Other important questions about Gertler and Karadi (2011)

Hi, the last topic I’ve made was about the possible implementation of the ZLB in Dynare of the Gertler and Karadi (2011) paper “A model of unconventional monetary policy”. Some people said to me that for that implementation I first have to log-linearize the model. Well, I tried to, but an error occured. I attach the mod file here GK_loglinear.mod (12.3 KB) . The error is as follows:

Error using print_info (line 37)
The generalized Schur (QZ) decomposition failed. For more information, see the documentation for Lapack function dgges:
info=30, n=28. You can also run model_diagnostics to get more information on what may cause this problem.

Error in check (line 76)
print_info(info, 0, options);

Error in GK_loglinear (line 611)
oo_.dr.eigval = check(M_,options_,oo_);

Error in dynare (line 235)
evalin(‘base’,fname) ;

The fact that the log-linearized model doesn’t work is very strange, considered the fact that the non linear model works without problems and it perfectly replicates the results of the third figure of the paper NonlinearGK11.mod (12.7 KB)

The other questions I’m going to ask are more conceptual. Since I’m currently working on this paper in order to use it for Japanese data, I need to know how two targeted Steady States (or Targeted Moments) have been computed for the US (since the paper is based on US data). The first one is the Steady State Leverage Ratio, based on the average of the averages of the Leverage Ratios of financial sectors in the US, and it is four, and the second one is the Steady State Interest Rate spread, based on the pre-2007 spreads between mortgage rates and government bonds and between BAA corporate vs. government bonds, and it is one hundred basis points. How did they compute them and from which source they took them? (OECD.Data site? World Bank site? IMF site?)

For the Leverage Ratio, I had an idea on how they could do that, but I’m not sure of it. Correct me if I’m wrong. Using this site Corporate sector - Banking sector leverage - OECD Data I downloaded the U.S. data of the items “Banking sector leverage” and “Financial Corporations debt to equity ratio”, from 1995 (the first year in which the data are available) to 2006 (since GK11 say they used pre-crisis Leverage, i.e. before 2007). Since GK11, as far as I’ve read, say they calculated phi_ss as the average of the pre-crisis financial sectors’ leverage, I did this experiment. I downloaded both datasets, averaged them by year, and calculated the average of these averages up to 2006, the result is approximately 4 (I came up with 4.2485), like the value chosen by GK11, but it could be just a coincidence, I don’t know. What do you think?

As for the targeted Steady State of the interest rate spread, I sincerely have no idea about they compute it, but I know that they set him at hundred basis points. So why in the files of Peter Karadi’s page that value is set at 0.01/4? Correct me if I’m wrong, but I don’t think that’s one hundred basis points, like they said in the paper. So why that value? I don’t know.

Thanks again in advance for the kind and useful replies. I wish you all a good day.

  1. Manual log-linearization is very error-prone and not recommended. Why do you need it?
  2. Regarding the 0.01/4 that is 25 basis points for the quarterly interest rate (the frequency of the model) and therefore roughly 100 basis points at an annualized level.
  1. I know. I did it on my own just because I’m relatively new on Dynare, so I have to learn to use all the tools properly (including the non-manual linearization).

  2. Thanks, now I get it. Do you know where I can find a database in which I could take the spreads in basis points of a particular country? I would need it in the case I’d want to use this model with data from other countries (other than just the U.S.). Regarding the traditional parameters I could use Bayesian Estimation or take them from other papers, no problem about that. But regarding the Steady State target parameters, like the Leverage and the Spread, I need to compute them on my own and, as for the spread, I don’t know where I can take it now.

No, usually you need to search for the data on a country by country basis.