The Markey-Gomez race by the numbers

It’s a week after Cong. Edward M. Markey defeated businessman Gabriel E. Gomez to succeed former Sen. John F. Kerry and a good time to examine the pre-election polls and see how accurate they were.

Polls, of course, are those statistical snapshots that every campaign uses to shape strategy and that pundits and pols use to track candidates’ progress and predict electoral outcomes.

According to Brent Benson, a political blogger who analyzes polling and public policy data in Massachusetts, the polls got it right: they basically predicted a 10 percent win for Markey, and that’s what he got.

Here’s how the polls worked:

Poll results can vary widely depending on who is being polled, how big the sample is, when the poll is taken and other factors.

Respondents are generally chosen in the same proportion as there are registered Democrats, Republicans and independents in the state.

Sometimes “any” registered voters are polled and other times it’s “likely” voters, those voters pollsters have identified as most likely to show up on election day, as determined by a respondent’s answers about his or her voting history.

Usually, the larger the sample the more accurate the poll, and the closer you get to the election, the better chance there is that the polling will match the actual election results.

There were 17 independent polls taken of the Markey-Gomez contest, with the first on May 2 and the final poll published on June 22, just three days before the special election.

All of the polls showed Markey winning.

The smallest margins were in the two earliest polls taken at the beginning of May, seven weeks before the election. They showed Markey ahead by only six and four points, respectively, and they also showed large numbers of undecided voters.

The biggest margin was a 20 point lead for Markey. But even though that Boston Herald poll was done just a week before the election, there were only 312 likely voters queried.  The small sample clearly skewed the results, since Markey certainly did not experience a groundswell of support and then turn around and lose 10 points in those final days.

Late trend for Gomez?

Several pundits, including the Spotlight, took notice of two polls taken back-to-back just two weeks before the election. They showed Markey with a mere seven point lead over Gomez, and together, these polls seemed to indicate a late trend that Gomez was finally getting some momentum.

But this trend, if it existed at all, soon disappeared and Markey’s lead was in the double digits in all but one subsequent poll.

The most accurate polls, not surprisingly, were the last ones to come out before election day. The Emerson College Polling Society, with a huge sample of 1,422 registered voters, and the Suffolk University poll, with 500 likely voters, both predicted the final spread, Markey by 10 points.

And this brings us back to Benson’s contention that the polls accurately predicted Markey’s 10 point win. He paid less attention to the 17 individual polls and instead applied a formula to the polls as a group and derived an average of all them combined.

So, in spite of the fact that the individual polls varied from a four- to a 20-point Markey advantage, the average shows he was leading by about 10 points throughout the campaign, with never more than a five percent variation above or below that average.

Pundits, polls and people who like to follow politics should all heed blogger Benson and remember to look at the big picture before touting any particular poll or two as showing a trend or predicting a winner.

Posted by Bruce Garlow on July 5, 2013. Filed under Berkshire Beacon Hill Spotlight,Columns,Opinion. You can follow any responses to this entry through the RSS 2.0. You can leave a response or trackback to this entry