In many areas, practitioners seek to use observational data to learn a treatment assignment policy that satisfies application‐specific constraints, such as budget, fairness, simplicity, or other functional form constraints. For example, policies may be restricted to take the form of decision trees based on a limited set of easily observable individual characteristics. We propose a new approach to this problem motivated by the theory of semiparametrically efficient estimation. Our method can be used to optimize either binary treatments or infinitesimal nudges to continuous treatments, and can leverage observational data where causal effects are identified using a variety of strategies, including selection on observables and instrumental variables. Given a doubly robust estimator of the causal effect of assigning everyone to treatment, we develop an algorithm for choosing whom to treat, and establish strong guarantees for the asymptotic utilitarian regret of the resulting policy.

This paper develops a generalization of the widely used difference-in-differences method for evaluating the effects of policy changes. We propose a model that allows the control and treatment groups to have different average benefits from the treatment. The assumptions of the proposed model are invariant to the scaling of the outcome. We provide conditions under which the model is nonparametrically identified and propose an estimator that can be applied using either repeated cross section or panel data. Our approach provides an estimate of the entire counterfactual distribution of outcomes that would have been experienced by the treatment group in the absence of the treatment and likewise for the untreated group in the presence of the treatment. Thus, it enables the evaluation of policy interventions according to criteria such as a mean-variance trade-off. We also propose methods for inference, showing that our estimator for the average treatment effect is root-N consistent and asymptotically normal. We consider extensions to allow for covariates, discrete dependent variables, and multiple groups and time periods. Copyright The Econometric Society 2006.

Researchers often report estimates and standard errors for the object of interest (such as a treatment effect) based on a single specification of a statistical model. We propose a systematic approach to assessing sensitivity to specification. We construct estimates of the object of interest for each of a large set of models. Our proposed robustness measure is the standard deviation of the point estimates over the set of models. Each member of the set is generated by splitting the sample into two subsamples based on covariate values, constructing separate parameter estimates for each subsample, and then combining the results.

This paper studies the problem of allocating a good between two players in each period of an infinite-horizon game. The players' valuations in each period are private information, and the valuations change over time. We analyze two special cases for the dynamics of valuations: correlated where players' valuations are exogenous but serially correlated; and learning by doing, where a player's past consumption improves his current distribution of valuations, but his valuations are otherwise uncorrelated. We analyze conditions under which there exists an efficient, Bayesian incentive-compatible (BIC), individually rational (IR), budget-balanced (BB) mechanism, when the mechanism designer has commitment power. We consider

Where should decision rights be lodged in organizations? Michael C. Jensen and William H. Meckling (1992) argue that moving a decision away from the inherently best-informed party involves costs in communication and garbling but may lodge it with someone who has better incentives to make good decisions. Generally, however, we expect that incentives are part of the organizational design. Why not just provide incentives to those with the best information so that they make the right decisions? One reason is that the available incentive instruments must serve multiple purposes, and designing them to induce better decisions worsens performance against other organizational objectives. Our experience suggests that this is a common situation in actual organizations: the means available to affect one sort of behavior or decision inevitably affect the incentives governing other choices. Then, the design of incentive schemes and the allocation of decision rights become interlinked. This paper looks at this idea in the specific context of a principal’s problem of inducing agents to provide unobservable effort while also motivating the efficient selection of investments. Each of these problems has been extensively studied in isolation (on inducing effort, see e.g., Bengt Holmstrom [1979] and Holmstrom and Paul Milgrom [1991]; on decisions, see e.g., Eugene F. Fama and Jensen [1983], Milgrom and Roberts [1990a, b], Philippe Aghion and Jean Tirole [1997], Matthias Dewatripont and Tirole [1999]). We thus know that motivating effort is done best by rewarding agents on precise measures of their effort, not necessarily on the total value created in the firm. At the same time, it is clear that getting the right investment choices may require that the decision-makers’ rewards be tied to total value created. The difficulty is that the available measures do not allow doing both. The only available performance measures are aggregates whose component pieces cannot be disentangled, while contracts must be written in advance of learning about investment possibilities. Including many contingencies in the contracts ex ante is impossible, and ongoing renegotiation in every ex post eventuality is prohibitively costly. More formally, we assume that it is not possible to contract on investment projects, nor can the principal bargain with the agents over the adoption of these projects once they are identified. Instead, returns to projects are reflected in the performance measures available for use in the effort-incentive contracting. Then the incentives for effort and for decisions are inextricably tied together. In this framework, we explore the interactions among the design of jobs and assignment of individuals to tasks, the shape and intensity of effort incentives, and the allocation of authority over project selection. We argue that it may indeed be optimal to assign decisions rights to someone other than the best-informed party. An authority-based hierarchy then emerges endogenously, with some agents being given the right to make organizational decisions over projects that others discovered. Moreover, as in Herbert Simon (1951), those in authority will make the decisions in a self-interested way. Simon emphasized that this † Discussants: Michael Riordan, Columbia University; Bengt Holmstrom, Massachusetts Institute of Technology; W. Bentley MacLeod, University of Southern California.

This paper analyzes the role of private information in U.S. Forest Service timber auctions. In these auctions, firms bid a per unit price for each timber species. Total bids are computed by multiplying these prices by Forest Service volume estimates, but payments depend on actual volumes harvested. We develop an equilibrium theory for these auctions. We then relate (ex post) data about volume to (ex ante) bids. We show that bidders have private information about volumes of species and use it as predicted by theory. Differences in bidder estimates appear to affect the allocation of tracts, but competition limits information rents.

We study entry and bidding patterns in sealed bid and open auctions with heterogeneous bidders. Using data from U.S. Forest Service timber auctions, we document a set of systematic effects of auction format: sealed bid auctions attract more small bidders, shift the allocation towards these bidders, and can also generate higher revenue. We show that a private value auction model with endogenous participation can account for these qualitative effects of auction format. We estimate the model's parameters and show that it can explain the quantitative effects as well. Finally, we use the model to provide an assessment of bidder competitiveness, which has important consequences for auction choice.

This is a submitted manuscript of an article published by the American Economic Association in the American Economic Review.

How much discretion should the monetary authority have in setting its policy? This question is analyzed in an economy with an agreed-upon social welfare function that depends on the economy's randomly fluctuating state. The monetary authority has private information about that state. Well designed rules trade off society's desire to give the monetary authority discretion to react to its private information against society's need to prevent that authority from giving in to the temptation to stimulate the economy with unexpected inflation, the time inconsistency problem. Although this dynamic mechanism design problem seems complex, its solution is simple: legislate an inflation cap. The optimal degree of monetary policy discretion turns out to shrink as the severity of the time inconsistency problem increases relative to the importance of private information. In an economy with a severe time inconsistency problem and unimportant private information, the optimal degree of discretion is none.

We study how diversity evolves at a firm with entry-level and upper-level employees who vary in ability and “type” (gender or ethnicity). The ability of entry-level employees is increased by mentoring. An employee receives more mentoring when more upper-level employees have the same type. Optimal promotions are biased by type, and this bias may favor either the minority or the majority. We characterize possible steady states, including a “glass ceiling,” where the upper level remains less diverse than the entry level. A firm may have multiple steady states, whereby temporary affirmative-action policies have a long-run impact.