I have a question for clarification of the timing of impulse responses in Dynare. I am currently using the statement
- Code: Select all
predetermined_variables k;
combined with a standard law of motion for capital
- Code: Select all
k(+1)=(1-delta)*k+i;
Now I let Dynare compute impulse responses to some contemporaneous exogenous shock called epsilon (think of it for example as a temporary shock to TFP that enters the production function) by saying
- Code: Select all
shocks;
var epsilon; stderr 1;
stoch_simul(order=1,irf=40) y k;
I realize that according to the manual, the shock hits at time t=1 on the x-axis in the resulting plots. However, I note that k also moves on impact (period 1), which should not happen as it is predetermined. Does Dynare internally shift predetermined variables to the old notation (i.e. k is used as k(-1)), so that the impulse response plotted is shifted by one period for predetermined states? I mean, does the impulse response to k at t=1 actually happen at t=2 or am I doing something wrong in my model? I would appreciate any help as I could not find any information in the documentation about the impulse response timing for predetermined variables.
Johannes