I have a very simple 2-country endowment model (by Devereux/Sutherland) that I have solved in Dynare (3 endogenous state variables (Z1,Z2,W), 2 control variables ( C1, C2), and 4 shocks (e1_y, e1_l, e2_y, e2_l).
When I solve the same model with the solution-algorithm by Binder and Pesaran (1997), for the variables C1, C2, W I obtain a different recursive law of motion compared to Dynare (Z1 and Z2 are the same). This suggests to me that I am making a mistake in defining the matrices (Chat, Ahat, Bhat, D1, D2, Gamma) for this specific algorithm:
Chat * x[t] = Ahat * x[t-1] + Bhat * E(x[t+1]|I[t]) + D1 * w[t] + D2 * E(w[t+1]|I[t]),
where w[t] = Gamma * w[t-1] + v[t]
The recursive law of motion is:
x[t] = Cx[t-1] + Hw[t]
Matrix C should have all eigenvalues within the unit circle, however, this is violated in my case. I have used the algorithm before for a simple stochstic growth model, which delivered the same recursive law of motion as in Dynare.
Any help or hints would be great!