 23, at 341 (explaining that because there are more "relevant" facts that must be determined before the standard of "reasonable behavior" can be applied, the rule applier may make an "error" with respect to one or more facts).
' See e.g., id. circumstances. The factors relevant to assessing reasonableness include (1) pavement conditions, (2) number of cars on the highway, (3) the speed of other cars on the highway, (4) visibility conditions, (5) eyesight and reflexes of this driver generally, and (6) the condition of the driver at the time.
The multifactored character of Model 3 rules may induce a better fit between rule application and the underlying norm than is possible with Model 1 or Model 2 rules. Moreover, the enumeration of application criteria in the rule itself apparently confines the rule applier's discretion. Thus, it is tempting to conclude that Model 3 rules are "better" than either polar alternative. Model 3 rules, however, will be as imprecise as Model 2 rules whenever the listed factors nearly exhaust the relevant possibilities and the rule maker does not attach weights to these factors or otherwise specify the relationship among them. If the weights are not specified in the rule, the rule in practice will differ little from a Model 2 rule, because the factors that determine the Model 3 rule's application may point in different directions. The Model 3 rules that have been designed for commercial law and contracts commonly contain large lists of unweighted factors. 26 For convenience, therefore, we conflate the two categories and refer to both Model 2 and Model 3 rules as either Model 2 rules or "vague rules." A legislator ideally should weigh the costs and gains of precision in rule statement when choosing the appropriate rule form in which to implement a policy, but legislators also consider the effect that rule form will likely have on the status quo. A Model 2 rule transfers much of the law-making authority from the rule maker to the rule applier whenever the appliers of the particular rule are numerous, diverse, and difficult for the rule maker to control. The state courts that apply ALI and NCCUSL products are numerous, diverse, and impossible for these law-making bodies to control. NCCUSL, for example, has no power to overrule a state court that gives content to a statutory phrase such as "reasonable" in a way that NCCUSL did not intend. It is tempting to conclude, therefore, that ALI and NCCUSL Model 2 rules do not change the status quo. Rather, these rules appear merely to confer on state courts a power to make law that those courts already possess. This view is too strong, however. Putting NCCUSL's stamp of approval on a vague rule may cause some courts to be less reluctant to regulate an area than they had been previously. Moreover, the choice of general predicates may occasionally matter. For example, requiring persons to drive "safely" may induce different judicial results than requiring persons to drive "reasonably." 2 7 Notwithstanding these caveats, a legislature will change the status quo much more significantly when it enacts a Model 1 rule than when it enacts a Model 2 rule. The clarity and salience of a Model 1 rule will influence the rule applier's decisions more significantly than if the rule merely directed the applier to implement underlying norms as she conceives of them. On the other hand, Model 2 rules grant broad discretion to the rule applier and permit her to make only modest changes in the status quo or none at all. We show below that the effect a rule will have on the status quo is a much more important factor in influencing the choice-ofrule form in the ALI and NCCUSL than the virtues and vices of precision in rule statement.

A. The Game and Its Outcomes
In this section, we develop a theory of private law-making groups such as the ALI and NCCUSL. These groups function similarly to legislatures. Rules are first proposed in committees that are dominated by members with technical expertise. The initial committee process results in a blueprint for reform that is delivered to a second committee which casts the blueprint into statutory form. The final product is then put to the larger body for a vote. 28 The models developed below incorporate the salient features of such a private legislature (or "PL").
Three types of participants act in PLs: (1) interest groups, such as banks; (2) reformers, such as law professors; and (3) PL members who attend annual meetings but do not participate in creating 27 In other words, vague rules can be transformative when they are cast in the form of attractive principles. It is widely believed, for example, that UCC § 2-302, which authorizes courts to ban "unconscionable" contract terms, has been transformative in this sense. This belief has never been systematically tested. 28 See Robert E. Scott, The Politics of Article 9, 80 VA. L. REV. 1783, 1803-06 (1994) (providing an informal model of the ALI and NCCUSL as private legislatures, with particular application to the Article 9 revision process). proposals. 9 We model the interaction of these participants as a single-shot stage game. First, a reformer or interest group member requests that a PL draft a proposed law. Second, the PL leadership decides whether the PL should create a study group to consider the matter. The probability that a PL will go forward increases with the likelihood of widespread enactment because creating laws is costly. We consider here the set of legal subjects that satisfies this "enactment constraint"-that is, the set of subjects the uniform laws of which are likely enough to be adopted to justify PL action. 30 Third, the study group creates a uniform law and reports its proposal to the PL under a closed rule.3' A closed rule is assumed because the size of PLs and the limited time they can devote to particular proposals precludes significant floor amendments. Fourth, the PL decides whether to adopt the proposal or reject it. 3 2 Fifth and finally, a proposal that passes is introduced as a bill in state legislatures. s The enactment constraint that purportedly operates at the second stage is seldom binding on PLs for three reasons. First, PLs exist to do law reform. This generates pressure to create uniform laws and model rules.
Second, the probability of widespread adoption is difficult to assess when a law-reform subject is proposed, because it would entail predicting the behavior of a large number of state legislatures and courts. Moreover, there is an extensive time lag between the decision to consider a subject and the promulgation of a law. For example, the UCC took twenty years to "A reformer or an interest group member also can be a member of a PL. Most PL member/legislators do not have a direct economic or policy-oriented interest in any particular proposed law. Thus, it is convenient to treat these three participant groups as if they were distinct. 50 We explain below why this enactment constraint is seldom binding on PLs. "Study groups in actual PLs almost always submit proposals to the larger body for a vote. Under a closed rule, the legislature can pass or reject a proposal but cannot amend it on the floor. " A PL can remand a proposal to the study group for further consideration. Remands delay final determinations but do not preclude them. Hence, no generality is lost by analyzing only the ultimate decision whether to accept or reject.
" When the ALI creates a restatement, the stage game differs in three ways from the model above. First, a reformer or interest group member does not have the option of avoiding the PL and introducing a proposed restatement directly into a legislature because legislatures do not enact restatements. The analogue to initial legislative introduction is to use test cases to establish the law. Second, and as a consequence of the first point, the ALI process ends with promulgation of a restatement. There is no later legislative stage. Third, the typical ALI committee that drafts a restatement has more academics and fewer practitioners than the typical NCCUSL study group.

PRIVATE LEGISLATURES
create. Such long delays make prediction difficult. As a consequence, the pressure to generate proposals is hard to defuse even with a plausible claim that a particular proposal has a low enactment probability. Third, reformers, we will argue, view participation in PLs as a gain and do not view the probability of ultimate enactment of a PL proposal as a constraint. Hence, PLs have a large supply of free labor to consider subjects and this inclines them to resolve doubts in favor of creating proposals.
The low enactment rate of NCCUSL uniform laws is consistent with this analysis. Because the enactment constraint is weak, we do not analyze the legislative role; that is, we suppose that the decision of a PL whether to accept or reject a proposal is not importantly influenced by the likelihood that the proposal will be widely enacted. 4 Our models of this game, developed and analyzed in the remainder of this Article, yield the following results: (1) When all of the participants in the PL process are symmetrically informed about the consequences of reform proposals, (a) a PL will produce many vague rules; (b) the ALI will produce more vague rules than NCCUSL; and (c) an interest group will attempt to participate on PL study groups and advisory boards and seek to generate Model 1 rules when its preferences over reform are similar to those of the PL itself, but interest groups will not lobby PL legislators.
(2) When information is asymmetric (the typical PL member/ legislator is poorly informed about the consequences of proposals while other participants know these consequences), the following can be expected: (a) the penchant of a PL to produce vague rules is either unaffected or may increase; (b) when only one interest group would be affected by a PL reform proposal, the group will attempt to participate in study groups and may also lobby, thereby having a greater effect on PL outcomes than on ordinary legislative outcomes; 3 5 (c) an interest group may have less power in PLs than in ordinary legislatures when interest groups compete over a reform proposal; (d) as a consequence of (c), the presence of competing interest groups before a PL should be less common than the presence of a single group or no group at all, but a reasonable G'Cf. Ribstein & Kobayashi, supra note 5, at 40-46 (analyzing legislative reception of NCCUSL projects).
" In the sense that it can get more Model 1 rules that advance its goals.

610
UNIVERSITY OF PENNSYLVANIA LAW REVIEW [Vol. 143: 595 equilibrium exists in which interest groups will compete; and (e) the institutional bias of PLs to behave conservatively, eschewing significant reform, is reinforced.

