This paper presents a meta-analysis of capital-energy substitution elasticities. We distinguish between Morishima elasticities, which measure technological substitution potential, and cross-price elasticities, which measure actual percentage changes in capital demand in response to energy price changes. We estimate a meta-regression model with separate coefficients for the two elasticity samples. The results show that the heterogeneity in both the cross-price and Morishima elasticity samples can to a large extent be explained by study differences in, among others, model specification, data characteristics, region and time period. Controlling for potential sources of misspecification and aggregation bias we subsequently calculate short- and long-run elasticities for different regions and time periods. The resulting elasticities show that technological substitution potential is large, especially in the long run for North America. Despite substantial differences across regions and time periods, the estimated cross-price elasticities suggest capital-energy substitutability without exception. © 2007 Elsevier B.V. All rights reserved.