The paper tests two popular asset based models of speculative attacks--Krugman and Rotemberg (1992) and Calvo and Mendoza (1995)--and in particular, their emphasis on the second moments of monetary aggregates. Analyzing monthly panels of appropriate countries in three regions, it finds evidence for the importance of money/reserve ratios predicted by both models, and their variance as predicted by C-M. The variance of velocity does not appear important, however, casting some doubt on the K-R target zone framework and the interpretation of the C-M results.