Fig. 4: Heterogeneity analysis by income.
From: Impacts of global warming on subnational poverty and inequality
a, Lower-income countries. b, Higher-income countries. Graphs present point estimates (regression coefficients) as the measure of centre, with error bars indicating 95% confidence intervals from regressions controlling for weather conditions (rainfall and humidity), GDP per capita, unemployment rates, the shares of added values for agricultural and manufacturing, population growth and FE. Robust standard errors are clustered at the subnational level. The headcount poverty is measured using the US$2.15 per day poverty line in 2017 PPP. Inequality is measured using the Gini coefficient. In a, lower-income countries (n = 2,220) include both low- and lower-middle-income countries, whereas higher-income countries in b (n = 3,834) include upper-middle- and high-income countries, on the basis of the World Bank income classification over time, available in ref. 45. The reference temperature bin is [18, 21) °C. Unlike the main analysis, which uses region year average temperatures, the temperature-bin approach relies on daily temperature data aggregated into annual counts of days in each temperature range.
