[Crosspost] Governors who refuse to expand Medicaid are doing so for political reasons, despite the needs of their state’s citizenry.

The following post was originally published on the LSE's USAPP blog, which you can follow here. The post discusses a paper of mine written with Charles Barrilleaux and published in State Politics and Policy Quarterly. You can find the paper here and all the data and code here.


Medicaid began in 1966 under President Johnson’s Great Society initiative, which expanded the national government’s role in health, education, and social welfare policies. Importantly, though, it relied upon state government’s accepting national financing–a carrot that induced states to provide services they would ordinarily not agree to provide.  Even with a generous national government match (at least dollar-for-dollar, and some states get more generous matches), Medicaid has long been among the most expensive programs for states. Medicaid played a critical role in President Obama’s signature health care reform legislation, the Patient Protection and Affordable Care Act (PPACA), or “Obamacare.” The original version of the law required states to expand their Medicaid programs, covering a significant portion of the 50 million US citizens who were without health insurance. This requirement, though, came at little cost to the states. The national government would initially pay 100 percent of the cost of the expansion and this would only shrink to 90 percent over several years.

In 2012, however, the Supreme Court stuck down this requirement, noting that states must be granted the opportunity to opt-out. This gave states a choice: expand their Medicaid program and receive the generous benefits or refuse to expand their program and reject the benefits. These benefits are not trivial. For example, by simply accepting these benefits, Texas could have expected to prevent between 3000 and 1800 deaths.

Yet of March 2015, Texas, along with 20 other state governments, has decided not to expand their Medicaid programs, establishing a new politics of Medicaid policy in the United States.  But why would states decline this money? In recent research, we ask why some governors opposed the PPACA Medicaid expansion in spite of the seemingly generous benefits. The law’s designers seemed to assume that even the most fiscally conservative states would expand their programs in order to greatly expand health insurance coverage in their states if the national government covered almost all of the cost.  But we point out that many governors face a tension between political expedience and citizens’ needs. Politics and need pull many governors in opposite directions. We ask: In this tug-of war between politics and need, which one wins? Is it politics? Or is it need? To get at these questions, we conduct several statistical analyses and the answer seems clear: politics wins, hands down.

Partisan politics fuel Governors’ Medicaid decisions 

As part of answering the question, we look at what variables best predict opposition. It turns out that variables capturing the political landscape best predict opposition: partisan control of the lower house, Obama’s vote share in 2012, the governor’s partisanship, and the average ideology of the states’ citizens. Measures of need, such as life expectancy, percent of the population uninsured, and heart disease death rates, are not helpful in predicting opposition.

But how big are these effects? We estimate that in otherwise “Republican” states (i.e., GOP-controlled legislature, 38 percent view ACA favorably), we find that a Republican governor is 49 percentage points more likely to oppose the expansion a Democratic governors. On the other hand, we estimate the percent uninsured in the state to have a small positive effect on the probability of opposition. This estimate suggests that governors of more needy states are more likely to oppose the expansion. While there is a large amount of uncertainty in this estimate, the data strongly suggest that partisanship has a larger effect than need.

Credit: TexasImpact (Flickr, CC-BY-SA-2.0)

In sum, our results reveal that partisan politics, not citizen need, explain governors’ decisions whether or not to accept the Obamacare Medicaid expansions as offered under the Affordable Care Act.  While it comes as no surprise that Republican governors were more likely to oppose the expansion than Democratic governors, it is surprising that governors were much less responsive to citizen needs. Indeed, we find no evidence that need matters at all.

The Implications

Our results may also point to a new state level opposition to federal spending that seeks to induce state and local government policy choices.  The original Medicaid policy was designed to lead the states to do what they would not ordinarily do.  The program’s initial design promised to pay physicians and hospitals “customary and prevailing rates” and allowed states to establish generous eligibility standards.  The national government has been trying to establish ways to provide high eligibility for Medicaid while controlling spending, but has not often succeeded.  This move by a number of state governments not to expand the program even under the most generous of federal payment schemes shows stronger resolve to hold down costs and access to services among some conservative state governments.

The choice to expand Medicaid has important implications for citizens.  The United States has reduced the number of its citizens who have no health insurance with the establishment and implementation of the ACA. The state governments that did not agree to expand Medicaid are receiving less federal money than they would spend to expand Medicaid.

