Does Health Insurance Improve Health?

The question sounds simpler than it is

Ask most people whether health insurance improves health and they will treat the answer as obvious. Of course it does. You get sick, you have coverage, you get care, you get better. But the more carefully you work through the economics, the more that clean story falls apart. Health insurance does something real and valuable, but the thing it does is not quite the thing most people assume. Its core function is financial: it redistributes the cost of health care across a pool of insured people so that any one person is protected against a catastrophic loss. Whether it actually makes the insured population healthier is a separate question, and the evidence gives a more complicated answer than the intuition wants.

I want to hold both of those ideas at once. Insurance can improve health outcomes by pooling risk and letting people who fall into an adverse state draw on the resources of everyone who didn't. But the act of having insurance, and even the act of using it, does not by itself cause health to improve. The reasons come down to efficiency and effectiveness — how much care gets used, by whom, and whether that additional care actually changes anything.

What insurance is actually for

Start with what insurance does well, because that is the part that is not in dispute. When you buy health insurance, you are buying protection against uncertainty, and the uncertainty runs in two directions. The first is whether an adverse event happens at all. You may know a great deal about your own risk profile — a family history of high cholesterol, high blood pressure, cancer, heart disease, or the lifestyle choices that raise your odds, like deciding to rock-climb as a passion instead of playing classical guitar in the safety of your own living room. That knowledge tells you something about probability. It tells you very little about the second uncertainty: what treatment will cost if the event does occur.

That second unknown is where insurance earns its keep. Many serious conditions turn out to be chronic, requiring long-term treatment, and the associated costs are large and recurring. Self-insuring against that kind of exposure — simply saving enough to cover whatever might happen — is difficult even for well-off households, and effectively impossible for anyone earning below the median or living within 400% of the Federal Poverty Line. By pooling risk, insurance lets people smooth their consumption across every possible state of the world. In the good state you pay a premium you barely notice; in the bad state you are shielded from a loss that would otherwise be ruinous. That is consumption smoothing, and it is the genuine, defensible core of what insurance provides.

Notice that none of this is a claim about health. It is a claim about money and risk. The value is real whether or not you ever get healthier, because the value is protection against financial catastrophe.

Having insurance is not the same as using it well

Once you accept that framing, the outcomes question gets sharper. Protection against an adverse event does not guarantee efficient use of the benefit, and inefficient use is where the link between insurance and health starts to weaken.

Consider someone on standard Medicare — Parts A, B, and D, with roughly 80% coverage and varying coinsurance rules for prescription drugs. Suppose this person has a history that puts them at risk for diabetes and wants to see a nutritionist to learn how to head it off through diet and exercise. They may well find that Medicare will not cover that visit. Facing the full out-of-pocket cost, they decide to skip it. That decision might have been exactly the thing that would have prevented the disease — but because the preventive service wasn't covered, the efficient outcome never happened. Later, once diabetes actually develops and gets diagnosed, Medicare covers all the routine and specialized care and the full drug regimen to manage it. The system pays generously for the expensive downstream disease and declines to pay for the cheap upstream prevention that could have avoided it.

That gap in covered preventive services doesn't just fail to help — it distorts behavior on the other side too. Because insurance lowers the price of covered services at the point of use, the insured person doesn't feel the full pain of payment. When something costs you far less than it costs to provide, you consume more of it than is optimal. More care gets used than the situation warrants, and that excess consumption raises total costs for the entire risk pool. In externality terms, subsidized care can function like a subsidy for a negative consumption externality: each person over-consuming imposes marginal damage on everyone else in the pool by driving up the shared bill.

More care is not the same as better health

If more consumption reliably produced better health, the over-consumption problem would at least buy us something. The evidence suggests it often doesn't.

