wpostServer: http://css.washingtonpost.com/wpost2

Most Read: Opinions

direct signup

Today’s Opinions poll

Should Congress deal with the immigration crisis -- tens of thousands of unaccompanied minors at the border -- before its August recess?

Submit
Next
Review your answers and share
Post Partisan
Posted at 03:29 PM ET, 04/23/2012

Be ready for goofy polls

You’ve heard this from me before, and you’re going to hear it from me again: we’re going to see more oddball polling results this year. Don’t be fooled by them.

Gallup recently ran a quite revealing chart of presidential approval ratings in the 13th quarter in office for all postwar presidents who initially reached office by election. What’s the most interesting number in the chart? The number of observations. Gallup ran five – five! – polls asking for approval of Bill Clinton during January through March, 1996. Nine approval polls for George W. Bush in the first three months of 2004. Richard Nixon? Only three. But for the current president, Gallup is running a daily tracking poll, and recorded 88 readings in January through March of this year.

Notice that this creates two kinds of traps for those interpreting the polls. On the one hand, any conclusions about how Obama compares, say, to Nixon in the months leading up to re-election may be misleading because we don’t really know how Nixon was doing. The infrequent polls may have missed fairly substantial real changes within that three-month period; we have no way of knowing, at least just from Gallup records. And if one of those polls was wrong for some reason – and statistics basics say that sometimes polls will be wrong just from luck – then the average that we’re reading could be way off.

On the other hand, that same logic means that with Gallup running a daily track, and other pollsters either running daily tracks or frequent polling, that we’re going to see quite a few bad numbers. Gallup, for example, has Obama’s approval spiking up to 50% today, but the odds are good that we’re just seeing a statistical blip, and his slump down to 43% late last month was also a meaningless blip.

Once again: look at the poll-of-polls averages. Mark Blumenthal’s Pollster trend line for approval sits at 47% and has hardly budget for weeks; the average over at Real Clear Politics is just barely higher.

And of course what’s true for approval is also true of head-to-head polling. Gallup had Romney up five last week; they have Obama up three now. Rassmussen has gone the other way, basically, over the last few weeks. There’s really nothing going on this month (beyond Romney perhaps consolidating some GOP voters who previously didn’t realize they were going to support him).

The bottom line is that all the data help us know more about what’s going on, as long as we use it well – which means focusing on the averages, and not individual, anomalous readings. Remind yourself: we’re expecting a lot of those anomalies. In both directions. Just get ready to ignore them.

By  |  03:29 PM ET, 04/23/2012

 
Read what others are saying
     

    © 2011 The Washington Post Company