The second column of the table estimates the second-stage regression using an alternative
dependent variable, the probability that the typical person in cell k has some type of
health insurance coverage. This coefficient is numerically close to zero and statistically
insignificant (the coefficient is −0.25, with a standard error of 0.60). The results, therefore,
indicate that a cutback in the probability of receiving Medicaid generates a completely
offsetting increase in the probability that a person is covered by employer-sponsored insurance,
thereby leaving unchanged the probability that the person has some type of health
insurance coverage.