I recently download the replication code for BGG model. In the paper, Bernanke et al. also provided the baseline model with no financial accelerator. How can I get the replication of the model with no financial accelerator? I would like to compare the two different models.

It seems you need to strip out the financial accelerator features yourself. It does not seem to be sufficient to just change the parameters, because the steady states are hardcoded. That complicates the issue considerably.

thanks, i will try that

Dear all,

I am facing the same issue. Here are my thoughts, any comments are very welcome:

The FA enters the model via: E_t(r^k_{t+1})-r_{t+1}=-nu[n_{t+1}-(q_t+k_{t+1})]

where nu is the elasticity of the premium with respect to N/K. In shutting the FA off I set nu to zero so that the premium deviations from SS are not affected by the leverage. However I subsequently get amplified IRF. So I assume something must be wrong.

There is also an equation governing the accumulation of wealth but that equation does not seem to belong to the FA directly.

I agree that the SS would change if the FA is turned off since the leverage essentially becomes irrelevant so firms would love to purchase more capital by borrowing more in steady state. In papers I studied so far however the steady state is kept at a constant level and they still get a smaller response.

Hey JMatschke,

I tryed to turn off the Financial Accelerator the same way. In my case it seems to work well. Maybe my dynare file can help you. FinancialAccelerator.mod (4.0 KB)

Your model delivers theoretical moments containing NaNs. How can this be corrected?

Indeed the provided mod-file has a unit root.

@Max1 It is easy to remove the unit root: Try nu=0.001 instead of nu=0 to switch off the financial accelerator. Furthermore, the autocorrelation of technology should be 0.999 instead of 1.

Thank you for the advice.

The unit root (eigenvalue =1) stems from the technology autocorrelation-parameter which is assumed to be 1 in the paper of BGG.

But what is the source of the NaN´s in the theoretical moments?

Maybe I can ask you another (stupid) question at this point; why is dynares preprocessor not interrupting the execution of the file? Together with the unit root there are 4 unstable EV and 3 forward looking variables. Can dynare distinguish between endogenous and exogenous eigenvalues?

Thanks in advance.

Dear community,

I also have a question regarding swithcing the financial accelerator in Bernanke, Gertler, Gilchrist (1999) on and off. On page 1370 the write:

“the “baseline” impulse response, generated by fixing the external finance premium at its steady state level instead of allowing it to respond to changes in the capital-net worth ratio. In other words, the baseline simulations are based on a model with the same steady state as the complete model with imperfect credit markets, but in which the additional dynamics associated with the financial accelerator have been “turned off”.”

How does fixing the external finance premium at its stst level generate a comparable model? If you fix the external finance premium (it is then like a wedge in required returns on capital stocks and returns demanded by lenders), the model will yield a steady state with entrepreneurs borrowing significantly more whilst having lower net worth, i.e. leverage increases. Output and other aggregates are higher as well. This is reasonable as the frictions do not increase the more entrepreneurs borrow. But it is not the identical economy.

Nevertheless, they write “based on the same steady state” and compare it with their baseline (financial frictions) model. How is this correct? It puzzled me for some time already. Any help is much appreciated.

It seems they have worked with the linearized model and then shut off the financial accelerator, keeping the linearization point constant. But that would not generate the issue you mention. You would have the same external finance premium in steady state in both scenarios. The only issue is that a higher leverage than in steady state would not result in a higher premium.

Dear jpfeifer,

thank you for your insights. I understand your point, but it is indeed the higher leverage that was the center of my question. It will yield higher capital stocks and higher output, thus a different steady state, eventhough the premiums are identical. Right?

And now my question is how legitimate is it to compare these two steady states through IRFs (naturally starting at identical points, i.e. 0% deviation from each steady state). Would it not be more transparent in this case to report absolute values and note that IRFs are for comparable but not identical economies? Maybe I am making too much a fuss about it, I dont know…

I don’t think you get my point. Say in the baseline model the external finance premium increases by 1% for every 10% of leverage and that the steady state leverage is 20%. In that case, the steady state external finance premium is 2 %. When computing the IRFs and leverage increases 10% above steady state, the premium will also rise 1% above steady state.

What the alternative model does is set leverage to 20% in steady state and the external finance premium to 2%. That means the steady state in the two economies is identical. What they now shut of is the response of the premium outside of steady state to increases in leverage. So if leverage goes up by 10% to 30%, the premium will stay at 2%.

Indeed, I did not quite get you point at the first read. I missed that they used the same point for linearization and therefore have identical steady states. Thank you for your help!