Monday, May 27, 2013

ROUNDHOUSE ROUNDUP: Crunching the Poll Numbers

A version of this was published in The Santa Fe New Mexican 
May 26, 2013

Is New Mexico suddenly getting more conservative?

You might think so based on recent polls from KOB-TV, performed by the SurveyUSA firm. But based on a conversation with one New Mexico pollster who looked at those numbers, the conservative trend looks to be a tad overstated.

Early in the week, KOB released a poll on the Albuquerque mayoral race that showed Republican incumbent Richard Berry ahead of Democratic challenger Pete Dinelli by a whopping 42 percentage points. I joked with a friend that the poll oversampled pendejos — Dinelli’s word for Democrats who vote for Republicans.

The next day, a statewide poll by SurveyUSA, also commissioned by KOB, showed Gov. Susana Martinez with a 66 percent approval rating. The poll showed that 70 percent — no, that’s not a typo — of women surveyed were positive about the governor, as were 64 percent of independents and 44 percent of Democrats.

I’m not at all surprised that the governor’s approval rating is more than 60 percent. Most polls put her around that figure. But 66 percent is the highest number I’ve seen for Martinez.

But what really stunned me was a third poll result released by KOB later in the week. According to this one, 51 percent of voters say same-sex marriage should be prohibited, while 44 percent said they support legalization of gay marriage. Previous polls have shown that a plurality of New Mexico voters favor marriage equality.

One part of this poll result that did ring true was that young people support same-sex marriage significantly more than the general population — 63 percent of young people support the idea, according to the SurveyUSA poll.

I asked Albuquerque pollster Brian Sanderoff to look at the poll numbers. The main thing that struck him was the percentage of Democrats vs. Republicans. According to SurveyUSA’s cross-tabs, 39 percent of those polled were Republicans, while 42 percent were Democrats — a 3 percent difference. But according to voter registration figures, the breakdown statewide is GOP 31 percent, Dems at 47 percent — a 16 percent difference.

Sanderoff said even though registered Republicans are better at turning up to the polls than Democrats, the percentages polled by SurveyUSA would skew the results. He said, based on turnout, a more realistic breakdown would be 45 percent Democrats, 33 percent Republicans.

Using that breakdown, Sanderoff crunched the numbers and determined that Martinez’s approval rating would have been 62 percent — still good enough to give Democrats nightmares.

On the same-sex marriage question, Sanderoff said using that 45-33 party affiliation ratio, the number of those wanting to prohibit gay marriage would be 47 percent, while those supporting it would have been 46 percent, a virtual tie.

Sanderoff said that the demographic mix in the Albuquerque poll seemed to be on target, except it’s “very light on cellphones.” Only 10 percent of the interviews in the poll were conducted on cellphones.

“Mayor Berry has a 44-point lead among people reached by landline and a 16-point lead among those reached via smartphone,” Sanderoff said. “Thus, the lead would narrow if they had the right mix of cellphones, however Mayor Berry would still have a very big lead.”

I don’t believe SurveyUSA has any ideological ax to grind. The firm’s results in the past have seemed reasonable.

According to New York Times poll guru Nate Silver, in 2012, SurveyUSA had a plus-0.5 percent bias toward Republicans in the presidential race — in other words, they tended to overestimate GOP performance by half a percentage point. But all but a few major polling companies tended to overestimate Mitt Romney’s numbers. Gallup had a Republican bias of 7.2 percent, Silver reported, while the Democratic-owned Public Policy Polling firm had a 1.6 percent GOP bias.

As far as accuracy goes, SurveyUSA had an average error of 2.2 percent in 2012, which was better than most.