The national government’s generous benefits to Medicaid programs are offered to state governments who cannot afford to provide Medicaid benefits, prefer not to provide the benefits, or otherwise would normally opt out of the program, but accept the federal payments because they expand coverage with low state cost.  The costs of non-participation in Medicaid are enormous:  the state of Florida, for example, turned away $50 million in subsidies in the 2014 fiscal year, leaving tens of thousands of person who might have had health insurance uninsured.  In 2015, the national government halts payments to hospitals that provide a large amount of service to the uninsured, services that were once paid by national government transfers to not-for-profit and/or public hospitals, leading a number of those hospitals to suffer financial crisis.  The rejection reveals states’ willingness to forgo national government financial contributions even if it results in less health coverage for the poor.  The demands by eleemosynary (charitable) hospitals for continued support for the uninsured may lead some state governments that previously refused to expand Medicaid to find a way to accept federal government money by redefining the meaning of Medicaid expansion.  For example, the national government may allow states to provide benefits that mirror Medicaid but are given a different name, thus allowing conservative state governments not to accept Medicaid expansion and the national government to expand coverage and retain support for the not-for-profit hospitals.

A 2014 report from the Urban Institute indicates that that cost to states of expanding Medicaid would be about $31.6 billion by 2016, but by doing so they would receive $423.6 billion in the form of Medicaid shares paid by the national government, as well as $167.8 billion from hospital payments that will cease after 2015.  In addition, citizens of states that enroll in the Medicaid expansion will get better health care than citizens in non-Medicaid expanding states.  Governors who do not expand Medicaid are doing so because of political belief, and the evidence suggests that those beliefs are not consistent with public preferences or public need.

On March 4, 2015 the U.S. Supreme Court heard another case that may further divide the U.S. along the lines of health care haves and have-nots.  That case’s petitioners claim that only persons who reside in states that have established their own health exchanges may receive federal health care subsidies as are provided under the ACA.  If the Court rules in favor of the petitioners, it will likely result in the collapse of the ACA in states that do not establish exchanges (which includes all of the non-Medicaid adopters as well as others that opted to use the federal exchange) and will create a truly divided system of health benefits in the United States.  The system for the states that have adopted the ACA and whose citizens are eligible for subsidies will have a much stronger network of health care access, coverage, and providers that will states that either refused the Medicaid expansion or chose not to establish a health exchange.  We will know in June 2015, when the Court decides, which of those systems we can expect to exist in the near future.


Software for Conference Posters

This summer, I presented poster of my paper on product terms in logit models at PolMeth and APSA. I've never felt really comfortable with making posters, but lately I used Apple's Pages to make this latest poster and it worked great. Here's a quick summary of the software that I've tried and why I love or hate it. I also added a few links to resources I found helpful near the bottom.

  • Inkscape: I did my first poster in Inkscape. It worked fairly well, but it has a steep learning curve. I already use it for post-editing R graphics, so I already know how to use it. Inkscape also does not do a good job of laying out text. For example, I believe I had to wrap my sentences by hand. Finally, it lacks spell-check, which I need badly. You'll notice a couple of ugly typos in my first poster. However, it does offer a lot of control, which I love. Adobe Illustrator serves as a proprietary, but similar, tool. I've read some nice things about it, but I don't have access to it. It might deal with some of the difficulties while offering similar control.
  • LaTeX: Many of my friends use LaTeX, but I just don't like the look of a LaTeX poster. The best I've found is from Nathaniel Johnston, but it is still to busy for me. While I love LaTeX for writing papers, I prefer WYSIWYG for posters (and presentation slides). LaTeX makes inserting equations really easy and perfectly justifies the text, but it doesn't give me amount of flexibility and control that I want (or need).
  • PowerPoint: I couldn't include .pdf graphics, so I didn't give it serious consideration.
  • Keynote: Keynote makes beautiful presentations, so I thought it would make beautiful posters. However, I couldn't get the text to wrap around the images the way that I really wanted.
  • Pages: The best software I've tried. Equations are not easy, but I use this site to export the few equations that I need to .pdfs and include these as images in the poster. Pages makes it easy to set the size of the poster to something unusual (i.e., 48" x 36"). Pages also handles .pdf graphics nicely. There's a copy of the poster that I made here in case you are interested.

Here are several links to information that I've found helpful in designing my own posters.

  • The Better Posters Blog has lots of interesting tips and even critiques of other's posters.
  • Colin Purrington gives lots of useful tips for designing and presenting posters.
  • GradHacker gives five tips for a better poster. My favorite: "Never underestimate the value of blank space." Most posters are too busy for my tastes.

