Can You Really Trust an Opinion Poll?

Probably not.  Well . . . it all depends.  In a recent blog I discussed my skepticism regarding opinion polls and my fear that most people trust them, making opinion polls the deciding factor in far too many cases.  I also suggested that many people think that legislators should pay close attention polls, which, I’m certain would be folly.

To my delight Dalia Sussman has an article in today’s Times in which she points out the major issue of difficulty in polling:  the questions.

Her first example of polling differences clarifies my skepticism so well, that I’ll print the paragraph: Consider the way Americans feel about changing the health care system. In July, with debate at its most heated, a New York Times/CBS News poll found that 66 percent of Americans supported the so-called public option. Around the same time, Time magazine reported that 56 percent supported it, a Pew poll found 52 percent in favor, and a Fox News poll reported the support at 44 percent.

Each organization had good data to support its findings, but it’s appropriate to dismiss them all.

What’s going on?  Sometimes inconsistent results suggest that the respondents haven’t thought through what they think.  But as I wrote in my post, the most obvious issue may well be the questions.  

That was Sussman’s approach.  So to get an assessment of the problems surround polling questions, she went to an expert: Jon Krosnick of Stanford.  With the polling questions in front him, Krosnick explained how different questions in different contexts result in different conclusions. 

For example, many September polls on health care asked people whether they supported or opposed Obama’s proposed changes, but the Times CBS poll offered respondents the choice of saying that they did not know enough about the changes yet to say.  Nearly half of the respondents, 46 percent, said just that.  The other polls showed different degrees of support or opposition to Mr. Obama’s proposals largely because they did not offer the “don’t know yet” option.  The right approach is debatable.

What’s the takeaway?  I’d be very skeptical of polling, polling results, and policy built solely on polling.  We expect more than that from our legislators. 

Link to original post

Leave a Reply