This research investigates the influence of methodological choices in portfolio sorts on the size of the carbon premium. By analyzing more than 100,000 portfolio construction paths, we find that differences in the construction of brown-minus-green portfolios create a substantial non-standard error. From 2009 to 2022, the mean carbon premium is −0.16% per month, with a non-standard error of 0.26%. Methodological choices regarding the carbon transition risk proxy, the portfolio weighting scheme, and double sorting induce the largest variation, while controlling for common risk factors reduces it. Estimates of the carbon premium from firm-level regressions are similarly sensitive to methodological choices. Finally, we show that carbon allowance prices are related to the level of the carbon premium, whereas unexpected climate change concerns help explain periods of lower methodological uncertainty.