A critical assessment of learning curves for solar and wind power technologies

The learning curve concept, which relates historically observed reductions in the cost of a technology to the number of units produced or the capacity cumulatively installed, has been widely adopted to analyse the technological progress of renewable resources, such as solar PV and wind power, and to predict their future penetration. The observed relationship has often been used as an input to energy system models and a justification for public spending on R&D and enhancing the scale of the technology. Learning curves have a place in research, but in this paper we argue that analysts often apply the concept, or make related assumptions, uncritically in their analysis of the technology. We make three observations. First, cost reduction can be driven by factors not correlated with current output, implying other factors as drivers of long-term learning effects. Second, despite the empirical observations, the theoretical foundation for learning curves is insufficiently established. The concept relies on historical development of the technology, that is, the result will be accurate if the future holds a path-dependent trajectory, whereas in reality there is a possibility of future breakthroughs as well as technological stalemates. Third, an observable cost reduction of a component in a given generation technology does not necessarily correspond with the trend in the total cost of deploying that technology. For example, module costs currently constitute a much smaller share of the total cost of solar PV compared with a few years ago. If the module’s rate of cost decrease is applied to the total cost of solar PV, it is highly likely to result in an incorrect prediction of future diffusion. Learning curves were originally developed as an empirical tool to assess the effect of learning-by-doing in manufacturing, and the jump to analysis of country-level technological change in renewable energy is an extension that requires careful consideration