B. The Participants' Utility Functions and the Information Structure
This section first describes the preferences of PL participants informally and then gives a formal representation. An interest group can only lobby an ordinary legislature; the group cannot officially participate in the process of creating legislation. In contrast, an interest group can participate in the process of creating PL legislation by having its members, supporters, or lawyers participate in PL study groups and on advisory boards.
An interest group has substantial incentives to participate in these ways. Legislatures may be more likely to pass a reform if it comes with the approval of the PL. Moreover, participation in the PL may reduce the group's total lobbying costs. 6 Finally, the payoff from successfully lobbying a PL can exceed the payoff from lobbying Congress or federal agencies because uniform laws are hard to alter. For example, suppose that the Federal Reserve Board can be persuaded to adopt a regulation favoring banking interests and that NCCUSL can be persuaded to propose a similar uniform law that would be widely adopted. The banks would prefer the uniform law because an administrative regulation can be repealed with a change of agency membership, while the approval of numerous states would be required to repeal the uniform law. 37 Reformers differ from interest group members in two relevant ways. First, a reformer derives utility from a PL's passage of a reform proposal independently of whether the proposal ultimately becomes law. The reformer is commonly a law professor who can write about and teach PL proposals because the proposals are plausible candidates for becoming law.
Moreover, status in academia attaches to one who causes a PL to adopt a proposal. In contrast, an interest group member derives utility from PL adoption of a proposal only when adoption increases the likelihood or reduces the costs of securing ultimate, stable legislative enactment. Second, participation in a PL is consumption to a reformer; she gets -' This would occur if the PL's own lobbying (if conducted) or the PL stamp of approval caused legislatures to require less persuading. " See White, supra note 15, at 2103. utility from being part of a law-reform movement. In contrast, participation in a PL and before a legislature is production to an interest group member. Because reformers benefit from PL participation and from PL adoption of a law, they seldom are constrained by legislative preferences. Reformers may not work for proposals that have no chance of becoming law, but almost any probability above zero will induce a reformer to participate in a PL. The third participant is the PL member/legislator. Ordinary legislators are interested in reelection prospects and in policy. PL legislators are not elected and so may be thought to have an interest only in policy, but this assumption is too strong. Law-reform proposals generate both economic and political consequences. An economic consequence is either the substantive effect a proposal would have "in the world" or the direct effect the proposal would have on a PL participant's income. As an example of economic consequences, a proposal to cap tort damage awards may reduce deterrence, shift wealth from the seriously injured to injurers, and reduce the income of the plaintiffs' lawyers who are PL participants. Commonly, only a small minority of PL participants have a direct economic stake in the fate of any proposal.
A political consequence refers to the effect that a proposal could have on a PL participant's reputation. Support for a proposal that is regarded as utopian or foolish could damage a participant's reputation for good judgment. A lawyer's income is a function of his reputation, but reputational effects are sufficiently influential to distinguish them from direct economic effects. Both the economic and political effects of a proposal will influence how PL legislators vote.
PL participants also derive utility from having a PL adopt a lawreform proposal, independently of the merits of the proposal in question. This is because participation in law-reform organizations implies the desire to do at least some law reform. Consequently, PL participants are inclined to reject the status quo. A participant nevertheless may prefer the status quo to any particular reform proposal that is presented to her, all things considered. We follow standard political science practice and assume that the participant preferences just described can be arrayed on a single linear dimension and that those preferences are "single peaked." Consider Figure 1.
only if y is closer to x than v is. 8 Reformers and interest group members have utility functions of the same form, but their ideal points may differ from those of typical PL participants. The players' ideal points are assumed to be exogenous; that is, a participant's preferences in policy space are formed before she plays the stage game. This means that while debate can influence which proposed law a participant will support, debate will not influence, for example, the participant's relative preference between stimulating commerce and protecting the environment. It is customary in political science models to assume risk aversion in order to motivate the legislator's desire to create institutions that reduce uncertainty concerning the effect of proposed laws. The legislators analyzed here (PL voters) cannot create such institutions (to do so would entail major constitutional reform in a PL). The voters, however, remain interested in the location of proposals (that is, their consequences). Assuming risk neutrality to analyze this concern does not affect the model's qualitative results. The effect of risk aversion on PL behavior is briefly considered infra note 57.
This assumption is made both because participant preferences seem exogenous in PLs such as NCCUSL and the ALI and because there is no endogenous mechanism for changing preferences. The median voter's ideal point in an ordinary legislature can be affected by logrolling, but there is no cross-subject logrolling in the ALl and NCCUSL. Consequently, debate in a PL can only illuminate the relation between proposals and their consequences; it cannot change the preferences that typical (that is, disinterested, nonreformer) participants have over those consequences. Logrolling does not occur initially because reporters and drafting committees are recruited for particular projects and principally concentrate on them: an ALL Article 9 reporter usually is unaware of what the Article 2 reporters are doing, and never consults with them officially." 0 Reporter and drafting committee projects thus are presented to the relevant PL as independent entities.
In order to logroll, therefore, the membership itself would have to condition passage of one project on the passage of another. The structure of the ALL and NCCUSL, however, makes logrolling on the floor impractical. The ALI membership is three times the size of Congress, heterogeneous, not organized in political parties, and meets annually for one week. 4 " NCCUSL also is large, heterogeneous, lacks internal structure, and meets annually for a short time. 41 Hence, it is difficult to make numerous, significant deals on the floor.
Logrolling also is risky because trades are hard to enforce. In an ordinary legislature, committees enforce trades. For example, consider a trade among legislators that would reduce corporate taxes in return for increasing environmental protection. In a later session, legislators who favor high taxes may wish to renege on the trade and repeal the tax reduction. The committee that initiates tax proposals can enforce the original deal, however, because amendments to statutes in its jurisdiction cannot be considered by the larger body without its consent-it is the gatekeeper for tax matters. The parties that appoint committees have reputations to protect, and so will choose committees that act as gatekeepers. Because deals are enforceable, pro-business legislators may be willing to trade with pro-environmental legislators. 2 PLs such as the ALI and NCCUSL lack enforcement power because they have no gatekeeping committees. Reporters and drafting committees are instructed to create single projects and are dismissed when projects are completed." Without gatekeepers, a promise by one NCCUSL or ALI faction to another not to change a proposed uniform law seldom will be credible. Hence, trading on important issues is risky. 4 In sum, if a PL cannot logroll, the assumption that PL participants have exogenous ideal points in policy space is realistic. Voting in the ALI and NCCUSL is by majority rule. This and the assumption of single peakedness imply that the equilibrium outcome of a PL vote will correspond to the preferences of the median PL participant. Thus, we follow standard political science practice in modeling a unicameral legislature in which only the preferences of the median legislator are considered. We also assume that study groups vote by majority rule. Hence, the preferences of the median study group member determine study group outcomes. When the median member is a reformer, we describe the group as a "reformer-dominated study group"; similarly, when the median member belongs to an interest group, the group is referred to as an "interest-group-dominated study group." The assumption that study groups (in the usual case, " Each draft of a NCCUSL study committee goes to a review committee before presentation to the general body. Were there a single standing review committee, it perhaps could arrange and police cross-subject trades. For example, such a committee could prevent a proposed law from coming to a vote if it upset a prior compromise. NCCUSL review committees, however, are appointed act by act and each is directed to "evaluate the draft of an Act proposed for submission"; there is no standing committee with the powers described here. HANDBOOK, supra note 8, at 402, 411 (citing the NCCUSL Constitution article 4, § 4.3, and article 29, § § 29.1, 29.2).
The ALI council reviews proposed laws and restatements before they go to the membership. The council is a standing committee and thus could enforce trades. There is no evidence that it does so.
"" Other evidence suggests that these PLs do not logroll. An overwhelming vote in support of a rule implies the absence of logrolling. This is because a coalition, if involved in logrolling, will try to give up as little as possible when compromising, thereby garnering only a narrow margin of votes necessary to win. Thus, a vote that reflects a trade should be close. Impressionistic evidence suggests that proposals commonly pass in the ALI and NCCUSL by large majorities.

PRIVATE LEGISLATURES
committees) can be modeled in this way is standard and does not affect the results. 4 5 Finally, obtaining information about a proposal's consequences is costly. This implies the possibility of information asymmetries: some players will find it worthwhile to become informed about the consequences of proposals while others will not. An informed player is said to have "expertise." We assume that study group members have expertise owing to their roles. An interest group that lobbies a PL is also assumed to have expertise (that is, to have incurred the cost of evaluating a proposal). In the following discussion, PL performance is analyzed first under the assumption that the typical PL legislator/participant has expertise, and second, under the assumption that she lacks it.

