Obamacare didn’t help

Vox has an article arguing that repealing or limiting Obamacare would worsen the opioid crisis, since you would be taking away insurance (mostly Medicaid) from people who desperately need it to handle their addiction.

Look, it may be that whatever health care bill the Republican Congress  passes  will make the opioid epidemic even worse, but it is highly unlikely that Obamacare did anything to help, and it probably dug us deeper in the hole. This is because, as Sam Quinones’s recounts in his excellent investigative history of the opioid crisis, Dreamland, Medicaid is one of the main mechanisms by which addicts became hooked, and public insurance is the main funder of prescription opiates. Medicaid paid for the Oxycontin, and the desire for a Medicaid card was a big driver of the increase in SSI applicants during the decade ending in 2008, when the opioid crisis first took off:


Given that we haven’t addressed the two fundamental issues driving the opioid crisis- the ease with which people are prescribed addictive narcotics and the ease and low cost with which Mexican black tar heroin is entering the country and being distributed to addicts- I find it highly unlikely that the expansion of Medicaid in 2014 and 2015 did anything to help.

We can also look at this question more directly. Here is the change in drug overdose death rates from 2013 to 2015 from the CDC, comparing the 26 states and DC who had expanded Medicaid by 2015 (most expanded on January 1, 2014) to the 22 states that did not expand Medicaid by the end of 2015 and to the two states (Indiana and Alaska) who expanded Medicaid during 2015.

Change in Deathrate Pop Weighted


Obviously this is hardly an open-and-shut case- it’s likely that the states which expanded Medicaid already had a more quickly worsening problem than those who didn’t. But again, we’re talking about an additional increase of over 1.5 per 100,000 in the death rate from drug overdoses in states that expanded Medicaid versus the ones that didn’t. This translates to about 2600 additional deaths in 2015.


Update: I changed the graphs and the calculation above to weight by the population of states in 2015 and by the 2013 death rates respectively. There’s an argument against weighting here (you’re discussing a state-level policy, so in some ways weighting states equally makes sense), but given that I’m interested in how this affected the country as a whole and given that I didn’t want to feel like I was choosing my assumptions to make my case (using population weights reduced the estimated effect from 2.5 per 100,000 to 1.5 per 100,000), it seems like the better choice. Here are the unweighted graphs:



And here are the state-level changes. Only one state (Maine) that did not expand Medicaid had an increase in drug overdose death rates of over 5 per 100,000 in the two years from 2013 to 2015. Ten states that expanded Medicaid had a change of 5 per 100,000 or larger:


Gabriel Rossman  developed a graph showing the full time course by using mortality data from CDC Wonder:, suggesting that there may have been a divergence in 2010 that was more important than ACA implementation after 2013 (when there was only an apparently small “kink” upward in the expanding states.)


And when I reproduced this without weighting it looks very similar, although with a more dramatic divergence in 2009-2010:


Were Medicaid expanders in states which were more likely to be hardest hit by the financial crisis? Perhaps, though this wouldn’t explain to me why states like Connecticut and New Hampshire and Massachusetts have had among the most rapid increases in opioid deaths in recent years:


Were the conservative governors who came into office in 2010 who turned down the Medicaid expansion able to slow down Medicaid or other public insurance enrollments even before the Affordable Care Act came into full effect, with some slowing of the opioid crisis as a result? Or, to put the shoe on the other foot, were liberal state governments that accepted the expansion also deliberately speeding public insurance enrollment in the years leading up to full ACA rollout, with the knock-on effect of even faster increases in overdoses?  This seems pretty plausible to me.

Second Update: 

Apparently Gabriel and I contributed to a showdown between German Lopez and Dylan Matthews of Vox and Senator Mike Lee’s communication director last night. That’s funny.

What’s not funny is that yesterday, after I wrote this post, I ran into the uncle of two former students. I asked how they were doing. When I taught them in high school, they were both nice, polite, reasonably bright underachievers who smoked too much pot. I figured they’d be working or in community college, muddling along like a lot of young guys.

Their uncle told me yesterday that they both had near-fatal overdoses in the last year. Both had started with prescription pills and moved on to heroin, with the predicted utter destruction of the rest of their lives. One seems to have moved back from the brink and is starting to get clean. The other is still in free fall.

Here’s a still image from a Slate animation of the age of overdose victims:


The average age at death of an opiate overdose is between 20 and 25. 

Now, do I think that the Medicaid expansion was itself the biggest driver of why ACA expansion states had larger growth in overdoses than non-ACA states? No.

I think that a full-court press to increase insurance rates across the society and particularly for the young had as an additional effect that a lot of young people decided to use that insurance for something that appealed to them: drugs. The fire had already been lit in the 2000s, and as suddenly everyone had a way to get drugs cheap, the fire could spread faster. This didn’t need to be just about Medicaid expansion.  The level of enrollment in Medicaid during 2014 ACA rollout is slightly associated with state-level changes in OD rates, but not enough to explain the divergence between ACA and non-ACA states:

Nor did the growth in SSI receipt diverge between expansion and non-expansion states:

Weighted SSI Graph

But the rate of uninsured, particularly among the young, did diverge, and contra Dylan Matthews and German Lopez, this didn’t need to wait until full ACA rollout to take effect. Thanks to the provision of the Affordable Care Act that required parents’ employer-based health insurance to cover children through age 26, there were significant changes in the percentage of young adults with health insurance just from 2009 to 2011:

Change in Uninsured Rates

If you just look at pre-ACA variables, it’s pretty hard to predict which states will take off in overdose rates. Even taking the baseline 2010 overdose rate, the percentage of the population receiving SSI in 2010, and the percentage of Medicaid recipients who were white together doesn’t do all that much to explain the states that really exploded in drug use since 2010:


The best single predictor I’ve found for the change in overdose rates by state from 2010 to 2015 is the percent of white uninsured in the state in 2015, with lower percentages uninsured having much larger increases in overdose rates.


