Flawed poll tainted our final-days election coverage
Few things are more nausea-inducing for an editor than discovering the lead story on your front page has factual errors. Much worse is learning that its central premise is wrong.
It doesn't happen very often, but it did in the lead-up to Election Day twice. I kept a garbage can close by.
When the numbers from our pollster Florida-based Mason-Dixon landed in my email box the Thursday before the election, I didn't expect many surprises. Hatch, Herbert in cakewalks. The 2nd Congressional District, predictable.
We anticipated two races could be close: Mia Love/Jim Matheson in the 4th Congressional District and Mark Crockett/Ben McAdams for Salt Lake County mayor.
The surprises: Our pollster had Love 12 percentage points ahead of Matheson, and Crockett leading McAdams by 10. A common thought in our newsroom: Is this the Mitt Romney effect?
Those numbers led to two big stories atop two consecutive front pages. To discover, days after publication, that the news was borne from seriously flawed data was humiliating.
We trusted our pollster. We've been running his surveys for almost a decade. If you look at all the races he has called for us, his track record is solid. This time was different.
As pollsters gather interviews, they must cross-check themselves to ensure their sample represents the greater whole. You want genders fairly represented. The sample should reflect political affiliation and religious makeup. It should include a wide geographical spread and diverse neighborhoods.
In Salt Lake County, for instance, that means polling in liberal Salt Lake City, more conservative Sandy/Draper, in West Valley City, Murray, Magna, Millcreek. Latinos, the county's largest ethnic minority, should be clearly represented. And a Salt Lake County poll should reflect the fact that there are more Democrats and independents willing to support a Democrat when compared with the rest of Utah. After all, Barack Obama carried the county in 2008.
That's where our poll failed.
Our numbers were met with skepticism right off the bat from the campaigns. Not unusual, as candidates don't like to hear they are losing. And their own polling can be flawed.
After our Love/Matheson and McAdams/Crockett surveys had been published, Utah pollster Dan Jones released his numbers for KSL and the Deseret News, showing Love with a slim lead and McAdams with a narrow edge. McAdams had also released his internal polling, showing a close race. We included the candidates' reservations in our stories and even added numbers from the competing polls in the county mayor story.
With skepticism building, political reporter Robert Gehrke tried to contact Brad Coker of Mason-Dixon.
When Coker responded via email late Sunday, he acknowledged undercounting Democrats in Salt Lake County, a mistake he attributed to the stress of a deadline when pressed on the issue by county reporter Mike Gorrell.
The situation left us with two bad options: Let the election play out, or run a six-column story on our Election Day front page correcting the numbers. It wasn't hard to decide. Gorrell's story led Tuesday's A-1.
Coker said his mistake significantly affected the McAdams/Crockett poll but was less important in predicting Matheson/Love.
Tuesday's results strongly suggest otherwise.
This whole episode leaves an acrid taste. The nationwide focus on polling in the last weeks of Election 2012 created a dull pain that made many voters wish the whole thing would just end.
In our newsroom, that general pain, and our specific insult, is leading to serious soul-searching and discussion on how to avoid such embarrassment in the future. Questions we must ask and answer: What do we really want to accomplish with polling? Is running the horse-race numbers so close to an election a good idea? What does any of this add to the body of knowledge?
One thing for sure: Our next poll, whoever does it, will pay attention to the specifics of Utah's political and demographic landscape and will reflect that which is unique to Utah and its various counties and cities. It will definitely register the willingness of Salt Lake County voters to support candidates from both parties.
Terry Orme is a managing editor at The Tribune. Reach him at firstname.lastname@example.org.