After the initial introduction of Medicaid to individuals eligible for AFDC, many states expanded their Medicaid programs to provide eligibility to pregnant women and infants during the 1980s-90s. Currie and Gruber (1996) exploit this variation across states over time to form a “simulated instrument” that attempts to isolate the impact of Medicaid eligibility, which they use to look at the impact on infant mortality; Cutler and Gruber (1996) study the impact on the crowd out of private insurance coverage; Dave et al. (2015) study the impact on labor supply of eligible women, and lastly Miller and Wherry (2018) study the impact on future earnings and health of children whose parents obtained Medicaid eligibility. Hendren and Sprung-Keyser (2020) combine these estimates and translate them into the implied MVPF of the Medicaid expansion.

Pays for Itself

Calculating the long-run net cost of Medicaid eligibility requires incorporating the upfront cost of the program and several key fiscal externalties.

Currie and Gruber (1996) report the upfront cost of expanding Medicaid eligibility is $202 (1986 USD) per eligible woman. Translating into 2011 dollars using the CPI-U-RS, this corresponds to $396. Scaling this estimate by the 11.4% of women in Currie and Gruber (1996) sample who are pregnant in any given year (footnote 17), suggests a total cost of of Medicaid eligibility of $3,473 per pregnant woman. Throughout their calculation Hendren and Sprung-Keyser (2020) follow Currie and Gruber and consider expanding eligibility, as opposed to take up. It is important to note, however, that only 34% of eligible individuals took up the policy. Calculating the cost of Medicaid per pregnant enrollee, requires scaling up the $3,473 using that 34% figure. That suggests the cost of providing a pregnant woman with Medicaid was $10,216 per birth.

Dave et al. (2015) provide evidence that the provision of Medicaid reduces female employment. This produces an important fiscal externality that Hendren and Sprung-Keyser (2020) incorporate in their MVPF calculation. Dave et al. (2015) estimate that Medicaid eligibility leads to a 21.9% reduction in employment. (In their MVPF calculation, Hendren and Sprung-Keyser (2020) draw this number from the specification in Table 2 corresponding to the regression without linear trends. This specification is chosen to be consistent with the original Currie and Gruber (1996) identification strategy.) Dave et al. (2015) note that 66% of their sample is employed with a mean earnings of $8,541, implying an average earnings of the employed of $8,541/.66=$12,941. Hence, the 21.9% impact on labor force participation implies roughly an earnings impact of $12,941*.219 = $2,834. Assuming a 19.9% tax rate, this implies a net cost to the government of $564 per eligible child. Hendren and Sprung-Keyser (2020) assume these effects last for a single year, roughly corresponding to the length of Medicaid eligibility offered through the expansion to the parents.

Calculating the cost of Medicaid eligibility requires considering the role that the government might play in providing health insurance for the uninsured. There is a large and growing literature suggesting that much of the uninsured end up not paying the full cost of their care when they go to the hospital. Finkelstein et al. (2019) suggest more than half of costs charged by low-income uninsured adults are uncompensated to the hospital. Moreover, the estimates in Gold et al. (1987) estimate that 50% of the cost of births by low-income mothers occur in public clinics financed by the government. This suggests that at least 25% of the upfront cost of Medicaid is recouped through reductions in other forms of uncompensated care spending. Taking this into account reduces the cost of Medicaid eligibility by $2,605 per enrolled pregnant mother.

Changes in long-run health outcomes also have a meaningful impact on the cost of Medicaid eligibility. Miller and Wherry (2018) document reductions in future hospitalizations of cohorts whose parents obtained Medicaid in states through these expansions. They document that a 1 percentage point increase in eligibility leads to a reduction in hospitalizations of 0.237% when children are 19 to 32 years old. The average resource cost of hospitalizations for these cohorts is $8,135 in 2011 USD per hospitalization. Using a 3% discount rate to account for the fact that these cost reductions occur roughly 26 years later, these costs would be $3,772 if they were paid at birth. Given a hospitalization rate of 374 per 10,000 individuals (0.0374), the discounted average cost of hospitalizations for these cohorts would be $141 (unconditional on hospitalization). This suggests that expanding Medicaid eligibility to the child’s mother reduces later life healthcare costs by $33 per year ($33 = 0.237% *100* $141). Hendren and Sprung-Keyser (2020) assume that 50% of these costs would have been paid by the government, which suggests an impact of roughly $17 on the government budget per year. Accounting for the fact that the effects are found in the data for children spanning 14 years, this suggests a reduction in costs of $239 per child enrolled. Assuming these yearly cost savings remain constant throughout the lifecycle up to age 65, the cost savings rise to $530.

In addition to cost savings from reduced medical expenses, Miller and Wherry (2018) also document impacts on high school graduation and future earnings. A 1 percentage point increase in eligibility leads to an increase in high school graduation of 0.028 percentage points and an increase in personal income of 0.116%. The sample mean earnings is $31,331 (in 2011 USD), implying that the counterfactual mean earnings is $30,246 per year. Thus the increase in personal income from a one percentage point increase in eligibility is $35.09. This implies a yearly increase in earnings of $3,509 in 2009 for those eligible and a discounted sum of earnings gains over the observed period of $21,860. Hendren and Sprung-Keyser (2020) use these estimates to lifecycle earnings gains and then discount those future earnings at 3%.They assume an effective tax rate of 18.9% based on Congressional Budget Office estimates of the rate at 200% of the Federal Poverty Line. Putting those estimates together implies that providing Medicaid eligibility increases future tax revenue by $10,024 per enrolled parent.

