Gallup, Abortion, and Shades of Gray

With the economy, health care reform, environmental regulation, and other important issues being widely discussed in policy circles, it would be easy for one to forget about wedge issues, such as abortion. However, with the news of the shooting of  Dr. George Tiller, among other recent acts of extremist right-wing violence, and the debate over a new Supreme Court nominee, abortion is back on the front pages.

In this light, I decided that I would take a deeper look into a recent poll that was conducted by Gallup that found changing attitudes toward abortion in the US. Gallup’s results showed that for the first time since they began polling the issue 14 years ago,  more Americans identify themselves as “pro-life” than “pro-choice”.  According to this new poll, virtually all movement in public attitudes toward the pro-life position has occurred within the past year.  After reading about these results, I had several questions, including:

1. The dramatic shift in the past year looked a bit odd to me.  Could Gallup expand upon the bottom-line reasoning from their reporting?
2. What was the party breakdown of the poll? It doesn’t mention weighting, but perhaps they did weight. (If I remembered correctly from the Presidential tracking polls in 2008, one of the big differences between Gallup and Rasmussen was that Rasmussen weighted and Gallup did not, leading to more swings in the Gallup tracker.) My concerns here were that a smaller, more extremist Republican tent, could indicate a misleading swing if they were still weighted at their 2008 levels.
3. Relating to #2, I recently read that Gallup had nearly a 50/50 split in Party ID in this poll. Was this correct?
   

Thanks to my former graduate school classmate, Cynthia English, a Gallup writer and researcher, I had the honor of having my questions answered by Lydia Saad, a Senior Editor at Gallup who worked on this poll. Ms. Saad gave very thoughtful answers to my questions and went above and beyond what I expected. Here are some of Ms. Saad’s responses:

• Kevin’s memory is correct; we do not weight our surveys by Party ID. Although some pollsters do it, weighting by Party ID is not the standard in national RDD surveys. Party ID is essentially a political attitude like every other that we measure; and while it is generally stable from one survey to the next, it does change over time and is susceptible to survey-to-survey variation due to the content of a given survey. Weighting by party ID on election polls, for example, can be problematic since it’s asked after the candidate preference ballot, and therefore largely mirrors the ballot. To weight by party ID on these surveys is to essentially weight by the ballot.

• We did obtain a near 50-50 split in leaned party ID on the 2009 May Values survey. Because this was unusual, we did two things to check the validity of the data. We re-ran the abortion questions on the G1K track two days later, and obtained nearly the same results. That survey had a 10-point advantage for the Democrats on leaned party ID. We also did a post hoc reweighting of the data by party ID, using targets giving Democrats/Dem leaners a 14-point advantage (typical of what we’ve been getting on recent stand-alone polls) and re-ran the survey results . (This was for internal analysis only; we are not publishing the reweighted figures.) The figures changed by only 1-2 points in most cases – indicating that the party distribution of the sample did not account for all or even much of the change seen in the abortion trends. However, as noted in point A, we don’t consider the party ID distribution we obtained in the survey “wrong” just because it was different from what we obtain on other surveys. Thus, we stand behind the published figures based on our standard Census-based demographic targets.

• As highlighted in the story, and expanded on in the “bottom line” analysis, the major change in abortion attitudes over the past year was seen among Republicans and independents who lean Republican. Thus, even if we were to hold party ID constant across the two surveys, attitudes would have become less friendly to abortion rights because Republicans moved to the right, while Democrats stayed the same. The question is, why did Republicans become more conservative in their views on abortion? The “pro-life” side has been eager to attribute it to the “success of their efforts” on the issue. I’m dubious about that. Without a high profile “pro-life” campaign over the past year to attribute this to (which I can’t),  I would expect to see that sort of attitudinal change happen more gradually. This was abrupt. The major change that’s happened is that Obama was elected, and since he is “pro-choice” and those views have been forefront in the news over the Notre Dame flap, I think it’s reasonable to hypothesize that this has compelled some.

