What Happens When Poor Pregnant Women Are Given Medicaid Coverage?

We’ll be putting out a new Freakonomics Radio episode later this week on the use of RCTs (randomized controlled trials) in healthcare delivery. It features the work of the MIT economist Amy Finkelstein and her colleagues at J-PAL, and it includes their analysis of what happened when Oregon expanded its Medicaid coverage.

If you want to get a head start on this topic, consider a new working paper (gated) called “Does Medicaid Coverage for Pregnant Women Affect Prenatal Health Behaviors?” The authors are Dhaval M. Dave, Robert Kaestner, and George L. Wehby. They didn’t use an RCT, but they did work hard to answer a tough and worthwhile question in the realm of healthcare delivery. Their conclusion, bolded in the abstract below, is quite counterintuitive (and surely, to some at least, depressing):

Despite plausible mechanisms, little research has evaluated potential changes in health behaviors as a result of the Medicaid expansions of the 1980s and 1990s for pregnant women. Accordingly, we provide the first national study of the effects of Medicaid on health behaviors for pregnant women. We exploit exogenous variation from the Medicaid income eligibility expansions for pregnant women and children during late-1980s through mid-1990s to examine effects on several prenatal health behaviors and health outcomes using U.S. vital statistics data. We find that increases in Medicaid eligibility were associated with increases in smoking and decreases in weight gain during pregnancy. Raising Medicaid eligibility by 12 percentage-points increased rates of any prenatal smoking and smoking more than five cigarettes daily by 0.7-0.8 percentage point. Medicaid expansions were associated with a reduction in pregnancy weight-gain by about 0.6%. These effects diminish at higher levels of eligibility, which is consistent with crowd-out from private to public insurance. Importantly, our evidence is consistent with ex-ante moral hazard although income effects are also at play. The worsening of health behaviors may partly explain why Medicaid expansions have not been associated with substantial improvement in infant health.

Just to be clear, weight gain is a good thing when you’re pregnant; and smoking 5 more cigarettes a day is not. Here’s how the authors attempt to explain their finding:

One possible explanation for this somewhat counterintuitive finding is that health insurance creates incentives to change health behaviors along with lowering the price of medical care. Insurance lowers the price of treating an illness, for example, an adverse medical outcome for either the mother or child, which may cause a reduction in maternal efforts to prevent the occurrence of such events (ex-ante moral hazard).

In addition, Medicaid coverage entails an income effect from the saved out-of-pocket expenditures and from spending on health insurance premiums (in the case of substitution of private for public insurance). This income effect may be used to purchase goods that improve infant health, but also goods that may harm infant health (e.g., cigarettes).

Incentives, incentives, incentives.

Oliver H

Seriously??? 0.7-0.8 percentage points??? And 0.6% for reduction in weight gain?

"Just to be clear, weight gain is a good thing when you’re pregnant; "

May I suggest you stick to topics you actually understand?

Yes, weight gain is a good thing when you're pregnant. But an 0.6% reduction in weight gain has pretty much no significance whatsoever.

Weight gain in pregnancy is about 27.6lbs. An 0.6% reduction in weight gain means gaining 0.166 lbs less.

Frankly, it's a joke to attribute that to anything. They could not possibly control for all confounders that affect female weight - in pregnancy to boot! - to such a degree that this was not within the confidence interval.

Frankly, this has nothing to do with science whatsoever. This is the cherrypicking of data to justify a moral hazard hypothesis on a level that can only be considered academic fraud. It is a complete and utter joke to claim changes of this "magnitude" to have any relevance.


Alex in Chicago

Umm, it doesn't matter if the negative affects aren't statistically significant, it means the Medicaid did not improve the health of the mother (and by extension the baby), so it was a waste of money. Which isn't unexpected if you read the Oregon studies.

Medicaid needs to demonstrate that it is unequivocally worthwhile in order to justify its very high costs to taxpayers. This is yet more evidence that it cannot live up to that standard.

Phil Persinger


It would be nice to see the Dave/Kaestner/Wehby paper so we can judge how pertinent it may be or how apposite our comments are. But it seems that D/K/W were playing w/ numbers, so the statistics cited do matter. If Oliver is correct, the numbers reported really mean that Medicaid-- and Medicaid alone-- did not budge the situation in the specific categories being investigated. That does not mean that the program is an overall failure.

Too many outside factors play into individuals' decisions concerning smoking and/or diet, just as many extra-medical factors lead to the decision-- especially early in the Oregon program-- of an individual to go to the ER rather than make an appointment to see a physician.

The D/K/W may not be worthless as a study on healthcare policy, but it's not worth much by itself taken out of larger socioeconomic and cultural contexts. The necessary narrowness of the paper-- insofar as we know what it's really about-- together w/ the importance Freakonomics has given it (perhaps for provocative purposes) are what have set Oliver's socks on fire.


Oliver H

Just adding to the other post - I won't give that trash enough importance to pay money to read the full study, but it would be nice to see the variance of the parameters at issue.

So let's look at some other data on prenatal smoking: In a study by the CDC, prenatal smoking rates were assessed over several years. http://www.cdc.gov/mmwr/preview/mmwrhtml/ss6206a1.htm?utm_source=rss&utm_medium=rss&utm_campaign=trends-in-smoking-before-during-and-after-pregnancy-pregnancy-risk-assessment-monitoring-system-united-states-40-sites-20002010#Tab1

Guess what: If we look at, say, 2006, the rate of smoking within the three months before birth was on average 22.0%. The confidence interval was 21.4–22.7%. Yowza. If we thus add the 0.7 percentage points from above to the mean, we are still within the confidence interval. Which begs the question: Aside from ignoring MEDICAL significance, did these amateurs actually check for STATISTICAL significance at all?

And of course, that's the confidence interval over the entire CDC study - the one for the individual states is of course significantly larger, as the sample size is smaller.

But let's look at changes, shall we? The mean over all states in 2000 was 23.0 (22.2–23.8), the result in 2009 was 25.2 (24.4–25.9) and in 2010, it was 23.2 (22.4–23.9). That is a change of a full two percentage points -sometimes within a single year -, much larger than the change in this so-called study in the article above.

So the scatter of the prenatal smoking rate is larger than the supposed effect of raising MediCare

Stuff like this is precisely why a lot of economists are considered little more than snake oil peddlers by other academic disciplines.

The only "incentives, incentives, incentives" here are incentives for economists to commit academic fraud to justify their doctrines. In other disciplines, conduct like this would lead to serious repercussions.



Well thats just a horrible use of statistical reasoning, and the fact that your first post loudly decried the lack of "statistical significance" without having actually even read the paper or having ANY idea what the statistical significance was would imply to a dispassionate reader that it is likely YOU that is defending preconceived ideology here.