This website uses cookies.

By clicking the "Accept" button or continuing to browse our site, you agree to first-party and session-only cookies being stored on your device to enhance site navigation and analyze site performance and traffic. For more information on our use of cookies, please see our Privacy Policy.

  1. Home
  2. ASSA Annual Meeting
  3. 2027
  4. Preliminary Program
  5. Model Misspecification: Prediction, Choice, and Equilibrium
« Back to Results

Model Misspecification: Prediction, Choice, and Equilibrium

Paper Session

Tuesday, Jan. 5, 2027 10:15 AM - 12:15 PM (EST)

Marriott Marquis Washington DC
Hosted By: Econometric Society
  • Chair: Laura Doval, Columbia University

Misspecified Model Estimation and Its Impact on Predictions

Anqi Li
,
University of Waterloo
Junnan He
,
Sciences Po
Lin Hu
,
Australian National University
Matthew Kovach
,
Purdue University

Abstract

"We study a linear statistical model where outcomes depend on regressors with

fixed population coefficients and observation-specific latent coefficients, along

with measurement errors. A decision-maker estimates population coefficients

and uses the estimates to predict the latent coefficients for a given observation.

We analyze how misspecification of some population coefficients distorts predic-

tions, investigating comparative statics with respect to: (1) residual information

in regressors associated with misspecified coefficients after projecting out those

associated with free coefficients, (2) alignment between misspecification vector

and latent-to-coefficient mapping. Applications include employee rating with un-

conscious bias and LLM-mediated consumer research."

Contracting with Simplified Models

Hassan Sayed
,
Center for Global Development

Abstract

A receiver contracts a sender to help her make a choice. The sender estimates a "model" mapping the receiver's choices to expected outcomes. The receiver does not know whether the sender estimated the model on data "locally" tailored to a specific choice, or "distantly" across many plausible choices. The sender's models are also simplified: they may understate or neglect tradeoffs the receiver thinks are important, or overstate tradeoffs she thinks are unimportant. I show that senders who overstate tradeoffs estimate models on "local" data, while those whose models neglect tradeoffs estimate models on "distant" data. The receiver has lower forecast error when provided a model estimated on "local" data, and thus is better off contracting with a sender who emph{overstates} tradeoffs. I characterize when senders gain from providing simplified models, and show that the optimal method of "oversimplification" involves choosing a fixed (but arbitrary) degree of modeling a tradeoff. Finally, I study competition for information provision between multiple senders. Senders spatially sort across the receiver's choice space: those who tend to underestimate tradeoffs locate themselves in the center, while those who overstate tradeoffs locate on the edge. When the number of potential senders is sufficiently high and costs of information provision are sufficiently low, the senders' "gains from simplicity" vanish, as the receiver is able to contract multiple senders and uncover the true data-generating process.

Monotone Comparative Statics with Misspecified Models

Abhishek Ananth
,
Emory University
Aniruddha Ghosh
,
California Polytechnic State University

Abstract

The theory of monotone comparative statics exploits ordinal structure to deliver robust qualitative predictions in decision problems under uncertainty. Existing results, however, typically assume that the true distribution governing uncertainty is known to the decision-maker. We relax this assumption by studying monotone comparative statics when a misspecified decision-maker evaluates uncertainty using a projection of the true distribution onto a restricted model class. This setup encompasses several prominent behavioral and robust decision frameworks, including models of misspecified beliefs and ambiguity aversion. Our main result characterizes intuitive conditions under which monotone comparative statics are preserved with respect to the true distribution, as well as other parameters of the decision problem. These conditions impose purely ordinal restrictions on the preferences as well as on the projection. We illustrate the usefulness of our results in canonical decision problems (including treatment targeting and auctions) and discuss implications for computation and model inference.
JEL Classifications
  • C1 - Econometric and Statistical Methods and Methodology: General
  • D8 - Information, Knowledge, and Uncertainty