For several months now, whenever the topic of enrollment in the Affordable Care Act came up, I’ve been saying that it was too soon to tell its ultimate effects. We don’t know how many people have paid for their new insurance policies, or how many of those who bought policies were previously uninsured.
For that, I said, we will have to wait for Census Bureau data, which offer the best assessment of the insurance status of the whole population. Other surveys are available, but the samples are smaller, so they’re not as good; the census is the gold standard. Unfortunately, as I invariably noted, these data won’t be available until 2015.
I stand corrected: These data won’t be available at all. Ever.
No, I’m not kidding. I wish I was. The New York Times reports that the Obama administration has changed the survey so that we cannot directly compare the numbers on the uninsured over time.
I’m speechless. Shocked. Stunned. Horrified. Befuddled. Aghast, appalled, thunderstruck, perplexed, baffled, bewildered and dumbfounded. It’s not that I am opposed to the changes: Everyone understands that the census reports probably overstate the true number of the uninsured, because the number they report is supposed to be “people who lacked insurance for the entire previous year,” but people tend to answer with their insurance status right now.
But why, oh, why, would you change it in the one year in the entire history of the republic that it is most important for policymakers, researchers and voters to be able to compare the number of uninsured to those in prior years? The answers would seem to range from “total incompetence on the part of every level of this administration” to something worse.
The new numbers will suffer, to some extent, from the same bias that the old questions suffered from: People are better at remembering recent events than later ones.
And what has been happening in the most recent months? A whole lot of change! Policies were canceled, benefits changed, people shifted around their coverage in anticipation of the new law. That doesn’t make for a very good baseline. It will be a very good measure of who has insurance right now, in 2014, but it’s not where I’d want to start my 2013 baseline for our new law. That’s why they should have done this for 2012 — or waited until 2016 — to give us actual comparable data for the transition period.
Heart of the debate
I find it completely and totally impossible to believe that this problem didn’t occur to anyone at Census, or in the White House. It would be like arguing that the George W. Bush administration might have inadvertently overlooked the possibility that when the U.S. invaded Iraq, there would be shooting. This is the biggest policy debate of the last 10 years, and these data are at the heart of that debate.
Even if the administration genuinely believes this is defensible, why would they give anyone reason to believe that it is cooking the books? I can certainly think of explanations, but I can’t quite bring myself to believe the worst of them. Which leaves me with the only slightly-less-utterly-appalling conclusion: At some point, very early on in the process, folks noticed that asking the new questions would make it difficult to compare Obamacare’s implementation year to prior years, and decided that assessing the effects of the transition wasn’t nearly as important as making urgent changes to … questions we’ve been asking basically the same way for a decade and a half.
No, wait, that doesn’t make any sense, either. Let’s go back to inexplicable, shall we?
If the administration is really serious about transparency and data-driven policy, as I’ve been told for a year now, then it will immediately rectify this appalling mistake and put the old questions back into circulation double-quick. But we’re more likely going to hear the most transparent and data-driven administration in history citing these data — without an asterisk — to tout the amazing impact of its policies.
Megan McArdle writes about economics, business and public policy for Bloomberg View. Twitter: @asymmetricinfo.