Hwang, Y. (2025) Journal of Business & Economic Statistics, 24,(25 (1),), 1–25 [ONS LS]

Other information:
Abstract:

This paper develops a method for estimating a superset that contains the linear projection coefficients of an outcome variable y onto a full set of covariates (𝑥,𝑧,𝑤), when the available data are split across two independent sources. One dataset identifies the joint distribution of y and a subset of covariates (𝑧,𝑤), while the other identifies the joint distribution of (𝑥,𝑧). This framework is motivated by applications where administrative data lacks information on individual-level heterogeneities—such as preferences—but auxiliary data (e.g., survey) provide such detail. I illustrate the application of the method in estimating neighborhood effects while addressing selection into neighborhood choice. The micro Census data contain rich longitudinal residence information but limited detail on individual heterogeneity, while the auxiliary cross-sectional survey provides richer information on heterogeneity but lacks residential histories. I find that growing up in an ethnic enclave substantially increases the likelihood of living in an ethnic enclave later in life, but I find little evidence of neighborhood effects on most other cultural assimilation outcomes.

Available online: Journal of Business & Economic Statistics,