本文系统梳理了结构流动研究在理论支撑、混淆因素控制、概念化、统计模型复杂度以及统计推断等方面的局限，并提出了调整—分解的分析策略。通过综合采用德明—斯蒂芬算法和流动表的三元分解法，这一分析策略直接针对流动表的边际分布构建“反事实”流动表，并以非建模对比的方式呈现结构流动、交换流动以及二者的交互作用。除此之外，潜在的混淆因素可以通过调整的方法予以控制，相关的统计推断可以通过参数分布建模或者自助法来实现。本文通过两个经验示例对这一分析策略进行了呈现。

This study systematically reviews the limitations of structural mobility research in terms of theoretical support, confounding factor control, conceptualization, complexity of statistical models, and statistical inference. It proposes a calibration-decomposition analytical strategy. By combining the Deming-Stephan algorithm and the trivariate decomposition method of mobility tables, this analytical strategy directly constructs a "counterfactual" mobility table based on the marginal distribution of the original mobility table. It presents structural mobility, exchange mobility, and their interaction through a non-parametric approach. In addition, potential confounding factors can be controlled through adjustments, and related statistical inference can be performed through parametric distribution assumption or bootstrapping. This study illustrates this analytical strategy through two empirical examples.

本文得到国家社会科学基金万人计划哲学社会科学领军人才项目（22VRC140）的支持。

欢迎阅读《社会学研究》！您是该文第110位读者！