This paper presents an empirical study of hedging the four largest US index exchange traded funds (ETFs). When hedging each ETF position with its own index futures we find that it is difficult to improve on the na�ve 1:1 futures hedge, that hedging is less effective around the time of dividend payments, and that hedged portfolio returns tend to have very large negative skewness and highly significant excess kurtosis. We also investigate the extent to which a long position on one ETF can be offset by a short position on another correlated ETF and consider how best to hedge portfolios of ETFs with one index futures. In these situations minimum variance hedging is clearly preferable to na�ve hedging, although it seems to matter little which econometric hedge ratio is used, and the cross-hedged portfolio returns are closer to normality than the futures hedged portfolios. The evaluation focuses on a very large out of sample hedging performance analysis that includes aversion to negative skewness and excess kurtosis as well as effective reduction in variance. Our results should be of interest to hedge funds employing tax arbitrage or leveraged long-short equity strategies. They will also be of interest to ETF market makers since hedging is the most cost effective way of reducing the market risk of inventories, thus hedging enables market makers to reduce bid-ask spreads in a competitive environment