Don't sweat individual polls

by: Chris Bowers

Tue Jun 15, 2010 at 16:15

The latest poll that is supposed to cause Democratic panic was released today by NPR.  This poll showed Republicans ahead by 8% in the 70 most heavily targeted congressional districts in the 2010 elections.  As with the Gallup poll from two weeks ago that showed Democrats down 6% in the generic ballot, the largest Republican lead ever in that poll, there is only one appropriate way to react to this poll: total, uncontrolled PANIC!

Or not.  The appropriate way to react to polls like these is, as always, to look at the broad context of all polling.  Here is why:

  • There is no One, True Poll There is a tendency among people who follow elections to look for The One, True Pollster that is more accurate than any others.  This is folly.  Every poll has error. No poll that is always right, nor will there ever be a poll that is always right.  The goal it to minimize the error as much as possible.

  • Polling averages more accurate than any poll: Simple polling averages of all available polls have less than 50% the error of individual polls. Across 144 statewiode, general election campaigns from 2004 to 2010, simple polling averages have produced a total error of just under 1.8% from the final projected margin to the final result.  By way of comparison, the average total error from individual polls to the final result was just over 3.6% for the same time period (the NCPP report linked in the previous sentence halves the total error, and refers to it as "candidate error.").  No pollster is ever going to beat the average of  all other polls by 50% across that many campaigns.  Poll averaging works.

  • The broad view of generic congressional ballot polling shows a toss-up.  Here are the 24 generic congressional ballot polls conducted over thepast month and released to public (Zogby Internet polls excluded).  When they are sorted by margin, the tightness of the 2010 campaign is clearly revealed:

    Generic Congressional Ballot, June 15th: GOP +0.9%
    Polls can be found at and

    Rasmussen's pro-GOP outliers aside, this is an extremely tight campaign. Of the 20 non-Rasmussen polls, 11 show the campaign either in a dead heat or with one party up by a single point. Only 4 of the 20 non-Rasmussen polls show a margin of more than 3%.  That is a pure toss-up.

  • Little has changed in five months: Since mid-January, the 30-day, simple mean polling average in the generic congressional ballot has shown little variance.  The high point for Democrats was a lead of 0.7%, and the high point for Republicans was a lead of 1.7%.  The current Republican lead of 0.9% is thus not only not cause for panic, but it is hardly newsworthy at all. The electoral environment remaains in a five-month period of near statis.
I understand entirely that writing blog posts about polls is both tempting and easy.  Not only are polls news, but there is an attractive, empirical side to them that can elide the oft-annoying spin tsunami of the political news cycle.  However, when looking at polls, it is always best to look at the entire picture.  While doing so makes for a less dramatic view of the electoral environment, it does make for a more accurate one.
Chris Bowers :: Don't sweat individual polls

Tags: , , , (All Tags)
Print Friendly View Send As Email

Too logical. (4.00 / 1)
Many political professionals have money and careers riding on coming to the rescue of Dems.  Thus, they push panic and their own solutions.    


Open Left Campaigns



Advanced Search

Powered by: SoapBlox