A local news article in a recent issue of the [Nashville] Tennessean initially caught my attention not because of the subject matter, but because of the way it was titled — NASHVILLE GENERAL Poll: Residents support more funding for struggling hospital. The question that immediately came to mind as I read that title was whether the wording of the question pollsters asked might have influenced the results, resulting in this headline. As I read the article, I was drawn more to what it revealed about polling than to the actual content related to the hospital.
The Role Of Polls
Polls, if taken following strict rules of statistical math to ensure the proper size and stratification of the pool of respondents, can be very useful. However, the slightest bend or twist from these rules can produce bias in the results. That why you’ll see certain polls emphasized in one set of media outlets and other polls emphasized in another set of media outlets. Bias can be introduced into a poll in at least two ways I can think of even if rules that ensure randomness [e.g., sample size] are followed.
First,, the polling medium [“land line” phones, cell phones, snail-mail mail-outs, email mail-outs, door-to-door, etc.] can introduce bias because different strata in the population are more apt to have access to — or to be responsive to pollsters using — various polling media [strata in this context meaning segments by race, age, sex, education, marital status, etc.].
Second, different ways of wording questions can cause people to feel “obligated” to pick a particular answer from a list of choices even if that answer is not fully consistent with what they would write as a free-form answer if given that opportunity. Surveys sent out by political parties [the main purpose of which is to solicit contributions, not to determine what voters think about issues] are the “poster children” of this source of bias — “__Yes __No: Do you support our strong position on a woman’s right to make her own decisions about her reproductive health and our support of Planned Parenthood as a key partner in helping them navigate through difficult personal decisions regarding unwanted pregnancies?” or “__Yes __No: Do you support our strong position that securing our borders is an absolute prerequisite to development of a successful Immigration Reform plan, and that building a wall on our sudden border is a necessary component of securing our borders?” Any Democrat would feel pressured to answer “Yes” on the first question, and any Republican would feel pressured to answer “Yes” to the second.
Many Democrats, if given the opportunity to write a free-form answer to the question to which “Yes” clearly means “I’m pro-abortion,” might write an answer that would not be so easily characterized that way. Many Republicans, if given the opportunity to write a free-form answer to the question to which “Yes” clearly means “Yes, I fully support the Republican stance on Immigration,” might write an answer that would not be so easily characterized that way.
The Toll Of Polls
Toll, in the context in which I’m using it here, refers not to the costs of conducting polls, but to the impact on public perception. As I’ve pointed out above, bias can — and I believe often does — play a significant role in interpretation and reporting of the results — and the results, particularly when presented through the biases of the media outlets through which they become public, have a tremendous impact on how the public’s view of the issues covered by polls is formed.
So What Conclusion(s) Can Actually Be Drawn?
I’ll conclude by getting back to the article that brought this concept to the forefront of my attention this week — the article that led with the headline “Residents support more funding for struggling hospital.” Well of course they do! Who is going to answer “No” to that question if the hospital is characterized as a “safety net facility for the poor and uninsured?”
The only conclusion that can actually be drawn from this particular poll is that people like the idea of government support for the underprivileged. It made no connection between the touchy/feely aspect and the more practical aspect that would likely have affected the answer of many of the people who were polled — i,e., the cost to taxpayers of this particular ongoing and constantly-increasing expenditure vis-a-vis the tradeoffs, like what other costs might ultimately have to be cut if we continue this funding. And I doubt if even a tenth of one percent of the people polled have any idea whether the $35 million current support level is adequate or not — or whether the “emergency” increase is a true need or one generated by less than optimal management over the years of funding support already received.
In closing, I should emphasize that the point I’m making here has nothing to do with this hospital or whether or not they need the increased governmental support they seek. The point is that polls need to be interpreted in light of the considerations I’ve brought out here.
Thanks for reading this post, and if you regularly follow my Blog, for that, too. Please consider sharing this or other posts with your friends, colleagues and associates.
Charles M. Jones
One thought on “The Role And Toll Of Polls”