Next to having one's book prominently displayed on the racks of every airport bookstore in the country, receiving recognition for an article written more than half a century ago is a close second. Indeed, for some of us it is the very highest form of intellectual acknowledgment. The Shapley–Shubik power index, invented more than 50 years ago, is a familiar concept in the analytical lexicon of political science. In brief, it is a measure of the ex ante likelihood that an individual will be pivotal in transforming a nonwinning coalition into a winning one. This depends on the voting or decision rule, on the one hand, and on the number of votes or decision weight possessed by an individual, on the other. The “power” of an individual, so determined, may be aggregated to measure the power of a faction, party, proto-coalition, or voting chamber (the latter in a multicameral setting).
Power is in quotation marks earlier because, as a discipline, political science has never completely resolved exactly what is meant by the concept. Indeed, Riker (1964), in an article that should be more widely acknowledged, concludes that the concept is so riddled with ambiguities that we should ban it from our vocabulary. (That advice has stuck with me all the years since I first encountered it as a graduate student, so that I still have trouble writing out the word.) Nevertheless, Shapley and Shubik's short Review article stimulated a light industry of efforts to refine an index of power. Early efforts by Dahl (1957, 1961) and March (1957), stimulated in part by Shapley and Shubik's essay and culminating in the oft-cited piece by Bachrach and Baratz (1962), represented an early trickle that became a steady flow for much of the rest of the century, with a recent surge as Europeanists became absorbed with reforms in European Union institutions.
Political scientists, enamored as they are of the study of power, have long sought a simple, compact measure. Shapley and Shubik, employing the logic of simple games in von Neumann and Morgenstern's cooperative game theory, provided just such a measure. What it provides is a mechanical characterization, stripped bare of “any of the sociological or political superstructure that almost invariably exists in a legislature or policy board” (Shapley and Shubik 1954, 791). That is, it takes seriously only the mechanism—voting rule, decision procedure, game form—by which official resources are transformed into outcomes. Thus, their measure of power provides a causal account of political results. This link between power and cause is, I believe, one of the very important reasons this paper and many like it have had such impact on the study of politics.
Reading it (again) a half-century after it appeared, I am struck by several things. First, there is humility. Shapley and Shubik believed, and many formal theorists still believe, that the early steps in attacking a problem should consist of simple, stripped down, almost toylike models of a process with the aim of pinning down the operating characteristics of a decision-making mechanism in its purest form. From this base one can add complications (“sociological or political superstructure”) or, in an equilibrium framework, do comparative statics analysis from which empirical expectations may be derived.
Second, there are nonobvious conclusions, one of the hallmarks of their formal approach. Consider three.
- Power as they measure it is not proportional to votes (or other political resources), for changes in the probability of being pivotal are not proportional to them.
- In their simple tricameral characterization of the American system of legislation, with the ratio of “votes” being (in 1954) 1:96:435 for the president, Senate, and House, respectively (ignoring overrides of presidential vetoes), and where the agreement of each body is required to pass legislation, the ratio of “power” nevertheless amounts approximately to 2:1:1—with the president as powerful as both legislative chambers combined.
- It is obvious that someone controlling a decisive proportion of votes (say, the majority of shares in a publicly traded corporation) possesses all the power. What is less obvious is that someone controlling some votes may have no power, even in less extreme circumstances. Such an agent is labeled a dummy player in the game. He or she is never in a position to transform a nonwinning coalition into a winning one. A four-player majority-rule game with votes for each player given by {3, 3, 2, 1} is such a case. The fourth player is neither sufficient nor necessary to the “winning” status of any coalition of which he or she is a part. (Riker [1964] points out that a weighted voting scheme among towns on Long Island in the 1960s actually created one town as a dummy in this sense.)
Finally, this article constitutes an effort to bring rigor to an ambiguous concept, not by replacing more conventional modes but instead by supplementing them with mathematical and other forms of deductive reasoning. By focusing analytical methods on a phenomenon arising in well-described empirical settings, Shapley and Shubik's article constitutes one of the earliest exemplars of formal political theory that has come to be a prominent feature of the modern scholarly landscape.
Lloyd Stowell Shapley is Professor Emeritus of Mathematics at the University of California, Los Angeles. His many honors include membership in the American Academy of Arts and Sciences and the National Academy of Sciences, along with the receipt of the John von Neumann Theory Prize.
Martin Shubik is the Seymour H. Knox Professor of Mathematical Institutional Economics at Yale University. Specializing primarily in strategic analysis, he is the author of approximately twenty books and over 200 articles.

Comments
No Comments have been published for this article.