forecasts
Posted: Mon Nov 26, 2012 11:04 amDear all,
I am trying to replicate the results (table 3) in Smets and Wouters (2007) by using 'forecast' in dynare. However, I have two questions:
1. when dynare do the forecast for periods, say 12 quarter, it estimates the parameters from the in-sample data, and then make the forecasts recursively by using the forecasts from the previous periods?
2. Smets and Wouters (2007) take the observables by using ln(X_t/X_t-1)*100, but interest rate is divided by 4. I am wondering, why do we need to multiply 100? As dynare would do loglinearization, so variables are supposed to be ln(X/X_ss). Are they also reported in %, so the numbers are multiplied by 100? Would it make difference in estimation and forecasts if we don't multiply the observables by 100?
Thank you so much for any reply.
Nymph
I am trying to replicate the results (table 3) in Smets and Wouters (2007) by using 'forecast' in dynare. However, I have two questions:
1. when dynare do the forecast for periods, say 12 quarter, it estimates the parameters from the in-sample data, and then make the forecasts recursively by using the forecasts from the previous periods?
2. Smets and Wouters (2007) take the observables by using ln(X_t/X_t-1)*100, but interest rate is divided by 4. I am wondering, why do we need to multiply 100? As dynare would do loglinearization, so variables are supposed to be ln(X/X_ss). Are they also reported in %, so the numbers are multiplied by 100? Would it make difference in estimation and forecasts if we don't multiply the observables by 100?
Thank you so much for any reply.
Nymph