On forward looking variables

Dynare help (legacy posts)
1

Hi,
I’m a bit confused on the definition of “forward looking variables” because when running my model, dynare just report

"There are 10 eigenvalue(s) larger than 1 in modulus
for 9 forward-looking variable(s)
Blanchard Kahn conditions are not satisfied: no stable equilibrium
"
To solve the problem, I need to figure out what is the forward looking variables defined by dynare.
According to message in About the rank condition
Then I understand "a forward-looking variable is a variable that appears in the model with a lead"
Yet, if I follow this definition, I find that my model have 12 forward looking variables. So I guess I misunderstand the definition. Is there anyone can help?
Besides,I have ever looked at GarciaCiccoetal2010.mod written by Johannes Pfeifer. In that mod.file, k(+1),k(+2),g(+1),d(+1),h(+1) are five forward looking variables and the model has exactly 5 eigenvalue(s) larger than 1 in modulus.

Thanks in advance.

2

This is tricky. In general, you are right. In more complex models, some variables are substituted out (e.g. lags and leads of order>1). Moreover, when you use the predetermined_variables command, some states can look as if they are forward-looking, but they are not, because e.g. k_t+1 is predetermined.

3

Thanks for prompt reply!

4

Hi, I try to solve this problem by simplifying the model. Yet, it really confuse me when the following error is report by dynare:
There are 6 eigenvalue(s) larger than 1 in modulus for 6 forward-looking variable(s)
The rank condition ISN’T verified!
Error using print_info (line 48)
Blanchard Kahn conditions are not satisfied: indeterminacy due to
rank failure
I will rather believe that 6 eigenvalue(s) larger than 1 in modulus for 6 forward-looking variable(s) means that Blanchard Kahn conditions are satisfied. Anyone here can help to explain what’s going on? If necessary, I can post my codes.
Thanks

5

This is not about Blanchard Kahn, but about the rank condition. Try running model_diagnostics.

6

Thanks for kindly suggestion. Problem solved.

7

Hi, everyone,

I am having a problem related to forward-looking variables and eigenvalues. After I started the .mod file, I got a message:
"There are 13 eigenvalue(s) larger than 1 in modulus
for 14 forward-looking variable(s)

The rank condition ISN’T verified!"

By reading comments here and other parts of this forum, I understand that I have a problem with leads and lags. But, something that has to be add:

  • I am using an alternative .m file to calculate my steady state;
  • All my .mod script is a linear model;
  • When I comment the command “steady;” the problem is gone and I have all the results. I noticed that my impulse-response are not convergent, but I do not know if it is a problem related to the periods.

I did not post my mod file because I do not have in this moment, but someone has a guide to solve this problem and learn about the question?
Thanks in advance

8
  1. Are you using the unstable version? 4.4.3 may accept wrong value as the steady state.
  2. What you describe sounds like a matter of parameterization. For the steady state provided, the model does not satisfy the BK conditions.
9

Jpfeifer,

I am using 4.4.3 version. Have I to use another version?

When you say: “matter of parameterization”, you are saying that the parameters that I chose are not getting SS, right?

Thanks!

10

Please try the unstable version.
Without having seen the mod-file, it is impossible to tell where the problem might be

11

Just a moment

12

Hi jpfeifer,

My .mod and steady state are now available to you. I returned to the basic model in order to correct some mistakes that I did. The basic model is not the original, but with just one difference.
I am a little confused about these points:

  • the IRF are weird;
  • the calculated SS is ok, but the results presented in console sound weird, because some of the variables are negative.

What do you think about? Is it corrected? Could I keeping going and add other variables and equations that I would like to?
SS_k.m (1.97 KB)
steady_state_k.m (1.58 KB)
BNDES_v12.mod (4.19 KB)

13

There must be something still wrong. Your model is linear, thus the steady state of the variables should be 0 as they are percentage deviations from steady state. That is not the case, because there is a constant in Equation 7. Putting resid
before steady yields

[quote]Equation number 7 : 0.016329
[/quote]

14

Yes, you are correct! I find a mistake

(1-nu)*R_fss*Rss*(R_f + R) + nu*R_fdss*Rss*(R_fd + R) = ((fi-1)/fi)*alpha*((P_1ss*Y_1ss)/K_1ss)*(P_1 + Y_1 - K_1(-1)); // Produtividade mrginal do capital 1

instead of

codeR_fssRss*(R_f + R) + nuR_fdssRss*(R_fdss + R) = ((fi-1)/fi)alpha((P_1ssY_1ss)/K_1ss)(P_1 + Y_1 - K_1(-1)); // Produtividade mrginal do capital 1
[/code]
However, when I corrected it, I do not have rank condition satisfied

[quote]The rank condition ISN’T verified!
Error using print_info (line 48)
Blanchard Kahn conditions are not satisfied: indeterminacy due to rank failure

Error in stoch_simul (line 98)
print_info(info, options_.noprint, options_);

Error in BNDES_v12 (line 318)
info = stoch_simul(var_list_);

Error in dynare (line 180)
evalin(‘base’,fname) ;
[/quote]

I checked the steady_state file and it is correct.
Do you have another suggestion?

Thanks again!

PS: I attached the new .mod corrected.
BNDES_v12.mod (4.2 KB)

15

Check all model equations again, particularly the timing. There may be more mistakes left.

16

Jpfeifer,

Yes, I found some differences in comparison to the article. However, the origina model is also different and still works! I am doing some tests in order to find the rank condition.

17

Jpfeifer,

I am checking. I noticed the original code is not exactly according to the text; I use the original one in order to apply my modifications. The equations are quite similar to original code. Should I change according to the text?

18

I don’t know which paper you are talking about, but you need to find out which version is correct. Either the paper or the code are correct.

19

Yes, you are right. The paper is Portuguese, but at the end, the author showed some of the equations he put in the code. I am thinking the code is the correct. Well, I will research more about with friends in order to find a solution! Thanks!
Dissertação - Rafael Mouallem Rosa.pdf (530 KB)