This project is read-only.

jforecast with more shocks open than exogenised variables

Topics: Models
Mar 5, 2015 at 11:18 AM

I'm using jforecast and would like to endogenise more shocks than the number of exogenised variables. How can I do this? Using the latest IRIS toolbox (20150227), I get an error message:

Error using inv
Matrix must be square.
Error in model/myforecastswap (line 117)
iM22 = inv(M22)

I know that it works perfectly with the command simulate. However, in contrast to jforecast, I don't get std of forecasts when using simulate. That's why I want to use jforecast.

Thanks in advance for suggestions!
Mar 5, 2015 at 8:27 PM
Try using plan/condition instead of plan/exogenise and plan/endogenise. Condition will solve for the shocks which are most likely to produce the conditioned path. You can exclude some shocks from being used by setting their standard deviations to zero over that part of the forecast horizon with the option 'vary='.
Marked as answer by michaeljohnston on 3/5/2015 at 12:33 PM
Mar 5, 2015 at 8:33 PM
Turning off some shocks when using plan/condition can be useful/advisable to exclude some undesirable explanations for the conditioning path. E.g., if you condition on growth which is low relative to your baseline then a lot of the explanation might be that the monetary authority undertakes some aggressive interest rate hikes, and you might want to force another explanation by setting the standard deviation of the monetary policy shock to zero over the conditioning horizon.
Marked as answer by michaeljohnston on 3/5/2015 at 12:34 PM
Mar 10, 2015 at 8:39 AM
This seems to work very well. Thanks a lot!