More eigenvalues (>1) than forward looking variables

Can one suggest any solution to this?

MSP_newest_try_14_without_CPF_dynamics.mod (16.8 KB)

As always in this case, check the timing for all variables, in particular for the predetermined states

I calculated steady state values based on calibrated parameter values and found, one value is coming negative.

So can wrong parameterization also result into more eigenvalues greater than 1? Or is it just a timing issue?

Plus I don’t have any predetermined variable in my model.

Hi, Looking at your mod file it is obvious you have predetermined variables in your model (not only the exogenous variables, for instance K_g). So timing, as always, should be checked. That said, it is true that the instability (more unstable eigenvalues than forward looking variables) issue may be the result of a calibration error. For instance, if your model is an AR(1) and if the autoregressive parameter is greater than one in absolute value, then you will have exactly the same error.


Thank you so much Stepan!!! After devoting so much time into the model, your comments engender a new ray of hope!
Yes one of my AR(1) parameter is greater than one!! And I am currently fixing it.

And regarding K_g (and other capital), they depend on last period un-depreciated capital and current (not last) period investment, so they are not predetermined…correct me if I am wrong…

Hey, after correcting I now have 7 eigenvalues greater than one for 7 forward looking variables.
But still the code doesn’t run.

The rank condition isn’t verified !
Error: Blanchard Kahn conditions are not satisfied: indeterminacy due to rank failure.

Now what is the problem?
MSP_newest_try_15_without_CPF_dynamics.mod (17.1 KB)

This one is harder to fix. Did you search the forum?


I think I should try two things now:

include output gap as well in Taylor rule to make it more effective to fiscal block…so that model can be closed more strongly…

Find some redundant equation that model diagnostic probably is unable to find …MSP_newest_try_15_without_CPF_dynamics.mod (17.1 KB)

Should a model have at least one predetermined variable?

No, a model can be purely forward… Though it will probably be at odds with the data, because the reduced form solution will be a white noise (ie no persistence in the generated data).


The point about capital being predetermined is that often the production function depends on the beginning of period capital stock, i.e. before today’s investment is realized.
If you have a multi-sector structure,

may also be relevant

Thank you Professor!! Please clear my another general doubt:

In a multi-sector model, is it ok to deal with every sector seperately, having separate market clearing equations etc.

Or one should aggregate whatever disaggregated output and consumption one has in the model?

I frquently encounters this error: you have more equations than number of variables!!! I am asking in this context…

Every variable in your model needs one equation. When your aggregation implies that a variable is removed from the model, then one equation in the model is redundant.

Ok, Thank you Professor for an another modelling tip!!

Hello professor and Stepan,

In my model, I have 3 sector (g, v, m). The household’s budget constraint is:

PC + PI = wL + r_gK_g + r_vK_v + r_m*K_m ,

where I (total investment) is CES aggregation of sectoral investments (I_g, I_v and I_m),

and sectoral capital have law of motion are:

K_g(+1) = (1-del)*K_g + I_g ;
K_v(+1) = (1-del)*K_v + I_g ;
K_m(+1) = (1-del)*K_m + I_g ;

Then whether capital would be pre-determined?

MSP_new_try.mod (12.4 KB)

That depends on your intended setup. It seems you are using three separate capital stocks with capital unable to freely flow between sectors. In that case, they should be predetermined.