We demonstrate that the correlation of saving and investment is measured best by an error correction model (ECM), because theory implies a cointegrating relation between these variables. The ECM comprises all previous specifications as special cases, which are shown to be potentially misspecified on theoretical grounds. We argue that the correlation can serve to reject the hypothesis of capital immobility, but not the one of capital mobility. Applying the ECM to Norway yields the following findings: First, the ECM outperforms prevailing specifications. Second, we detect structural breaks, which underpins the need for careful diagnostic testing and, third, the correlation's time profile is consistent with other indicators of capital mobility. The Feldstein-Horioka puzzle does not exist for Norway.