The day in conversation for 2026-11-09
discussion
Transcript
The discussion: Iris: I’m still wrestling with how convincingly the decision models reproduce actual human responses, especially across different types of spending surveys. It seems like while they capture heterogeneity better than LLMs, we haven't quite quantified *why* the specific structure of these DMs yields that improved fidelity; is it purely the functional form or something deeper about how they internalize preferences? Ruth: I agree about the fidelity, but what concerns me is extrapolating those results. If a decision model performs exceptionally well in a controlled simulation environment, does that guarantee its utility when we introduce genuine market shocks or non-linear feedback loops that weren't explicitly built into the DM structure? We need to test if this predictive power holds when the underlying economic regime shifts. Warren: That ties into my concern about identification challenges. When we look at discrete choice under congestion, we find an equilibrium exists even with those externalities, but how robust is that equilibrium when we move away from perfect information? Specifically, how much does the welfare calculation change if the congestion parameter itself becomes stochastic rather than fixed? Iris: The stochastic aspect is tricky because it interacts with our identification methods. When we look at sharp identification for missing covariates, the method relies on a specific structure for those gaps—intervals or perhaps censoring—and if we introduce that randomness into the congestion term, does our interval-observed framework still provide
More episodes
- Daily Summary for 2026-11-19
- 2608.21274-Recommendation Quality and the Concentration of Consumption: Experimental Evidence from Netflix
- 2408.02757-A nonparametric test for diurnal variation in spot correlation processes
- 2306.16393-High-Dimensional Canonical Correlation Analysis
- 2605.17299-Geometric Brownian motion with intermittent entries and exits
- 2607.04753-Fooling Yourself: how narratives shape beliefs
- 2408.14671-Inference on Treatment Effects with Repeatedly Mismeasured High-Dimensional Controls
- 2201.06898-Difference-in-Differences Estimators for Treatments Continuously Distributed at Every Period
- 2501.15307-Influence Functions: A Unified Differentiation-Based Framework
- The day in conversation for 2026-11-19