Saturday, May 16, 2009

Majority Now Pro-Life!!!!

That would be the majority of Americans, 51%, according to a recent Gallup poll. I'm sure you have heard about this already:

The Gallup Poll reported Thursday that 51 percent of Americans now call themselves pro-life rather than pro-choice on the issue of abortion, the first time a majority gave that answer in the 15 years that Gallup has asked the question.

The findings, obtained in an annual survey on values and beliefs conducted May 7-10, marked a significant shift from a year ago. A year ago, 50 percent said they were pro-choice and 44 percent pro-life _ in the new poll, 42 percent said they were pro-choice.

The new survey showed that Americans remained deeply divided on the legality of abortion _ with 23 percent saying it should be illegal in all circumstances, 22 percent saying it should be legal under any circumstances, and 53 percent saying it should be legal only under certain circumstances.

The findings echoed a recent national survey by the Pew Research Center, which reported a sharp decline since last August in those saying abortion should be legal in all or most cases _ from 54 percent to 46 percent.

Polling Report collects polls about abortion. It shows that the latest polls (April and May) have been all over the place.

This most recent Gallup poll demonstrates a flip of the percentages in the pro-choice and pro-life categories from just a year ago. Such a big change would usually have something to do with recent news (and not necessarily part of a long-term trend), but it's hard to see what those news might have been.

A different hypothesis is also possible:

Gallup said its new poll showed an increase in the pro-life position across Christian religious affiliations, including an eight-point gain among Protestants and a seven-point gain among Catholics. It also reported a 10-point shift toward the pro-life category among Republicans but said there was no significant change among Democrats.

"It is possible that, through his abortion policies, Obama has pushed the public's understanding of what it means to be 'pro-choice' slightly to the left, politically," according to the Gallup analysis. "While Democrats may support that, as they generally support everything Obama is doing as president, it may be driving others in the opposite direction."

This could be the explanation, I guess. But let's take a step back and ask what it is exactly that these polls do. First they contact a certain number of people and ask them about their opinions (with questions that are sometimes not very well formulated). THEN they generalize the findings of that sample to all American adults or voters or whatever the group of interest might be.

How did the Gallup poll do that generalization? Did they standardize the various opinions using stated political affiliations? For instance, suppose that Republicans last year were 30%, Independents 30% and Democrats 40% of the electorate (just as an example). Then those might be the weights used to calculate the final 51% finding. I suspect that they did something like this, given that we are told the shift is caused by Republicans and Republican-leaning Independents and that Democrats have not changed their opinions.

Now, I've read that the Republican Party is shrinking towards its base. If this is true and if the poll uses old percentages for each political affiliation we could get these results even without any real change in the American public opinion. I'm not saying that this is what happened. Further research is necessary. But it's worth pointing out.
Addendum 1: Check out the results over time in this Gallup summary. What's very odd about it is where the changes in the most recent polls are: at the extreme tails.

Addendum 2: Gallup really should give more detailed information. I cannot find out if the sample is a simple random one, if they use post-stratification, what the exact percentages for Republicans, Democrats and Independents are, or anything else that would let me test my hypothesis that the results are driven by how the political groups are weighted.

Addendum 3: The most likely hypothesis seems to be a sampling error. The sample has "too many" Republicans and "too few" Democrats, as shown in this post.