Author:Kaplow, Louis

INTRODUCTION 1376 I. BALANCING VERSUS STRUCTURED DECISION PROCEDURES 1381 A. Final Decision Rule 1381 1. Analysis 1381 2. Additional Considerations 1384 B. Information Gathering 1387 1. Analysis 1387 2. Conduct of Legal Proceedings 1389 II. ANTITRUST 1391 A. Rule of Reason 1391 B. Mergers 1410 III. TITLE VII DISPARATE IMPACT 1420 IV. CONSTITUTIONAL LAW 1438 A. Strict Scrutiny 1438 B. Proportionality Analysis 1451 CONCLUSION 1459 INTRODUCTION

Balancing is a familiar mode of decisionmaking in the law and beyond. When one consideration favors a particular decision (say, liability) and another opposes it, it seems to be the essence of reason that the superior decision reflects the balance of the competing forces, taking into account the weight of the evidence and the importance of each factor. Many legal rules, such as the negligence test for tort liability, operate in this fashion.

Sometimes, however, structured decision procedures are used instead for these types of decisions. As a benchmark for comparison with balancing, this Article takes the following stylized version as a point of departure:

(1) The plaintiff must show that the harm of the defendant's act exceeds some threshold. If not, there is no liability. If so:

(2) The defendant must show that the benefit of its act exceeds some other threshold. If not, there is liability. If so:

(3) The harm and benefit are balanced, and there is liability if and only if the harm is greater.

Similar or related schemes are thought to characterize some existing legal decision procedures or have been proposed in antitrust law (rule of reason, mergers), discrimination law (Title VII disparate impact), and constitutional law (strict scrutiny, proportionality analysis). Yet structured decision procedures of this sort are neither used nor advocated in most other areas of law.

It is natural to ask when and why legal rules should employ structured decision procedures in lieu of balancing. And, when they are used, it is necessary to set the two decision thresholds, for their height is critical to the procedure's bite and, in particular, its difference from unconstrained balancing. This Article aims to answer these questions and others, with a particular emphasis on the descriptive and normative implications in the aforementioned legal domains.

Part I presents a general, abstract comparison of structured decision procedures to unconstrained balancing in order to frame the analysis of the substantive doctrines examined in the next three parts. (1) Part I begins by comparing the two approaches as final decision rules when all of the information that will be considered is before the decisionmaker. In many cases, including all those that reach step 3 under the structured protocol, the decisions will be the same. In important settings, however, the outcomes will differ. Moreover, in all such cases the outcome under the structured rule is necessarily inferior in the sense that such cases involve either the assignment of liability when the benefit exceeds the harm or a failure to assign liability when the harm exceeds the benefit. In addition, the purported virtue of structured rules in avoiding difficult balancing turns out to be misleading. They avoid balancing in many easy cases, but neither effort nor error is reduced. And they avoid balancing in some hard cases, but that is precisely when they stop short of the balancing performed in step 3 even though the resulting outcome from step 1 or step 2 may well be incorrect. Moreover, they sometimes require close comparisons with the thresholds even though balancing would have been easy. On reflection, it is remarkable that structured decision procedures are believed to prevail in important areas of law and are advanced as replacements for balancing in some others without even having asked the basic question of how outcomes under the two methods differ.

