Subscribe via RSS Feed

Determinants of State Level Support for the ACA

[ 10 ] July 1, 2012 |

Now that the constitutional questions regarding the ACA have been surprisingly determined, attention is shifting to the political implications that this has for November.  I’ve built a little (emphasis on little) dataset to casually explore this issue.  In the near future I hope to both examine this on a case-by-case basis as well as add a few additional variables that have come to mind, data availability permitting.

The two key variables in the dataset are state level support for the ACA and the percentage uninsured in the state.  My source for the former is from a paper written by Richard Gonzales, a doctoral candidate at Harvard’s Department of Health Care Policy, which was discussed here at The Incidental Economist.  Gonzales estimates state level support from national Gallup data over a six month period (September 2009 to March 2010).  As this is an estimate, it does introduce an additional layer of uncertainty into the model, but it’s the best data I could find.  The point estimates and error bands appear sound in terms of face validity; the highest levels of support are found in New York, Hawaii, and Vermont, with the lowest in Oklahoma and Wyoming.  The public opinion data are old, but Monkey Cage suggests (and we all pretty much agree on) public opinion on the ACA has been relatively stable.  Percentage uninsured by state level is courtesy of Gallup, which are available here.

The first pass at the data is to run a simple bivariate correlation.  Gallup give us uninsured rates for 2008, 09, and 10; for simplicity I take the average of the three.  The mean is 16.2%, the range 4.6% to 26.6%.  The (state level, N=50) mean of support for the ACA is 48%, range 32% to 63%.  Support for the ACA and percentage uninsured are correlated at -.50.  Meaning, as uninsured goes up by state, support for the ACA declines.  The regression estimate is -0.93 (essentially for every one point increase in uninsured, there is a one point decrease in support).

 

This is obviously counter intuitive, and not the full story.  Hence, I’ve added the change in uninsured from 2008 to 2010 (percentage point change, thanks to commenter Fake Irishman for the observation), the PVI of the state (Cook Partisan Voting Index, a measure of how partisan a jurisdiction is and in which direction), and the percentage of the vote received by Obama in the 2008 election.  Shockingly, the latter two enjoy a very close relationship (correlated at .98) as the PVI is in part calculated from the 2008 election results, so in running the multivariate regression models, I ran one with PVI and one with percentage vote for Obama.  Models were estimated with both the average uninsured from 2008-10 as well as each specific year’s data, and it makes no difference.  The first model below includes PVI, the second the percentage vote for Obama in 2008.

The dependent variable is the percentage by state in support of the ACA.  PVI ranges from -20 to +13 with a mean of -2.5 (I’ve normed Republican PVIs as negative not because, well, I’m not a Republican, but rather it’s more intuitive: I’m assuming Republican states are going to be less supportive of the ACA than Democratic states.  It’s sharp insights like this that a Ph.D. in political science equips me to make).  Chg is the percentage point change in uninsured from 2008-10, mean 1.9%, range -2.3% to +4.9%.

Here we can see the original bivariate results have been flipped.  First, you’ll note only one of the three variables is significant.  I make an argument I’ve made countless times before: it doesn’t matter.  I’m dealing with the universe of data in these models, so slavishly bowing to the cult of significance is not relevant, but for those who care the standard errors and t-tests are included.  Percentage of uninsured, and the percent change in uninsured both have positive relationships with support for the ACA.  Put another way, as the percentage of uninsured increases, so too does support for the ACA.  Likewise, as the percentage of people uninsured grows, so does support for the ACA . . . while controlling for the partisanship of the state in question.  The partisan inclination of the state is the single strongest predictor in this model, which is informative.

This model replaces PVI with state level vote for Obama in 2008 (mean 50.52%, range 32.54% to 71.85%).  This model is consistent with the PVI model, so doesn’t require additional discussion.

These are not groundbreaking findings.  With the usual caveats about the ecological fallacy, support for the ACA is mediated through a pre-existing political prism: Democrats are more likely to support it, and Republicans less likely.  The lesson for the Obama administration should be, post Supreme Court ruling, seize the initiative to frame the issue among independents, but simultaneously focus on the base.  Because of the aforementioned case, the administration has been unable to frame the issue, and while running out of time, now is the time.  The second lesson I take away from this is that, as much as Dave Noon and I mocked the ignorant on facebook using this little quiz, outside of independents who think health care is a salient issue (I’m betting that this is a small percentage), the facts of the issue are not going to sway a significant percentage of the electorate.

As we know about the Republicans, this isn’t about the facts or about good public policy.  It’s politics, about winning and losing.  This is how we have the hilarious situation of Mitt Romney running against what is, essentially, his own policy.  Not ironically, Massachusetts anchors the low end of the range on percentage uninsured at 4.6%.  That’s a talking point worth exploiting.

