Remember that big drop in approval numbers Barack Obama suffered over the weekend? You’re remembering a phantom.
The poll that got all the chatter going about Obama’s alleged drop was published on Monday by the New York Times. It found Obama’s approval rating at 41 percent, in polling taken from March 7-11.
Here, via Pollster, is the wrap up of all the polling taken roughly during the same time period as the Times poll:
CBS/NYT — March 7-11 — 41%
Bloomberg — March 8-11 — 48%
Ipsos/Reuters — March 8-11 — 50%
ABC/WaPo — March 7-10 — 46%
Pew — March 7-11 — 50%
Gallup — March 10-12 — 47%
Rasmussen — March 11-13 — 47%
The Times poll is an outlier. The overall Pollster trend is now at around 47 percent — put all the polling together and the trend is either slightly up or slightly down over the last month, depending on how one reads the data. (The key variable is whether you want your trend line to overreact or underreact to each poll — whether you place more or less emphasis on the most recent polls — something you can change by adjusting Pollster's "smoothing" function.)
Either way, there's no getting around the fact that the big Obama polling drop last week never happened. Never happened.
One of the marvels of mathematics is that you really can use a startlingly small sample size and learn what the whole nation would say if asked the same question. That’s only true within limits. And one of those limits is that polls have a margin of error. Another limit is that one out of twenty polls will land outside of that margin of error. Which means you’re inevitably going to get some duds.
Let me demonstrate it another way. Gallup has been doing a daily tracking poll for Obama’s approval ratings ever since he took office. I count 10 times that the Gallup track has bounced either up or down by at least 5 points and then returned back to where it was within no more than 10 days. These very likely are random fluctuations.
So be careful about putting very much trust in any single poll, no matter how methodologically sound it might be. My advice is to use Pollster or another poll-of-polls average. Be skeptical, too, of stories about “rising” or “falling” poll numbers. News organizations tend to compare their own polls to each other, rather than comparing each poll to an average of all recent polls, which would give us a clearer picture of whether there have been any meaningful changes.
We shouldn’t automatically dismiss the evidence found in a survey that seems very different from what we expect. But a little patience and a little prudence go a long way in understanding what the polls are really telling us.