Extended Data Fig. 3: Both distributed lag models and local projections show evidence of growth effects.
From: Quantifying climate loss and damage consistent with a social cost of carbon
a–c. Each panel shows estimated marginal effects (∂growth/∂Temp) from a global pooled regression of growth in GDP/capita on temperature, with 0, 5, or 10 lags of temperature. Light shaded regions are bootstrapped 95% confidence intervals, darker regions are 90% CI (1000 bootstraps). Dotted vertical lines show average temperatures at end of the sample (2016-2020) for select economies globally. Marginal effects are noisier but more negative with increasing numbers of lags, consistent with temperature affecting the growth rate of GDP. d Marginal effects for 5-lag model estimated separately in three 20-year periods are not statistically different from each other and have not flattened over time; point estimates are more negative in later periods. e–f Estimates from local projections model plot the impact on growth in the 15 years following a temperature shock in year j = 0, for economies with average temperatures at 5, 15, and 25 °C; point estimates shown for each economy in e and confidence intervals in f–h. Consistent with distributed lag models, a one year temperature shock has a persistent negative effect on output for economies with average temperatures above around 15 °C, although estimates are somewhat noisier at longer time lags.