• The external validation component is very important. We are not alone in showing a shift toward the “pro-life” position (or anti-abortion position, in the case of legality questions). Aside from Gallup, four other organizations have come out with abortion data in recent weeks, and all of them show a more “pro-life” stance than they did in their last measurement in 2008 (all pre-election).

• “PRO-LIFE” V “PRO-CHOICE”
  Gallup Values survey shows a 7 point increase in “pro-life” and an 8 point decline in “pro-choice” (May 08 vs. May 09) SWING=15 POINTS.    Gallup G1K survey shows a 6 point increase in “pro-life” and a 7 point decline in “pro-choice” (May 08 vs. May 09) SWING=13 POINTS.    Fox News shows an 8 point increase in “pro-life” and 6 point decline in “pro-choice” (September 08 vs. May 09) SWING = 14 POINTS.    CNN shows a 1 point increase in “pro-life” and a 4 point decline in “pro-choice” (Aug 08 vs. April 09) SWING=5 POINTS.

• LEGAL/ILLEGAL
  Gallup Values survey shows a 3 point increase in “legal in only a few/illegal in all” and a 4 point decrease in “legal in all/most” SWING=7 POINTS.    Gallup G1K survey shows a 5 point increase in “legal in only a few/illegal in all” and a 6 point decrease in “legal in all/most” SWING=11 POINTS.    Quinnipiac shows a 3 point increase in “always/usually illegal” and a 5 point decline in “always/usually legal.” (July 08 vs. April 09) SWING=8 POINTS.    Pew shows a 3 point increase in “always/usually illegal” and an 8 point decline in “always/usually legal” (Aug 09 vs. Apr 09) SWING=11 POINTS.

Given these responses, I think it is fair to say that Gallup and others are on to something. There does seem to be a change, albeit possibly temporary, in attitudes toward abortion. Given this change, several questions come to mind:

1. Is reporting about abortion with only two binary options the most appropriate way of showing public opinion? What are some other options to polling about abortion?
2. Will this change be permanent or just a temporary blip in public opinion due to reasons that Gallup points out?

As far as reporting such polling results as binary options (”pro-choice” v. “pro-life” or always/sometimes legal vs. always/sometimes illegal), Gallup also thankfully breaks down its legal/illegal question into four categories. Granted, opinion on abortion is probably more nuanced than four categories, but it is encouraging that Gallup offers these details:

• Legal under any circumstance (change from 2008 to 2009):  -6 points
• Legal under most circumstances: +2 points
• Illegal under most circumstances: – 3 points
• Illegal under all circumstances: +6 points

While it is possible and in fact likely that many Republicans who once had a nuanced position on abortion now identify themselves in the extreme given the polarization of the GOP tent and the fear of an “abortion-loving president” from the talk radio set, the movement away from the “legal under any circumstances” category is still a bit  perplexing by the “Republican Party being more extreme” movement theory.  Are there really many Republicans who just one year ago thought abortion should be legal under any circumstance who now are 1.) Still Republicans AND 2.) No longer hold this position?  It’s possible, but definitely not as likely or as easily explainable as the movement toward the “illegal under any circumstances” camp.  Perhaps this cross-tabulation is just random noise, which wouldn’t be surprising since the margin of error is going to be much higher among these subgroups.

What are some other approaches for asking about abortion?

While I commend Gallup for asking about this question in more than a strictly binary fashion, it’s important to point out that there are other possible ways of asking about abortion that could possibly lead to very different baseline conclusions.  Paul Rosenberg does a nice job of summarizing the findings of the General Population Survey (GPS), which gives survey respondents three different abortion scenarios and asks them to indicate whether they think abortion should be illegal in:

1. None of these cases
2. One of these cases
3. Two of these cases
4. All of these cases

Since many people may have a hard time defining exactly what “pro-choice” and “pro-life” are (Gallup admits that it doesn’t necessarily endorse these terms, and hence uses them in quotes), this approach is nice because it conceptualizes the issue in three nuanced situations, ranging in acceptability.  By using this approach, the GPS finds that only 9% of respondents believe that abortion should be illegal in all three given cases.  This is not to say that this bottom-line result is more accurate than other polls, as it is  asking about slightly different things. (It’s also important to note that these numbers are not meant to in any way dispute Gallup’s trend, but rather to show that a different interpretation of baseline values could be made by using a slightly different methodology.)

