P. Nelson C.M. Sprecher, “Where will new nuclear power plants be constructed? Validation of a predictive model”, Proceedings of 31st Annual Meeting of ESARDA, May 26-28, 2009, Vilnius, Lithuania.
Nuclear reliance (percentage of electrical generation from nuclear) has been shown to be satisfactorily predicted by a simple linear regression on various characteristics of states. It is hypothesized that nuclear deficit := actual electrical generation by nuclear minus predicted can be used to predict the degree of intent of states to rely on nuclear generation of electricity. This hypothesis is validated against two different measures of (relatively near-term) nuclear intent, in terms of an error measure developed for this nuclear deficit performs marginally satisfactorily as a predictor, but there is room for improvement. Possible sources of improvement that are suggested by a detailed analysis of the predictive errors include incorporation of a ‘wealth effect’ in the underlying regression, and incorporation of anticipated growth in demand for electricity in the predictions from current nuclear deficits.