One Way to Look at the Presidential Polls

by Stuart Rothenberg April 25, 2012 · 8:30 AM EDT

It’s really amazing how some people with years of political experience change their opinions about the political landscape to match the latest poll.

It’s not that poll results shouldn’t affect our understanding of politics or inform us about what people are thinking. It’s that too often people behave as if the most recent poll they encounter has enormous predictive value. At this point in the cycle, it probably doesn’t.

The best example of this myopia is the way that folks on “Morning Joe,” the MSNBC morning show that still sometimes tries to be analytical, have been discussing polls.

For months, most of the regulars on the show beat up on Republican Mitt Romney, buying into the developing narrative that he and his party have dug themselves into such a deep hole that he may not be able to get out.

Then — presto! — a CBS News/New York Times poll comes out showing the presidential race as a dead heat, and the next day the folks at Morning Joe are shaking their heads about how President Barack Obama is in as equally bad shape as Romney and that the election is up for grabs. The regulars on that program aren’t alone in doing this, of course.

In fact, the latest CBS News/New York Times poll, conducted April 13-17, isn’t very different from the previous one, conducted March 7-11, which showed Obama holding a narrow 3-point lead, 46 percent to 43 percent, over Romney. Statistically and analytically, those two results are the same.

Of course, the April CBS News/New York Times survey shows a dramatically different race from CNN’s poll, which has Obama leading Romney, 52 percent to 43 percent. Both use registered voters. It isn’t up to me to decide which one is “right.”

To be honest, I’m not thrilled with the last couple of CBS News/New York Times surveys, which I usually consider one of the more credible polls.

The April CBS News/New York Times poll found Obama’s job ratings as 48 percent approve/43 percent disapprove. A month earlier, Obama’s ratings were a much worse 41 percent approve/47 percent disapprove.

Think about it. Obama’s job approval increased by 7 points during the month, but the general election ballot test got closer. Does that seem likely? Obama’s 41 percent approval in March in the CBS News/New York Time poll was much lower than most other reliable surveys, so that number also seemed odd.

Anyway, not everybody falls into the trap of overgeneralizing from a single survey or of comparing surveys conducted by different polling firms. Mark Blumenthal, a former partisan pollster turned polling analyst for the Huffington Post, never fails to be smart, measured and modest in his analysis.

And most professional pollsters, Democrat and Republican, are much more cautious than the average casual political observer of reading too much into one poll number or in projecting forward on the basis of a single survey.

It’s no wonder that there is some confusion about the race for the White House, given the wide range of polling results from different outlets.

Gallup once again shows Romney ahead, most recently by 3 points (but a couple of days earlier by 5 points), while Fox has him up by 2 points. The CBS News/New York Times poll has the race even, but Quinnipiac University and Pew have Obama up by 4 points, NBC News/Wall Street Journal has the president’s margin at 6 points — the same margin as its late February/early March survey — and CNN has Obama ahead by 9 points (down from 11 points less than a month ago).

According to Pollster.com, all surveyed registered voters. While most are within the “margin of error,” the results lead to very different conclusions. Some have Romney ahead narrowly, while others have Obama with a substantial lead.

Given the range of results, not all of these polls can be conveying the state of the race accurately. Two recent Fox News polls in the swing states of Ohio and Florida might be able to help — if, of course, you believe they present an accurate snapshot of the contests in their states. Conversely, if the national numbers are right, then the state numbers can’t possibly be accurate.

The April 15-17 Anderson Robbins Research (D)/Shaw & Company Research (R) poll for Fox News shows Obama leading Mitt Romney by 6 points (45 percent to 39 percent) in the Buckeye State.

If those numbers are correct, then the Gallup and national Fox numbers must be wrong. In addition, if the state survey accurately reflects the state of the presidential contest, then the CBS News/New York Times numbers can’t be right either.

Ohio is a swing state, and it is unlikely that Romney will win nationally but lose the Buckeye State by 6 points. In fact, he’s likely to do better in Ohio than he is nationally. After all, Obama won Ohio in 2008 by about 4.5 points while he was winning nationally by just more than 7 points.

The Fox News state survey in Florida, conducted by the same two companies at the same time as they were polling in Ohio, shows Obama leading Romney by 2 points in Florida, 45 percent to 43 percent. Again, the national Gallup and Fox News numbers and the Fox Florida numbers can’t both be right.

Obama carried Florida four years ago by less than 3 percentage points while he was winning nationally by more than 7 points. He will surely underperform in Florida again later this year, so if he wins Florida, he certainly will win nationally (by a larger margin).

My guess (and it’s purely a guess and therefore not worth much) is that Obama holds a narrow lead over Romney, probably in the low-to-middle single digits. That lead could well shrink during the next couple of months, but even if it doesn’t, the general election is likely to be close if jobs and gas prices continue to be problems for the White House.

In any case, my advice is clear: Don’t treat any survey as if it has a monopoly on the truth.