C. PLs Under Symmetric Information
We begin with the assumption that all PL participants are equally (that is, "symmetrically") informed about the consequences of any proposal. We do so in order to explore PL performance under a variety of conditions. It is more plausible to suppose, however, that the typical PL participant knows less than reformers and interest group members do. Moreover, such an asymmetric information model, which is set out below, explains more of the data than the symmetric information model. Nevertheless, the two models share many of the same predictions, a reassuring coherence on the theoretical level.
First, assume that a study group proposes a related set of rules to a PL and that all of the participants are symmetrically informed as to the consequences of adopting those rules. Then, (i) a large number of the rules that pass will be Model 2 rules; (ii) many of the 4 Study group members, as noted above, are chosen because they have expertise and are credible with the larger body. See supra part II.A. These criteria also influence the makeup of congressional committees. But see Keith Krehbiel, Are Congressional Committees Composed of Preference Outliers?, 84 AM. POL. Sci. REV. 149, 149 (1990) (arguing that congressional committees can be more extreme and homogeneous than the legislature as a whole). The policy preferences of potential members sometimes influence appointment to legislative committees, but this criterion seems less significant than the others. See KEITH KREHBIEL, INFORMATION AND LEGISLATIVE ORGANIZATION 108-34 (1991) (discussing the various influences on forming legislative committees). For this reason, we do not pursue the prospect that a PL leadership will appoint a particular committee simply in order to get a particular proposal. There is also little evidence that this happens.
Model 1 rules that pass will advance the interests of industry groups; and (iii) interest groups will not lobby PL legislator/members. Four reasons support result (i). First, the description of the participants' utility functions above implied that a vague rule that leaves the status quo relatively intact is preferred, cet. par., to doing nothing. 46 Second, Model 2 rules are less likely to create reputational losses for participants and may actually create reputational gains. Reputational losses are unlikely because these rules delegate much of the legislative power to courts; thus, a PL participant cannot be embarrassed by the adoption of a Model 2 rule in the way she can be embarrassed by the adoption of a clearly directive Model 1 rule that actually accomplishes something. Reputational gains derive from the fact that Model 2 rules commonly are couched in phrases with positive affect (such as "good faith" or "reasonable") or appear to consider all relevant factors; hence, a participant may be thought to be well-motivated or sage when her PL adopts such a rule. Third, vague rules can create direct economic gains for PL participants. These rules increase or maintain uncertainty, and thus increase, or do not reduce, the occasions on which lawyers will have to give advice or be involved in litigation." Thus, these three reasons imply a PL preference for Model 2 rules, but this preference can be outweighed by a participant's preference for the consequences that a particular Model 1 rule would create.
The fourth reason why vague rules are common in a symmetrically informed world is that, under this condition, study groups often have an incentive not to propose Model 1 rules. As a consequence, the other reasons that induce PL participants to prefer Model 2 rules will go unchallenged. The argument that supports this result requires some elaboration, but it is a significant implication of the model. Moreover, this argument also implies that when study groups do propose Model 1 rules, these rules commonly will advance the interests of industry groups (result (ii)).
To understand why study groups will function in this way (under conditions of symmetric information), it is necessary to 46 See supra part 1l.B. consider when a PL will pass a Model 1 rule. Point x in Figure 2 is the median PL participant's ideal point. The status quo is represented by the point sq. In a symmetricinformation world, the median PL participant would support any rule in the interval (sq, bi: x is equidistant from sq and a proposal whose consequences are reflected in b. Hence, any proposal of a Model 1 rule to the right of sq and to the left of b is preferred to sq by the median participant and would pass. The set of rules in the interval {sq, bi is the "win set." A study group will anticipate the median PL participant's behavior, and will then make one of three kinds of proposals. First, if the study group's ideal point is in the interval {b, z}, the study group will propose a Model 1 rule that is to the left of, but near, b. This rule is better for the study group than sq and is closer than sq to the median PL participant's ideal point; hence, the rule is better for all and will pass. Second, if the study group's ideal point is in the interval (sq, b}, the study group will propose the Model 1 rule that is its ideal point. This rule is in the win set and so also would pass. If the study group's ideal point is to the left of sq (for example, at y), the study group will not propose a Model 1 rule. Any proposal in the win set is further from the group's ideal point than the status quo, so it prefers the status quo; any Model 1 rule outside the win set would not be proposed because it would be defeated.
A study group's best alternative to proposing a losing Model 1 rule often is to propose a Model 2 rule rather than none. To see why this is so, consider the decision of an interest group to participate in the process. In a symmetric-information world, a study group will get its Model 1 rules adopted only if its preferences are close to the preferences of the median PL voter. Interest group members will not incur a participation cost unless the expected gain is greater, that is, unless participation will sufficiently increase the probability that the study group will propose a winning rule. Hence, interest groups participate on study groups if .(a) they will have influence and (b) their preferred policies are in or, in this illustration, to the right of the win set. When these groups do participate, a study group probably will propose Model 1 rules. This rule-type best constrains courts to follow the industry's policy rather than their own. Reformers also want to influence study groups when their preferences are near the preferences of the median PL participant. Reformers, however, prefer anything that can plausibly be called a reform to the status quo. Thus, a reformer will participate on a study group although the most likely consequence of participation is that the PL will adopt Model 2 rules. This consequence actually is likely because, as we argue below, reformers often have ideal points that lie far from those of median PL participants. 4 " Hence, reformers are reluctant to propose Model I rules in a PL's win set. Because reformer-dominated study groups are common, and because these study groups are preference outliers but want "reform," study groups often will propose Model 2 rules. This, then, is the fourth factor inclining PLs to the production of Model 2 rules.

9
The preceding analysis also explains why, in a symmetricinformation world, the Model 1 rules that PLs pass often will advance the goals of interest groups. Industry groups will play only when they can pass Model 1 rules (otherwise, their participation is not cost-justified). Hence, given the incentive of reformers not to propose Model 1 rules, the presence of Model 1 rules in a world of symmetric information is a strong signal of industry influence.
Another testable implication of the conclusion that reformerdominated study groups will propose Model 2 rules is to compare ALI restatements to NCCUSL uniform laws. Unlike NCCUSL uniform laws, the ALI restatement projects do not employ broadly 48 See infra note 51 and accompanying text. 49 Our formal model assumes that the costs of creating Model l and Model 2 rules are the same. This assumption sometimes is strong when a law is broadly applicable. The more heterogenous the parties are to whom the law applies, and the greater the variety of contexts in which the law is to apply, the more convenient it is for the lawmaker to draft on a high level of abstraction. It is less costly for her to tell persons to behave "reasonably" than to draft clear, sensible rules for a large number of contexts. Because some uniform laws and restatements, such as UCC Article 2 (Sales) and the Restatement of Contracts, are broadly applicable, one would expect them to contain at least some vague rules independently of the factors considered in the text. A richer model would explicitly include rule-creation costs in the players' utility functions.
representative study groups, but rather rely on appointed academic reporters and academic assistants to consider subjects." The professors that constitute these bodies commonly have ideal points that differ substantially from those of typical ALI participants. 51 ALI reporters thus should propose many Model 2 rules; that is, restatements should contain more vague rules than do uniform laws.
Our final result in the symmetric-information world is that while interest groups may participate on study groups, they will not lobby the PL in its role as a legislative body. Lobbyists attempt to get results either by providing legislators with information or by altering legislator preferences. In this model, the legislators are assumed to be informed so there is no need to tell them anything, and their preferences-their ideal points-are assumed to be exogenous so lobbying cannot affect them. Interest groups thus will not lobby because lobbying is costly. 52 We conclude this section with a remark about the symmetricinformation assumption. It seems implausible to suppose that the typical PL participant knows the consequences of proposals because, as the descriptions of the ALI and NCCUSL show, PL participants lack the resources and time to make independent investigations of these proposals. A participant can conveniently read most proposals, but reading seldom is enough. For example, the ALI's new standard for how corporate directors should respond to unsolicited takeover bids can be read in five minutes, but the economic and political consequences of adopting that proposal are unclear without doing a serious investigation of the subject. If typical PL participants will not make such investigations, how are they to become informed?  Sci. 835, 858-61 (1993). Lobbying sometimes consists of offering legislators tangible benefits in order to change their preferences. Our assumption that preferences are exogenous precludes such behavior. Also, we have no evidence that tangible benefits are conferred on PL legislators. Thus, lobbying in this Article refers to sending messages to these participants.

620
UNIVERSITY OF PENNSYLVANIA LAW REVIEW [Vol. 143:595 When certain conditions are satisfied, "passive" PL participants actually can be well informed about the consequences of proposals, but these conditions are very demanding. To see why, suppose that a study group has strictly monotone preferences over the median PL participant's choice set. The policy issue is how much sellers should be required to disclose about product safety, and an interest-groupdominated study group wants as little disclosure as possible. The PL's task is to choose a disclosure standard. If (i) the median PL participant knows the study group's preferences precisely; (ii) the study group can credibly disclose the truth about the consequences of any particular disclosure standard; and (iii) the median PL participant can conveniently (formally, costlessly) process all information conveyed to her, then the participant will be fully informed about the consequences of any study group disclosure proposal.
The optimal strategy of an uninformed PL participant in this game is to "assume the worst"; that is, to make the inference, on the basis of what she does know, that is least favorable to the study group. In the illustration here, the PL participant thus would assume, contrary to the actual desire of the study group, that every study group proposal requires the minimal amount of disclosure. This assumption is least favorable to the study group because it induces the participant to reject every proposal unless she prefers minimal disclosure. As a consequence, the study group will always waste the costs of creating proposals. This result is suboptimal for study groups. A study group's best response to the participant's "assume the worst" strategy thus is to choose among the following alternative strategies (depending on the study group's preferences): (a) make a disclosure proposal that is in the win set and tell the truth about it; (b) propose a vague rule; or (c) propose no rule and thus save rule-creation costs. Response (a) ensures that the median PL participant will be well informed about the consequences of meaningful reform proposals. Response (b) also is informative because all participants can equally evaluate the consequences of vague rules.