While the increased tax revenue reduces the effective cost to the government, some of these savings are offset by the fact that the government subsidizes college attendance. Miller and Wherry (2018) find an increase in college attendance of 3.5 percentage points. Using the national average public expenditure on college in 1997 of $6,448, and assuming those induced to enroll attend for two years this implies a cost increase of $371.

Summing together all of these components suggests a net cost of $-7,014.

-$7.0K
Net Cost

-1.2K Upper Margin
-13.0K Lower Margin

Hendren and Sprung-Keyser (2020) include three components in the willingness to pay: the transfer to parents, the WTP for reductions in infant mortality, and children’s WTP for increased earnings in adulthood.

First, those who previously paid for insurance who now obtain Medicaid coverage instead. This constitutes a transfer from the government. Cutler and Gruber (1996) estimate that as much as 50% of women who become eligible for Medicaid choose to drop their private insurance policy. Assuming 50% of the market is crowded out of private insurance and that the cost of the private insurance is roughly equal to the resource cost of the expansion ($3,473 per year), the policy reflects a $1,737 transfer per eligible pregnant mother.

In addition, there is a willingness to pay from reductions in infant mortality. Currie and Gruber (1996a) estimate that a 1 percentage point increase in Medicaid eligibility results in .028 fewer infant deaths per thousand births. To translate these into a WTP statistic relevant for the MVPF, one needs to multiply this by the parents’ WTP for a decrease in the odds of losing a child (alternatively the child’s own hypothetical willingness to pay for their life). There is of course considerable debate about this parameter, Hendren and Sprung-Keyser (2020) assume here that individuals have a value of a statistical life (VSL) of $1M in 2012 dollars. (These results are similar with VSL assumptions between $0 and $9M). This means that parents are willing to pay $10K for a 1% reduction in the infant mortality. This suggests a willingness to pay of 1M*2.8/1,000=\$2,800. In 2011 dollars, the willingness to pay is $2,763.

Third, Hendren and Sprung-Keyser (2020) consider the WTP by the children for improved labor market prospects in adulthood. While some of the the 11.6% increase in gross earnings accrues to the government in terms of increased taxes, individuals should be willing to pay for the remaining after-tax increase in income. (This calculation assumes that higher earnings reflects an increase in wages, not an increase in the utility cost of effort.) Over the observed 14 years of earnings this suggests a willingness to pay of roughly $16,775. Extrapolating the earnings effects out to age 65, Hendren and Sprung-Keyser (2020) calculate that the policy results in an additional $26,236 in gains. Summing the components results in net willingness to pay of $47,400 ($43K in earnings, plus $2,763 from infant mortality, and the $1,737 transfer from crowding out of private insurance).

$47.4K
WTP

72.8K Upper Margin
22.0K Lower Margin

The resulting MVPF is infinite because costs are negative with a 95% confidence interval that has a lower bound of \infty. This occurs because the 95% CI for net costs is entirely contained below zero.

The infinite MVPF estimate is true both for the baseline specification in Hendren and Sprung-Keyser (2020), and for their lower bound WTP specification which does not incorporate any infant mortality reduction benefits. The MVPF is also infinite if one does not project costs beyond the observed time frame. Program costs are fully recouped by age 32.

MVPF

Congressional Budget Office (2016). Effective Marginal Tax Rates for Low- and Moderate-Income Workers. Accessed 06/28/19. https://www.cbo.gov/publication/50923

Currie, Janet and Jonathan Gruber (1996). “Health Insurance Eligibility, Utilization of Medical Care, and Child Health*.” The Quarterly Journal of Economics, 111(2), 431-466. DOI: https://doi.org/10.2307/2946684

Cutler, David M. and Jonathan Gruber (1996). “Does Public Insurance Crowd out Private Insurance?” The Quarterly Journal of Economics, 111(2), 391-430. DOI: https://doi.org/10.2307/2946683

Dave, D., S. Decker, R. Kaestner and K. Simon (2015). “The Effect of Medicaid Expansions in the Late 1980s and Early 1990s on the Labor Supply of Pregnant Women,” American Journal of Health Economics, 1(2), 195-193. DOI: https://doi.org/10.1162/AJHE_a_00011

Gold, Rachel Benson, Asta M. Kenney and Susheela Singh (1987). “Paying for Maternity Care in the United States.” Family Planning Perspectives, 19(5), 190-206. DOI: https://doi.org/10.2307/2134964

Finkelstein, Amy, Nathaniel Hendren and Erzo F.P. Luttmer (2019). “The Value of Medicaid: Interpreting Results from the Oregon Health Insurance Experiment,” Journal of Political Economy, 127(6). DOI: https://doi.org/10.1086/702238

Hendren, Nathaniel and Ben Sprung-Keyser (2020). “A Unified Welfare Analysis of Government Policies.” The Quarterly Journal of Economics, 135(3): 1209–1318. DOI: https://doi.org/10.1093/qje/qjaa006

Miller, Sarah and Laura R Wherry (2018). “The Long-Term Effects of Early Life Medicaid Coverage.” Journal of Human Resources, January 30, 2018 0816_8173R. http://jhr.uwpress.org/content/early/2018/01/25/jhr.54.3.0816.8173R1.abstract

- Category
- Social Insurance
- Sub-Category
- Child Health
- Beneficiary Type(s)
- Adults, Children and Youth, Children under five, Mothers and pregnant women, Parents
- Average Age
- 1
- Average Income
- 31930
- Country of Implementation
- United States
- Year of Implementation
- 1986
- Empirical Method
- Difference in Differences
- Research Type
- Secondary
- Peer Reviewed
- Yes
- MVPF Publication Link
- academic.oup.com/qje/article/135/3/1209/5781614