There are 360º to the truth – and to the American health insurance debate. For every study that exhibits positive effects of extending public health insurance, two more will be published showing that insuring even the very poor has no significant impact on their health. This post will focus on a 2013 study that looked at the Oregon Health Insurance Experiment, its loopholes and the policy implications that have been taken from it.
In 2008, Oregon decided to expand their Medicaid program (a federal-and-state co-funded health insurance program for the poorest fractions of the American population). Since the state couldn’t accommodate everyone who was conceivably eligible for this expansion, they had a lottery. The people who won the lottery would then able to apply to Medicaid (provided they were eligible for it). The people who didn’t win, couldn’t apply.
This was a once-in-a-lifetime opportunity for economists to carry out a real-life controlled experiment, and so the state of Oregon called in a bunch of first-class health economists to study what happened before and in the 2 years after the lottery, both to the people who had won (the treatment group) and to those who hadn’t (the control group).
The study found no improvements in measured physical outcomes (cholesterol, blood pressure, blood sugar levels); a considerate increase in the use of health care services, and a significant reduction in financial strain (Medicaid virtually wiped out the chance that one would go bankrupt in the face of unexpected medical expenses). Although these findings can be both contradicted and supported by previous literature on the subject, the debate that followed the publication of the Oregon Experiment touched some methodological nuances that, interestingly enough, made the case both for and against public health insurance in the United States. We shall look at the two.
Let’s first take the argument against public health insurance. In rough numbers, only 60% of the winners actually submitted their application for Medicaid after having won, half of which were considered eligible. Then, goes the argument, it’s possible that, out of the winners, those 60% were simply more prudent than the 40% who didn’t apply. Since the authors of the study were interested in the effect of being granted Medicaid rather than the effect of winning the lottery, they compared a subset of non-winners (a mix of prudent and imprudent) to a subset of the winners (assumed to be prudent because they registered themselves, versus the subset of winners who didn’t). Therefore, it is possible that some of the effects the authors interpret as being ‘effects of being on Medicaid’ can be a mix of the effects of both prudence and of being on Medicaid, and not Medicaid itself. This would imply that the effects of being on Medicaid, as presented in the study, are actually magnified.
On the other side of the coin lie the advocates for public health insurance. And they believe the improvements produced by Medicaid on health outcomes are understated. Why? Because the study was underpowered (i.e., the sample wasn’t large enough to convey significant results) and also because it just isn’t reasonable to expect that a 2-year study will find much of an effect on the physical outcomes that could indeed be telling of the improvement of the population’s health.
I will say this: there is a lot of romance going around when it comes to randomized control trials: take one control group and a treatment one and you will find something that resembles enough the truth. But in reality, lots of things change at the same time and there are a lot of unobservable differences between groups, so we should lower our expectations. Even if an experiment can find the true answer to a given question about causality, that will necessarily be an extremely narrow question. Instead, we should be aiming at painting a pointillist picture of knowledge, where lots and lots of particular experiments add up to build useful, truly general conclusions.
 Access to Medicaid granted people all basic medical care services – hospitalizations and doctor’s visits, preventative care, prescription drugs – for a monthly fee of $0-$20, where the amount one paid depended on how close they were to the poverty line.
 Hanratty, 1996; Lurie et al., 1984 showing that insurance improved health in a cost-effective manner
 Levy and Meltzer, 2008; Kronick, 2009 suggesting that health insurance does not have much of an impact on health.
 Empirical work primarily aims at tracing the causal effects of one factor on another. For that, it needed to make sure that all other factors that could influence the outcome of joining Medicaid were equally distributed through winners and non-winners of the lottery. That would guarantee that the final results wouldn’t be biased.
General Reference: Baicker et al, 2013, The Oregon Experiment – Effects of Medicaid on Clinical Outcomes.