Migration and the epidemiological approach: time and self-selection into foreign ancestries matter

The epidemiological approach in comparative development uses data on individuals of immigrant origin to study cultural persistence, the determinants of cultural norms, and the effects of genetic traits. A common assumption of this methodology is its susceptibility to attenuation bias. We challenge it by demonstrating how the increasing reliance on foreign ancestries to identify respondents’ origins can introduce confounding biases. Specifically, self-selection in reporting foreign ancestry and unobserved variation in ancestral migration timing may lead to inflated estimates. We formalize these mechanisms through a theoretical framework and illustrate their empirical significance by reassessing key findings from influential studies by Fernández and Fogli (2006) and Giuliano and Nunn (2021).

比较发展领域中的流行病学研究范式，以移民背景个体的相关数据为研究样本，旨在探究文化持续性、文化规范的决定因素以及遗传性状的影响。该研究方法的一项通用预设为其易受衰减偏误（attenuation bias）的干扰。本文对此预设提出挑战，论证了愈发依赖外来祖先信息以界定受访者出身背景的做法，可能引入混淆性偏误（confounding bias）。具体而言，受访者在申报外来祖先身份时的自我选择行为，以及祖先移民时间维度上的未观测异质性，均可能导致估计结果被高估。本文通过构建理论框架对上述作用机制进行形式化刻画，并通过重新评估费尔南德斯与福利（Fernández and Fogli, 2006）、朱利亚诺与纳恩（Giuliano and Nunn, 2021）两项经典研究的核心结论，阐明了此类偏误的实证显著性。