CLAM Infers Local Causal Effects from Coarse Spatial Data

Inferring fine-grained spatial patterns from aggregated, coarse-resolution data presents a significant challenge, especially when trying to understand causal relationships. This is particularly relevant in fields like public health or environmental policy, where interventions are often applied broadly, but their effects vary significantly at a local level. Traditional methods struggle to capture these localized causal effects when only coarse observations of interventions and outcomes are available. This research introduces CLAM (Causal Spatial Disaggregation to Infer Local Effects From Coarse Data), a novel method designed to address this problem. CLAM works by jointly learning both the underlying causal mechanism and a disaggregation mapping. It effectively exploits high-resolution contextual covariates, which are factors that modulate causal effects, to infer these fine-grained spatial patterns. The method supports localized effect estimation, counterfactual reasoning, and principled outcome disaggregation, demonstrating its ability to reliably capture spatially varying causal effects across diverse settings.