(7) The second problem is associated with assortive mating. (8) If the estimated equation suffers from unobserved variables that affect compensation packages of both spouses, then assortive mating will cause a positive correlation between Hs and the error terms in Equation (2), [v.sub.i], and Equation (1), [[epsilon].sub.i].
Again, firm size, the same as spousal health insurance, may produce biased results due to assortive mating. In this case, if couples with low unobserved productivity factors take low-paying jobs with small firms that do not offer health insurance, then there will be a positive correlation between spousal firm size and the error term in the earnings equation.