Assessments of social and economic impacts of climate change are primarily based on the results of biophysical climate impact models, which are aggregated, extrapolated and/or valued in monetary terms. Another potential source of information on climate impacts are spatial and/or temporal analogues, such as Ricardian analysis of climate impacts on agriculture. Another recent effort to this end involves the development of the G-Econ database (Nordhaus, 2006), which describes the relationship between climatic and geographic factors on the one hand and regional economic productivity on the other. A multivariate regression derived from this database has been used to estimate global economic impacts of climate change in a recent version of the DICE model. The reanalysis presented here was motivated by some counterintuitive results in Nordhaus (2006). I have developed a modified version of the GEcon database, which corrects several inconsistencies in G-Econ, and which is available at two different spatial resolutions (grid cells and subnational administrative units). This database is applied to reanalyze key results in Nordhaus (2006) and to perform additional analyses, focussing on the influence of climate on population density, density of economic output, and output per capita. I discuss the implications of several statistical problems in the G-Econ data (skewness, heteroskedasticity, excess zeros, different weights of data points, different spatial resolutions of predictors and predictands) for statistical analysis and assess the sensitivity of results to variations in statistical estimators, aggregation units, and weighting schemes. This reanalysis finds that the counterintuitive results in Nordhaus (2006) can be largely explained by flawed methods for data aggregation and analysis, and by incomplete data in the G-Econ database. Given that several statistical problems have been inadequately addressed by Nordhaus (2006), the validity of estimates of global climate impacts based on G-Econ remains doubtful.