PLRD: Partially Linear Regression Discontinuity Inference -- by Aditya Ghosh, Guido Imbens, Stefan Wager

TEXT START: Regression discontinuity designs have become one of the most popular research designs in empirical economics.

The Dissection

This paper is an engineering upgrade to econometric inference. It proposes PLRD to reduce estimation error and produce shorter, valid confidence intervals, using simulations calibrated to twelve real applications through Wasserstein generative adversarial networks. Its subject is measurement quality, not economic survival.

The Core Fallacy

The relevant DT error is category confusion: sharper causal estimates do not restore the mass employment → wage → consumption circuit. PLRD may reveal more accurately whether a policy changes outcomes near a threshold. It cannot restore human productive necessity, bargaining power, ownership of AI capital, or institutional capacity under P1–P3. The paper itself makes a narrower methodological claim; the fallacy appears when improved inference is mistaken for improved systemic control.

Hidden Assumptions

• Local regression-discontinuity effects remain relevant to decisions beyond the cutoff.
• Threshold rules, treatment implementation, and institutional conditions remain stable.
• GAN-generated data preserve every feature relevant to comparing estimators.
• Large-sample guarantees translate to messy, finite, shifting real-world datasets.
• Statistically shorter confidence intervals produce better policy decisions.
• Policymakers retain the authority and resources to act on better estimates.

Social Function

Classification: partial truth, prestige signaling, and transition management. The method may genuinely improve empirical credibility. It also supplies institutions with a more precise instrument for measuring and administering programs during decline. Used as proof that better evidence can preserve the existing order, it becomes ideological anesthetic: a cleaner dashboard over a failing machine.

The Verdict

PLRD is potentially useful econometric infrastructure, but it is irrelevant to the terminal mechanism identified by the Discontinuity Thesis. It can sharpen the autopsy of local policies; it cannot prevent the death of the system those policies inhabit.