Are temperature trends affected by economic activity? Reply to Benestad (2004)

Abstract

factors. This makes it sound like the White’s matrix is the model. It is not: it is just a weighting applied to the variance-covariance matrix of the coefficients of the model, and it is derived from the regression residuals, not from the non-climatic subset of the explanatory variables as Benestad says. He then adds: Ordinary Least Squares (OLS) models may produce biased estimates, and the presence of heteroskedasticity in the residuals may be an indication of model misspecification such as incorrect functional form. The SHAZAM model therefore ought to give unbiased estimates of the coefficients describing the relationship between a number of factors and the temperature trend. Using OLS in the presence of heteroskedasticity does not cause biased parameter estimates, nor does using the White’s matrix adjustment change the functional form. Heteroskedasticity can produce inefficient estimates in small samples (i.e. the estimated variance can be too high). The correction applied has nothing to do with bias or functional form. These items are all discussed in the Kmenta book referenced in the paper itself.