Iacoviello/Neri (2010), German Data
Posted: Sat May 23, 2015 6:49 pmHi everyone. I'm trying to use the model from Iacoviello/Neri (2010) "Housing market spillovers" to estimate parameters for the German economy from 2000 Q2 to 2013 Q4. However, I'm having trouble getting the calibration right, since I have no prior experience with dynare or economic models.
Mainly, I'm having trouble with the weight on housing in the utility function (JEI in the model file). I couldn't find any hints as to how exactly it was calculated.
One possible calculation method I found was using a linear regression of the quarterly log-difference of consumption over the quarterly log-difference of lagged housing wealth (both in per capita terms, divided by the population of people between 16 and 67 years of age):
In the 2010 paper: delta_log(C,t) = a + JEI * delta_log(HW,t-1)
My understanding of it: log(C,t)-log(C,t-1) = a + JEI * (log(HW,t-1) - log(HW,t-2))
However, plotting in Excel showed my result had too high variance (R²=0.027), with a value of JEI=0.06.
I also tried to use the formulae for qh/Y from Iacoviello (2005), but I have my doubts that it is applicable in this case, as there is no entrepreneur and ALPHA, an calibrated parameter in the 2005 model, is one of the estimated parameters in the 2010 model. Using the ALPHA from the definition of the priors and some abstraction on the parameters for entrepreneurs (namely entrepreneurs as patient households) yields a JEI of 0.07-0.08.
I've also done some trial and error. Any value for JEI above 0.2 seems to yield an error message and the calculation works the longest with a value JEI = 0.135. Either way, the error messages I usually recieve are about there not being a steady state, the hessian not being positive definite and/or an element in the Jacobi matrix being infinite or NaN.
Also, I am using the chained index for GDP as the necessary deflator in the data. Using the same method with the U.S. data for housing price indices as stated by Iacoviello/Neri (2010) yielded similar results to the data files in the original model (similar shape and size), so I assume this is the correct approach.
I'd be thankful if anyone has any hints, especially regarding the weight on housing.
Mainly, I'm having trouble with the weight on housing in the utility function (JEI in the model file). I couldn't find any hints as to how exactly it was calculated.
One possible calculation method I found was using a linear regression of the quarterly log-difference of consumption over the quarterly log-difference of lagged housing wealth (both in per capita terms, divided by the population of people between 16 and 67 years of age):
In the 2010 paper: delta_log(C,t) = a + JEI * delta_log(HW,t-1)
My understanding of it: log(C,t)-log(C,t-1) = a + JEI * (log(HW,t-1) - log(HW,t-2))
However, plotting in Excel showed my result had too high variance (R²=0.027), with a value of JEI=0.06.
I also tried to use the formulae for qh/Y from Iacoviello (2005), but I have my doubts that it is applicable in this case, as there is no entrepreneur and ALPHA, an calibrated parameter in the 2005 model, is one of the estimated parameters in the 2010 model. Using the ALPHA from the definition of the priors and some abstraction on the parameters for entrepreneurs (namely entrepreneurs as patient households) yields a JEI of 0.07-0.08.
I've also done some trial and error. Any value for JEI above 0.2 seems to yield an error message and the calculation works the longest with a value JEI = 0.135. Either way, the error messages I usually recieve are about there not being a steady state, the hessian not being positive definite and/or an element in the Jacobi matrix being infinite or NaN.
Also, I am using the chained index for GDP as the necessary deflator in the data. Using the same method with the U.S. data for housing price indices as stated by Iacoviello/Neri (2010) yielded similar results to the data files in the original model (similar shape and size), so I assume this is the correct approach.
I'd be thankful if anyone has any hints, especially regarding the weight on housing.