The remaining rows of Table 8 re-estimate the regression models using alternative definitions
for the vector I. As noted earlier, the citizenship status of a person may be partly
endogenous, as many immigrants become naturalized citizens to escape the impact of the
PRWORA cutbacks. One simple solution to this problem is to use the immigrant’s year
of arrival in the United State to define the various groups. In the second row, the dummy
variables in the vector I indicate if a person is native, has lived in the United States for more
than 10 years, has lived in the United States for fewer than 10 years, or is a refugee. The
coefficient δ reported in the first column is −1.01 (0.87). Moreover, note that the coeffi-
cient is 0.28 (0.74) in the second column, where the dependent variable is the probability
of being covered by some type of health insurance. Therefore, there seems to be a negative
structural relation between the probability of receiving Medicaid and the probability
of having some type of health insurance coverage, again suggesting a strong crowdout
effect