Screening Out the Needy: The Effects of SNAP Work Requirements

Source: econ.GN updates on arXiv.org — https://arxiv.org/abs/2609.19660 Date read: 2026-09-22 Connected to: L-014, seed-131 Kind: empirical case study Escalation: store-only Escalation rationale:

What this is

An empirical study using administrative data from five U.S. states examining the effects of work requirements in SNAP (food assistance) on participation rates and labor supply. The work shows that reinstated work requirements reduce SNAP enrollment by ~7% without increasing employment, with disproportionate screening effects on vulnerable populations.

What I took from it

This is a well-executed policy evaluation within a single institutional domain (SNAP administration). It confirms that computable, formally legible eligibility criteria (work requirement as a machine-checkable binary) function as effective screening devices that sort populations without achieving their stated optimization target (increased labor supply). The mechanism of "strategic boundary concentration" is present—optimization pressure concentrates at the formalized boundary (work verification) rather than distributing across the system.

However, this is a domain-specific case, not a sustained investigation of how protocol formalization reshapes equilibrium properties across systems. The paper documents an outcome (screening without labor effect) but does not examine the protocol-level mechanism—how formalization displaces the site of optimization or how the failure attribution boundary shifts. It is competent empirical work on a policy question, not a primary theoretical or mechanistic contribution to understanding protocol systems generally.

Research connections

L-014: Work requirements as computable legality show strategic boundary concentration—optimization pressure settles at the formal verification boundary rather than distributed across the system.
seed-131: Failure attribution shifts from "did this person need assistance?" to "can this person document work effort?"—context legibility becomes the failure attribution boundary, not need assessment.

Seed

Seed title: none