Results from cross-sectional studies that compare pointin-time
labor market outcomes across individuals with different
levels of education time may be biased by unobservable
factors that affect both the decision to attend college and labor market outcomes (Card, 1999). One approach to addressing
with this type of selection bias is to use repeated observations
of earnings and employment for individuals who
are employed both before and after they enroll in college
and include individual fixed effects in the empirical speci-
fication. With a sufficiently long panel of data, each student’s
pre-college earnings potentially serves as a counterfactual for
what she would have earned had she not gone to school,
eliminating concerns of selection on time-invariant characteristics
such as ability or motivation. This strategy necessarily
excludes “traditional” students who matriculate immediately
following high school graduation; however, community colleges increasingly serve older, non-traditional students.13
Furthermore, many federal job training programs target older
workers who have been displaced from previous jobs. The average
age of women in my sample who enroll in college following
welfare entry is 28