# Large Coefficients of AUTOCORRELATION

Dear All,
I have simulated a model an I found large coefficients of autocorrelation. I would like to know if this a sign of a possible problem in my model.
More precisely, I got e.g,
AUTOCORRELATION OF SIMULATED VARIABLES
VARIABLE 1 2 3 4 5
lPIB_qc 1.0028 0.9954 0.9880 0.9802 0.9722
ly_qc 0.9412 0.9448 0.9444 0.9429 0.9389
ly_rc 0.9491 0.9482 0.9495 0.9492 0.9447
lw_qc 0.9900 0.9869 0.9835 0.9800 0.9766
lw_rc 0.9852 0.9814 0.9778 0.9752 0.9721
Regards,
Ful

1 Like

Hi,
as always, this depends on your model. A high persistence per se is not a problem as some shocks like technology show usually show a high persistence. This persistence often transfers to other variables. Moreover, if you only have one shock in the model, let’s say a supply shock, this shock moves all variables always in one direction, leading to a high correlation between all endogenous variables.
However, lPIB_qc has an autocorrelation larger than 1 which signals an explosive process. This is clearly a warning sign. However, without knowing your model it is hard to judge. Is your model stationary, i.e. did you detrend it by e.g. using first differences?
Note that the estimated autocorrelation may also be a result from a short simulation sample. You should also look at the theoretical moments.

1 Like

Thanks