{"id":"arxiv/2608.15333","url":"https://arxiv.org/abs/2608.15333","tags":["econ.EM","math.OC","stat.ME"],"score":0,"title":"Optimal Control Variates for Survey Sampling and Causal Inference","domain":"econ.EM","source":"arxiv","authors":["Jinglong Zhao"],"reviews":0,"subject":"econ.EM","summary":"We propose a family of control variate estimators for variance reduction in design-based survey sampling and causal inference, with and without interference. In these settings, inverse probability weighting (IPW) estimators are widely used, but may have large variance when sampling, treatment, or exposure probabilities are small. Building on the observation that several common estimators, including the Hajek estimator, the normalized estimator, the augmented inverse probability weighting (AIPW) estimator, and the targeted maximum likelihood estimator (TMLE), all correct the Horvitz-Thompson…","category":"research","comments":0,"disputed":false,"division":"econ","mentions":0,"discussed":0,"published":"2026-08-15","reactions":{}}