1

What is the minimum number of time periods required to reliably conduct a Granger causality test? I have a time series with only six data points, and I would like to know if it is possible to conduct Granger causality tests on it.

Specifically, I want to examine cultural values from the World Value Survey, which collects data in "waves" that span five years each. So, the time points are:

Wave 6 (2010-2014)
Wave 5 (2005-2009)
Wave 4 (1999-2004)
Wave 3 (1995-1998)
Wave 2 (1990-1994)
Wave 1 (1981-1984)

Given that there are only five or six data points (there's a gap between Wave 1 and Wave 2), is the Granger causality test at all feasible here?

I know that this is a slightly different question, but if six time periods is too few, are there any appropriate causality tests I could run?

Tripartio
  • 2,176
  • 1
    Granger tests operate over a number of lags, w/ only 6 time points available, how many lags did you want to include? – gung - Reinstate Monica Jul 03 '15 at 19:32
  • With my particular data set, I guess one lag would make sense, since each time point is five years apart. – Tripartio Jul 03 '15 at 19:49
  • 3
    With this little data there is no way of getting reliable inference even though technically you might be able to execute the test. You should better look for other data sources or another research question. – Richard Hardy Aug 04 '15 at 16:44
  • @RichardHardy, my research is on cultural factors. So, "You should better look for other data sources": the World Value Survey is probably the only truly international longitudinal study of cultural variables that exists. And "look for ... another research question": perhaps the best I can do is ask more modest, correlational questions, rather than causal questions. Since your comment is the most detailed response I've received, please submit it as an answer, and I will accept it as the best answer. I can accept "You can't do it" as a best answer, even though it's not what I would like to hear – Tripartio Aug 04 '15 at 20:27
  • 1
    Thanks, but honestly I don't think my comment qualifies as an answer (let alone a good one). I wish I had an idea how to solve your problem, but for now it seems there is just not enough data to do any kind of econometrics. – Richard Hardy Aug 05 '15 at 06:04
  • Thanks anyway. As I said, it looks like I'll have to drop that specific causal question for lack of data. – Tripartio Aug 07 '15 at 00:03

0 Answers0