What does this mean?

The book Myth of a Polarized America further explores some these issues and argues that most of the “pro-choice” versus “pro-life” debate is in fact media driven and that most Americans actually lie somewhere in the middle, holding a nuanced opinion about abortion and other social issues.  In this light, is it possible that media reports that only repeat binary results of such wedge-issue polls encourage the narrative of a divided, polarized America?

Although recent right-wing extremism may be afoot, it is important to remember that most of those that are pro-life are anything but extremists.  Despite what they may tell pollsters, one can legitimately argue that most Americans hold nuanced views that deserve nuanced reporting that respects the complexities that are inherent in such social issues. Given the apparent sudden change in attitudes, it will be interesting to look at this issue again in five or six months or in a year to see whether this is a short-term blip in response to the first brand-new Democratic President in 16 years or a sudden, sustainable change in public opinion.  The best period to which to compare this recent movement would be 1993, when President Clinton first took office. Unfortunately, 1993 was two years before Gallup began polling this issue.

Bottom Line

I applaud Gallup and others for looking in-depth at these issues and hope that the mainstream media can begin to report such public attitudes and beliefs with the nuance and respectful tone that they deserve.  Like most things in life, abortion does not involve mutually exclusive sets of ideas and values for most individuals.

Final Tracking Poll Update

I know that on Election Eve everyone is nostalgic and filled with memories of tracking poll composites gone by. With that said, get the Kleenex boxes ready and behold the last tracking poll update of this election season:

Overall, Obama maintains a significant advantage in the tracking polls. As you see below, his lead has been steady for the past eight days. So much for the McCain surge.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/27	10	5	8	8	5	3	7	6.7
10/28	7	5	8	7	4	4	7	6.0
10/29	7	3	7	6	5	3	8	5.5
10/30	7	5	6	5	7	4	8	6.0
10/31	9	4	7	6	7	4	9	6.6
11/1	10	5	7	7	5	5	9	7.0
11/2	9	5	5	7	6	2	11	6.5
11/3	11	6	5	6	7	5	9	7.3

What should be noted here is that five of the seven tracking polls now show Obama’s lead to be between five and seven percentage points. This is a remarkable convergence and gives me pretty good confidence in making an educated guess of the actual election outcome. Yes, McCain might win the majority of the undecideds, but Obama is above 50 percent in all of these polls that push undecideds and therefore in the safe range. Granted anything can happen, but with numbers like these, a McCain win would be a big upset.

Sunday Night Tracking Poll Update

With less than two days to go until Election Day, the tracking poll composite remains relatively stable. Taking a look at the last seven days, we see that the Obama’s lead has fluctuated around six-percentage points.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/27	10	5	8	8	5	3	7	6.7
10/28	7	5	8	7	4	4	7	6.0
10/29	7	3	7	6	5	3	8	5.5
10/30	7	5	6	5	7	4	8	6.0
10/31	9	4	7	6	7	4	9	6.6
11/1	10	5	7	7	5	5	9	7.0
11/2	9	5	5	7	6	2	11	6.5

Overall, you can see that four out the seven trackers show the difference in the race to be either 5, 6, or 7 percentage points. As predicted, there are some signs of weekend tightening. However, Obama maintains a clear advantage through Saturday polling.

Tracking Poll Update: 3 Days To Go

With less than 72 hours to go until Election Day, Senator Obama holds a commanding seven percentage point lead in the composite of the tracking polls. Going into the weekend (the results released today include interviews through Friday), Obama has recaptured at least half of McCain’s two point surge from last weekend. Most of this movement back toward Obama appears to have occurred after his Wednesday night infomercial. Is McCain in a better position than he was heading into last weekend? Yes, but barely. Overall, heading into the final 72-hour stretch, it’s Obama’s election to lose.

