Initial values in shock decomposition

[Aldo]

I am estimating a version of Christiano, Trabandt and Walentin (2011), and when I compute shock decomposition I get initial values very persistent for data_RstarU which is interest rate of EE.UU, you can see this in page 121, for my other variables the initial values is not important.

Equations where interest rate of EE.UU appears is in:

Code: Select all

Current Account, Interest Rate Parity I think most important here, a Foreign VAR:

// ************************************* Foreign VAR: lystarU lpistarU lRstarU ******************************************
//****179
lystarU-log(ystar)=a11*(lystarU(-1)-log(ystar))+a12*(exp(lpistarU(-1))-pistar)+a13*(exp(lRstarU(-1))-Rstar)+ystar_eps/100;

//****180
exp(lpistarU)-pistar=a21*(lystarU(-1)-log(ystar))+a22*(exp(lpistarU(-1))-pistar)+a23*(exp(lRstarU(-1))-Rstar)
                     +a24*(lmuzU(-1)-log(muz))+a24*alpha/(1-alpha)*(lmupsiU(-1)-log(mupsi))
                     +c21*ystar_eps/100+pistar_eps/100;

//****181
exp(lRstarU)-Rstar=a31*(lystarU(-1)-log(ystar))+a32*(exp(lpistarU(-1))-pistar)+a33*(exp(lRstarU(-1))-Rstar)
                     +a34*(lmuzU(-1)-log(muz))+a34*alpha/(1-alpha)*(lmupsiU(-1)-log(mupsi))
                     +c31*ystar_eps/100+c32*pistar_eps/100+Rstar_eps/1000;

Is this some problem? How could I fix this?

[jpfeifer]

Some of your autocorrelation coefficients are very close to a unit root. That large persistence can explain why the initial values do not die out. You need to find out where this persistence comes from.