Global Non-Linear Effect of Temperature on Economic Production: Comment on Burke, Hsiang, and Miguel

The most influential paper attempting to show that higher temperatures will reduce economic growth was written by Marshall Burke, Solomon Hsiang, and Edward Miguel and was published in Nature. The authors used annual data by country on temperature and GDP per capita growth to estimate an optimal temperature. They then estimate how much, in terms of percentage annual growth, deviations from the optimum cost the country. From these results they calculate the cost of the warming that is predicted by the IPCC and conclude that warming will reduce world GDP per capita by 23 percent. I find problems with their work, including the following: (1) The statistical significance of their results is diminished when any one of the following three adjustments are made: (i) residuals to be minimized are weighted by country size and growth volatility, (ii) when adjustments are made for temporally and spatially autocorrelated growth, and (iii) when a few outlying observations are removed. (2) When all of these adjustments are made, the results disappear. (3) Their results are not consistent by world region. (4) Any effect that exists appears to be reversed in the following year. (5) Any effect that exists shows some evidence of mitigation over time. (6) Their main result is that warming will reduce world GDP by 23 percent. Their bootstrap estimates of this result show signs of unreliability, but these warning signs are not discussed in the paper.