3
The strength of the conditions that generate this full-information result suggests that PL participants seldom will be fully informed. In particular, preferences over legal rules commonly are "s The result that the informed party will disclose fully in the circumstances assumed above is proved formally in Paul Milgrom & John Roberts, Relying on the Information of Interested Parties, 17 RAND J. ECON. 18, 24-29 (1986). not strictly monotone (that is, rules often cannot be evaluated on the simple dimension of more or less, as in the disclosure example, but rather tradeoffs must be made); PL participants may know study group preferences over the choice set only probabilistically; 4 study groups sometimes cannot credibly communicate the consequences of complex proposals; and PL participants often would find it too costly to process all the relevant information.
When these common difficulties exist, PL participants who will not make independent investigations of proposals will be poorly informed about those proposals' consequences. In the next section, therefore, we analyze PL behavior under the more realistic assumption that PL participants commonly cannot distinguish among substantively different proposals.

D. PLs Under Asymmetric Information
We generate three principal results in an asymmetric-information world. First, the incentive of study groups to propose vague rules in lieu of unpassable precise rules remains, and may even be stronger. Second, interest groups may have more power or less power in PLs than in ordinary legislatures, depending on whether one interest group is interested in a proposal or whether interest group preferences conflict. Third, when interest groups lack power or there are no interest groups involved (only reformers are active), the existence of asymmetric information reinforces the tendency of PLs to behave conservatively.

Conservatism and the Uninformed PL's Decision Rule
An institution behaves conservatively if it commonly prefers the status quo to reform. The status quo is not a PL participant's ideal point, however. We thus call a PL conservative if it commonly prefers the status quo to Model-l-style reform proposals. 55 In order to see whether such a status quo preference exists, and to understand the median PL participant's response to the efforts of reformers and interest groups to exert influence, we begin by " If preferences are not monotone and preferences are known only imperfectly, then PL participants would be prevented from knowing which outcome is worst for a study group. As a consequence, the median PL participant could not use the "assume the worst" strategy.
--To prefer vague new rules to the status quo, in our view, is not to be a serious reformer.

1995]
assuming that the participant is substantially uninformed. To be precise, suppose (a) the median PL participant knows the location of the status quo (that is, she can evaluate the consequences of doing nothing); (b) she knows, in an approximate way, the proposal space (the set of possible proposals that could be made on a subject); (c) she has no independent knowledge of the consequences of any proposal that is made to her; and (d) she can infer nothing about a proposal's consequences from the composition or actions of study group members.
The median PL participant's optimal decision rule for voting on proposals in these circumstances is derived formally below, but the rule has a simple intuitive version: if the median PL participant likes the status quo, she should reject every Model 1 rule proposal that is made to her; if she dislikes the status quo, she should accept every such proposal that is made to her. The logic of this conclusion follows from the fact that the median PL participant knows her ideal point and the location of the status quo. Because she also knows the proposal space, she can make a (rough) estimate of the likelihood that she will prefer any particular proposal to the status quo. In particular, if the participant likes the status quo, then there is a fairly high probability that every meaningful reform proposal (that is, Model 1 rule) will make matters worse for her; hence, she should vote no. On the other hand, if the participant dislikes the status quo, then there is a fairly high probability that every meaningful reform proposal will make matters better for her; hence, she should vote yes. The strategy of voting no if one likes the status quo and yes otherwise we call the "uninformed decision rule." A PL that uses this rule thus will have a conservative bias if the median PL participant likes the status quo.
To derive the uninformed decision rule formally, we make two further simplifying assumptions: (e) the set of possible proposals is uniformly distributed on the unit interval; that is, each proposal in the set of possible proposals is as likely to be made as any other, and the proposal space equals the probability space; (f) the median participant's ideal point, x, is to the right of the status quo, sq. 56 ' These assumptions are innocuous. The first is standard and simplifies the analytics. The argument made here applies to any symmetric distribution such as the normal. As to the second assumption, we could derive the same uninformed decision rule by fixing the median participant's ideal point to the left ofsq; the signs on the initial equations would change, but would cancel in the solution.

PRIVATE LEGISLATURES
The PL participant derives utility U,q = -Isq -xJ from sq (see Figure 3). This inequality is not solvable because the median participant is assumed not to know the location of y (that is, she cannot evaluate the consequences of proposals). The participant, however, prefers proposals that generate more utility for her than the status quo. Thus, her optimal strategy is to vote yes if the probability that she will prefer every proposal that could be made to her exceeds the probability that she will not. Every proposal in the interval {sq, bi generates more utility for the median PL participant than the status quo. The probability that any such proposal will be made is 2(sqx) (assuming that proposals are drawn from a uniform distribution on the unit interval). Proposals in the intervals (a, sq} and [b, zi generate less utility for the participant than the status quo. The probability that a proposal will lie in the former interval is sq; the probability that a proposal will lie in the latter interval is 1 -[sq + 2(xsq)]. This leads to the formal uninformed decision rule: vote yes when the probability that every proposal would be better exceeds the probability that it would not, or when 2(xsq) > sq + 1 -[sq + 2(xsq)]. This simplifies to vote yes when xsq > 14. The median PL participant thus will vote no to every proposal unless her ideal point lies more than twentyfive percent of the proposal space away from the status quo. 5 " In " The analysis assumes that PL participants are risk neutral. A risk-averse participant would reject proposals more frequently than the decision rule derived above would direct because her utility declines at an increasing, rather than at a uniform, rate as proposals move further from her ideal point (that is, referring to supra note 38, u/"(x) < 0). Those who believe that PL participants are risk averse rather than risk neutral thus should think that the argument for PL conservatism made here understates the case.

1995]
the language used above, a PL will behave conservatively unless its median member is quite dissatisfied with the status quo.
This degree of dissatisfaction does not seem to exist for the PLs we study. Observers often remark that the senior lawyers and judges who constitute such PLs as NCCUSL, the ALI, and the American Bar Association are unsympathetic to radical reform. This preference is as much institutional as personal. Significant reform requires a change in the underlying normative framework that supports the legal regime. The ALI's founding documents and NCCUSL's constitution, however, commit these PLs to reject controversial reform and to restrict themselves to "technical" improvements in the law. 8 That commitment to incremental change implies that the median PL participant's ideal point lies near in policy space to the status quo. And this in turn makes the condition for voting yes hard to satisfy. As a consequence, these PLs tend to behave conservatively.
Finally, study groups anticipate this voting behavior. Hence, when a study group cannot credibly inform a PL about a proposal's location, its decision rule is as follows: (a) when the median PL participant's condition for voting yes is satisfied, propose the Model 1 rule that is the study group's ideal point; (b) when the condition is not satisfied, propose a Model 2 rule or no rule at all.

a. Factors That Lead PL Participants to Update Beliefs About a Proposal's Location
An interest group that may be affected by a proposal will wish to participate in a study group in order to become informed and to help shape events. If the study group creates a proposal that the median PL participant would support under the uninformed decision rule, however, the interest group will not lobby. There is no need to incur lobbying costs for proposals that will pass anyway. The discussion below thus assumes either that the interest group dislikes the study group proposal, or that the PL would reject a favored proposal unless the interest group persuaded the PL to vote yes. The interest group is likely to lobby PL legislators in either of these events because it may well be persuasive. 5 9 See supra notes 19-21 and accompanying text. 59 The analysis in this section draws from Arthur Lupia & Matthew D. McCubbins, To see why, we retain all the assumptions that generated the uninformed decision rule except (d): the median PL participant now is assumed able to draw inferences about the location of a study group's proposal on the basis of what interested parties say and do. She uses these inferences to update her prior beliefs about a proposal's location. The median PL participant uses three factors to update her beliefs: (1) whether the proposal, called y, is costly for the study group to make; (2) whether the PL can impose a penalty for misrepresenting the location of y; and (3) whether reports ofy's location are made by persons with preferences that are similar to those of the median PL participant. We discuss each factor in turn.
First, an interest group member incurs costs in helping to create proposals. Interest group members thus will generate proposals only if these would alter the status quo sufficiently to justify the effort. Hence, if interest group members are observed to participate extensively in the affairs of a study group, 6 " the PL participant knows that the resulting proposal is a nontrivial distance from sq. This result is illustrated by Figure 4. The median PL participant infers that the study group proposal is either in the interval {a, p1 or {r, zi. A proposal would have to be at least that far away from the status quo (sq) in order to justify the costs of creating it. The second factor leading the PL participant to update her beliefs is the ability to impose a penalty on parties who misrepresent a proposal's location. Study group members will represent the location of a proposal to the PL. Moreover, reformers and interest o It is plausible to assume that the identity of study group members and advisors is known to the PL because the PL leadership appoints the membership of these groups, and because group members commonly appear before the PL to explain proposals. group members who do not participate on a study group may lobby the PL, and in doing so they will make representations as to the location of proposals. Suppose that these persons will pay a penalty for lying if they are found out. Then they either will tell the truth or misrepresent sufficiently to justify the expected cost of the penalty. For example, recall our assumption that the median PL participant's ideal point is to the right of the status quo. Now let a player inform the PL that a particular proposal would be better for the median participant than the status quo. 61 The participant then will infer that the proposal cannot lie in the interval fg, sql in Figure 5. To see why, consider the possibilities. The speaker could be telling the truth, in which case the proposal actually is better than the status quo and, therefore, lies to the right of sq. On the other hand, if the proposal is worse for the median PL participant than the status quo (that is, to the left of sq), then the proposal must change the status quo sufficiently to justify the lie. Hence, when there is a reputational cost to lying, a player will falsely claim that a proposal is better than the status quo only when it is considerably worse (that is, in the interval {a, g}).
The third criterion that leads the PL participant to update her beliefs concerning the location of any proposal is whether the report comes from someone whose preferences are close to those held by the median PL participant. Simply put, the more closely another person's preferences resemble yours, the less incentive that 61 The consequences of a proposal may be difficult to state precisely. Hence, the parties will make approximations. This fact can be captured by supposing that the median PL participant interprets a message respecting the location of a proposal to mean that the proposal is better (or worse) for her than the status quo. person has to mislead you. 62 As a consequence, the probability that the median PL participant will believe a study group member's or another person's description of a proposal is a function of the probability that the speaker has the same preferences as the median PL participant. 3 If the speaker is known to have exactly the same preferences, then the speaker's statement will be considered true. In Figure 6, the speaker is believed to have the same ideal point as the median PL participant and claims that a proposal is better for the participant than the status quo. The participant thus believes that the proposal is in the interval tsq, b} and will support it.' '6 Austen-Smith thus shows that "the closer are the lobbyist's preferences over consequences [of legislation] to those of the House,... the more likelyit is that there can be influential vote stage lobbying." Austen-Smith, Information, supra note 62, at 811. '4 Again, a similar analysis can be made with respect to the message "worse" (than the status quo). before the message and will be supposed to fall anywhere in that interval after it. 65

