We consider how a firm dynamically allocates business among several suppliers to motivate them in a relational contract. The firm chooses one supplier who exerts private effort. Output is non-contractible, and each supplier observes only his own relationship with the principal. In this setting, allocation decisions constrain the transfers that can be promised to suppliers in equilibrium. Consequently, optimal allocation decisions condition on payoff-irrelevant past performance to make strong incentives credible. We construct a dynamic allocation rule that attains first-best whenever any allocation rule does. This allocation rule performs strictly better than any rule that depends only on payoff-relevant information.

Policy makers, firms, and researchers often choose among multiple options based on estimates. Sampling error in the estimates used to guide choice leads to a winner’s curse, since we are more likely to select a given option precisely when we overestimate its effectiveness. This winner’s curse biases our estimates for selected options upward and can invalidate conventional confidence intervals. This article develops estimators and confidence intervals that eliminate this winner’s curse. We illustrate our results by studying selection of job-training programs based on estimated earnings effects and selection of neighborhoods based on estimated economic opportunity. We find that our winner’s curse corrections can make an economically significant difference to conclusions but still allow informative inference.

A researcher can use a tightly parameterized structural model to obtain internally consistent estimates of a wide range of economically interesting targets. We ask how reliable these estimates are when the researcher’s model may be misspecified. We focus on the case of multivariate, potentially nonlinear models where the causal variable of interest is endogenous. Reliable estimates require that the researcher’s model is flexible enough to describe the effects of the endogenous variable approximately correctly. Reliable estimates do not require that the researcher has correctly specified the role of the exogenous controls in the model. However, if the role of the controls is misspecified, reliable estimates require a property we call strong exclusion. Strong exclusion depends on having sufficiently many instruments that are unrelated to the controls. We discuss how practitioners can achieve strong exclusion, and illustrate our findings with an application to a differentiated goods model of demand for beer.

We propose a local measure of the relationship between parameter estimates and the moments of the data they depend on. Our measure can be computed at negligible cost even for complex structural models. We argue that reporting this measure can increase the transparency of structural estimates, making it easier for readers to predict the way violations of identifying assumptions would affect the results. When the key assumptions are orthogonality between error terms and excluded instruments, we show that our measure provides a natural extension of the omitted variables bias formula for nonlinear models. We illustrate with applications to published articles in several fields of economics.

We propose a way to formalize the relationship between descriptive analysis and structural estimation. A researcher reports an estimate ĉ of a structural quantity of interest c that is exactly or asymptotically unbiased under some base model. The researcher also reports descriptive statistics<a:math xmlns:a="http://www.w3.org/1998/Math/MathML" display="inline"><a:mover accent="true"><a:mi>γ</a:mi><a:mo>ˆ</a:mo></a:mover></a:math>that estimate features γ of the distribution of the data that are related to c under the base model. A reader entertains a less restrictive model that is local to the base model, under which the estimate ĉ may be biased. We study the reduction in worst‐case bias from a restriction that requires the reader's model to respect the relationship between c and γ specified by the base model. Our main result shows that the proportional reduction in worst‐case bias depends only on a quantity we call the informativeness of<d:math xmlns:d="http://www.w3.org/1998/Math/MathML" display="inline"><d:mover accent="true"><d:mi>γ</d:mi><d:mo>ˆ</d:mo></d:mover></d:math>for ĉ . Informativeness can be easily estimated even for complex models. We recommend that researchers report estimated informativeness alongside their descriptive analyses, and we illustrate with applications to three recent papers.