I am trying to stochastically simulate the SW2007 paper using the Pfeifer code available here. I would like all parameters to be calibrated to their posterior mean from Table 1 in the SW2007 paper.
I understand that the replication code involves a conversion of the risk premium process with the multiplier. I want to be sure I have done the conversions correctly. As far as I understand, in the Pfeiffer code the conversion is as follows b_pfeifer = c3*b_sw2007.
This would mean that there is no change to rho_b which should still be calibrated to 0.22. However, the st dev of this shock would need to be calibrated to a different value. In SW2007, this value is 0.23. Here it would need to be c3*0.23. Given c3 = 0.1244 at posterior mean, this would mean sigma_b = 0.0286. This is the only change that would be required to the calibration. Did I do this correctly?
This is confusing because the st dev for risk premium in the Pfeifer code is calibrated to 1.85 which would roughly correspond to (1/c3)*0.23 instead of c3*0.23.
You are correct. It looks to me as if the reported posterior in the paper is actually c_3*sigma_b, . The standard deviation in the mode-file is 0.2425, which is the value reported in the paper’s table 1B. This suggests that the reported value is actually the c_3*sigma_b, not sigma_b alone.
Thanks for your response! That’s interesting. Had never noticed that before. So, to use your code to match the posterior mean, I would actually calibrate sigma_b with the exact value from the paper. That is, I would set sigma_b = 0.23 since the paper seems to report c3*sigma_b and not sigma_b alone? And your code already has everything set up for c3*sigma_b to be used directly?
Yes, my code is original replication files. Thus, the _mode file is the one used in the paper and the parameter set there is the one actually used. Given the implementation, the sigma_b = 0.23 set corresponds to the c3*sigma_b used in the paper.