I subject the estimates to a variety of robustness tests.
First, I examine whether my estimates are affected by four
separate sample restrictions. To eliminate concerns that my
results are driven by a negative correlation between employment
and post-entry college attendance, I limit my sample
to women with a strong pre-welfare labor market attachment.
Second, I eliminate women who were already attending
college before entering welfare. Third, I eliminate individuals
who are still enrolled in college at the end of my
sample period. Finally, I drop observations from the quarters
surrounding welfare entry in which earnings display a dip. As
shown in Table 3, these estimates are quite consistent with
those generated from my preferred specification in Table 2
(columns 3 and 6). I find little evidence of earnings or employment
gains from earning additional community college
credits, large impacts of certificates and AAS degrees on earnings,
small and largely insignificant earnings gains following
AA/AGS degree receipt, and large increases in employment
following the receipt of any community college credential