1
$\begingroup$

I've got two conflicting answers when I search the internet for my question. Since cointegration is sensitive to maximum lag length, it is important to choose maximum lag length wisely. According to one answer, maximum lag length in the underlying VAR should be 1 for annual data, 4 for quarterly, and 12 for monthly data. I use annual data, so I should use 1 for maximum lag length.

But if I use 1 for maximum lag length, why and how can I use any information criterion to obtain the optimal number of lags? I mean, if I only have one option, 1 lag, information criteria become useless, right?

Also, when I proceed to VECM, only one lag will not show any short-run relations between the variables, just to the cointegration coefficient, huh? I'm a novice at econometrics, so please bear with me if I'm asking silly questions.

$\endgroup$
  • $\begingroup$ Look here for a related question and a nice answer to it. Also look here, especially if you use R. And even if you do not use R, it is still a good presentation of VECM modelling. $\endgroup$ – Richard Hardy Oct 30 '14 at 20:00
  • $\begingroup$ You can use a sensitivity analysis of your model under different numbers and/or lengths of lags, too. $\endgroup$ – Alexis Feb 5 '16 at 1:28
1
$\begingroup$

That recommendation (1 for annual, 4 for quartely, etc) is simply a rule of thumb that was obtained by other people who probably used some information criteria so that now you don't have to. That is, if your time series is much longer or shorter than a typical case, you should determine the optimal lag yourself, otherwise relying on the rule of thumb is ok.

$\endgroup$
  • 1
    $\begingroup$ Thanks for your answer! What do you think would be the length of a typical case? Also, if my time series is shorter/longer than that, and I would determine the optimal lag myself, I´d use information criteria such as AIC or BIC? Or just be parsimoneous and use one lag (which I red somewhere)? $\endgroup$ – Lars Ahnland Nordfors Oct 1 '14 at 12:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.