Page 1 of 1

What prior for shocks ?

PostPosted: Thu Jul 14, 2011 4:38 pm
by Lau
Dear all,

I am working on a symmetric two-country monetary union model, and I am concerned about the size of the shocks estimated in my model. The variance of my data are about 3 to 5 times bigger for the data of my country compared to the one of the euro area. Should I consider that the variance of my estimated shocks to be multiplied by a value between 3 to 5 too ?
I saw that Kolassa(2008) for example assumed that his prior are 3 times bigger for Poland than for the Euro Area, and is this is a relevant approach? Most of monetary union model assumed just the same prior for shocks from the two countries ?
Thanks

Re: What prior for shocks ?

PostPosted: Thu Jul 21, 2011 2:29 am
by jpfeifer
In principle, you can use any prior you want - as long as it is independent of the data you are using to form the likelihood. So if you have outside information that the shock variances are larger for one country, you should use a wider prior.

But it is not admissible to look at the data, see a higher variance and then impose a wider prior.

What you could do is use an relatively uninformative prior like a uniform distribution with a sufficiently high upper bound.