中文摘要
我们将开创性的 Schelling (1969) 隔离模型核心中的内生反馈循环纳入一个邻里选择的动态模型，并利用该模型研究 1990 年至 2004 年间塑造旧金山湾区种族隔离与收入隔离的力量。此类分析需要对家庭如何回应其邻居的社会经济构成进行因果识别。我们借助新颖的工具变量实现了这一点，这些工具变量可以在一个含摩擦的动态选择模型中得到合理化论证。这些工具变量具有潜在的广泛应用：研究沿任何可观测人口维度进行的分类、估计邻里构成对房价等结果的影响，或识别其他网络外部性。我们发现，基于种族和收入的歧视性（偏好性与统计性）分类普遍存在且错综复杂：几乎所有家庭对相似的邻居作出正面反应，对不同的邻居作出负面反应，尽管反应强度各不相同。尽管存在这些歧视性反应，摩擦——搬迁成本与不确定性——减轻了它们对隔离的影响。这意味着基于其他邻里便利设施进行的分类可能对隔离产生重大影响，并为基于地域的去隔离政策提供了依据。由于这些摩擦的存在，基于家庭重新配置的去隔离政策也存在取得成功的空间。
Abstract
We incorporate the endogenous feedback loop at the core of the seminal Schelling (1969) model of segregation into a dynamic model of neighborhood choice and use it to study the forces that shaped racial and income segregation in the San Francisco Bay area from 1990 to 2004. Such an analysis requires causal identification of households' responses to the socioeconomic composition of their neighbors. We achieve this with novel instrumental variables that can be rationalized with a dynamic choice model with frictions. These IVs have potentially broad application: studying sorting along any observable demographic dimension, estimating the effects of neighborhood composition on outcomes such as house prices, or identifying other network externalities. We find that discriminatory (taste‐based and statistical) sorting by race and by income is widespread and complex: almost all households respond positively to similar neighbors and negatively to different neighbors, although at varying degrees of intensity. In spite of these discriminatory responses, frictions—moving costs and uncertainty—mitigate their impact on segregation. This implies that sorting on the basis of other neighborhood amenities may have a large impact on segregation and may justify place‐based desegregation policies. Because of these frictions, there is also scope for desegregation policies based on the reallocation of households to succeed.