Responsible researcher: Bruno Benevit
Authors: Amy Finkelstein, Nathaniel Hendren and Erzo FP Luttmer
Original title: The Value of Medicaid: Interpreting Results from the Oregon Health Insurance Experiment
Intervention Location: United States
Sample Size: 75000 adults
Sector: Healthcare
Variable of Main Interest: Health coverage
Type of Intervention: Eligibility for health insurance coverage
Methodology: OLS
Summary
The analysis of well-being in relation to health programs is a crucial topic in the context of health policies. Therefore, observing beneficiaries' willingness to pay for purchased services is essential for evaluating health programs. In this context, this study addressed the relevance of expanding these policies by examining an experiment carried out in the state of Oregon to expand the Medicaid program, aimed at low-income and uninsured adults. Medicaid transfers to beneficiaries were identified through different approaches, while estimates of the additional value for protection against risks showed greater variability. Furthermore, the research revealed that most of the program's expenses are not directly related to operational costs, but rather to the transfer of resources.
Several studies in the literature have sought to evaluate the reduced form of Medicaid on several aspects relevant to well-being, including the use of health care, health itself, and exposure to risks. However, there have been few attempts that have sought to estimate the effect on well-being directly, disregarding a relevant value associated with the program. In this way, the assessment of how much beneficiaries would be willing to pay for services and the measurement of possible monetary transfers to beneficiaries end up being neglected.
Faced with a non-functional market, the empirical analysis of well-being resulting from public health insurance for low-income adults can become challenging since we do not observe contract prices. This limitation prevents welfare analysis based on estimates of prior willingness to pay, derived from contractual choices, as is common in private health insurance markets, requiring the use of other empirical strategies.
Medicaid is the largest health insurance program for socioeconomically vulnerable populations in the United States, operating in partnership with all states in the country . In terms of public spending, resources allocated to the program exceeded US$550 billion in 2015. According to TANF ( Temporary Assistance for Needy Families, 2016), Medicaid has a significantly larger budget than several other programs in the United States, such as the food stamp program – SNAP ($70 billion), the Low Income Tax Credit – EITC ($70 billion), the Income Supplemental Security – SSI (US$60 billion) and financial assistance in cash (US$30 billion).
Specifically in relation to Oregon, the state's Medicaid encompassed low-income uninsured adults, that is, all individuals below 100% of the federal poverty line between the ages of 19 and 64 who were not already categorically eligible for the program. The expansion offered comprehensive medical benefits with no cost-sharing for the beneficiary and no or low monthly payments.
The Oregon Health Insurance Experiment was conducted in early 2008, followed by the expansion of Medicaid in the state. This expansion included the opening of a waiting list, with 30,000 of the 75,000 people on the waiting list who requested to join the program being randomly selected, allowing the identification of a group of treated and controls, both compliers to the program. Both groups presented balance between their characteristics. The selected beneficiaries now have access to coverage of health services provided by Medicaid with the payment of premiums or shared costs with zero or insignificant prices.
This experiment provided estimates from a randomized trial of the effects of Medicaid on low-income and uninsured adults in the state, providing a variety of information potentially relevant to measuring well-being (Finkelstein, Hendren, and Luttmer, 2019). .
The absence of consumer research in the Oregon context was addressed using proxies , such as the difference between average consumption for a low-income uninsured population and self-reported medical expenditures by study participants, with consideration of a minimum consumption threshold . Additional analyzes considered consumption data from a low-income sample from the Consumer Expenditure Survey.
In the first two years, key results revealed that Medicaid increased overall utilization of health services, including outpatient care, preventative care, prescription medications, hospital admissions, and emergency room visits. Furthermore, it was observed that the program contributed to improving self-reported health and reducing depression. However, there was no statistically significant impact on mortality or measures of physical health. Additionally, Medicaid reduced the risk of significant medical expenses for beneficiaries but had no economically or statistically significant impact on employment, earnings, or private health insurance coverage.
The study employed two main analytical approaches to estimate beneficiary willingness to pay for Medicaid . Both approaches were applied to Medicaid in the Oregon Health Insurance Experiment, using direct data from study participants to measure out-of-pocket medical spending, health care utilization, and overall health. The study's random lottery selection provided the possibility of estimating the causal effects of Medicaid on various health measures. To this end, an individual utility function model was initially established so that it was increasing (positively affected) according to (i) consumption of non-medical goods and services and (ii) health, where health is affected by the consumption of healthcare.
