Modeling spatial processes of extreme heat impacts on global economy: a multi-scale spatio-temporal approach

Rising frequency, intensity, and geographic scope of extreme heat profoundly impede global sustainable economic development. However, existing climate econometric models are limited in capturing the spatial processes through which extreme heat affects the global economy, often resulting in downward-biased estimates of total economic losses. This study develops a novel multi-scale spatio-temporal model that integrates classic multi-level modeling with spatial statistics, explicitly addressing key challenges faced by climate econometrics. A Bayesian Markov chain Monte Carlo simulation algorithm is derived for model implementation. Using this model, we present the first quantitative assessment of the impacts of extreme heat on global economic production and their scale-dependent spatial processes. Our findings reveal that, at the national scale, economic losses caused by input-output economic linkages initially decline slowly, then drop sharply with increasing connectivity, with an inflection point around 0.1. When accounting for spatial propagation effects, a 1 ℃ increase in extreme heat intensity leads to an average loss of 2.54% [0.90%, 4.19%] of annual GDP per capita-substantially higher than estimates assuming economic losses are locally confined. Moreover, the economic impacts of extreme heat exhibit significant spatial heterogeneity, with positive marginal effects detected in colder regions and negative effects in warmer regions, with a turning point around 33.7 ℃. This study offers a new methodology to evaluate the impact of climate change from a multi-scale and spatial perspective.