This article investigates the issue of international portfolio diversification with respect to the three largest financial markets in the world-namely the US, Japan and the UK. In addition to making use of traditional portfolio analysis, we also suggest a procedure to calculate bootstrap correlation coefficients that can take into account the dynamic structure between the markets as measured by bootstrapped causality tests. Weekly data is used. The results from the first approach are supporting international diversification. The bootstrapped causality tests provide additional empirical support for this conclusion since the size of the causal effects is negligible and the bootstrap correlations are similar as the standard ones. (c) 2006 Elsevier B.V. All rights reserved.