A Project of The Annenberg Public Policy Center

62%: Conventional Wisdom or “Dubious Research”?


The liberal group Health Care for America Now has come out with another ad attacking the health insurance industry. The ad is running as part of a million dollar ad buy, according to the group. It repeats a claim that is being made repeatedly as the debate over overhauling the health care system heats up.

The ad says "62 percent of personal bankruptcies are caused by medical debt," and the group’s backup points to a study conducted by Harvard researchers, led by David Himmelstein, that was released in June of this year.

To start, the study doesn’t quite back up the ad. The researchers found that "illness or medical bills contributed to 62.1% of all bankruptcies," but that’s not the same as saying that bankruptcies were caused by "medical debt," as the ad claims. To say that something is one of a number of contributing factors falls short of saying it "caused" something. And the study counts loss of income due to illness as a contributing factors. Inability to work isn’t the same thing as a "medical debt."

Furthermore, there’s healthy disagreement among scholars about whether the 62 percent figure is correct. The study’s methodology and results are being challenged by others — and not for the first time. The Harvard researchers released a similar study in 2005 which concluded that medical costs contributed to about half of all bankruptcies. The statistic became a talking point for a number of Democrats during the Presidential primaries and we addressed it in a piece that went over criticisms of the findings, the author’s responses and other relevant scholarship. Ultimately, we concluded that it would be "difficult to say for certain whether the much-touted finding – that half of all bankruptcies are due to medical issues (though not necessarily medical bills) – is correct."

The new figure of 62 percent has also been criticized. Aparna Mathur, a fellow at the conservative American Enterprise Institute, wrote that the Himmelstein survey "clearly states that only 29 percent of the respondents believed that their bankruptcy was actually caused by medical bills," and that the authors used their own criteria to label bankruptcies as medical. Marthur authored a separate paper that that came up with a much lower figure, saying that "nearly 27 percent of filings are a consequence of primarily medical debt, while in approximately 36 percent of cases medical debts co-exist with primarily credit card debts."

Now a fresh review of the financial impact on those who become seriously ill is on the way from a Northwestern University academic who has been critical the Harvard team’s work in the past.  David Dranove, director of the Center for Health Industry Market Economics at Northwestern University’s Kellogg School of Management, co-authored a lengthy critique of the original Himmelstein paper and concluded that medical bills were a contributing factor in only 17 percent (not half) of personal bankruptcies. Now Dranove is working on a new study about the relationship of medical bills and life savings. It will not be released for another two weeks, but when we asked for his comment on the 62 percent figure, Dranove said:

Dranove, Oct. 2: The Harvard doctors can tell us over and over again that 62% of bankruptcies involved folks who had medical spending, but their research cannot tell us whether any of those individuals would have avoided bankruptcy without medical spending. Perhaps all of them would have been bankrupt anyway, but it would have taken a month longer. Or maybe they would have been just fine financially without health spending. We just don’t know.

Numbers should matter to academics. As we have seen, even dubious research becomes conventional wisdom when the researchers are credentialed, the journal’s referees and editor fail to do their job, and the result bolsters a populist political viewpoint. It is also important to note that different journals have different standards and that editors sometimes accept papers of questionable statistical veracity in order to promote more work in the field.

We won’t try to judge which scholar is correct. What we can say is that the 62 percent figure rests on assumptions that are very much in dispute among experts who study such things.