If you have a different take on any of the software I discussed above, or have experience with other software, let me know in the comments or on Twitter (). Also, if you have any favorite resources that might help me or other improve our posters, let me know.


Journey: "Arguing for a Negligible Effect"

I'm always trying to convince my graduate students that it takes a long time to publish a paper. Motivated by this post by Nathan Jensen, I thought I'd share the timeline for my paper "Arguing for a Negligible Effect" that just came out in AJPS.

The Journey

Started fourth year of graduate school.

2011-10-05: Finished first draft.

One year passed. Rewrote, defended prospectus, received comments, etc.

2012-10-07: Submitted to AJPS.

2012-10-09: Failed technical check, but successfully resubmitted.

Six months passed. Got a job at UB in the meantime.

2013-04-23: Received invitation to revise and resubmit. Accepted invitation.

Four months passed. Started new job at UB. The paper got a lot better during this time, thanks mostly to several great anonymous reviewers.

2013-08-15: Resubmitted.

Three months passed.

2013-11-21: Accepted unconditionally.

2013-12-10: Submitted final version. Received another technical reject. Resubmitted successfully.

2013-12-30: Final version accepted and sent to press.

2014-02-11: Received proof from Wiley.

2014-02-18: Sent corrections back to Wiley.

2014-03-07: Published online.

Seven months passed.

2014-10-13: Published in print.

Summary

This was a very smooth process. It never faced any major obstacles and it was accepted at the first journal I sent it to. It still took three years to go from first draft to in-print and 13 months to go from submission to acceptance.

Takeaway point to graduate students: I had a rough draft of a great idea at the start of my fourth year in graduate school. It wasn't accepted until after I got a job. It is the papers that you write in your second and third years that make you attractive on the job market. Make your second and third years count.

 

 


"Statistically Insignificant" or "Non-Statistically Significant"

I use the phrases "statistical insignificance" and "statistically insignificant" often, but I was recently informed that these terms are not correct. Instead, I was told to say something like "non-statistically significant." In light of this, I'm careful to say "not statistically significance" or "a lack of statistical significance" in my forthcoming AJPS article.

Since then, though, I've been paying attention and notice that researchers smarter than me use both, so I'm not too worried about the distinction.

In Bayesian Methods and "The Insignificance of Null Hypothesis Significance Testing", Jeff Gill uses the phrase "non-statistically significant."

In his blog posts and articles, Andrew Gelman seems to prefer phrases like "statistically insignificant" and "nonsignificant."

I think I prefer "statistically insignificant," since the negation is more clearly on the significant. "Non-statistically significant" makes it sound like we're talking about some other kind of significance, such as substantive significance.

The danger is that the word "insignificant" implies there is "no effect."

I'd be curious to know what others think.


Some Initial Observations on Replications as Class Projects

I taught the graduate course in linear models at UB last semester and a major portion of the course was a replication project. Here are a few quick observations.

  1. Building the course around a replication project has made organizing the course a lot easier. After all, my ultimate goal at the end of the semester is that students be able to run their own regression. Since students are working on their replication projects and asking questions, I have a pretty good sense of what I should be talking about next.
  2. Based on my discussion with some of the students in the class, the replication projects gives the in-class discussions and readings a sense of purpose. I usually try to set up the readings during class, have the students do the necessary reading, and then apply those ideas to their projects. I've given them a clear target to reach by the end of the semester (a high-quality quantitative analysis) and they can see how each topic we discuss helps them get closer to that goal. It's still early, but I think it's been effective so far.
  3. I'm encouraged by the availability of data--it seems like researchers are doing a better job of that. Under Rick Wilson's editorship, AJPS has become an example for other journals to follow.
  4. We've had a couple of cases, though, in which the posted data was incomplete. In each case, a variable was missing.
  5. I'm a little surprised at how difficult it is to understand what is going on in the analysis from the paper. I just try to imagine how hard it would be to replicate someone's results if they did not provide the computer code. I encouraged students to focus their efforts on papers that provided data and code. In two cases, we still weren't able to replicate due to missing data. In a separate cases, the students didn't have a script and were totally lost. They had a complete and well-documented data source, but we eventually had to e-mail the authors for a Stata .do file. With this, we were able to replicate the results.

In the end, these replications were very popular with the students and they managed to write a few great papers. I'm so satisfied with the outcome, I'm doing it again this semester in my advanced methods class.


For more posts, see the Archives.