The first approach, called full information, required a detailed specification of a normative utility function and estimates of the causal effects of Medicaid on the distribution of all elements of that function, necessitating the observation of all arguments of such a function with and without insurance. This approach did not require accurate modeling of the budget set created by the program, allowing for the incorporation of frictions such as behavioral biases or information frictions. However, the high demand for information has made a comprehensive specification of Medicaid on all elements of the utility function necessary for accurate measurement.
The second approach, called optimization, reduces the implementation requirements considered in the full information approach by making two new assumptions: Medicaid affects individuals exclusively through its impact on their budget constraint, parameterizing this factor, and assuming that individuals had the ability and information to make decisions to optimize your behavior. By specifying the marginal utility function with respect to a single argument, it was possible to evaluate the marginal impacts of the program on other potential arguments of the utility function. For inferences about non-marginal changes in the budget pool, such as covering an uninsured individual with Medicaid , an additional statistical assumption was used to interpolate between local estimates of the marginal impact of program generosity.
As parameters for these models, parameters relating to hospital expenditure on medical costs (drug prescription and hospital services), own medical expenses and the price of own medical expenses were estimated. The primary health measure was a conversion of self-rated health into quality-adjusted life years (QALYs), based on existing estimates of QALYs associated with different levels of self-rated health.
Thus, two terms were identified: (i) a resource transfer term perceived by the beneficiary received by him/herself and (ii) a “pure insurance” term measuring the benefit of a neutral resource reallocation in relation to the restriction budget considering different states of the health parameter. All estimates of the impact of Medicaid were local average treatment effects (LATE) for program enrollees who were selected by lottery.
The results of the utility function estimates revealed that the gross cost of Medicaid in Oregon to beneficiaries was $3,600 per year. In practical terms, the effect of the program on the net cost was U$1,448, resulting from the sum of the average increase in medical expenses (U$879) with the reduction in own medical expenses (U$569). These amounts revealed that the out-of-pocket price was 0.21, and that approximately 60% of government Medicaid represents a transfer to third parties. Using the optimization approach through the linear approximation, a transfer term of U$661 was estimated, varying between the lower limits of U$569 and the upper limit of U$752 when the linear approximation was not considered.
Medicaid beneficiaries would be indifferent between the benefits of the program and the consumption of US$1,675 in other goods and services, establishing their willingness to pay for the benefits. Additionally, when decomposing this amount between the value operated in consumption and the value associated with health gains, it was identified that only 80% comes from the impact on consumption (U$ 1,381 in relation to the U$ 294 associated with health). Depending on the value of the health-related component in Medicaid , the transfer component represents between one-third ($569) and one-half ($863) of its value under the full information approach.
The values of these estimates varied little when considering different consumption measures. Estimates of the sensitivity of results considering different parameter measures revealed that beneficiaries' willingness to pay for health services varied between $0.5 and $1.2 per dollar of the net cost of these services.
The analysis of the well-being of non-traded goods is fundamental for measuring the effectiveness of a public policy. Such analysis involves application difficulties, meaning that the benefits of Medicaid for its beneficiaries have often been ignored in academic literature. In this article, we proposed to evaluate through parameterized models such as the expansion of Medicaid in the state of Oregon for low-income and uninsured adults.
The study revealed that Medicaid is best understood as having two distinct parts: a subsidized health insurance product for low-income individuals and a transfer to external parties that would otherwise subsidize medical care for the low-income uninsured. It was identified that 40 cents of every dollar of government spending on Medicaid represented coverage of costs that these beneficiaries would incur if they were uninsured, and that the remainder (60 cents) represented a transfer to these external parties. This evidence highlights the importance of future work that studies the immediate and final economic impact of these transfers.
References
FINKELSTEIN, A.; HENDREN, N.; LUTTMER, EFP The Value of Medicaid: Interpreting Results from the Oregon Health Insurance Experiment. Journal of Political Economy , vol. 127, no. 6, p. 2836–2874, 2019.
US DEPARTMENT OF HEALTH AND HUMAN SERVICES. FY 2015 Federal TANF & state MOE financial data , 2016
Check the Institution's registration in the e-MEC System here
© 2025 All Rights Reserved