Here’s a look at the trends:

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/24	7	7	7	12	10	4	9	7.8
10/25	8	8	7	12	9	4	9	8.1
10/26	9	8	8	11	5	4	7	7.6
10/27	10	5	8	8	5	3	7	6.7
10/28	7	5	8	7	4	4	7	6.0
10/29	7	3	7	6	5	3	8	5.5
10/30	7	5	6	5	7	4	8	6.0
10/31	9	4	7	6	7	4	9	6.6
11/1	10	5	7	7	5	5	9	7.0

Tracking Poll Update: 5 Days To Go

With five days to go until Election Day, it’s official that McCain is surging. Surging to a five or six percentage point deficit that is. The movement appears to have occurred Saturday, Sunday, and Monday. As you see in red below, Obama was up by about eight percentage points in the sample (10/25) that included last Wednesday, Thursday, and Friday. Next, also in red, you see that Obama was up by six percentage points in an independent sample (10/28) from Saturday, Sunday and Monday. Based on my massaging of the data, the worst day of polling for Obama appears to have been Sunday. Over the course of the campaign, there has been some convincing evidence that Obama polls slightly worse over the weekend when many of his younger supporters are not as likely to be home to answer pollster’s calls. However, since today’s results include Monday, Tuesday, and Wednesday interviews, they should be a fairly accurate gauge of the current state of the race.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/20	9	4	5	8	6	6	9	6.7
10/21	10	4	6	8	8	6	9	7.2
10/22	8	6	5	10	10	4	11	7.6
10/23	6	7	5	10	12	1	11	7.3
10/24	7	7	7	12	10	4	9	7.8
10/25	8	8	7	12	9	4	9	8.1
10/26	9	8	8	11	5	4	7	7.6
10/27	10	5	8	8	5	3	7	6.7
10/28	7	5	8	7	4	4	7	6.0
10/29	7	3	7	6	5	3	8	5.5
10/30	7	5	6	5	7	4	8	6.0

What does this mean? Well, Obama was up by about seven or eight percentage points last week. This week he is up by about five or six percentage points. I think it’s likely that McCain will pick up another percentage point or two this coming Saturday, Sunday, and Monday, which will make the difference three or four percentage points in the final tracking poll numbers. If I had to guess the popular vote, based on current trends, I’d have to go with a four-percentage point Obama victory. However, your guess is as good as mine.

The next update will be Saturday.

Tracking Poll Update: 7 Days To Go

With one week to go until Election Day, the trackers are showing some movement toward Senator McCain. Let’s take a look at the last nine days:

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/20	9	4	5	8	6	6	9	6.7
10/21	10	4	6	8	8	6	9	7.2
10/22	8	6	5	10	10	4	11	7.6
10/23	6	7	5	10	12	1	11	7.3
10/24	7	7	7	12	10	4	9	7.8
10/25	8	8	7	12	9	4	9	8.1
10/26	9	8	8	11	5	4	7	7.6
10/27	10	5	8	8	5	3	7	6.7
10/28	7	5	8	7	4	4	7	6.0

These numbers tell us how important context is. First, look at the movement only in the last four days (numbers in red). From this perspective, it appears that McCain is gaining ground at a steady pace. Next, look at the last nine days as a whole. Here we see that the race has been relatively stable, albeit with an overall .7 percentage point movement toward Senator McCain.

Therefore, what these numbers mean depend largely on your perspective. If you view the results simply in the vacuum of the last four days, then yes, McCain is surging. However, if you view these numbers in the context of the last nine days, the race is relatively stable. Statistical training teaches me to err toward the latter, unless shown otherwise. This is because you probably need at least a week of data in order to truly flush out any noise.

Another reason not to make any assumptions quite yet is that the state polling results released today show a steady race and/or a slight movement toward Obama.

In other words, if you want to know whether this movement toward McCain in the composite is real movement, check back in a couple days.