Part I then compares the two approaches as guides to information gathering, which is sometimes advanced as a virtue of structured decision procedures because stopping early (at step 1) economizes on effort. This advantage proves to be largely illusory. Optimal, unconstrained information gathering, which is what one ideally would do under a balancing approach, involves a number of principles that are sharply violated by the structured protocol, stemming from the latter's sequential separation of the investigation of the harm and the benefit of a challenged practice. First, much evidence is expressly comparative; indeed, characterization evidence is relevant precisely to the extent that it bears differentially on competing understandings of the defendant's alleged act. Attempting to separate the two is artificial and fraught, somewhat like using scissors, disjoined, one blade at a time. Second, evidence often naturally clumps by source rather than by subject: internal documents, witnesses, and experts; not harm and benefit. Third, even if all evidence bore only on harm or only on benefit and naturally clustered in single-issue bundles, it is a priori unlikely that the optimal order of gathering and assessing evidence would be to do first all of one type (harm), followed by all of the other (benefit). Instead, it is (roughly) sensible to collect first, second, and so forth whatever bundle has the highest diagnosticity to cost ratio; at any given point, the most promising bundle may just as plausibly involve benefit as harm. Indeed, since there tend to be diminishing returns with respect to the exploration of each issue, it is unlikely that all of the most promising avenues would concern only one and all of the least promising only the other. A final subsection explains that neither of the two approaches actually governs the conduct of U.S. civil litigation, which has a structure of its own--a fact that casts a curious light on many doctrinal and policy discussions of structured decision procedures.

Parts II-IV, the core of this Article, apply this general analysis of balancing versus structured decision procedures to three areas of law. The reader should note that the analysis is qualified in a number of ways: Constraints of space result in partial analysis, mainly to illuminate core ideas. Each area of law is different from the others and from the stylized three-step protocol that is analyzed in Part I. Moreover, doctrine in each area is to a degree murky, and the foregoing analysis sometimes calls into question whether common descriptions of the rules reflect actual practice. Rather than repeating these caveats throughout, I will simply proceed, often presenting points in blunt terms for purposes of brevity and clarity, at some expense to fidelity. Despite these qualifications, the general framework and analysis powerfully illuminate and in important ways reshape our understanding of the law in each of the three domains.

Part II examines antitrust law. The Supreme Court's canonical statements of the Sherman Act's rule of reason, which span a century, present it as a pure balancing test: liability turns on whether the challenged practice overall suppresses rather than promotes competition. Setting to the side some categorical carve-outs--such as the per se illegality of price fixing--commentators and courts increasingly restate this rule as a structured protocol that resembles the three-step stylization examined here. To that extent, the foregoing criticisms are apt.

For mergers, the structure is similar, although not often explicitly stated as such. Even though most nontrivial horizontal mergers a priori generate some upward pricing pressure, few mergers are challenged and some challenges fail in court, and this is so (according to conventional lore) without reaching the question of whether there are offsetting efficiencies (akin to step 2). This suggests a distinct step 1 with a high threshold. On those occasions in which that hurdle is overcome, government guidelines and practice then move to step 2 but in turn find that it usually fails, suggesting a significant threshold there as well--particularly because it is commonly accepted that many mergers are motivated by greater efficiency (which is the rationale for challenging so few). Once placed in this Article's framework, merger assessments can better be understood, criticized, and improved. One suspects, however, that regarding both information collection and decisionmaking, government agencies' internal analysis reflects a less constrained balancing process than official pronouncements suggest.

Part III turns to disparate impact cases under Title VII. Under prevailing Supreme Court precedent and the 1991 Amendments to the Civil Rights Act, they are governed by a structured decision procedure that has some resemblance to that examined here. (2) The first step of disparate impact inquiries, which focuses on the plaintiff's prima facie case (typically proved with statistics), suffers greatly from the problem of attempting to separate harm from justification, in a way that few have recognized. The second step (in some respects collapsed with the third, as will be discussed), on the defendant's business justification, is understood to raise a serious question regarding the threshold: specifically whether it is essentially zero ("job related" is taken to mean that any connection whatsoever between the discriminatory employment practice and productivity is sufficient), massive ("business necessity" is the operative phrase, and "necessity" means necessity!), or somewhere in between (which could be more akin to balancing). These steps, the role of alternative employment practices, queasiness about balancing in this context, and other matters are illuminated by this Article's framework.

Part IV considers the doctrine of strict scrutiny--and proportionality analysis (employed in many countries and proposed by some for the United States)--which likewise has a structure that bears some similarity to the stylized inquiry examined...

To continue reading