0
$\begingroup$

Dear StackExchange community,

I'd have a question on impulse responses that I have not found an answer to in econometrics textbooks. Specifically, I would want to know how to interpret impulse responses if the shocks are based on historical data (for instance monetary policy shocks that have already been identified) and not identified by sign restrictions or simple one standard deviation 'simulations'. The time series of these shocks obviously have varying signs over time and therefore when I generate the impulse responses (through local projections or a VAR), I have trouble interpreting the IRF given that the underlying shocks are negative and positive depending on the time period.

What I know is that linear impulse response functions are insensitive to the size of the shocks (if we ignore scale for a moment) and that they are symmetric. Maybe I am thinking about this wrong, in any case, I would appreciate your advice!

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Not sure if I understand your question correctly. But IRFs show what happens to the system up to a horizon h if the dynamics is only driven by 1 shock at time t. Yes, these are linear so if you multiply them by -1 you can see what happens under a shock with a reversed sign (e.g. interest rate cut/hike).

Edit: Ok, I'm less familiar with LP than with VAR. But with a reduced form VAR you are not retrieving the IRF to a single specific shock. You are calculating the response to 1 standard deviation innovation of a specific variable. So if you have a narrative series, you are calculating the response to that exogenously identified innovation. In LP-IRFs you have, the identified shock is the shock regression coefficient at horizon 0. Hence you know from your very estimates the sign and the magnitude of that retrieved shock.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thanks for your reply. My point is that I don't know whether the shock at time t is positive or negative. This is because my impulse responses are based on a time series of pre-identified shocks (for instance oil price shocks or monetary policy shocks). In some econometrics textbooks, this is called the narrative approach. So the shock in Jan. 2010 could be -0.1 and in Feb 2010 it could be +0.5. Using local projections I get positive impulse responses over a time horizon of 12 months. But given that the sign of the underlying shocks varies over time, I don't know how to interpret them. $\endgroup$
    – laliberte
    Commented Nov 21, 2021 at 12:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.