Jeffrey S. Banks

A Bargaining Model of Collective Choice

We provide a general theory of collective decision making, one that relates social choices to the strategic incentives of individuals, by generalizing the Baron-Ferejohn (1989) model of bargaining to the multidimensional spatial model. We prove existence of stationary equilibria, upper hemicontinuity of equilibrium outcomes in structural and preference parameters, and equivalence of equilibrium outcomes and the core in certain environments, including the one-dimensional case. The model generates equilibrium predictions even when the core is empty, and it yields a “continuous” generalization of the core in some familiar environments in which the core is nonempty. As the description of institutional detail in the model is sparse, it applies to collective choice in relatively unstructured settings and provides a benchmark for the general analysis of legislative and parliamentary politics.

Sophisticated voting outcomes and agenda control

Necessary and sufficient conditions for an alternative to be a sophisticated voting outcome under an amendment procedure are derived. The uncovered set, as first defined by Miller (1980), is shown to be potentially reducible, and conditions are determined for which this reduction equals the set of sophisticated voting outcomes. In addition, simple methods are given for calculating both the uncovered set and its reduction.

Allocating uncertain and unresponsive resources: an experimental approach

We identify an important class of economic problems that arise naturally in several applications: the allocation of multiple resources when there are uncertainties in demand or supply, unresponsive supplies (no inventories and fixed capacities), and significant demand indivisibilities (rigidities). Examples of such problems include: scheduling job shops, airports, or supercomputers; zero-inventory planning; and the allocation and pricing of NASAs planned Space Station. Using experimental methods, we show that the two most common organizations used to deal with this problem, markets and administrative procedures, can perform at very low efficiencies (60-65% efficiency in a seemingly robust example). Thus, there is a need to design new mechanisms that more efficiently allocate resources in these environments. We develop and analyze two mechanisms that arise naturally from auctions used to allocate single-dimensional goods. These new mechanisms involve computer-assisted coordination made possible by the existence of networked computers. Both mechanisms significantly improve on the performance of administrative and market procedures.

A model of electoral competition with incomplete information

A model of two-candidate electoral competition is developed in which voters are uncertain about the policy either candidate would implement if elected. Candidates simultaneously announce policy positions, from which voters attempt to infer the true positions the candidates would adopt. Announcing a position different from the true position is costly to the winning candidate, with these costs increasing as the difference between the true policy and the announced policy increases. A refinement of the sequential equilibrium concept is used to describe the behavior of candidates and voters.

The Political Control of Bureaucracies under Asymmetric Information

We analyze a model of interest group influence where this influence arises from a groups ability to provide relevant information to political actors about regulatory performance. The model rests on three premises: (I) bureaucrats possess an informational advantage, vis-a-vis their political overseers, concerning policy relevant variables; (2) this informational advantage is potentially exploitable by the agency; and (3) politicians anticipate this exploitation and adapt to it in their design of agencies. We show that a major consequence of this anticipation and adaptation is service to organized interest groups.

Explaining Patterns of Candidate Competition in Congressional Elections

The low probability of defeating incumbent members of Congress deters potentially strong rivals from challenging them. Yet almost all incumbents are challenged, usually by opponents who lack previous experience in office and run underfinanced, ineffectual campaigns. But if strong challengers are deterred from challenging incumbents, why are not weak challengers, who have even less chance of unseating an incumbent? The model developed in this paper indicates that there is a simple reason why weak candidates choose to run against incumbents: they do so in order to maximize their probability of getting elected to Congress. Together with the findings of previous researchers, the results of our analyses of congressional primary data from 1980 through 1984 provide strong support for the major hypotheses derived from our model.

Theory, experiment and the federal communications commission spectrum auctions

Abstract The FCC uses the simultaneous multi-round auction (SMA) to assign spectrum licenses. Congress mandated that the SMA be compared to an auction that allows “combinatorial” bids. We review the theoretical and experimental evidence relevant to the SMA procedures and their inherent defects. Results from experiments evaluating the SMA in comparison to a combinatorial auction designed by Charles River and Associates show that several of the SMA rules hinder efficiency and create a trade-off between efficiency and auction length. The combinatorial auction outperforms the SMA when license values are superadditive, but requires more time to complete and is not robust.

Agency Budgets, Cost Information, and Auditing

A model of legislature-agency interaction is analyzed where the agency possesses an informational advantage in that only it knows the cost of its services. The legislature has the ability to audit the agency, where auditing is a costly means of verifying the agencys information. Two different procedures are analyzed for determining the agencys budget: in one, the agency makes a budget request, after which the legislature can either accept or reject the request, or audit the agency and impose a budget equal to the true cost of services. In the other procedure, the legislature can follow a request with a counterproposal to the agency, which can then either accept or reject. Since under both procedures auditing is costly, it will be optimal for the legislature to refrain from auditing a request if the perceived benefits do not outweigh this cost. At issue is the ability of the legislature to impose discipline on the agencys request and final budget through an optimal choice of audit and counterproposal strategies, the extent of the information transmitted through the agencys budget request, and the efficiency of the resulting outcomes. A refinement of the sequential equilibrium concept provides the behavioral predictions for the procedures.

Switching Costs and the Gittins Index

The Theorem of Gittins and Jones (1974) is, perhaps, the single most powerful result in the literature on Bandit problems. This result establishes that in independent-armed Bandit problems with geometric discounting over an infinite horizon, all optimal strategies may be obtained by solving a family of simple optimal stopping problems that associate with each arm an index known as the dynamic allocation index or, more popularly, as the Gittins index. Importantly, the Gittins index of an arm depends solely on the characteristics of that arm and the rate of discounting, and is otherwise completely independent of the problem under consideration. These features simplify significantly the task of characterizing optimal strategies in this class of problems.

Probabilistic Voting in the Spatial Model of Elections: The Theory of Office-motivated Candidates

We unify and extend much of the literature on probabilistic voting in two-candidate elections. We give existence results for mixed and pure strategy equilibria of the electoral game. We prove general results on optimality of pure strategy equilibria vis-a-vis a weighted utilitarian social welfare function, and we derive the well-known “mean voter” result as a special case. We establish broad conditions under which pure strategy equilibria exhibit “policy coincidence,” in the sense that candidates pick identical platforms. We establish the robustness of equilibria with respect to variations in demographic and informational parameters. We show that mixed and pure strategy equilibria of the game must be close to being in the majority rule core when the core is close to non-empty and voters are close to deterministic. This controverts the notion that the median (in a one-dimensional model) is a mere “artifact.” Using an equivalence between a class of models including the binary Luce model and a class including additive utility shock models, we then derive a general result on optimality vis-a-vis the Nash social welfare function.