Bayesian estimation under disinflation

[rain]

Hi everyone,
I wonder how one can take the disinflation process into account while estimating a small open economy model? I mean in the disinflation period, inflation clearly has a downard trend therefore one cannot use demanded inflation in estimation. Currently, I am using hp filtered inflation however I want to estimate the model with raw data, not hp filtered data. I wonder your opinions about what is the best practice for constructing and estimating a model under disinflation.

thanks in advance for any help,

rain

[coot]

you write your measurement equation like:

Code: Select all

pi_obs=pi_ss*pi_hat+pi_ss

where pi_obs is quarter on quarter price index (P(t)/P(t-1)) and pi_ss is price index at steady state and pi_hat is log liniarized price... every where here inflation is a price index like pi=1+inflation.

for me this works

[rain]

Thank you very much for your answer,
However, I could not get it. What is the reason of multiplying pi_hat with s.s. value of inflation? Could you please explain this?
Thank you.

[jpfeifer]

The reason is that observed inflation rates can be written as steady state value times one plus the percentage deviation:
pi_obs=pi_ss(1+pi_hat)=pi_ss*pi_hat+pi_ss
For example, if pi_SS is 1 and pi_obs=1.1, inflation is ten percent above its steady state and hence pi_hat is 0.1.This is just an example of an observation equation (see the lower part of http://www.dynare.org/phpBB3/viewtopic.php?f=1&t=2490&p=5165&hilit=observation+equation#p5165)

But more importantly, what are you trying to do exactly? What do you mean with disinflation process? Do you simply mean a period where inflation is below its steady state value due to a series of shocks? Or do you mean a structurally different phase where the inflation target was lower than before?

[coot]

I think he means that "inflation" has a downward path, a downward trend.... there is no problem with this... in my model there is a clear disinflation process and nothing happened... actualy I think he considers since inflation has a downward path it may have a unit root... try modeling Quarter/Quarter inflation not Year/Year. ON Q/Q basis you shouldn't have any problems.

[rain]

Hi guys,
Thank you for your answers.

What i mean by "disinflation" is that inflation and inflation target has a downward trend through time. That is to say, inflation itself is non stationary. Therefore, I guess, I need to introduce a unit root process for inflation target. My question is how should i transform my observable consistent with my model? For example should I use change in inflation as an observable or inflation itself?