2025-01-19-世界银行-天气的错误测量_在经济背景下使用遥感地球观测数据（英）_70页_3mb

Summary

This paper highlights the challenges of using remotely sensed Earth observation (EO) data for weather variables in economic models. Despite the availability of EO datasets, the mismeasurement varies across sources, with differences extending beyond simple scaling (cardinality) to affect the ordinality of coefficients in regressions. The study, using data from six Sub-Saharan African countries and multiple EO products, finds that coefficient ordering changes significantly based on the choice of EO dataset, country, and model specification. This lack of robustness means researchers can obtain vastly different results simply by selecting different EO products or metrics.

Key recommendations include exercising caution in EO data selection, conducting robustness checks by testing alternative datasets, and avoiding over-reliance on these data for economic analyses due to potential measurement errors and sensitivity to data choice.

Economists should recognizescience as a social phenomenon and avoid treating EO data as error-free measures of truth, ensuring results are carefully validated.