b. The Relative Credibility of Interest Groups and Reformers
The three factors that lead the median PL participant to update her beliefs as to a proposal's location imply that interest group members can send more credible messages than reformers can. Conventional analyses have assumed the contrary: that reformer messages are more credible than interest group messages because interest group members have a pecuniary interest in PL outcomes while reformers are disinterested. Even on its own terms this assumption is too broad. Reformers also may have pecuniary interests: a professor who is known to have formulated a law can earn substantial consulting income following the law's adoption. But even when such economic concerns are set aside, reformers remain less credible than interest group members.
' This last statement should be qualified. An interest group's message can be illuminating to the PL even though its preferences differ from those of the median PL participant, if these preferences also differ in a particular way from those of a study group and if the PL also knows the study group's ideal point. This qualification is illustrated in Figure 7, where i is the interest group's ideal point and a study group's ideal point p lies to the right of b. The interest group and the median PL participant apparently have opposing preferences because the interest group wants to deviate from the status quo to the left while the median PL participant wants to deviate to the right (her ideal point is x). Nevertheless, the interest group can send credible messages about a proposal's location. The PL participant knows that the study group will not make a proposal that is to the left of sq given its ideal point. With respect to proposals to the right, the interest group will say "worse" (than the status quo) only if a proposal is to the right of b (for example, y), and "better" if it is to the left (for example, r). These messages are credible because the interest group and the PL have relevantly identical preferences given the study group's ideal point; any proposal in the interval fsq, b) makes them both better off than proposal y, and any proposal to the right of this interval makes them both worse off. Knowing this, the study group will propose Model 1 rules in the win set {sq, b). It is unnecessary to pursue other configurations of ideal points and proposals to illustrate that it is the relevant similarity of the message sender's preferences to those of the median PL participant that importantly determines PL outcomes.

PRIVATE LEGISLATURES
People become reformers partly because they have strong policy preferences. Some reformers will overstate or misrepresent a proposal's consequences in order to implement those preferences. Hence, the median PL participant has to choose which possibly biased message to believe. Recalling the three criteria developed above, participation in study groups is a net gain, not a cost, for reformers. Consequently, the median PL participant cannot learn much about the location of a proposal when reformers are observed to play the major role in the proposal's creation, except that the proposal is not at sq.
Further, reformers have a weaker incentive than interest group members to tell the truth because they incur lower penalties from lying. Interest groups and PLs play a repeated game. For example, banks have been attempting to influence NCCUSL and ALI outcomes for five decades. When parties play a repeated game, early bad behavior leads to later sanctions. Thus, banks have a strong incentive to tell the truth. In contrast, reformers and PLs commonly play the game only once: few commercial law professors have worked on more than one set of Article 4 revisions. Thus, reformers who are discovered to have slanted a proposal's location cannot be punished.
Finally, interest group member preferences commonly are closer to those of the median PL participant than are reformer preferences. Reformers are academics, who hold political views that usually are to the left of the average citizen, and are considerably to the left of the elite senior lawyers who constitute large majorities in such PLs as NCCUSL, the ALl, and the ABA. 6 " Because PL participants commonly assume that interest group members are more likely than reformers to have similar preferences, participants are more likely to believe an interest group member's statement than a reformer's statement about the consequences of a proposal.
To summarize, interest group member statements are more credible than reformer statements for three reasons: it is more costly for an interest group member to participate in creating proposals than it is for a reformer; the interest group member incurs a greater penalty for lying; and the interest group member's preferences commonly are doser to those of the median PL participant.

c. The Relative Influence of a Single Interest Group in a PL and in a Typical Legislature
When only one interest group is active, it will have more power in a PL than in a typical legislature because credibility is more important to PL participants than it is to legislators. A typical legislator can learn the location-the consequences-of a proposal in two ways: she can receive messages from knowledgeable parties, or she can hold a hearing. Holding a hearing is more reliable (documents can be subpoenaed, witnesses sworn and cross-examined), but it is more expensive because hearings take time. Listening to knowledgeable parties is cheaper, but is less reliable because these informants may have an incentive to misrepresent the truth. Thus, messages and hearings are substitutes in legislatures. A sophisticated legislature will use both tools so as to maximize the probability that it is well informed.
In contrast, a PL participant must rely only on messages. Because most PL participants have no direct economic stake in particular proposals, are not paid for time spent on PL business, and are given no resources, they do not make personal investigations. Also, PLs lack the means, and in consequence the institutional structures, to hold hearings. Messages are more important to PL participants than they are to ordinary legislators because messages are the only source of information available to PL participants as to the consequences of proposals. 68 Therefore, the conclusion that an interest group message commonly is more credible than a reformer message implies that interest groups have more power in PLs than in ordinary legislatures.
To see how this power is exercised, consider Figure 8.
FUDENBERG & JEAN TIROLE, GAME THEORY 448-51, 453-54 (1993). In order to understand this assertion, one must realize that an interest group would deviate from the pooling equilibrium if it had preferences similar to those of the median participant and if it would be disadvantaged by application of the uninformed decision rule. This is because such an interest group wants to be recognized for what it is-a group whose preferences are similar to those of the median participant and that the PL can punish-because then its messages will be trustworthy. Thus the interest group will deviate from the uninformative pooling equilibrium to send credible messages.
' The Director of the ALI has said that an important criterion for appointment to a study group is that the person have credibility with the ALI leadership. See supra part I.A. This is consistent with the analysis here: credibility is important in a PL because verifiability is not an option. Suppose that an interest-group-dominated study group's ideal point is in the interval {sq, b). The study group will propose a Model 1 rule in that interval and truthfully claim that the rule is better than the status quo for the median PL participant. 6 9 For the reasons given, this claim is likely to be believed and the proposal will pass. A reformer-dominated study group whose ideal point is in the interval {sq, b) also would like to propose a Model I rule in that interval. However, this study group's claim that the proposal is better than the status quo for the median member is less credible. 70 This lack of credibility has two consequences. First, if an affected interest group's ideal point is to the left of sq, the group will lobby in opposition to any Model 1 rule in the interval [sq, b). This opposition probably will kill the rule because the reformers will not be believed. If reformers anticipate this, they will propose a Model 2 rule instead. Second, reformers have an incentive to propose Model 1 rules that the interest group likes, within limits. To illustrate why, suppose (in Figure 8) that a reformer-dominated study group's ideal point is r, which is equidistant between sq and g, while the interest group's ideal point is at i. The interest group would support any proposal to the right of sq. Thus, a reformerdominated study group will propose a Model 1 rule just to the left of g. This rule will be better for the reformers than sq and will also be acceptable to the interest group. With interest group lobbying support, the rule has a good chance to pass (that is to say, true statements by the interest group that the rule is better for the median PL participant than the status quo are credible).

