Thank you for your kind reply. Yes. I run the code with their data. I also checked whether subtracting a preference shock affects results, in that case correlation matrix is the following. It seems strange, does not it? I understand when the correlation is less than 0.5, but the numbers(0,86,-0,81 ) are so high. May I miss something ?
e_a e_g e_mu e_s
e_a 1.00 -0.81 0.46 0.16
e_g -0.81 1.00 -0.35 -0.36
e_mu 0.46 -0.35 1.00 0.04
e_s 0.16 -0.36 0.04 1.00
Bahar.