This paper presents new estimates for 21 OECD countries covering the period 1960-2001, focusing on two questions: To what extent does the impact of public capital on output differ across countries? And to what extent does it differ over time? Using vector autoregressions (VARs), we find that in some countries a shock to public capital has a positive long-run impact on GDP while in others the longrun impact is zero or even negative. We also find that variability of public capital and its long-run impact on output are negatively correlated. Furthermore, when the public capital stock is large relative to the private capital stock the long-run impact of public capital is lower. Our results on 'recursive' VARs suggest that in the majority of countries the effect of a public-capital shock on output has decreased over time. Countries where the impact of public capital decreased during the 1990s have a declining public-capital-to-GDP ratio, and vice versa. Estimates based on a panel VAR for the OECD area confirm the declining long-run impact of public capital.