AEA Papers and Proceedings
ISSN 2574-0768 (Print) | ISSN 2574-0776 (Online)
Causal Inference with Satellite Imagery: A Comparison of Methods for Forest Conservation Data
AEA Papers and Proceedings
(pp. 87–91)
Abstract
Can satellite-based predictions substitute for traditional outcome measurements in program evaluation? Using forest cover data from the Democratic Republic of the Congo, Brazil, and Indonesia, we conduct semisynthetic simulations comparing estimation methods. The RSV estimator (Rambachan, Singh, and Viviano 2025) formalizes the postoutcome structure of remotely sensed data—changes in forest cover cause changes in satellite imagery—delivering approximately unbiased treatment effects with correct coverage. Some alternative approaches exhibit significant bias despite highly accurate pretrained predictors. When limited validation data are available, the RSV estimator efficiently incorporates observational samples while remaining robust to distribution shift, achieving meaningful reductions in standard errors.Citation
Alsharif, Haya, Ashesh Rambachan, Rahul Singh, and Davide Viviano. 2026. "Causal Inference with Satellite Imagery: A Comparison of Methods for Forest Conservation Data." AEA Papers and Proceedings 116: 87–91. DOI: 10.1257/pandp.20261019Additional Materials
JEL Classification
- C45 Neural Networks and Related Topics
- O13 Economic Development: Agriculture; Natural Resources; Energy; Environment; Other Primary Products
- Q23 Renewable Resources and Conservation: Forestry
- Q28 Renewable Resources and Conservation: Government Policy