Tracking Poll Update: 9 Days To Go

With nine days to go until Election Day, the race remains relatively stable. Overall,  there is not much change from yesterday’s composite. Today, McCain improved in three of the trackers, Obama improved in two of the trackers, and two of the trackers remained unchanged. Any day that does not see significant movement toward McCain is a victory for Obama. As we mentioned yesterday, McCain must gain, on average, around .75 percentage points a day in order to draw even. Here is a look at today’s numbers and recent trends.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/20	9	4	5	8	6	6	9	6.7
10/21	10	4	6	8	8	6	9	7.2
10/22	8	6	5	10	10	4	11	7.6
10/23	6	7	5	10	12	1	11	7.3
10/24	7	7	7	12	10	4	9	7.8
10/25	8	8	7	12	9	4	9	8.1
10/26	9	8	8	11	5	4	7	7.6

We’ll do updates again on Tuesday, Thursday, Saturday, Sunday, and Monday.

Tracking Poll Update: 10 Days To Go

With 10 days to go until Election Day, Senator Obama maintains a significant advantage in the polls. Not only is Senator McCain not gaining ground, but he’s actually in worse shape today than he was even a few days ago. McCain must gain on average .75 points a day to have any shot. In comparison, Obama gained about .15 points a day from mid-September to mid-October during his rise in the polls. In addition, voters are much more committed than they were even a few weeks ago. Rasmussen’s tracker shows that Obama has “solid support” from 48% of the electorate compared to McCain’s 40%. Three weeks ago, Obama had roughly the same bottom line lead in Rasmussen’s tracker, but only had the solid support of 43% of the electorate. In other words, McCain now faces a scenario where he must win the support of virtually all truly undecided voters to have any shot. Good luck.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/20	9	4	5	8	6	6	9	6.7
10/21	10	4	6	8	8	6	9	7.2
10/22	8	6	5	10	10	4	11	7.6
10/23	6	7	5	10	12	1	11	7.3
10/24	7	7	7	12	10	4	9	7.8
10/25	8	8	7	12	9	4	9	8.1

Tracking Poll Update: 7’s are Wild

Not only does Barack Obama continue to have a lead of around seven percentage points,  but there are now seven trackers included in our composite.  The latest tracker comes from ABC/The Washington Post, which started reporting a three-day average on Monday.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	ABC	7-Poll Adjusted Average
10/20	9	4	5	8	6	6	9	6.7
10/21	10	4	6	8	8	6	9	7.2
10/22	8	6	5	10	10	4	11	7.6
10/23	6	7	5	10	12	1	11	7.3

It’s important to note how unstable Zogby and IBD/TIPP have been recently. Wild swings in only one or two of the trackers (especially in opposite directions) is concerning. In addition, take a look at the following TIPP cross tab:

Age 18-24: McCain 74%-Obama 22%. Yeah, I know that some hippsters are voting for McCain to be ironic, but this is over the top. Even a subgroup from this age group shouldn’t have a margin of error large enough to explain this.  Nate Silver believes that they must be under sampling young voters by a very tight likely voter screen. Either that, or they realized everyone would be talking about them if they reported a result like this (damnit all!) Nonetheless, they’ve secured a spot on The Drudge Report’s front page for the last two days (Zogby had been there before).

With all that said, we’ve decided to keep TIPP/IBD and Zogby in the composite. With seven trackers, their swings are largely mitigated. However, we are considering a slightly different methodology to deal with some of these issues. We’ll have more on that this weekend.

Tracking Poll Update: Obama +7

Election Day is exactly two weeks away, and Senator Obama has a solid seven point advantage in the composite of the tracking polls. The mainstream media is trying desperately to push a “McCain surge” narrative (close race = ratings). However, the raw numbers don’t show much of a narrowing at all. In fact, the race looks very stable between Obama +6 and Obama +7 over the past week to ten days. Yes, that’s off Obama’s high of +8-9 points of two weeks ago, but a solid lead nonetheless.

Date	Gallup	Rasmussen	Hotline	Research 2000	Zogby	IBD/TIPP	6-Poll Adjusted Average
10/13	10	5	6	12	4	2	6.7
10/14	9	5	6	11	6	3	6.7
10/15	7	5	8	11	4	3	6.3
10/16	6	4	8	11	5	4	6.1
10/17	6	4	10	10	5	5	6.4
10/18	4	5	7	7	4	7	5.4
10/19	7	6	7	7	3	5	6.0
10/20	9	4	5	8	6	6	6.4
10/21	10	4	6	8	8	6	7.0

Next Page »