The day in conversation for 2026-11-03

discussion

Transcript

The discussion: Ruth: I’m focusing on this paper on encouragement designs—the claim about deriving sharp inequalities based on conditional distributions—it sounds powerful because it promises testable implications where we usually get fuzzy boundaries in these types of experiments. Warren: Exactly, the sharpness is what makes it useful for pinpointing substitution patterns, but I wonder how robust those conditional distribution assumptions hold up when we move from simulated environments to real-world settings where the underlying distributions are likely misspecified? Iris: That’s a key point. And it makes me think about Paper 3, the one on within-household designs with interference. If that paper uses game theory to establish sharp bounds for effects even when spillovers are involved, does that sharpness necessarily break down if the households aren't operating under the specific interaction structure assumed by the game? Ruth: I’m less concerned about the interaction structure and more about how those bounds behave when we introduce external constraints. For example, if we look at Paper 2 on affirmative action assignments, where they found an impossibility regarding cumulative fractional counts based

More episodes

← Home