The 2023 Uruguayan Census recorded a population of 3,444,451 with an estimated undercoverage of 10.3%. Post-enumeration evidence shows that omission was non-random, concentrated in vulnerable areas, rural territories, and among young adults. Integrating administrative records (AR) recovered aggregate counts but did not resolve selection bias in outcome variables, as AR lack core census variables, exhibit urbanicity and institutional-visibility biases, and do not reconstruct households. Estimates derived from enumerated microdata remain biased. We treat effectively enumerated households as a non-probability sample with an unknown selection mechanism and construct weights using a doubly robust (DR) estimator. This framework combines a segment-level response-propensity model, using the web linkage rate as a contact proxy, with calibration to combined-census demographic totals (sex, age, department). Because the DR estimator is consistent when either model is correctly specified, it provides robustness against undercoverage misspecification. We describe the application at a scale of three million records, document its effect on social indicators, and present a variance approximation based on an equivalent stratified cluster design. Finally, we establish a methodological framework to guide national statistical offices on optimizing non-response adjustments based on their available registers and paradata.

翻译：2023年乌拉圭人口普查记录人口为3,444,451人，估计漏报率为10.3%。普查后证据表明，漏报并非随机，而是集中在脆弱区域、农村地区及年轻成年人群体中。整合行政记录（AR）虽恢复了汇总计数，但未能解决结果变量中的选择偏差问题，因AR缺乏核心普查变量，存在城市性与机构可见性偏差，且无法重建家庭单元。基于列举微观数据的估计仍然存在偏差。我们将有效列举的家庭视为具有未知选择机制的非概率样本，并采用双重稳健（DR）估计量构建权重。该框架结合了以网络链接率作为联系代理的区段级响应倾向模型，以及对联合普查人口统计总数（性别、年龄、省份）的校准。由于DR估计量在任一模型正确设定时均具有一致性，因此能针对漏报误设提供稳健性。我们描述了在三百万记录规模上的应用，记录了其对社会指标的影响，并提出了基于等效分层整群设计的方差近似方法。最后，我们建立了一个方法论框架，用以指导国家统计机构根据其可用的登记数据和辅助数据优化无响应调整。