Let’s put some recent ideas together.
Last month, we saw that drug overdose rates were rising particularly quickly for people in states that fully implemented the Affordable Care Act and accepted the Medicaid Expansion.
Even since 2013, there was a faster increase in overdoses in the Medicaid expansion states.
Moreover, even if you controlled for some relevant economic and demographic characteristics of individual counties, counties in states that expanded Medicaid still had larger increases in overdose rates.
This pattern didn’t appear to be true of the other “deaths of despair“- alcohol-related deaths and suicides weren’t rising faster in ACA states, despite their increases for middle aged whites nationwide.
Nor was obesity– a corollary of general ill health and contributor to chronic pain- rising faster in ACA states.
Instead, insurance itself appeared to be closely connected with where overdose rates were rising.
Earlier this week, however, we looked at another possible contributor to rising mortality (or perhaps simply an indicator of general economic and social stability in the community)- reduced marriage rates among middle aged women:
Perhaps because they are more politically liberal, this is one characteristic in which the Medicaid expansion states differ dramatically from the Holdout states. Medicaid expansion states have significantly lower marriage rates, particularly for young white women:
For both Medicaid and non-Medicaid expansion states, the increase in drug-related deaths has also been concentrated among white women- if we reference to the rate within each age group in 2000, we can see a rapid increase for both types of states, with another jump since 2014 (ACA implementation) in Medicaid states (y axis shows percent of year 2000 value)
This same increase has not been duplicated for black women, for whom overdose rates have been generally stable since 2000 (y axis shows percent of year 2000 value):
You can summarize these graphs by looking at the average change for black and white women in overdose rates from 2005 to 2015 for holdouts and Medicaid Expansion states:
A reasonable question, therefore, is whether marriage is a “missing mediator” of the Obamacare effect– maybe we’re seeing not the direct effect of health care policy on overdose, but in part an effect of changing family and household structure, which is accelerated in more liberal states? Or, for that matter, since Obamacare creates larger disincentives to marry for people eligible for the exchanges or Medicaid (since marriage would involve losing a subsidy or Medicaid eligibility), could the way the ACA boosted overdose rates be partly by accelerating decreases in marriage rates? Contrariwise, if you don’t believe marriage drives anything directly, but you do think that political or culturally liberal tendencies are important, then controlling for levels and changes in marriage should help you isolate the true policy impact of the Affordable Care Act.
Looking cross-sectionally across all years for all states, an effect of marriage on overdose is at least plausible. Marriage rates within age groups, by state, are more closely associated with overdose rates within those age groups than they were with all-cause mortality:
Looking at changes over time within states and age groups appears less convincing, however. For white women there is a slight negative association between change from 2005 in marriage rates and change in overdose rates, particularly among young white women in the ACA states.
For black women, there doesn’t appear to be any such association at all (although because the data tends to be censored for small values, we don’t see all the age groups for all the years):
It appears, then, that marriage continues to be an extremely strong predictor of the overall health of a community (spilling over into such indicators as overdose rates), for both black and white women, but that short-term changes in marriage rates do not have a measurable association with overdose rates.
In fact, in an OLS regression of state-level overdose rates among white women in 2015 on 2005 baseline characteristics, only the 2005 overdose rates and Medicaid expansion status were significant predictors of where overdose rates were higher in 2015.
|OLS Regression||White Female Overdose Rate 2015||White Female Overdose Rate 2015|
|White Female Overdose Rate (same state and age group) 2005||0.712***||0.798***|
|Percent with post-secondary education, 2005||-0.0514|
|Percent married, 2005||0.707|
|Percent with post-secondary education, 2015||-0.342|
|Percent married, 2015||-0.408|
|Did the State Expand Medicaid?||7.753***||5.029**|
|Average earned income, 2005 ($1000s)||0.612|
|Average earned income, 2015 ($1000s)||-0.560|
Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
It still looks to me like Obamacare made a difference.
Data files and Stata do file for additional analyses in this post here. (Warning, the individual-level American Community Survey data is pretty yuge.) MarriageAndObamacare