Dynare forums

I am simulating a NK model which has capital in it. Since capital is predetermined, I took the capital equation one period backward. Taking the capital equation one period backward, I get: k=(iss/kss)*i(-1)+(1-delta)*k(-1). But the model does not solve since I am getting indeterminacy when I run it.

However, if I also take the return on capital equation ,rk, one period backward the model solves. When I take the rk equation backward, I get:
rk=((ilss*ibss)/rkss)*(ib(-1)+il(-1)-pi+theta_k*(k-k(-1)))-(ilss/rkss)*((1-delta)*il+theta_k*(k(+1)-k));

Why is it that I get indeterminancy when I run the model as is (only taking the capital equation one period backward), but when I take both the capital and return on capital (rk) equations one period backward, the model solves?

Can anyone please tell me why? Also, is this strategy for getting a solution correct?

What do you mean with taking the rk equation backward? Your timing need to be consistent. If you have

Code: Select all

k=(1-delta)*k(-1)+i

and

Code: Select all

rk=alpha*k(-1)^(alpha-1)*l^(1-alpha)

the Euler equation must be something like

Code: Select all

1/c=beta*1/c(+1)*(1-delta)+rk(+1))

Thanks for replying jpfeifer,

The original rk equation is:
rk(+1)=((ilss*ibss)/rkss)*(ib+il-pi(+1)+theta_k*(k(+1)-k))-(ilss/rkss)*((1-delta)*il(+1)+theta_k*(k(+2)-k(+1))); (1)

If I take the above equation one period backward, I get:
rk=((ilss*ibss)/rkss)*(ib(-1)+il(-1)-pi+theta_k*(k-k(-1)))-(ilss/rkss)*((1-delta)*il+theta_k*(k(+1)-k)); (2)

When I run the model with rk one period backward using equation (2), it solves.

My euler equation is coded in dynare as:
c(+1)=c+sigma*(ib-pi(+1)).

Should the rk equation (1) or (2) be used and is my overall timing consistent? Please let me know. Thanks.

Sorry, but I do not get, why rk only appears in expectations. Usually, only the second equation should appear.

Thanks jpfeifer. In my model, the expected rental rate of capital depends on the current and expected loan rate, inflation and the bond rate.
I have tried entering rk as stated in (1) and using the 'predetermined_variables k;' notation and the model solved. Is this ok? Thanks again.

Considering your confusion with the timing, it is best to use the predetermined_variables command (which by the way does nothing else than shift the timing of capital by one in all equations) together with the usual begin of period-notation that is regularly used in papers.

Okay, thanks jpfeifer.