correlation coefficient in recursive bivariate probit

[Varsha Vaishnav]

Hello,

I have estimated a recursive bivariate probit model to study the association between women's bank account ownership and poverty using -rbiprobit-. The correlation between the residuals is coming out to be positive and significant, but I was expecting it to be negative as unobservables that increase poverty should also limit women's access to financial resources. I'm confused as to how I should go about it. Please help.

Thanks