The Census Bureau is Probably Not Cooking the Books

Writers at the Wall Street Journal (and, as far as I can tell, ObamaCare haters everywhere) are angry that the Census Bureau has changed the way it measures how many people are covered by health insurance, saying that

[T]he Census revamped the CPS household insurance questions, muddying comparisons between the pre- and post-ObamaCare numbers. The results of the new method will be disclosed this fall […]

Robert Pear of the New York Times obtained internal Census documents that note that the new CPS system produces lower estimates of the uninsured as an artifact of how the questionnaire is structured. One memo refers to the “coincidental and unfortunate timing” and that, “Ideally, the redesign would have had at least a few years to gather base line and trend data.”

The insinuation in this article is obvious: that the Census Bureau timed this change so that it would exaggerate the benefits created by ObamaCare. But I’m not convinced.

First of all, it is misleading to say that the new method producing lower numbers of uninsured is an “artifact” of the new questions. The new questions intend to produce lower numbers of uninsured Americans because research at the CB shows that the old questions they used create an upward bias. They are correcting an artifact, not  creating one.

Second, the change was announced in September 2013, before the botched ObamaCare roll out. If the CB meant to cover up the bad numbers, it must have been clairvoyant about the fact that the launch of would encounter so many problems, and it should be obvious that this is not the case.

Moreover, do we really have any reason to think that the CB biases their results in Obama’s favor? Sure, one could claim that the statisticians at the CB have a liberal bias (though I see no evidence of this). But do they really have an incentive to construct an elaborate plan that will potentially muddy numbers so terribly that Obama looks good when he should look bad? If they do have such an incentive, I do not see what it possibly could be.

That said, this probably was not the best time to roll out a new measurement system, considering that health care coverage is such a charged topic right now. Still, we want the best estimates that we can get, and in the link provided above, the CB cites several sources that show that the new question formats provide more accurate data. The alternative to changing the method would be keeping a system that gives biased results for several more years.

Anyway, I imagine that once people cool down a bit and realize that the change is probably not a sign of corruption, this story will fade fast. But it gives us a good chance to reflect on the role government statisticians and whether we accuse certain organizations of being corrupt because it bodes well with our political ideology.