Earlier this month a poll from Christopher Newport University’s Wason Center has the Commonwealth’s political class in a “twitter.” Republicans argue that the poll is conclusive proof that Virginians do not favor Medicaid expansion, and some Democrats are choosing to ignore the poll on the theory that “if you are explainin’, you are losin’.” My take is that all polls are “data points” that need to be assessed at face value, and we have an obligation, as elected officials, to explain what policy choices are before us and how citizen views should be taken into account in making them.

All polls are surveys based on samples from a larger population. Polling is not a perfect science, and is subject to considerable error. A good poll should be both scientifically valid and reliable. To be so, social scientists first need to be sure that the questions test what they are designed to test. Second, the results should be capable of being replicated through another sample. So does the recent CNU poll pass the test? In some ways, yes; and in others, no.

First, to ensure a scientifically valid poll, the questions must first be worded in a clear and neutral fashion. Without clarity or neutrality, the results will be slanted. This is clearly where the CNU poll has problems. As Republican Sen. Emmett Hanger of Augusta pointed out, the poll’s statement “Democrats propose to subsidize…” is factually incorrect; there are three Republican Senators who developed and support “Marketplace Virginia” and that approach is not “strict Medicaid expansion”. There are several other words that could influence the outcome of a poll such as “subsidize, fear, waste, and abuse.”[1]

Second, the subject in the sample must be randomly selected; in this poll, that appears to be the case.

Finally, the survey must be sufficiently large to reduce the “built-in sampling error.” That usually requires about 400 or more subjects for a population the size of Virginia, something that apparently the CNU poll has met.

But the biggest issue in the use of polling data, and thus the biggest problem with the CNU poll, is the extrapolation from one data point of the survey results into suggesting the results suggest a “trend.” When the CNU pollster concludes, as a result of the survey, that “Democrats are losing the debate on Medicaid expansion,” he is making a jump not justified by the poll.

Although this is not really a “push poll,” you cannot infer a trend simply on the basis of these results. The only way you can discuss a trend is if the exact same question was asked to a randomized sample from the same universe at two different periods of time.[2] The CNU poll did not do this and any conclusions about “trends” are meaningless as a result.

As always, it is a pleasure representing you in the General Assembly and I hope that you will contact me with your views and comments about issues affecting the region and the Commonwealth.

[1] The actual poll question reads, “In [the Medicaid] debate, the Democrats propose to subsidize private insurance for 400,000 uninsured and low income Virginians by using federal Medicaid money that would otherwise not come to Virginia. Republicans oppose this expansion because they fear the federal Medicaid money will not come as promised, and also say the current Medicaid program has too much waste and abuse and needs reformed [sic] before it is expanded. I’d like to know where you stand, would you say that you generally [RANDOMIZE: “support using federal Medicaid money to expand health coverage” or “oppose using federal Medicaid money to expand health coverage”]?
 
[2] The February 3, 2014 CNU poll asked the question: “Medicaid is a health care program for families and individuals with low income that is funded by both federal and start tax dollars. Currently, Virginia is faced with decision about whether to expand the Medicaid program to cover an additional 400,000 mostly working poor Virginia who are uninsured. In general, do you support Medicaid expansion or oppose it?” 56% of the respondents said they supported expansion.