significantly and the magnitudes are smaller, suggesting that the increase is
primarily on the extensive margin.
In Table 4 we examine the effect on women’s wages using a tobit
regression that adjusts for topcoding.14 Unfortunately, the question of usual
weekly wages was only asked to two of the CPS rotation groups, and sample sizes
for this variable are thus much smaller than for the other outcome measures. The
coefficients of interest are not statistically significant as shown in columns (1) and
(2). However, log weekly earnings in Table 4, Panel I, columns (3) and (4)
increase for all women (a 6% to 8% change in wages) and are somewhat larger
(an 11-16% change in wages compared to 8-9%) for more educated women
(Table 4, Panel III, columns (3) and (4)) than for less educated (Tables 4, Panel
II, columns (3) and (4)). Because increases in weekly earnings may result from
increased hours worked and/or increased wages (which may be affected by
choices in benefits), we are reluctant to put too much theoretical emphasis on the
earnings results.
Women who had health insurance coverage from their employers prior
to the policy implementation may be more likely to stay employed than those who
did not in order to keep that coverage. However, women who did not have their
own health insurance coverage may need to earn more money to self-insure if
14 Results trimming the top 5% of wages in an OLS regression framework are very similar.
19