Consider G-Causality on two stationary time series vectors (call these variables $X$ and $Y$), each with 100+ observations. It's daily financial market time series data. I have reason to believe that there's reverse causality between these two variables (i.e., $X$ causes $Y$ as well as $Y$ causes $X$).

I want to know the current consensus (or nearing consensus) regarding the best method of selecting lag length.

A Google search has only revealed some 1984 and 1985 papers. Scholarpedia's entry on Granger causality says that it's O.K. to minimise AIC or BIC, but no references were provided for this claim so I don't want to code that up until I get confirmation. Or is selection based on qualitative reasoning?

Disclaimer: Cross posted on talkstats.com.

  • 2
    $\begingroup$ Disciplinary jargon alert: I would say that "X causes Y as well as Y causes X" is an example of reciprocal causation, rather than reverse causation. In my area, reverse causation means "We postulate that X causes Y, but in actuality Y causes X." $\endgroup$
    – Alexis
    Mar 11, 2015 at 17:39

1 Answer 1


The question here is really about the best way to select lag length for a VAR, as I noted in this answer. Granger causality doesn't even enter into it until your model for the time series is selected, which is why you may not see many papers specifically concerned with lag order for Granger causality tests. It's more about lag order selection for vector autoregressive models. I'd take a look at this paper for a relatively recent reference on which criteria (AIC, BIC, SIC, HQC) are most appropriate, though they may largely agree for your application.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.