While it may seem clear that health insurance can help you with health, there are few causal conclusions that are obvious to economists. For example, let’s say you compare the health of everyone who has health insurance and everyone who does not. It’s not surprising that people with health insurance find healthier, but the two groups also differ in many other ways. For example, given that many Americans have health insurance through their employers, it is possible that health insurance people will be hired and earn more on average. How can I solve the effects of health insurance from other possible confounding factors?
Or imagine comparing the health of people before and after they signed up for health insurance. This approach has several promises, but again, if you have health insurance, it is also related to gaining profits, higher income, and perhaps other ways to achieve a more calm life, then the task of separating the effects of health insurance from other confounding factors.
Alternatively, you can imagine a large group being randomly divided, with some of the groups receiving health insurance and some of them not present. You can then track two randomly selected groups over time and see what happens. This is essentially an approach used, for example, to test the safety and efficacy of new drugs. So social scientists keep an eye on the situation where this kind of random choice has been enrolled in health insurance, but perhaps not a policy, but a coincidence.
In their essays, “The impact of health insurance on mortality,” Helen Levy and Thomas C. Buchmuller We focused on some of these situations where access to health insurance was determined by a high degree of randomness (Annual Review of Public HealthApril 2025).
One of the clearest examples in Oregon happened in 2008. The state wanted to expand Medicaid eligibility, but there was no money to expand it for everyone. The results, as the author explains, “A 2008 Oregon Health Insurance Experiment studied roughly 75,000 low-income adults under the age of 65. Therefore, some received health insurance at random, while others did not.
Another truly randomized study considered the “IRS initiative in which in early 2017 information about Healthcare.gov sent letters to a randomly selected sample of 3.9 million households that were covered by the ACA.” [Affordable Care Act] Even if you fail to compensate the previous year, you will still be subject to individual delegation penalties. This study found that the text was significantly increased with a small range. “In this case, some people randomly received letters that added the group’s share to their health insurance, while others did not.
Yet another approach looked at people who were under the age of 65 and were not eligible for Medicare, or who were over 65 years old and admitted to a California hospital eligible for Medicare. The idea here is that just over and just over should be a very comparable group. After all, the only way they differed was to be born months apart. With this “discontinuity” approach (in this example, the discontinuity is 65 years old), whether the share of health insurance between groups is increased or less is randomly similar.
Other examples include Medicaid coverage, which is a joint federal state program, so the program was often introduced statewide in a way that shifted over time. This was the case in the 1960s when Medicaid was first enacted, and even in the 2010s when states were allowed to expand Medicaid coverage, but did so for years. Researchers can look at this data and see if, once a group of people qualifies for Medicaid, their patterns of health outcomes shift from the previous patterns, and the patterns of health outcomes for groups that were not eligible at that time. Here, the random ingredients are the staggered period when health insurance was introduced.
My theme here is that there are plausible ways for researchers to study the causal relationship between health insurance and health. Of course, not all of these studies cover the same age group or find the same results. But my guess is that many readers don’t care much about the way they do their research. It also concerns how the authors of this review summarise the overall results. Here I quote from the summary of their paper:
2008 Review Annual Review of Public Health We took into consideration the issue of whether health insurance would improve health. The answer was cautious as few studies provided compelling causal evidence. This question is revisited by focusing on a single outcome: mortality. The answer is more crucial as multiple high-quality studies that have been published since 2008 utilize new sources of quasi-experimental variation and new empirical approaches to assess old data. Research using a variety of data sources and research designs provides reliable evidence that health insurance coverage reduces mortality. The most prone effects that tend to be stronger for middle-aged adults and children are generally evident immediately after benefits of coverage and growth over time. The evidence now clearly supports the conclusion that health insurance improves health.