I do want to add a few more variables to this little dataset as well as look closer at the states in question, especially swing states.  But right now,there’s the small matter of Spain v Italy to attend to.

Share with Sociable

Comments (10)

Trackback URL | Comments RSS Feed

  1. beckya57 says:

    I got 10/10 right on your quiz, thereby proving once again that I’m a complete nerd. Your model makes sense to me, though the stats were a bit much to wade through (and I have a PhD in social sciences too!). What this illustrates for the millionth time is the what’s the matter with Kansas problem: the uninsured in red states have absolutely no clue that the law could help them quite a bit, presumably because all of their news comes from Rush and Fox, not to mention their tribal identities being with the other side. This does give us some cause to hope: if the law, having survived the SC, can also survive the upcoming election, more people will begin to discover that the law actually will help them, that their Medicare isn’t being cut, and no one’s sending Granny to the death camps. This, of course, is precisely what the right wing is so afraid of, which is why they’ve been fighting it so hard.

    • chris says:

      I got 8/10 — I thought there were too many exceptions to the mandate to qualify as “nearly all”, which I guess is arguable, but I think it was a trick question. The explanation says that “some exceptions will be granted” but doesn’t estimate how many people will fall within the exemptions or how it still qualifies as “nearly all”.

      Of course, if you count Medicare and veterans as exceptions (they don’t have to “obtain” insurance because they already have it, and are at no risk of the penalty), let alone people covered under employer plans that fit the standard, then there’s no way to defend the “nearly all” claim.

      Also, I answered “don’t know” on the Medicare cuts question because I didn’t know. But I’m not really surprised.

  2. It might be interesting to break down the public opinion data by income and see what the trends are. Post Red State, Blue State, Rich State, Poor State it seems like that’s a common place to start breaking down state-level findings.

    And getting data on this might be tough, but it’d also be interesting to try and get a measure for relative percentages of uninsured in a smaller geographical area, and see if that makes a difference. (You’d maybe be able to find Congressional district numbers on that, maybe. Census districts would be pretty good but probably hard to come by). Similar things to throw in there along these lines that might be interesting would be proxies for how people interact with the medical system: number of hospitals within 100 miles, number of insurers in the area, number of people on medicare/medicaid, etc. Just throwing stuff out there.

    • Fake Irishman says:

      I suspect what you’d see is that poorer people tend to be somewhat in favor of reform while richer people are more solidly set against it. I seem to recall Scott Althaus (among others) has some nice articles discussing how lower information voters (who tend to be poorer) tend to be more in favor of redistribution that their high information counterparts (who tend to be richer), but the overall effect is muted because lower average knowledge translates into more diffuse opinion (fewer people with strong views.

      Anyhow, nice post Dr. B., Some interesting food for thought.

  3. Yo Dawg says:

    Yo dawg,

    You know you can change the name of variables and the axis names so it is easier to read? Come on!

  4. Incontinentia Buttocks, FILLED TO THE BRIM WITH "ART" AND "THEATER" COLLEGE STUDENTS AND HIP-HOP THUGS says:

    Your quiz is incorrect: we found out this week that, under tha ACA, people who don’t carry insurance won’t get fined…they’ll get taxed! ;-)

  5. Fake Irishman says:

    One quick question for the Professor: When you say “percentage change uninsured” do you mean “percentage” or “percentage point”? For example, let’s say a state had 10 percent of its residents uninsured, and between 2008 and 2010, that jumped to 15 percent. Would the change in your variables be regarded as a 50 percent increase (50) or a 5 percentage point increase (5). That little thing makes a big difference for the ramifications of how large an effect these coefficients produce. Under the percentage point scenario, a state like Texas (20 to 27 percent uninsured I believe) would see a change of about 2.8 points in favor of the ACA, (7 point change X .4 coefficient) a noticeable but modest bump. In contrast, under the percentage scenario, Texas would see a shift of about 14 points in favor of the ACA (7/20) X 100 = 35 percent increase in uninsured 35* .4 coefficient= 14, which even Rick Perry might notice. (Granted, he can’t even count to three, but I’m sure his pollster can….)

    Thanks!

    • dave brockington says:

      Good observation. I did the write up in a bit of a hurry this morning prior to the Spain v Italy match. Percentage point, not percentage of. I’ll edit in a comment to that effect.

  6. [...] the Supreme Court decision are heavily mediated by pre-existing partisanship, as highlighted by my post a couple weeks ago on state level support for the ACA (which I’m going to follow up on soon as I’ve added [...]

  7. [...] (typeof(addthis_share) == "undefined"){ addthis_share = [];}Three months ago I mashed together a little data set in order to examine variance in state level public support for [...]

Leave a Reply

You must be logged in to post a comment.