When researchers examine lead-lag relationships between credit default swaps and (as an example) stock markets, many use Vector Autoregressive Models (VAR). They want to find out what market "is leading" the other. But when looking at financial returns, one would not expect to find autocorrelation. My question: Am I missing something here? I have read tons of papers using this approach and I do not get why it seems to work quite well. I have seen it so often now that there must be something specifically useful about VARs or something wrong with the conclusion I draw (or both). Why is this approach still being used? Does the assumption that markets are not fully efficient enables people to use it?