For those that don’t know Gretl is a freeware econometrics package. Despite not costing anything I’ve found it to be a very useful econometrics program that can do pretty much anything — or, at least, anything that I’ve ever wanted it to do.
Gretl can be a bit daunting to use, however. This especially so given it’s ‘stripped down’ presentational format (which I rather like, but others may not). Anyway, the author Hishamh over at the Economics Malaysia blog has put together a series of post that guides the user through all the major uses of Gretl. The posts, complete with screenshots, are indispensable and I will here run quickly through what they show.
In the first post the author shows how to input and format data.
In the second post the author shows how to run and interpret a regression.
In the third post the author shows how to ensure that the test that has been run is robust — in this he shows the reader how to test for independence and normal distribution.
Finally, in the fourth post the author shows how to introduce dummy variables to control for seasonal variation in the data.
The series is not a substitute for actually understanding the issues involved in econometric modelling; these need to be understood independently. But I’m sure that the resourceful reader can make ample use of Google, Wikipedia and, if necessary, an econometrics textbook to supplement the series.
Of course, as readers of this blog will know, I’m rather skeptical of econometrics techniques when used on economic data. I have not since changed my opinion and continue to think that these techniques do more harm than good.
Just as Keynes, however, in his critique of Tinbergen suggested that the econometric method was useful for finding out the effect that an increase in railway traffic, the rate of profit on railways, the price of pig-iron and the rate of interest had on net investment in railway rolling-stock (that is, railway vehicles), I would suggest that there are some rather obvious relationships that can be tested. (Although the issue then arises as to how useful such tests are when the relationships reach a certain level of obviousness…).
As well as this there some more contentious relationships that can be tested and used very provisionally in making forecasts. The author of the series’ own example — that is, the lagged effect of exports on imports in a small open economy — is a good example of this. As would be, for example, the import/export elasticities of demand (i.e. how exports/imports are affected by income and the exchange rate) or multiplier estimations. But again, such relationships likely miss more than they capture and they are sure to be unable to incorporate anything of real interest. For that reason I would continue to strongly endorse David Freedman’s ‘shoe leather’ approach to economic statistics.
But if you’re going to do econometrics, and you’re going to use Gretl in particular, the above series is great for either the beginner or those that want a refresher course.
Update: The new version of the excellent Gretl manual by Lee Adkins is out now and can be downloaded here.