The clearest illustration comes from research on regional variation in medical practice. Song and colleagues, in their work on regional variations in diagnostic practices, compared Medicare recipients across geographic lines to ask whether the regions that use more medical services actually produce healthier patients. Their design was clever: they watched what happened to people who moved from low-utilization regions to high-utilization ones, observing the same individuals before and after the move. What they found was that "Residents of higher-intensity regions generally had more office visits, underwent more diagnostic tests, had a higher number of diagnoses, and had higher risk scores than did residents in lower-intensity regions." In other words, moving into a high-spending region made you look sicker on paper — more diagnoses, higher risk scores — without any evidence that you were actually worse off before you arrived. As they put it, "With few exceptions… among the beneficiaries who moved, the number of diagnoses, the risk scores, and the number of diagnostic tests and imaging services were similar among persons within each quintile during the period before their move."¹

The takeaway is uncomfortable but hard to dodge: additional services and additional spending do not necessarily translate into improved health outcomes. A lot of what higher-intensity regions produce is more diagnosis and more testing, not more health. If you were looking for proof that pouring more covered care into a population automatically makes it healthier, this is precisely the evidence that undercuts you.

Cost-sharing is a blunt instrument

None of this is an argument against insurance, and it is important not to let it slide into one. Inducing people to seek care when they genuinely need it matters, and it matters most for the people who are actually sick. The problem is that the main tool we use to control over-consumption cannot tell the difference between necessary and unnecessary care.

The RAND Corporation's health insurance experiment on cost-sharing is the reference point here. It showed that raising the coinsurance rate — making people pay a larger share at the point of use — was effective at deterring them from using benefits beyond what they themselves judged necessary. And crucially, the people who cut back their utilization were, on average, no worse off for it. That is a real finding, and it led to a reasonable conclusion: coinsurance can be an effective and reasonably safe way to hold down insurance costs, because a lot of the care it discourages was care of little value.

But the average hides the people who matter most. Among those who were already sick or unhealthy, higher cost-sharing was associated with higher incidences of mortality and worsening health.² Coinsurance doesn't discriminate. It discourages the healthy person from an unnecessary visit and discourages the sick person from a necessary one with the same price signal. That is why I'd resist calling it a sophisticated instrument for cost control. It works in the aggregate precisely because most people are mostly healthy most of the time, but it does its damage on the tail of the distribution — the unhealthy population that insurance is supposed to protect in the first place. A tool that saves money on average by quietly harming the sickest is not one to reach for casually.

This is also where I'd part ways, at least partly, with the standard prescription that the optimal policy makes individuals bear a large share of costs within some affordable range and only fully insures them once costs become unaffordable. That logic is sound for routine care among the healthy. It fits much less well for populations — like people managing a chronic disease — whose predictable, affordable-looking care is exactly the care that, left unpurchased, snowballs into expensive complications later.

So, does it improve health?

Putting it together, the honest answer is: it depends on what you are asking insurance to do. Health insurance reliably provides coverage and protection against adverse events. It smooths consumption by guarding against the spending and lost earnings that come with getting sick. Those are genuine benefits, and they justify the institution on their own terms without any appeal to health outcomes at all.

What insurance does not provide is a clean causal link to improved health. It subsidizes clinical care and therefore increases how much care people consume — but healthy people consuming more than they need generate a negative consumption externality that raises costs for the whole pool, and the regional-variation evidence shows that the extra consumption often buys diagnosis and testing rather than health. Meanwhile, the blunt tools we use to rein in over-consumption end up deterring the sick along with the well.

I think getting this rationale right is more than an academic exercise, because it changes how you design the thing. If you believe insurance exists to make people healthier, you optimize for maximum utilization and treat any barrier to care as a failure. If you understand that insurance exists primarily for financial protection, and that the health payoff depends entirely on whether the right care reaches the right people, you start asking better questions: which services actually change outcomes, which populations are harmed by cost-sharing rather than merely nudged by it, and where a cheap covered preventive visit today avoids an expensive covered disease tomorrow. Insurance is a mechanism for managing risk. Whether it also improves health is something we have to engineer deliberately — it does not come free with the policy.