How Not to Poll

The Quinnipiac University poll usually is pretty good. But a recent (late February) poll on funding for Planned Parenthood had a startling textbook example of a bad polling question:

"Do you support or oppose cutting off federal government funding to Planned Parenthood?"

I support it-- no, wait, I support funding Planned Parenthood, so I oppose cutting off funding...

In the end, 53% opposed cutting off funding, while 43% supported it.

In contrast, here's the question from a CNN/Opinion Research Corporation poll in early April:

"As you may know, the federal government currently provides funding to support some programs run by the organization called Planned Parenthood. Do you think federal funding for Planned Parenthood should be eliminated entirely, or do you think the federal government should continue to provide funding to Planned Parenthood?"

This time, 65%  said the government should continue to provide funding, while 34% said funding should be eliminated. Now, opinions may have changed a little over March, but if you take the numbers from the second poll as true, and assume that a significant part (say, one-third) of each group says the opposite of what it intends to say, you get something like the numbers in the first poll.

What's sad here is that polling is expensive, and the Quinnipiac poll's sample was more than twice as big as the CNN poll's. Seems like a waste of money to me. The bigger sample gives them a more precise estimate, but a more precise estimate of a meaningless number is still meaningless.