[Limdep Nlogit List] Bivariate ordered probit with sample selection of one of the ordered probits

[Richard Phillips]

I have firm level information where all of the firms in my data set have a
credit rating from an incumbent rating agency and some of the firms have
chosen to receive a second rating from a new agency that entered the market
and began issuing credit rating opinions.  The ratings for both agencies
follow an ordinal scale and there is a one-to-one mapping between the
ratings between the two agencies.

Let Y*1 be the underlying latent rating for the incumbent agency:
Y*1 = a¹X + u1
where the observed counterpart of Y*1 is Y1, X are observable firm
characteristics, and u1 is the error term for the incumbent agency.
 
Similar let Y*2 be the underlying latent rating for the new agency:
Y*2 = b¹X + u2
where the observed counterpart of Y*2 is Y2, X are the same observable firm
characteristics, and u2 is the error term for the new agency.

Let d* be the underlying latent variable for the decision to select a rating
from the new agency:
d*=c¹Z + d¹X + e
where d=1 if d* > 0 and 0 otherwise, and Z is a set of addition observable
firm characteristics that identify the selection equation, and e is the
error term for the selection equation.

[Y1, X, Z] is observed for all observations; Y2 is observed iff d=1.
u1, u2, e are distributed multivariate normal N3(0,0,0,1,1,1,p12,p1e,p2e).

Has anyone seen this model written up in a paper anywhere and/or does Limdep
9.0 estimate this model?  Any help appreciated.