The optimal currency hedging policy for portfolios with foreign asset exposure is easily determined by mean-variance analysis, yet many investors choose inefficient hedging policies. Some investors reject the mean-variance approach because they believe the required inputs are not sufficiently stable to generate satisfactory out-of-sample results, whereas other investors blindly choose a 50 per cent uniform hedging policy in order to minimise ‘regret risk’. We present results that are relevant both to foreign equity managers and to plan sponsors with liability obligations, and these results demonstrate that naively determined inputs to mean-variance analysis are sufficiently stable to reduce out-of-sample portfolio risk. We then challenge the minimum regret solution for three reasons: it implies an implausible utility function; its empirical validity is confounded by errors in means and even if investors have such an implausible utility function, a variant of mean-variance analysis yields a solution that is superior to a constant 50 per cent hedge ratio. Finally, we investigate the economic rationale that gives rise to varying optimal hedging polices across countries.