PRIVATE LEGISLATURES
This analysis has three implications: (a) an interest-groupdominated study group will get the Model 1 rules it likes if its '9 Recall that b is as far from the median participant's ideal point x as is sq. Hence, any proposal to the left of b and the right of sq is better for the median participant than sq. Proposals to the right of b and the left of sq are further from the median participant's ideal point than sq and are, therefore, worse for the participant. preferences are reasonably close to the preferences of the median PL participant (that is, in the interval tsq, bl); (b) an interest group has a good chance to block a proposed Model 1 rule by a reformerdominated study group even though a well-informed PL would have adopted the rule (the interest group would attempt to block if its ideal point is to the left of sq); and (c) a reformer-dominated study group will propose Model 1 rules that an interest group prefers if both the reformers and the interest group hold preferences that are relatively close to those of the median PL participant (that is, in the  interval {sq, bl). These three implications support our result that interest groups have more power in PLs than in ordinary legislatures (when there is only one active group). This result, in turn, implies that reformers will make strong efforts to enlist interest group members to support reform proposals or will try hard to defuse interest group opposition. The reformers need interest groups more than the interest groups need reformers. 7 1 We conclude this aspect of the analysis by considering the case when a study group's ideal point is to the left of sq-that is, far from the median PL participant's ideal point. In this asymmetricinformation world, an interest-group-dominated study group would: (a) propose a Model 1 rule that is in the interval {a, sql and then lie about its location, if it would gain enough from passage to justify the possible reputational costs; or (b) propose a Model 2 rule or no rule at all. Study groups lack this choice when information is symmetric because then proposals that a PL dislikes always fail; hence, these proposals are not made. A reformer-dominated study group seldom would propose a Model 1 rule to the left of sq, even though its ideal point lay there, because the reformers' false claim that the proposal lay in the interval {sq, b) would not be believed. 7 Because the median PL participant learns little from reformers, an apparent implication of the analysis is that when only reformers are active, they cannot achieve anything: the participant will follow the uninformed decision rule, voting yes when the rule directs and no otherwise. A more accurate conclusion, however, is that reformers cannot achieve anything significant enough to attract public notice. The PL participants' general preference for reform inclines them to support proposals that cannot create negative reputational effects. Thus, when interest groups are absent, successful NCCUSL and ALI proposals unsurprisingly tend to be innocuous on their face. For example, a uniform law to compel the attendance of out-of-state witnesses may, or may not, further the administration of justice, but the rule is unlikely to embarrass the PL participants who voted for it. In short, reformers sometimes can cause reforms to be adopted that no group could be mobilized to oppose. Such reforms are commonly called "technical." Rather, because of the pro-reform bias that reformers have, such a study group would propose a Model 2 rule instead.

PLs with Competing Interest Groups
In a world of asymmetric information, interest group power is substantially diluted when groups compete. Further, the institutional bias of PLs to behave conservatively is reinforced. In explaining these results, we retain the assumption that an interest group will always participate in a study group if it can. 2 The analysis here thus focuses on an interest group's decision whether to lobby PL legislators when another group might lobby them also, and on the effect of that decision.
To begin, assume that a study group proposes a Model 1 rule that affects two interest groups. The two groups are competitive: one group, Gq, prefers the status quo, while the other group, G 1 , prefers the proposed rule. The groups decide simultaneously whether to lobby; that is, each group makes its lobbying decision without knowing what the other will do. This is a reasonable assumption because many proposals are originated by reformers, who then promulgate the reforms to potentially interested parties. 73 Interest groups thus commonly learn about the existence of proposals at about the same time. 74 Let the uninformed decision rule direct the median PL participant to vote no to any Model 1 rule proposal. Then G.q will not lobby unless G 1 does, because if both do nothing, the proposal fails. 5 If G, lobbies alone, the criteria of costly participation, penalty for lying, and similar preferences imply that G 1 might persuade the median PL participant to support the proposal. The median participant is persuaded to change her vote to yes with probability q (0 <_ q <_ 1). The more credible G 1 is, the higher q will be. G, realizes returns (in present value terms) of r 1 if the proposal passes. 75 Interest groups sometimes are not competitive in the strict sense that one prefers the status quo while the other prefers reform. Both may prefer the status quo to any reform, or both may prefer any reform to the status quo but, at the same time, prefer different reform proposals. When interest groups have "similar" preferences, in this sense, little generality is lost by assuming that there is only one group. Thus, we focus here on the more important case-when groups compete.

1995]
UNIVERSITY OFPENNSYLVANIA LAW REVIEW [Vol. 143:595 The expected value of lobbying for G 1 thus is qr. and its lobbying cost is c. Hence, GI's condition for entering to lobby is qr > c, or q > (c/rI).
If Gq lobbies, it will block the proposal with certainty. When interest groups send conflicting messages concerning a proposal's location, the median PL participant infers that one of them is lying, but does not know which one. Since she cannot investigate, her best response is to act as if neither message had been sent. Thus, she will reject the proposal because, we assume, the uninformed decision rule directs her to do so. 76 Let rq be Gq's return from preserving the status quo and let its lobbying cost also be c. Its gain from lobbying is rqc because it blocks the proposal with certainty. The proposal, however, fails with probability (1q) if G,, does nothing. Hence, its payoff from doing nothing is (1q)r,, + q(-r) = r,,(1 -2q). Gsq's condition for entry then is rqc > r1qU( -2q), or q > (c/2r, 1 ). It is easier to satisfy Gq's entry condition than GI's, because if Gq enters it gets a positive payoff for certain, while if G 1 enters it gets a positive payoff with probability q.
There are three unique equilibria in this model, two in pure strategies and one in mixed strategies. 7 If G,'s entry condition is 6 Recall that a PL participant, unlike an ordinary legislator, cannot make an independent investigation. This distinguishes the model used here from the model In their model, the legislator learns the truth with certainty when both groups lobby. She then knows that one group is lying-hence, she infers that her uninformed decision how to vote may be wrong, and she will certainly investigate. As a consequence, her threat to investigate if both groups lobby and send conflicting messages is credible. Both groups are aware of this threat. Because there is a penalty for lying, each group thus sends the same message when both lobby-the truth. See id. at 233-45. In the model here, the PL legislator cannot check facts and so acts as if no messages were sent when both groups lobby; that is, she votes as her uninformed decision rule directs her to. When only one group enters, and the legislator can verify messages at a cost, there may not be an equilibrium in pure strategies: the group will tell the truth if the legislator checks; but the legislator then will not check because the group will tell the truth; but then the group will lie; and so forth. For an analysis of mixed strategy equilibria in this case, see Eric Rasmusen, Lobbying When the Decisionmaker Can Acquire Independent Information, 77 PUB. CHOICE 899, 901-07 (1993). We need not explore these possibilities here.
" A strategy in a game is "a complete plan of action-it specifies a feasible action for the player in every contingency in which the player might be called on to act." ROBERT GIBBONS, GAME THEORY FOR APPLIED ECONOMISTS 117 (1992). A Nash equilibrium in pure strategies exists when each player's strategy is a best response-that is, the utility maximizing response-to the strategies the other players have chosen. In this event, no player has an incentive to change her strategy, so the game is in equilibrium. To understand mixed strategies, suppose that a particular not satisfied, neither group lobbies and the proposal fails; if G 1 's entry condition is satisfied but Gsq's entry condition is not, then G 1 lobbies and the proposal passes with probability q." There are no pure strategy equilibria in which both groups enter. This is because G, wants to lobby only if Gq does not, while G,, wants to lobby only if G 1 does. 9 There also is a mixed strategy equilibrium in which Gq enters with probability Pq = (qrlc)/qr and G enters with probability P, = (rSq -2 qr,/(q -C). 80 This equilibrium is reasonable. The entry probabilities imply that G,, is more likely to lobby if G, has a good chance to persuade the PL to vote yes; G 1 would realize a large payoff from passage of the proposal; and Gq's lobbying costs are low. The first two factors increase the likelihood that G, will enter and so increase the payoff to Gq from entering as well. The effect of the last factor is obvious. More subtly, G, is less likely to lobby the better are its chances to persuade the PL and the lower are its entry costs. These results occur because the higher q is, and the lower c is, the more likely Gq is to enter and block the proposal with certainty. At that point, GI's lobbying costs would be wasted." mixed strategy equilibrium exists when each player's mixed strategy is a best response to the mixed strategies of the other players. This means that no player has an incentive to change the probability with which she plays particular pure strategies given the probabilities of the other players. Game theorists interpret mixed strategies in this way: player A actually prefers to play one of the pure strategies available to her-say sL. The other players do not know just which pure strategy player A prefers and so each believes that she plays particular pure strategies with probability p,,.
When the beliefs that each player holds about the others are consistent-when no player has an incentive to revise her probability assignments-a mixed strategy equilibrium exists in the game. o This equilibrium also is established in the Appendix. s The particular results change (but not the general ones) if the median PL participant's uninformed decision rule would induce the PL to support a proposed Model 1 rule. Then G 1 would not lobby unless G,q lobbied, because the proposal would pass were no lobbyists to appear. G, thus has an incentive to lobby if G lobbies, because then there will be competing messages that cancel and the PL will use its uninformed decision rule to vote yes. However, there is no pure strategy equilibrium in which both enter. With respect to the assumption that lobbying

