中文摘要：

虽然双系统估计量目前是人口普查质量评估领域估计总体实际人口数的主要方法，但其内在固有的缺陷却依然存在，即由于人口普查与其质量评估调查不独立引起的交互作用偏差使其低估或高估人口数。独立性假设失败源于在普查及其质量评估调查中登记的因果相关性，以及在普查及质量评估调查中登记概率的异质性。Bell模型是当前公认的测算交互作用偏差的有效方法。该方法建立在0-17岁及成年女性的双系统估计值不存在交互作用偏差的假设条件下。利用美国普查局提供的2010年资料全面展示了双系统估计量交互作用偏差的测算过程。实证结果表明，黑人成年男性的双系统估计值存在显著的交互作用偏差。研究有助于我国在未来人口普查质量评估工作中意识到交互作用偏差的存在，把测算的交互作用偏差添加到双系统估计量估计的人口数中，并依据修正后的双系统估计值计算人口普查净误差。

英文摘要：

Although dual system estimator is now main method to estimate the true population in the field of census quality assessment, its inherent defect still exists, that is that correlation bias resulted from nonindependence between census and its quality assessment survey makes dual system estimator systematically underestimate or overestimate the true population. This independence assumption can fail due either to casual dependence between the act of inclusion in the census and the act of inclusion in the quality assessment survey or due to heterogeneity across persons in the probabilities of being included in the census and in the quality assessment survey. Bell model is now regarded as effective method to measure correlation bias. This method is based on the assumption that dual system estimates of 0-17 and adult female have no correlation bias. Using the 2010 data provided by the Census Bureau of the United States fully demonstrates the process of estimating correlation bias of dual system estimates. The empirical results show that dual system estimates of black adult males have significant correlation bias. This research helps our country to realize correlation bias in the future of census quality assessment, to add estimated correlation bias to dual system estimates, and then figure census net error according to being adjusted dual system estimates.