Empirical challenges in the use of learning curves for assessing the economic prospects of renewable energy technologies

In bottom-up energy models endogenous technical change is introduced by implementing technology learning rates, which specify the quantitative relationship between the cumulative experiences of the technology on the one hand and cost reductions on the other. The main purpose of this paper is to critically analyze the choice of modeling and estimation strategies in learning curve analyses of power generation costs. We identify and discuss a number of theoretical and econometric issues involved in the estimation of learning curves. These include the presence of omitted variable bias and simultaneity, but also methodological problems related to the operationalization of theoretical concepts (i.e., learning-by-doing) and the associated use of data. We illustrate the importance of these issues by employing panel data for wind power installations in four western European countries, which are used to compare the results from different learning curve model specifications. The results illustrate that the estimates of learning rates may differ significantly across different model specifications and econometric approaches. The paper ends by outlining a number of recommendations for energy model analysts, who need to select appropriate energy technology learning rates from the empirical literature, or who choose to perform the empirical work themselves.