To address this concern, I follow an empirical approach
similar to Jacobson et al. (2005), taking advantage of the
fact that I observe earnings for women both before and after
welfare receipt. If selection bias is only driven by timeinvariant
unobservable characteristics, then including an
individual fixed effect will deal with the endogeneity of college
attendance and credential receipt. Essentially, this approach
differences out a person-specific mean level of wages
from observed quarterly wages, using observations of individuals’
wages prior to college attendance as a counterfactual
for current wages in the absence of college-going. Additionally,
I include observations of individuals who do not
attend college to identify year and quarter fixed effects and
the impacts of other observable characteristics. The key identifying
assumption is that there are no time-varying unobservable
shocks correlated with educational attainment and
labor market outcomes that affect college-going women differentially
than other recipients.