Implementing LRE in King-Watson form

Dear all,
I am relatively new to Dynare, which may be the source of some confusion.

I am trying to replicate results from a GAUSS code solving a linear rational expectation model in King-Watson form, but run intro trouble. My hunch is that, like many in this forum, I am violating dynare’s timing convention.

Loglinearizing a set of FOCs gives the following set of equations in KW form:
AE_t(y_{t+1}) = By_{t} + C_0 \delta_t + C_1E_t(\delta_{t+1}) \\ \delta_t = \delta_{t-1} + e_t
Note that if y_t = \begin{bmatrix} \lambda_t\\k_t \end{bmatrix},
where \lambda_t and k_t are nx1 vectors, then A, B and C parameters take the following form:
A = \begin{bmatrix} a & 0\\ 0 & a \end{bmatrix}, B = \begin{bmatrix} b& 0\\ b& b \end{bmatrix}, C_0 = \begin{bmatrix} 0\\ c \end{bmatrix}, C_1 = \begin{bmatrix} c\\ 0 \end{bmatrix}

Assume we have n=2, then I have modelled those equations as:
a11*lambda1(+1)+a12*lambda2(+1)= b11*lambda1 + b12*lambda2 + c13*delta1(+1) ;
a33*k1(+1)+a34*k2(+1)= b31*lambda1 + b32*lambda2 + b33*k1 + b34*k2+ c31*delta1 + c32*delta2;
delta1 = delta1(-1) + e1;

which results in an error (Blanchard Kahn conditions are not satisfied: indeterminacy).

Pre-dating the second set of equations to
a33*EXPECTATION(-1)(k1)+a34*EXPECTATION(-1)(k2)= b31*lambda1(-1) + b32*lambda2(-1) + b33*k1(-1) + b34*k2(-1) + c31*delta1(-1) + c32*delta2(-1);

does not solve the problem either as:
The following endogenous variables aren't present at the current period in the model: k1 k2

What am I doing wrong?
KW_replic.mod (3.9 KB)

Predating would imply that you set those variables as predetermined using the predetermined_variables command, but leave the timing of the other variables (and thus the equation) unaffected.

1 Like