UNIVERSITY OF PENNSYLVANIA LAW REVIEW
[Vol. 143:595 Two additional implications of the model are significant. First, when the stakes involved in reform are high, an interest group may have less lobbying power before a PL than it would have before an ordinary legislature; and second, PLs will behave more conservatively. To understand the first implication, suppose that a proposed Model 1 rule would work a substantial reform. In the usual case, this would impose substantial costs on persons who prefer the status quo; that is, rq would be high. In consequence, G,q's entry condition becomes easy to satisfy, so the PL is more likely to observe competing messages. Because these messages cancel, the PL will apply the uninformed decision rule. Thus, when a proposed Model 1 rule would substantially deviate from the status quo and affect more than one interest group, the PL is likely to act as if no lobbying had occurred. The more significant the issue, in short, the less power competing interest groups have.
The second implication-that PLs will behave more conservatively when interest groups compete-follows from the first. If a proposal would change the status quo substantially and affect only one interest group, that group may cause the PL to change its vote to yes. Major reform proposals, however, tend to create winners and losers; that is, they will engender competitive lobbying. In this event, as we have just stated, the PL would ignore the groups' messages in favor of using its uninformed decision rule. This strategy, much more often than not, would induce the PL to choose the status quo. PLs such as NCCUSL and the ALI thus react more conservatively to proposals that would work significant reform than to proposals that would alter the status quo only slightly. 8 2 decisions are made simultaneously, G, has an incentive to delay its lobbying decision as long as possible (if the uninformed decision rule directs a no vote). This is because if GI entered early, G, would observe its entrance and thus enter to block its proposal (if G,,'s entry condition is satisfied). In contrast, G, might force G, to make its entry decision simultaneously (that is, late in the game) if G I waited. This is to GI's advantage: when GI's and G,,'s entry conditions both are satisfied, and both must decide whether to lobby at the same time, G. enters only probabilistically, while ifG, entered early and alone, G,, would enter and block G,'s proposal with certainty. Thus, a second motivation for treating lobbying as a simultaneous-move game is that interest groups often have an incentive to force each other to play that game.
' The conservative bias of PLs also implies that a PL is more likely to approve Model-i-type revisions of an existing statute or uniform law than to adopt a Model 1 rule covering a previously untreated area. Revisions of existing statutes are seen as "technical" exercises-correcting minor flaws or updating a statute. From this it follows that revisions are less likely to engender competitive lobbying than other proposals presented to the PL, cet. par. See Scott, supra note 28, at 1820 (arguing that, for these reasons, a single active interest group will have greater influence on Conservatism commonly means that a legislature will adopt no new rules at all. Conservatism can take this form in PLs, but also can result in the adoption of vague rules that appear to accomplish something, but in fact do not.
We can summarize the models' principal results in three propositions: Proposition 1: A PL will produce many Model 2 rules. The penchant to produce these rules exists when all of the actors before a PL are symmetrically informed about the consequences of proposals. It is reinforced by the more realistic assumptions that information is asymmetric and that competing interest groups can arise. Model 2 rules are not chosen just when and because the PL is committed to the virtues that flow from stating rules in abstract fashion; rather, the penchant of PLs to adopt vague rules is a function of the political configuration of the relevant actors and the PL's institutional structure.
Proposition 2: Interest groups may have a larger or smaller influence on PL outcomes than they have on ordinary legislative outcomes. When voters in the PL itself are poorly informed, a dominant interest group will have power because it is best able to make credible representations about the consequences of proposals. This influence will result in many Model 1 rules. Again, these rules are not chosen just when and because the PL is committed to the virtues of precision in rule statement. Rather, the choice-of-rule form reflects the relevant actors' political power. On the other hand, when interest groups compete, the messages they send will be too noisy to influence PL outcomes.
Proposition 3: PLs have a status quo bias that induces them to reject reforms that any cohesive interest group would oppose.

III. AN ANALYSIS OF ALI AND NCCUSL PROJECTS (WITH PARTICULAR EMPHASIS ON COMMERCIAL LAW)
The ALI and NCCUSL consider the Uniform Commercial Code to be their major success. The UCC is the most influential and widely adopted uniform law. There has been little rigorous data collection with respect to the results developed above, but much is known about the composition of the study groups that created the the revision process in securing adoption of bright-line rules that advance the group's interests).
UCC and that are active in revising it. Much is also known about the participation of interest groups in the Code-creation process. Also, the results of that process-the character and specificity of UCC rules-are observable. There is also considerable evidence of this character with respect to other recent ALl and NCCUSL projects. We summarize much of this evidence below. 8 s Before doing so, we develop two predictions concerning the character and rule form of the UCC itself: (a) Article 9 and its current revision, and Articles 3 and 4 and their recent revisions, will contain many Model 1 rules, while Article 2 and its current revision will contain many Model 2 rules; and (b) the Code's Model 1 rules will be the product of industry influence.

A. Article 9
There was extensive interest group participation, largely by asset-based financers and banks, in the drafting of Article 9. The principal reporter of the Article 9 project, Grant Gilmore, reported on the events that led banks and finance companies to support the UCC project that they had earlier rejected as a radical reform. 8 4 This support developed after Homer Kripke, then a legal counsel to CIT Financial Corp., became one of the key advisors to Gilmore and the other drafters. 8 5 Kripke subsequently described how, during their drafting deliberations, banking interests blocked proposed clauses that would have imposed on them the costs of various consumer-protection provisions. 8 6 Kripke reported that it was 83 We test our theory primarily by retrodiction-the ability of a theory to predict prior events: here, the shape of the UCC. Steven Weinberg recently emphasized the value of this method: It is widely supposed that the true test of a theory is in the comparison of its predictions with the results of experiment. Yet ....
in the case of general relativity a retrodiction, the calculation of the already-known anomalous motion of Mercury, in fact provided a more reliable test of the theory than a true prediction of a new effect, the deflection of light by gravitational fields.
. Often it is a successful prediction that one should really distrust. In the case of a true prediction.... the experimentalist does know about the theoretical result when he does the experiment. And that can lead, and historically has led, to as many wrong turns as overreliance on successful retrodictions. STEVEN WEINBERG, DREAMS OF A FINAL THEORY 96, 97 (1992). 84  ' See Homer Kripke, The Principles Underlying the Drafting of the Uniform necessary to avoid arousing the opposition of banks and finance companies in order to ensure passage of the UCC project." 7 Thus, the original Article 9 was the creation of an "interest-groupdominated study group." We predict that such a study group will propose Model 1 rules that are in its interest. In a world of symmetric information, the interest group will participate on a study group only if it can get its way because participation is costly. In a world of asymmetric information, the study group will also propose Model 1 rules if its ideal point is relatively close to that of the median PL participant and if, because it is the only group, its messages concerning the location of proposals will be credible. The impressionistic evidence suggests that these conditions were met in the original Article 9 project.
The business lawyers who served on the Article 9 study group had preferences concerning the regulation of commercial practice that appeared to be close to the preferences of the business lawyers who then dominated the ALI and NCCUSL." 9 Further, Article 9 largely affects finance companies and banks, parties whose interests concerning the regulation of asset-based financing coincide in many (if not most) respects. 90 Moreover, Article 9 affects no other cohesive interest group. 91 The conditions that lead us to suppose that the original Article 9 study group would propose business-oriented Model 1 rules also leads us to predict that those rules would be adopted. This prediction seems confirmed in the original Article 9 statutory scheme. Article 9 explicitly "purports to promote the interests of those industries that helped create and lobby for it. Not only does [Article 9] seek to lower the costs of asset-based financing, but the statute also creates rules (such as repossession) that many believe disadvantage consumer debtors." 92 In addition, the original Article 9 did not regulate railroad car trusts and insurance in order to diffuse opposition from the railroad and insurance industries.

