Negative Standard Deviation MLE


Hope all is going well!

I am estimating the BGG with the financial accelerator using MLE. In particular, I am trying to estimate the parameter of the financial accelerator, the reaction of the Taylor rule and the (rho,sigma) parameters governing the three shocks.

However, after doing the ML estimation, I get a negative standard deviation for one shock. It is displayed that Dynare works with the variance so, there is not a big problem tho but I am puzzled how could I get a negative value.


The likelihood values for a positive and a negative standard deviation are identical due to being squared in the variance. If it bothers you, you can specify a lower bound of zero.