Last week, the New York Times reported that the Census Bureau would be significantly changing the questions and methods it uses to determine who has health insurance. The redesign is an attempt to address some of the flaws in the current design that have long troubled the agency. A working paper from the Census Bureau had found that it provided an “inflated estimate of the uninsured” and was prone to “measurement errors” that diminished the reliability and usefulness of the measure.
The timing of this change could hardly be worse. The massive coverage provisions of the health care reform have just taken effect, and these new changes could make comparisons to past years difficult, or meaningless. Another document from the agency explains that the questions would elicit such different responses that “it is likely the Census Bureau will decide that there is a break in the series for the health insurance estimates.”
As the Times reports, the differences in responses between the two sets of questions are significant; in a trial run last year, the percentage of people without health insurance was 10.6 percent with the new questionnaire, compared with 12.5 percent using the old version, with similar effects across all demographic groups.
Some defenders of the decision have pointed out that these new questions will also give data for 2013, so there will be at least one year of pre-ACA data to compare to. This is true, and having at least one data point will be helpful to some extent, but what we really want to evaluate when analyzing the law would be the longer term trend, for two reasons. One, there is a decent amount of variation in these surveys that make single data points less informative. Two, while the major coverage provisions of the law take effect in 2014, the law has already been influencing the insurance market in smaller ways since its passage, and more than half of the reduction in the uninsured will occur after 2014, according to the Congressional Budget Office. This is why having a stable baseline would be useful, so we could examine the longer term trends in insurance coverage, and why now is close to the worst time to incorporate this change. The Census Bureau acknowledged as much in a paper, admitting that “[i]deally, the redesign would have had at least a few years to gather base line and trend data.”