Accounting for Unobservable Heterogeneity in Cross Section Using Spatial First Differences
We develop a novel cross-sectional research design to identify causal effects in the presence of unobservable heterogeneity without instruments. When units are dense in physical space, such as in geocoded administrative data or gridded datasets, it may be sufficient to regress the “spatial first differences” (SFD) of the outcome on the treatment and omit all covariates. SFD aligns the identifying assumptions for simple cross-sectional samples with the assumptions of other quasi-experimental designs, in terms of mathematical structure and plausibility. We demonstrate SFD in simulation and by estimating the returns to schooling. We then use SFD to obtain new estimates for the effects of time-invariant geographic factors, soil and climate, on long-run agricultural productivities — relationships crucial for economic decisions, such as land management and climate policy, but notoriously confounded by unobservables. We also demonstrate three empirical tests that can suggest whether SFD is likely to be successful at identifying causal effects in new contexts.
-
-
Copy CitationHannah Druckenmiller and Solomon Hsiang, "Accounting for Unobservable Heterogeneity in Cross Section Using Spatial First Differences," NBER Working Paper 25177 (2018), https://doi.org/10.3386/w25177.Download Citation
-