VAR LAG SELECTION

[jimm306]

Hi Tom,

I am estimating a VAR with 5 daily financial return series.I am thinking whether to use BIC (which suggests 1 lag) or AIC (which suggests 10 lags) for determining the lag length when performing Granger causality tests. Which criterion would you recommend in this context?

More generally, is it valid practice to rely on information criteria such as AIC or BIC to select the appropriate number of lags in a VAR?


Thanks

[TomDoan]

There's a discussion of that on the @VARLAGSELECT procedure page. It's rather odd for a set of return series to come up with any large lag number. (They are typically either zero or one).