Dynare forums

Dear all,

I have a few questions about the way to (correctly) transfrom data when trying to estimate a dsge model in dynare.
To be specific I'm not sure of how to treat varibles that are themselves given in percentage points (such as inflation and interest rates) versus varibles that are given in dollar/euro or whatever currency term (such as gdp, consumption etc).
Guide on declaring observation equations in dynare states for example:

Consider the case of output. Our model variable ˆyt represents log output deviations from the
long-term trend and has mean 0. Thus, it exactly corresponds to log empirical output per
capita detrended using any of the above filters that also takes out the mean. We denote this
detrended log output variables with yobst . Think of yobst for example as one-sided HP-filtered
log GDP per capita. In this case, specifying the observation equation is basically redundant
as we directly “observe” ˆyt:
yobst = ˆyt (36)

but this transformation would convert original gdp series to it's absolute (dollar term) deviation from trend, not to it's percentage deviation from steady state (assuming that this one corresponds to HP trend).

I am also not sure on what is ment by "quarterly interest rate", did the author mean quarter on quarter (qoq) interest rate, and if so, why is inflation not transformed in a same way as interest rate (by dividing with 400)? Both are commonly reported in annualized terms.

For output you are missing that the log deviation is taken, not the absolute deviation. Log deviations are percentage deviations.

Regarding the "quarterly interest rate". This is the interest in the model that has to be paid between two periods, which are quarters. If that interest rate is 2%, the annualized interest rate that financial markets are going to quote is (1.02^4-1)*100.
For inflation, it depends on how it is computed in the data. If you look at annual CPI inflation, you would need to use the same transformation. But if you consider the change in the GDP deflator between two quarters, this will not be an annualized measure.