1995]
This description of the revision process leads us to predict what is observable on the face of the Article 9 Study Group's report: (1) the Study Group recommends many reform proposals; (2) these proposals are overwhelmingly framed as Model 1 rules; and (3) the Model 1 rules are facially congruent with industry preferences. 99 When one interest group is active in a PL, it often can block proposals it dislikes and secure passage of proposals it wants. Banks and secured financiers commonly have similar interests and seldom are opposed in the ALl and NCCUSL by competing groups. Thus, we also predict that the ALI and NCCUSL will adopt the Article 9 Study Group's revisions without significant amendment. 0 Further evidence from other business-law projects also supports the prediction that a cohesive interest group can (and does) block proposals that only have reformer support. For example, NCCUSL directed a drafting committee to create a proposed uniformpayments law, and the principal proponent behind the proposal was a law professor from Harvard. The banks disliked the law and that their statements and votes are publicly made in the glare of the interest groups. To some, that may be somewhat daunting and intimidating. Drafting committee members whose practice, employment, or academic consulting is for or on behalf of an interest group may be hard pressed to take an action contrary to that group. Rapson, supra note 21, at 263-64. Rapson says that the pressures he describes affected the Article 9 revision process. See id. at 264, 259-60. " See id. at 1822-47 (analyzing the 36 specific areas in the Final Report, each of which contains numerous proposals to amend either the statutory language or the official comments to Article 9). This analysis shows that many of the key revisions are framed as bright-line or Model 1 rules; these rules appear on their face to promote the interests of specific classes of secured creditors in contexts where competing interests may well be disadvantaged; and these effects are designed to be hidden from public view behind cosmetic efforts to maintain the efficacy of the filing system. See id. A large majority of these participants will not derive direct economic benefits from the passage of any one proposal, nor do Model 2 rules produce results specific enough to advance a participant's policy preferences. Under these circumstances, an interest group sometimes can argue credibly that a particular Model 2 rule will create excessive uncertainty, so that support for the rule would be perceived as inconsistent with having good judgment. Such an argument is likely to be advanced in deliberations over Articles 3, 4, and 9, where cohesive interest groups dominate, but it is unlikely to be advanced in connection with Article 2 because the costs and benefits of sales law Model 2 rules are spread widely and shallowly. Hence, no group has an incentive to lobby in order to block them. Article 2, therefore, is more likely than Articles 3, 4, and 9 to have Model 2 rules. The study group to revise Article 2 seems reformer-dominated. So far as it appears, no commercial interest group has lobbied to change the statute. Because academics were and are in charge, we predict that both the original Article 2 and the revisions will contain many vague rules. The former prediction is confirmed on the face of the statute. Almost everyone who has studied the subject agrees that the original Article 2 has many Model 2 rules.
A salient example of the continuing predominance of Model 2 rules in sales law can be found in the Article 2 Drafting Committee's approach to the issues of contract formation and enforceability. According to the Reporter of the Drafting Committee, the Committee "has preserved the original approach to contract formation and modification attributable to Karl N. Llewellyn.... This approach minimizes formality, but when necessary, expands rather than limits the opportunity to contract. The emphasis is uponflexible standards, mutual conduct, and the intention of the parties." 15 Specific examples of the Model 2 rules that follow from this "emphasis" include proposals to repeal the Statute of Frauds requirement of section 2-201,116 to revise section 2-204(3) so as to link contractual obliga-tion primarily to the existence of "a reasonably certain basis for giving an appropriate remedy,"' and the following proposed revision to section 2-207: Varying terms contained in the writings and other records of the parties do not become part of a contract unless the party claiming inclusion proves that the party against whom they operate (i) expressly agreed to such terms, or (ii) assented to such terms and had notice of themfrom trade usage, prior course of dealing or course ofperformance. Except between merchants, the burden of proof under this subsection is satisfied by clear and convincing evidence." 8 We also predict that reformer-dominated study groups will attempt to enlist interest group support and diffuse interest group opposition. The only groups that were possibly cohesive enough for the reformers to consider in connection with the Article 2 revisions were those that represented consumers.
Representatives of consumer groups have.been included in the drafting process. The revision ofArticle 2, however, is blessed with not only a new concentration on consumer issues because of... past difficulties but additionally involves persons on the Drafting Committee with considerable consumer law experience and several advisors or observers that represent consumer interests in their work. To date, this has given rise to strong discussion and a much improved atmosphere for reaching consensus. Id. at 1573 n.33, Evidence of attempts by reformers to diffuse industry opposition can be seen in other NCCUSL projects. On occasion, industry groups themselves may not know that NCCUSL is drafting a proposed uniform law. In this circumstance, NCCUSL representatives seek interest groups out in order to diffuse opposition and enlist support. Two academic reporters, one a commissioner, thus wrote: The tradition is to reach out to groups that are likely to take an interest in a proposed uniform act, and to secure the participation of such groups in an advisory capacity throughout the drafting process. The ideal is to identify relevant interests and to satisfy their concerns to the extent possible in order to have a legislative product that takes account of the real problems and that does not beget needless opposition ....

D. Retaining the Status Quo: The Effects of Interest Group Competition
We predict that when interest groups compete, a PL is likely to retain the status quo. We give three examples here. The first involves the attempt to revise UCC Article 6, regulating bulk sales. The attempt ended in failure. Three interest groups competed: the unsecured lenders, represented by the Credit Men's Association, wanted a strong law; 2 ' the secured lenders wanted to be exempt from any law or to have no law at all;. 2 ' and the auctioneers, represented by the Association of United Auctioneers, wanted to escape liability for their participation in sales of debtors' estates. 2 2 NCCUSL could neither unite behind a serious revision of Article 6 nor repeal it. Either alternative would have changed the status quo substantially. Instead, NCCUSL recommended that each state be given the option of repealing its version of Article 6 or adopting a revised version, a power that the states already possessed. 123 Our second example concerns products liability and medical malpractice law. The scope of a firm's liability for defective products and of a doctor's liability when patients are injured have been among the most debated and significant issues in American private law since the 1970s. The issues are important, and strong interest groups take different sides concerning them. For example, the trial lawyers prefer the status quo while the insurance industry wants substantial reform. This conflict implies, under the analysis to create a Revised Article 6 which answered some fundamental criticisms and which could be supported by the various interested groups from the bar and credit industry." NATIONAL CONFERENCE OF COMMISSIONERS ON UNIFORM STATE LAWS,

REVISED UNIFORM COMMERCIAL CODE ARTICLE 6-BuLK SALES 1 (Draft for Discussion
Only 1986). 120 See Winship, supra note 119, at 688-89. 121 See id. at 690. When they cannot make a policy choice, the ALI and NCCUSL sometimes pass menus that do not restrict state discretion rather than vague rules. For example, when the original UCC was written, an important question was whether a consumer who was injured by a defective product could sue the manufacturer directly or whether she was restricted to a suit against the retailer. The original UCC § 2-318 offered the states a choice of three governing provisions, which spanned the positions courts and commentators had taken. These alternatives thus achieved nothing.
here, that the ALI will adopt vague rules or none at all. In 1986, the ALI council commissioned a serious study of products liability, environmental liability, and medical malpractice law with the goal of developing principles that would be the basis of ALI action. In 1991, its study group produced a large number of principles, some of them controversial.' 24 The ALI has taken no action on the basis of any of these principles. Instead, it commissioned a Restatement of Products Liability Law. Because restatements are directed to courts, this choice precluded significant reforms such as using administration compensation schemes or scheduling damage awards.
The nature and progress of the Products Liability Restatement constitute our third example of PL performance when issues are controversial. The ALI appointed two academic reporters to draft the Restatement.' 25 We predict that a reformer-dominated study group will propose vague rules when its ideal point lies far from that of the median PL participant. The writings of the reporters reveal them to hold views on products liability reform that seem far from the likely preferences of many ALI members.1 26 As an illustration of what has happened, perhaps the most controversial area of products liability law involves design defects. The substantive issue concerns when a product that performs as intended is nevertheless defective because a different design could have produced more safety. Resolving this issue requires a jury to make difficult tradeoffs between reducing accident costs and reducing the amenities a product can produce (a completely safe car is little fun to drive and would be quite expensive). The difficulty of these tradeoffs, especially when made in the context of litigation by lay persons, has caused considerable unpredictability in application of the law. In brief, it is difficult to predict when a jury will find a design defective.1 27 The reporters' first draft responded to this problem by providing that "a product is defective in design when the foreseeable risks that a private legislature with a membership similar to that of the ALI and NCCUSL and procedures similar to theirs will have a strong status quo bias and sometimes will be captured by powerful interests.
In addition, the products of these private legislative processes will sometimes be characterized by vague and imprecise rules and other times by crude but precise bright-line rules. On its face this choice-of-rule form does not seem normatively objectionable. Precise (Model 1) rules are useful in some contexts, while vague (Model 2) rules delegating broad discretion to courts are useful in others. Our analysis shows, however, that groups such as the ALI and NCCUSL produce these rules in consequence of a particular institutional dynamic and not because of their intrinsic virtues as instruments for social control. We suggest that Model 1 rules result from the desire of a dominant interest group to preserve its victory in the legislative process (by confining the discretion of the rule applier) and not because they are socially desirable. On the other hand, Model 2 rules result, we believe, because reformers propose them when they are unable to get Model 1 rules enacted. The impressionistic data that we marshal is consistent with this theory.
Our analysis does not establish conclusively that the rules produced by the ordinary legislative process would be less driven by these institutional factors.
We suspect, however, that typical legislatures perform well relative to PLs for two reasons. First, legislatures have mechanisms for agreement that permit normative debate to reach a resolution-a resolution that will be more clearly reflected in the resulting statutory product. Second, ordinary legislatures have mechanisms for finding facts that are unavailable to PLs, and are exposed to many more sources of information concerning the effects of the proposals that they consider. Truth is a likely corrective to outputs that are skewed by the process itself. In any event, our purpose in this Article is to advance a much more modest claim: whatever the relative merits of private and public legislative bodies, the complacency that has heretofore marked the academic attitude toward the private law-making groups is not warranted.
There are two lessons to draw from this. The first is that academic attention should focus on inputs as well as outputs: there should be more theory and more evidence relating to how private law-making groups function. It may be possible (we are dubious but far from certain) that PLs such as the ALI can be reformed. The second lesson is that the ALI and NCCUSL, at least provisionally, should no longer be immune from critical investigation. A concrete implication of this view is that debates about whether a subject is best regulated by a uniform law or a federal act should be influenced by a perception of how uniform laws are actually made. Because this Article makes strong claims about PLs, we close by noting our sympathy with the views of the founders of the ALI and NCCUSL. The founders intended their proposed organizations only to deal with issues that satisfied two "jurisdictional" requirements: first, that society had reached a consensus concerning the relevant values; and, second, that those values could be translated into laws solely with the use of traditional legal expertise. The organizations would perform poorly, the founders believed, were they instead to attempt the typically legislative tasks of harmonizing value conflict and regulating complex economic activity. This Article is evidence of the founders' wisdom. If G 1 enters with this probability, then a mixed strategy is optimal for Gq. Hence, there is a mixed-strategy Nash equilibrium in which both groups enter with the probabilities just shown.13