This paper tests the Feldstein-Horioka "puzzle" for the two richest countries of the world: Japan and the USA. For this purpose it employs three different cointegration tests that are applied to the Feldstein-Horioka long-run investment-saving equation in conjunction with the Lumsdaine and Papell (1997) unit root test considering two structural shifts. A novel aspect of the paper is that it determines thenumberof breaks solely from a scrutiny of the data and that in constructing the dummy variables for the breaks it uses the endogenously determined break dates. It shows that allowing for structural shifts eliminates the "puzzle" both for Japan and the USA. (C) 2010 Published by Elsevier B.V.