Econometric estimates of the determinants of current account (im)balances invariably start with the accounting ‘identity’ that the current account is equal to the difference between (national) savings and (national) investment. However, this accounting relationship does not hold in the official data. The measurement errors are small but still significant for G-7 countries and for other countries often large (>5 % GDP), especially for less developed countries. Larger samples imply thus also larger measurement errors. Moreover, there are also important differences among the data reported by different widely used international sources, like the World Bank or the IMF. We illustrate the importance of these measurement errors in two ways: (1) the simple correlations between savings and investment often used to gauge capital mobility (the Feldstein-Horioka (1980) paradox) appear very different when different data sources are used, and (2) in a standard model for current account determination, using savings minus investment instead of the current account leads to some significantly different results, regarding the significance of the level of GDP per capita or banking crisis.