2016 County-Level Estimates of the Obamacare/Overdose Effect

The CDC has made county-level mortality and OD rates available for 2016, so we can replicate the finding that  we observed last year: that overdose rates have risen more since Obamacare was passed in counties in states that expanded Medicaid than in counties in states that did not. Counties in Medicaid expansion states have increased overdoses by 10 per 100,000, counties in non-Medicaid expansion states (Holdouts) have increased overdoses by 5 per 100,000.

You might be concerned that this is due to the counties in Medicaid expansion states being poorer, having a worse overdose problem previously, or otherwise not strictly comparable. However, looking at American Community Survey data about these counties, it appears that they are in most respects, fairly similar:

American Community Survey Characteristics of County (2010)
	Holdout Counties			Medicaid Expansion Counties
Percent Receiving SSI	424	4.768		525	5.174
County Population	424	231,557		525	332,473
Percent Employed	424	56.73		525	56.91
% non-Hispanic White	391	78.58		493	80.09
Median Income	424	$48, 590		525	$51,590
Overdose Rate	342	17.56		462	19.62

Controlling for baseline variables that influenced 2016 overdose rates,  such as 2010 overdose rate, population, median income, % White Non-Hispanic, and rate of SSI receipt, 2016 overdoses were still generally higher in counties with expanded Medicaid (blue dots below)  than counties without (red dots):

The difference in logs is roughly the difference in how much larger percent growth in overdoses there was in Medicaid expansion counties than you would expect from 2010 county characteristics.

Treated as a regression estimate, the coefficient (impact estimate) of being in a county that expanded Medicaid is between 5 and 7 per 100,000, slightly larger than the estimate for 2015 was.

(I discuss these methods and data sources in more detail in this earlier post.)

There are a number of arguments against treating this as a “true” impact estimate, some good, some bad. My own view is that it did represent a “real” effect, but that it was historically contingent on other factors at work from 2010 through 2016.

As the New York Times observed today, part of the rapid increase in overdoses in 2016 was a crossover of the opioid epidemic from rural white communities to black and urban communities. We’re all in this together, it would seem.