Page 1 of 1

Problem with non-stationary series

Posted: Thu Aug 11, 2011 2:56 am
by mbabst
Hi everyone,

i'm trying to estimate the long-run elasticities between over-night stays (logier) and the real exchange rate index (index). I have monthly data. Both series are non-stationary and the over-night stays show high seasonality.

However I'm not sure how to do this. I first built the log's and estimated the following LS-regression:

logier_ln = c + a*logier_ln(-12) + b*index_ln

but then I have problems with autocorrelation and seasonality.

Then I built the growth rates as follows since I have monthly data > log(x)-log(x(-12)). When I now look at the Scatter Plot of logier_lnd12 and index_lnd12 I can see a linear dependence. So I estimated simply logier_lnd12= c + a*index_lnd12 but i think this is too easy.

Last I built a simple error correction model, would this be the right way?

Could someone please help me and show me a better way to estimate the elasticities

The Data set for germany you'll find as attachment

Thank you very much in advance!
germany.wf1
Data set germany
(10.6 KiB) Downloaded 181 times