Mills of progress grind slowly? Estimating learning rates for onshore wind energy

Estimated learning rates for onshore wind span a large range of about 40 percentage points. We propose a multi-factor experience curve model with a new economies of scale measure and estimate learning rates for onshore wind using country-level data from seven European countries. We find learning by doing rates of 2%-3% and learning by searching rates of 7%-9% in terms of LCOE. When decomposing LCOE, we find no significant learning in installed costs but significant learning in capacity factors. Accounting for improvements in capacity factors and modeling learning by searching can hence be promising for energy models that endogenize technological change. We confirm our results in several robustness checks, and show that depreciation rates of the knowledge stock have large effects on estimated learning rates.