• The graph we would have made (Massachusetts vs. nation)

    Last week Aaron and I contemplated making a graph that compared Massachusetts’ uninsurance rate with that of the nation over time. We were busy. We didn’t make the graph. But Ezra Klein included just that in his column on Saturday.

    His piece credits the Bay State’s individual mandate for the ultra-low rates of uninsurance. In part, you can credit the natural low rate the state enjoys as well. It was lower than the national average before the 2006 health insurance reforms too. And, Klein also mentions all the ways in which Massachusetts’ health system is not a bed of roses, the high costs, the provider market power, and so forth. I acknowledge all that.

    What I want to point out about the figure, though, is not the difference in rates, but the trend. Massachusetts’ rate of uninsurance has been going down, during the Great Recession mind you. Meanwhile, that of the nation is going up, as would be expected in a struggling economy. This comparison of rates (difference-in-differences) controls for many of the idiosyncrasies of Massachusetts. If you object to straight comparisons because you think Massachusetts is not representative of the nation (and you’re right about that) comparing rates knocks out some, though not all, of the confounding factors.

    Even if the individual mandate can’t make the nation look like Massachusetts (I can almost hear sighs of relief to that), if it can reverse the upward trend in uninsurance rates, that’s something in its favor, at least as far as I’m concerned.

    Comments closed
    • Hopefully, if you had done such a chart, you could have at least made it accurate although it is unlikely I would have agreed with any conclusion you drew from the data.

      I am looking at the DHCFP report and cannot find anything that looks similar to these numbers. But there were much more egregious errors in Klein’s blog post Saturday that I missed this one (particularly the claim that the individual mandate reform in Massachusetts depressed premiums 40% — HA HA)

      As you say, the real key is the rate of change. The fact is that even by giving insurance away for the last 5 years, Massachusetts cannot get universal health insurance coverage. Whether you use the Census bureau numbers, the old DHCFP numbers, or these Liberal Institute numbers — it is clear that the idea of taking the Free Care Pool money that used to go to hospitals and giving it instead to insurers didn’t work.

      • @Dennis – The DHCP report has loads of figures on non-elderly adults and they tell the same story, that the 2010 uninsured rate is below that of 2008 and 2009.

        You’re right that Klein’s graph has different (lower) numbers. But it says it is for adults, probably including the elderly. He probably did it this way because that’s what he had national data on. There is likely enough information in the DHCP reports to reproduce Klein’s figures, but I haven’t done that.

    • Would be nice to see this chart matched up with one on costs.

      “(particularly the claim that the individual mandate reform in Massachusetts depressed premiums 40%”

      Only for individual insurance. Is there data showing otherwise?


    • Even if the individual mandate can’t make the nation look like Massachusetts (I can almost hear sighs of relief to that), if it can reverse the upward trend in uninsurance rates, that’s something in its favor, at least as far as I’m concerned.

      I think that over insurance is a bigger problem than uninsurance though both may be a problem at this time.

      I would like to see fewer people with employer funded insurance and fewer people with low deductible insurance. I wonder how MA compares to the rest of the states in that regard. Further I (like David Goldhill) would like to see people get very high deductible insurance.


      IMO it is good that MA has moved on Health Insurance I would like to see other states go in various directions including single payer and 2 tiered. One of the worst parts of Fed Gov acting on health insurance is that it will reduce the scope of experimentation at the sate and local level.

    • I just had a thought, I wonder if we could all agree that it would be better if the rich (those with net worth above say $2 million) did not have health insurance since they can pay out of pocket.

    • How do the premiums in Massachusrtts compare to premiums of other states, for comparable plans?
      My understanding is that MA has some of the highest premiums in the country.
      Don Levit

    • @Austin – I am not arguing that the rate did not go down. How could they not go down when the insurance is given away free? But why did he make up numbers that did not appear in the report and claim the report was the source? And if he wanted to compare US and Mass numbers, the intellectually honest thing to do would have been to use the Census numbers (I don’t know if there are Gallup numbers for Mass.?) The fact is that all these changes since 2006 no matter what the source are statistically insignficant.

      • @Dennis – I did my best to explain. You’ll have to ask Klein if you want to know more. How do you know the changes ar not statistically significant? Does the report say that? (I don’t have time to look, but maybe you have.)

    • @ austin

      As I said above, I am not arguing that the number of insured did not go up. How could it not have? The state gives the insurance away for free. [But note that the number of people actually buying insurance rather than getting it free has dropped according to the state since Romneycare’s mandate kicked in.]

      All my original comment said was that if YOU had made such a graph (which was the headline of your blog post), you would have at least done it accurately. Klein said to me in an email that he got to his numbers by doing what you said he did. But he didn’t do it using any numbers I can figure. And there was no need to come up with his own numbers because the report included a number for non-elderly adults.

      And — more important to rational discussion — he purposely avoids the study that would fairly compare U.S. with Mass (the Census) without him having to fabricate numbers. For him to compare one set of contrived Urban Institute-based numbers with Gallup numbers is nonsensical when a methodologically consistent set of comparatives exist from the Cenus Bureau..

      To answer your question, even though I don’t think it useful to get into wonky statistical significance stuff, yes the report specifically says the data Klein is talking about (adults) is not statistically significant. The state’s version of the Urban Institute report reads as follows:

      “In some cases, what appear to be relatively large differences in estimates between years are not statistically significant. This arises because estimates based on small subgroups of the overall population have larger variances, making point estimates less precise.”

      (It’s kind of a vanilla description of non statistically significant so it is interesting that the Urban Institute felt compelled to say it. I say “state’s version of…” because there may be a different Urban
      version out there somewhere.)

      Also according to the state version, the change in the different numbers over time for all insured between 2009 and 2010 — from 2.7 to 1.9, — including children and elderly (not what Klein is talking about), is statistically significant at a 95% degree if confidence PLUS OR MINUS 5% using the “two-tailed test.” But the change in the numbers from 2008 to 2010 is more actionable and it is trivial.

      Getting really wonky, I would not depend on a two-tailed test for directionality. And of course the 2009-2010 change for the total is driven by the decrease in the children number from 1.9 to .2 — PLUS or MINUS 1% — with a 95% degree of confidence.

      There are real suspect deltas, small absolutes with wide variances for the size of the absolutes based on small n’s that are intended to represent 6,500,000 people. That’s why the best data is the actual
      bottoms-up numbers as provided in the state’s Quarterly Key Indicators reports. Those are real people numbers, not estimates.