If you put this variable- that the Affordable Care Act did affect, together with the baseline variables (2010 OD Rates, SSI receipt, and percent of Medicaid recipients white)- suddenly those outliers don’t look like outliers any more.prediction4

That’s what I think caused at least part of the divergence between blue (ACA expansion) and red (no ACA expansion) states in the overdose rates. Not the Medicaid expansion in of itself, but the dominant political party in these states taking reducing the percentage of uninsured as a principal goal, without realizing that a new population of younger insured carried huge new risks of abuse.


20 thoughts on “Obamacare didn’t help

  1. The zeal with which the anti addiction special interests attacked prescription opioids had a bad outcome. Namely switching from legal and (lightly) regulated substances to black market, unregulated substances.

    One of the drivers of the policy argument was overdoses from prescription opioids. However, it is my opinion that most of the deaths involved multiple substances. Alcohol for one. And benzodiazepines. And then there are suicides. My data is purely anecdotal, but we were at a point with ‘pill mills’ where some opioids were quasi deregulated. And as someone crazy enough to read the footnotes, there seemed to be room for subjectivity in terms of listing a cause of death in situations with multiple drugs involved. The CDC or DEA study (I can’t remember) stated that they associated each death with a single cause. My guess was that up to half of the overdoses could have been multiple drug or suicide or both. I don’t understand the actual mechanics of overdosing, but they seem to have a hard time actually killing someone in capital punishment situations.

    As far as the drugs costing thousands of dollars per month, that is a different problem, and could have been more directly addressed.

    We seem to be able to accept lifestyle related deaths from alcohol and tobacco. The United States seems to regularly get into situations where the best is enemy of the good. Pill mills weren’t the best. But they avoid the worst problems of criminalization. There are records of purchases. They take place indoors. The substances are correctly labeled. They are safe to use as prescribed. &c. As far as suicides, an OD seems better than a gunshot to the head, and is at least no worse.


  2. Have you tried controlling for racial makeup? It’s pretty widely reported that black people were hit much less hard by the opioid epidemic (because doctors are more reluctant to prescribe painkillers to black people), and the Medicaid expansion rejection states include a lot of states with a below-average black population. According to the demographic information at KFF.org, 12% of the US population is black, and after Texas (which has a 12% black population) the next five most populous non-expansion states had above average black population (FL, GA, NC, VA, TN).

    Not being a very statsy guy I looked at Maryland–the Medicaid expansion state with the highest black population (except Louisiana, which didn’t expand Medicaid till 2016, so Medicaid expansion was outside the data set)–and it’s hard for me to interpret. The death rate climbed pretty drastically from 2010 to 2013 before really taking off in 2014 and 2015–but 2010 had the lowest death rate in this whole data set, so the 2012-3 numbers aren’t that much higher than the pre-2010 norm.

    Liked by 1 person

    1. Correction: “the Medicaid expansion rejection states include a lot of states with an *above*-average black population.” Specifically the more populous ones that I’d expect to drive the unweighted average.


    1. Dishonest in what sense? As I said, you can’t predict change from 2010 on from 2010 OD rates or other obvious baseline characteristics alone. Obviously, the “red” states have been increasing, too, so this isn’t the main cause. Quinones’s book ends largely before the period I’m talking about, and tells about a devastating crisis well before it reached current levels. I am saying that the Affordable Care Act definitely didn’t help (in contrast to what Vox claimed), and likely worsened an already fast-deteriorating situation.


      1. This only works if you take into account the fact that the populations differed to begin with. Comparing, say, white people from West Virginia and white people from Virginia is going to produce differences completely independent of Medicaid expansion. Maine appears to have been the only New England state to not expand Medicaid and it has similar OD rates to the neighboring states that did.

        If you really want to test this, you need to compare states with similar demographics – not just between the races (black, white, hispanic, etc.) but also within them. What seems to be the case is that certain groups, like the Scots Irish, seem far more at risk of developing problems with opioids than others.

        What that means is that you’re not wrong, exactly, but you don’t seem willing to stare your data in the face and say things explicitly. For some people, Medicaid expansion may be a bad idea, or may require safeguards to stop them from killing themselves off. But that doesn’t mean that it’s a bad idea in general, as you’re heavily implying. It might be good for the more disciplined middle and working classes.


      2. Some good points. It’s clearly regional (Northeastern) but it doesn’t any longer seem to be about poverty or education in any real sense- NH, CT, MA, and MD are among the richest and most educated states in the country. I wouldn’t have put them in the same bucket as WV on almost any metric a few years ago.


  3. My personal experience with homeless/near homeless and long-term unemployed folks on Medicaid/SSI/whatever is they actually receive a great deal of free or nearly free medical care. I don’t know how good it is, but they sure get a lot of it.

    Liked by 1 person

  4. I want to site this data in a paper I am writing, however some of your charts do not lists where they are from, and there is not author data nor a date of upload. Can I have the name of the author, their qualifications, as well as the sources of the tables that do not have a source listed?


      1. Also, I include the address of the CDC Wonder mortality data and the KFF description of which states were in the Medicaid expansion by date in the graphs above.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s