I’m trying to implement the model from “TECHNOLOGY INNOVATION AND DIFFUSION AS SOURCES OF OUTPUT AND
ASSET PRICE FLUCTUATIONS” by Comin, Gertler and Santacreu. In the steady state I get non-zero residuals in 2 equations. I know what it means. Obviously, there is a mistake in my steady-state, but I can’t find it. Remark: some parameters are chosen in an another way than in the paper. I ask you to look at the mod-file and I would be very thankful if you could tell, whether there is and where is the mistake. Thanks in advance!!! cgs.mod (6.5 KB) w15029.pdf (539 KB)
% stochastic processes
log(chi) = rho*log(chi(-1)) + eps; %31
log(x) = eta*log(x(-1)) + sigma; %32
log(g) = nu*log(g(-1)) + varrho; %33
log(p_k_st) = rho_st*log(p_k_st(-1)) + eps_st; %34
means they will all have steady state 0. That is not what you use for x and g.
I finally could derive the steady state equations to get all the residuals = 0. I’m getting 8 eigenvalues greater than one for 12 jump variables. So I use the model_diagnostics command. And I get the following output:
model_diagnostic: the Jacobian of the static model is singular
there is 7 colinear relationships between the variables and the equations
Relation 1
Colinear variables:
y
c
g
p_k
j
i
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_et
n_k
j_y
j_k
i_s
i_e
lambda_y
lambda_k
p_k_bar
Relation 2
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 3
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 4
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 5
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 6
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 7
Colinear variables:
c
g
p_k
j
z_k
z_y
v_k
v_y
n_y
a_k
a_y
h_k
h_y
x
k
l
o_k
o_y
pi_k
pi_y
p_k_st
p_k_et
n_k
j_y
j_k
lambda_y
lambda_k
p_k_bar
Relation 1
Colinear equations
6 7 9 10 12 13 14
Relation 2
Colinear equations
6 7 9 10 12 13 14
Relation 3
Colinear equations
31
Relation 4
Colinear equations
32
Relation 5
Colinear equations
33
Relation 6
Colinear equations
34
Relation 7
Colinear equations
6 7 9 10 12 13 14
The presence of a singularity problem typically indicates that there is one
redundant equation entered in the model block, while another non-redundant equation
is missing. The problem often derives from Walras Law.
What does this colinearity mean? How could it be, that only one equation at a time (No. 31 - 34, stochastic processes) is colinear? Is the eigenvalues problem arising from the steady state or does it depend on the starting values for k, lambdas and other? Varying this I get from 5 to 8 eigenvalues > 1, bot not more. cgs.mod (6.53 KB)
Your exogenous processes specify unit roots. That’s why there is collinearity in these equations. Given the many missing explosive eigenvalues, I would think that you are having a systematic timing issue in your model.
I rewrote three of four exogenous processes, such that they are sorting with the paper now. But in the paper one process looks like
log(X_t) = log(X_t-1) + sigma; %31
Can it be a problem?
Furthermore I get 3 colinear equations (number 6,7 and 30). In order to have the equation
log(Chi_t) = rho*log(Chi_t-1) + eps; %30
fulfilled in the steady state I choose Chi = 1. But then the equations 6 and 7
z_y(+1) = (chi_y_bar*(chi^ksi_y) + phi)*z_y; %6
z_k(+1) = (chi_k_bar*(chi^ksi_k) + phi)*z_k; %7
imply that in the steady state chi_k_bar + phi = 1 and chi_y_bar + phi = 1. But that contradicts the calibration of the model (chi_k_bar = 0.0304, chi_y_bar = 0.0202, phi = 0.99). What is the right way? Thanks in advance! cgs_main_steady.mod (6.87 KB)
The first one will give you a unit root, which is a feature of the model as far as I can see. But i
I cannot work myself into the calibration of the model. So you need to figure out how to make the steady states consistent.
But I would like to point out that the last two equations are not conforming to Dynare’s timing convention. They are laws of motion that describe predetermined variables.
in Dynare means
I guess that z_y(+1) must be predetermined here like capital.