Abstract: We propose a framework to adjust the predictions based on an economic model, and the estimates of the model parameters, when the model may be misspecified. Our approach consists in minimizing the sensitivity of the estimates to the type of misspecification that is most influential for the parameter of interest. We rely on a local asymptotic approach where the degree of misspecification is indexed by the sample size, and derive formulas to construct estimates whose mean squared error is minimax, based on simple one-step adjustments. We construct confidence intervals that contain the true parameter under both correct specification and local misspecification. We calibrate the degree of misspecification using a model detection error approach, which allows us to perform systematic sensitivity analysis in both point-identified and partially-identified settings. As main illustrations we study panel data models where the distribution of individual effects may be misspecified and the number of time periods is small, and we revisit the structural evaluation of a conditional cash transfer program in Mexico.