Articles and Analysis


Update: Pennsylvania Results by Race and Education

Topics: 2008 , Barack Obama , Doug Schwartz , Gallup , Hillary Clinton , Pollster , Quinnipiac

I got a bit sidetracked yesterday, so apologies for the delay in posting this additional data from the Quinnipiac University Poll of Pennsylvania released yesterday. In a post last week, I noted that Barack Obama's modest gains on the last Quinnipiac poll had come mostly from college educated white voters, essentially matching the levels he had reached in mid-February on previous Quinnipiac polls. That trend continues on this latest survey.

Given the wide variation in recent Pennsylvania results, why focus on data from just one pollster? Well, first, the Quinnipiac University poll has been conducting very large samples of likely Democratic primary voters lately (ranging from 1,304 to 1,549 interviews on the last three surveys) that reduces the sampling error within smaller subgroups. Second, Quinnipiac has long and extensive experience in Pennsylvania. The pollsters I know who have worked in the state hold the poll in high regard.

Third, perhaps indicative of the first two factors, the current Quinnipiac results -- showing Hillary Clinton leading Barack Obama by six points (50% to 44%) -- track fairly closely with our overall trend estimate based on all polls (51% to 43%). Finally, Quinnipiac is one of the only polls that regularly tabulates its results by education and Doug Schwartz, the director of the poll, has been kind enough to share additional tabulations that do not appear in their public release.

So on to the data. With this latest update, Obama now leads Clinton among college educated white voters by 12 points (54% to 42%), representing a net 19-point reversal since mid March. However, while Obama's performance among white voters without a college degree is slightly better than in mid-March, Clinton continues to lead among non-college educated whites by a two-to-one margin (62% to 31%) -- roughly the same margin as in mid-February.


With this survey, Doug Schwartz has also shared data for the Clinton and Obama favorable ratings by these same subgroups. As the following table shows, Barack Obama's favorable rating has increased modestly among all four subgroups since a dip in mid-March in the midst of the first wave of stories about the Jeremiah Wright controversy. The improvement follows Obama's speech on Wright and race, a bus tour of rural Pennsylvania and his first wave of television advertising in the state. Note that while his rating among non-college whites is weaker than among other voters, it is still mostly positive (55% favorable, 26% unfavorable).


Hillary Clinton's favorable rating has been mostly stable among white voters over the last month. Note that while Clinton's favorable rating among college educated white voters (69%) is lower than Obama's (78%), most of these voters rate both candidates favorably, which accounts for the relative volatility of their vote preference over the last few months. Many of these voters face a tough choice between two candidates they like. Also note that Hillary Clinton's unfavorable rating among African Americans (34%) is now greater than Obama's unfavorable rating among non-college educated whites (26%).


Thus, while these new results show modest progress for Obama in recent weeks, they also demonstrate why Clinton remains the favorite to win Pennsylvania and why Obama faces a big challenge gaining additional ground. Yes, Obama has gained significantly among college educated white voters, but that subgroup is relatively small (roughly a quarter of the Quinnipiac samples). Also, Obama's current margin in Pennsylvania among college educated whites not only exceeds his exit poll performance in Ohio (where Clinton won the subgroup, 52% to 45%) but also in most other states (he did better only in Illinois, Missouri, Virginia and Wisconsin). Further progress over the next two weeks may be tougher, given a debate and a greater television advertising presence by the Clinton campaign.

To overtake Clinton in Pennsylvania, Obama will need to put a bigger dent in her lead among the non-college educated white voters that are likely to be just over half of the electorate on April 22. If he merely holds his current leads among college whites and African Americans, Obama would need roughly 38-40% of non-college educated whites to win.

UPDATE - Gallup has a new analysis (and video report) out today that shows the same pattern by race and education since mid-March in their national Gallup Daily tracking survey:

The results show little change in the pattern of support among white Democrats with high school educations or less. This group supported Clinton over Obama by 33 points in the March 9-15 period, and supports her by a 30-point margin now. This suggests that as race arguably became a bigger factor in the campaign, less-educated white Democrats were not affected.

The biggest change, in fact, came about among those with postgraduate educations, who went from an 8-point margin for Obama in the March sample to a 29-point margin now. In other words, Obama's increase in support overall among whites during this period looks like it occurred for the most part among those with college and in particular postgraduate educations.



Not sure if it's on your end or mine, but the "fav/unfav" graphs don't show up.


Mark Blumenthal:

The glitch was on our end and my doing -- sorry about that. Should be fixed now.




How about the AA vote? It seems to me that Obama might still gain some points by increasing his numbers among AAs. Recent trends of previous contests put him close to 90%. This poll is at 75%.



I want to repeat an observation I made on the eve of the NH primary - we're again seeing a remarkably consistent take on Obama's level of support, and more variation in Clinton's - and propose an explanation.

First, the data. For Obama, we've seen a fair amount of consistency (InsiderAdvantage: 42%; PPP: 43%; SV: 42%; Rasmussen: 43%; Quinnipiac: 44%). Even the outlier � SurveyUSA � puts Obama at 38%.

On the other hand, when it comes to Hillary, we see more variation: Insider Advantage: 45%; PPP: 46%; SV: 47%; Rasmussen: 48%; Quinnipiac: 50%; SurveyUSA: 56%. Why the spread? Well, there�s a fairly strong inverse correlation with the number of voters selecting neither candidate (undecided or other) and Hillary�s level of support. IA: 13%; PPP: 11%; SV: 11%; Rasmussen: 9%; Quinnipiac: 6%; SurveyUSA: 6%.

The demographic breakdowns are interesting, and I'm sure they're playing a role in the varied results. But I strongly suspect that the controlling variable here is how hard voters are being pushed to choose one of the two leading candidates. We've seen this before. The polling this cycle, particularly in states without large black populations (and thus little effect from Reverse Bradley/Wilder), has been pretty good at predicting Obama's actual levels of support. In Ohio, your average put him at 43.6%, and he garnered 44.1%. In New Hampshire, it was 36.7% vs. 36.4%. Even in states where "everyone knows the polling was off," like Massachusetts, we find 39.5% vs. 40.8%

So I'd suggest we adopt a new predictive measure based on the polling data. Let's take the synthetic average to be accurate for Obama, and put Hillary's level of support at something like (98%-Obama%). If we did that, I think we'd find that the polling this cycle has actually been remarkably accurate. Or am I totally off base? I'd love your thoughts on this.


Another Mike:

The Gallup analysis showing Obama gaining in support among college educated voters, gives at least some hope that Obama can continue to make progress with this group, even though his college educated vote share in Pennsylvania is already higher than most prior states, including Ohio.



Is college education just a predictor of internet usage? Have there been polls done, which ask voters how often they use the internet, and where they get their news?



Notice in the data that among whites without a college degree 19% are reporting that they haven't heard enough about Obama to have formed a favorable or unfavorable opinion of him, whereas only 5% of them are saying that about Hillary. (Those numbers include a small few percent who refused to answer the question). Obama's campaign is innundating the airwaves in PA this week and next. Some of these people will be converted to Obama in all likelihood.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.