I have small questions regarding NK_baseline.mod in Dynare examples.
In taylor rule some shocks are written as AR(1), but espm is stated as idd shock. Why it is not AR(1) process? And why do we need to take exponential of epsm?

Typically, monetary policy shocks are conceptualized as policy mistakes. As such, they should not autocorrelated, i.e. predictable. There is also the identification issue given that there is already interest rate smoothing in the equation.

The exp() comes from the fact that the equation is written in levels and is multiplicative.

I am sorry, @jpfeifer, I forgot to ask here about the sign of the shock. Since this Taylor rule is coded as a multiplicative one, is exp(epsm) considered a positive monetary policy shock (if we log-linearize, it would be with a “+” sign)? I understand the sign of shock when we use AR(1) process.