I'm running a heckman model however some issues have arisen:

1. The log likelihood function is not concave and does not converge after a reasonable number of observations, I've tried up to 400 iterations. Why might this occur?

2. While I drop one of the the variables I used in the selection equation to identify the conditional equation, I end up with a rho of 1; perfect correlation between error terms. This does not make sense to me, any suggestions why this might be the case?