In this article we evaluate the most widely used spread decomposition models using Exchange Traded Funds (ETFs). These funds are an example of a basket security and allow the diversification of private information causing these securities to have lower adverse selection costs than individual securities. We use this feature as a criterion for evaluating spread decomposition models. Comparisons of adverse selection costs for ETF's and control securities obtained from spread decomposition models show that only the Glosten--Harris (1988) and the Madhavan--Richardson--Roomans (1997) models provide estimates of the spread that are consistent with the diversification of private information in a basket security. Our results are robust even after controlling for the stock exchange.