Resounding silence on a poll gone wrong

Posted by mediadog

What lessons have Maine's news outlets learned from the voters' veto of the gay marriage bill? Nearly three weeks after the referendum it's difficult to find evidence that they've learned anything.

Consider this example: In the week just prior to the election, the state's largest newspapers front-paged a poll that showed the "No on 1" (against repeal of the pro-gay law) ahead by 11 points.

That margin was so big that many readers may have decided that the result was a foregone conclusion. People have considerable confidence in polls so this one had the potential to discourage many "Yes on 1" (in favor of repeal) voters.

And because both of the state's largest newspapers had editorialized in favor of "No on 1," some folks could be forgiven for believing that discouraging "Yes" votes was the actually the media's intent. (A large segment of the reading public often suspects media of maliciousness when incompetence is the real culprit.)

In any case, how did this poll of 401 voters turn out so badly? It was the work of Pan Atlantic SMS, a Portland-based firm that has had some polling success in the state. Its pollsters were right about Obama over McCain and Collins over Allen, last year but those were easily predictable races.

Not so the gay marriage question. An early poll done in April by Pan Atlantic called the race a dead heat with the "No on 1" side ahead by just a couple of points. Except for predicting a "No" win, this was pretty much in line with polls done just before the referendum by other firms.

These included a telephone poll of 600 Mainers by the far-left Daily Kos website in September. Surprise of surprises, it predicted a win for the "Yes" side, 48% to 46%. Yipes, the lefties at the Kos didn't get the margin quite right, but they did make the right call.

After that, a telephone poll of 1,000 voters by Public Policy Polling of North Carolina, released on Oct. 20, predicted a dead heat , 48 % "Yes" to 48 % percent "No."

So what happened to Pan Atlantic's Oct. 14 poll -- the one most prominently displayed in the papers and quoted on the air -- and its final "tracking" poll released Oct. 26? Both predicted an astounding 52% to 43% win for the "No" side.

Furthermore, Pan Atlantic, which is run by Patrick and Victoria Murphy -- both of whom are prominent Maine Democrats and major donors to liberal candidates and causes -- said its poll had a margin of error of 4.9% and 95% confidence level.

If that's the case, how could its poll possibly have forecast a win for "No on 1" by 11 points when "Yes" actually won by about 6 points -- and 32,000 votes? That's a huge miss for an outfit like Pan Atlantic that has not been not shy about trumpeting its capabilities.

But several weeks have gone by since the referendum, and it's impossible to find any explanation by either Pan Atlantic or the media about why this poll was so wrong. The pollsters have been silent, which is understandable given the embarrassing magnitude of the error, but the media have been quiet too. The Press Herald, for example, has not offered a mea culpa for its over-display of the poll results.

The failure of anybody to provide an explanation -- or even an excuse -- surely provides ammunition for those who suspect manipulation and express distrust. The fiasco also should be a caution to the media about prominently displaying poll results on Page 1 right before an election.

Polling can work, but it is far from an exact science and when it's wrong the media gets blamed. In this case, rightly so.