Answers are grounded in this paper and cite their sources.
Understanding how place influences human behavior is a central and enduring question in geography, yet quantifying spatial contextual effects remains challenging due to unmeasured confounders and endogeneity. Multiscale geographically weighted regression (MGWR) can measure behavioral and intrinsic contextual effects but assumes exogeneity, overlooking how spatial context could also influence covariate levels—an indirect contextual effect that could bias parameter estimates. This study introduces multiscale geographically weighted fixed effects regression (MGWFER), a local panel modeling framework that integrates MGWR with individual fixed effects to address all three contextual influences while allowing coefficients to vary across space and scale. A two-stage estimation algorithm is developed to first remove time-invariant spatial confounding from covariates and then recover local intrinsic contextual effects. In simulation, MGWFER outperforms MGWR and a pooled MGWR (PMGWR) and yields locally differentiated estimates that admit a causal interpretation under the stated assumptions. A case study on educational attainment in the state of Georgia further demonstrates that MGWFER reveals substantially stronger spatial contextual effects and corrects misleading conditioned relationships from standard local models. Findings suggest that place might exert a far greater causal influence on human behavior than previously recognized, underscoring the importance of integrating causal inference into spatial analysis.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: