WASHINGTON — Opinion surveys were a huge topic of discussion across the country this election year.

Prediction models like New York Times’ FiveThirtyEight and the Princeton Election Consortium said President Barack Obama was a heavy favorite to win a second term based on the trends of polls in key swing states. Despite drawing heated criticism from traditional political pundits and partisans alike, the models and the polls they were based on were vindicated in a major way Tuesday night.

In Hawaii, the polls were fewer and farther between. Each was a snapshot in time, with the overall trendlines more important than any individual topline number. Put together, they gave readers a pretty good roadmap of what they could expect election night.

The final version of The Civil Beat Poll was conducted Oct. 24 to Oct. 26, after early voting had begun and about two weeks before Election Day. Here’s our postmortem report card for where our survey hit and missed. It’s organized just like your ballot was Tuesday.


Final Civil Beat Poll Snapshot: Barack Obama 61 percent, Mitt Romney 34 percent

What We Wrote: Obama Lead Narrows, But Hawaii Still Very Safe — “Obama had already built a substantial advantage at the time of the survey, leading 66 percent to 31 percent among the third of respondents who said they’d already voted. … Romney was closer, down 59 percent to 36 percent, among those who said they haven’t voted yet but definitely intend to. But those votes are not yet in the bank.”

Actual Election Results: Obama 70 percent, Romney 28 percent

What Happened: Nationally, enthusiasm for Romney dipped in the final days as Obama responded strongly to Superstorm Sandy. Locally, blowouts tend to expand rather than contract close to Election Day as supporters of the losing candidate lose interest. There was a reduced turnout among conservatives and right-leaning independents who said they hadn’t voted yet but intended to, leaving an electorate that looked more like the Obama supporters who voted early. Even still, Romney was more competitive in Hawaii than John McCain had been in 2008.

U.S. Senate

Final Civil Beat Poll Snapshot: Mazie Hirono 55 percent, Linda Lingle 40 percent

What We Wrote: Hirono’s Double-Digit Lead Over Lingle Holding — “The gap in the new poll was particularly large among the third of respondents who said they’d already voted, with Hirono leading 61 percent to 37 percent versus 52 percent to 42 percent for the two-thirds who said they haven’t yet voted but definitely intended to do so. … That means there’s room for Lingle to narrow the gap slightly. Pro-Lingle groups are spending big money on TV ads for the final days of the race to help make that happen. But she’d need a major turnaround to make the contest competitive.”

Actual Election Results: Hirono 62 percent, Lingle 37 percent

What Happened: Just like in the presidential race, a blowout expanded as the clock ran out. Turnout among would-be Lingle supporters dropped, leaving final election results that mirrored, nearly perfectly, the 61-37 Hirono advantage Civil Beat found among early voters.

1st Congressional District

Final Civil Beat Poll Snapshot: Colleen Hanabusa 54 percent, Charles Djou 43 percent

What We Wrote: Hanabusa Expands Lead Over Djou In Race for Congress — “For Djou, there’s not much time left to turn things around and not much flexibility left in the electorate. … Only 3 percent of voters surveyed were still undecided, and 93 percent of Hanabusa voters said they would definitely back her versus 7 percent who said they were only leaning toward her. ‘He’s hanging in there. He’s hanging tough. He’s got good name recognition. But it definitely does appear that Colleen Hanabusa has stretched it out a little bit over the last poll,’ said Merriman River Group Executive Director Matt Fitch, who partners with Civil Beat on its polls.”

Actual Election Results: Hanabusa 54 percent, Djou 44 percent

What happened: Hanabusa won by nearly the exact amount she led in the final Civil Beat Poll, well within the margin of error.

2nd Congressional District

Final Civil Beat Poll Snapshot: Tulsi Gabbard 73 percent, Kawika Crowley 15 percent

What We Wrote: “In the 2nd Congressional District, Gabbard was on her way to a landslide victory over the little-known Kawika Crowley. That domination included: An 81 percent to 14 percent lead among those who said they’d already voted.”

Actual Election Results: Gabbard 77 percent, Crowley 19 percent

What happened: Gabbard’s 58-point advantage in the survey matched her 58-point win on election night.


Final Civil Beat Poll Snapshot: Ben Cayetano 50 percent, Kirk Caldwell 45 percent

What We Wrote: Cayetano Peaks, Caldwell Surges — “Kirk Caldwell has continued his steady surge in the Honolulu mayor’s race, closing to within striking distance of Ben Cayetano just two weeks before Election Day … the gap was much narrower among the 70 percent of respondents who said they hadn’t yet voted but definitely will. Caldwell was down just 48 percent to 46 percent. The race is also slightly tighter among those who say they’re locked in for their candidate, with Cayetano having more of the softer support among those who are leaning one way or another. A lot of it is going to come down to turnout,’ said Matt Fitch, executive director of Merriman River Group, which partners with Civil Beat on its polls. ‘We know that Caldwell does better among the likeliest voters, and Cayetano’s best voting footprint is with a large turnout of casual voters who only vote sometimes.’ … 45 percent is the high water mark this year for Caldwell, the 60-year-old former acting mayor. He’s gained in every survey of the race we’ve done.”

Actual Election Results: Caldwell 53 percent, Cayetano 45 percent

What happened: As in the primary, Cayetano underperformed his polling numbers — or Civil Beat oversampled his supporters.

“Basically what we saw there was exactly what we saw in the primary,” Fitch told Civil Beat Wednesday. “Cayetano underperforms his polls because he has less resources. Caldwell does best with the likely voters and Cayetano needs a big turnout in order to do his best, and he didn’t get a big turnout. The turnout was down.”

Specifically, the turnout Tuesday was down from 2008, both in terms of total voters and percentage of registered voters.

On the surface, the Hawaii Poll sponsored by the Honolulu Star-Advertiser and Hawaii News Now was closer to the final result, showing an 11-point Caldwell lead three weeks before Election Day. But a lead that big that early would presumably mean, in absence of major changes in the dynamic of the contest, a trendline resulting in a 15- or 20-point Caldwell win, not a single-digit victory.

Civil Beat’s topline number was off from final results, but the trendline of Caldwell gaining ground on Cayetano in every survey this year pointed astute readers to what was a predictable single-digit Caldwell win.

Mainland V. Local Voice

After the primary election in August, when The Civil Beat Poll foretold a close Democratic showdown between Hirono and Ed Case that didn’t materialize, we explored the possibility that the voice we used in our survey had impacted the responses.

Here’s a summary of that hypothesis:

First, it’s important to know that we use interactive voice response (IVR) technology, also called touch-tone polling or, to critics, “robo-polling.” There are pros and cons to robo-polls, and one of the purported benefits is that using pre-recorded prompts is supposed to eliminate any bias issues created by the human being asking the questions.

In this case, that might not have worked perfectly.

That’s because the voice used in Civil Beat’s poll might have inadvertently created what’s known in the industry as a “social desirability bias” where poll respondents feel pressured to give answers they believe are socially acceptable.

But the bias only comes into play if the survey somehow creates the impression of a socially preferable response. In the case of The Civil Beat Poll, the voice used in all our surveys so far — a Caucasian woman in Ohio — could subconsciously cue respondents to opt for Caucasian candidates, like Case, over Japanese-American candidates like Hirono.

To test that hypothesis, we used two voices — the same Caucasian woman in Ohio and Civil Beat Assistant Editor Sara Lin, who’s of Chinese and Hawaiian ancestry and was born and raised on Oahu — in our September and October surveys.

We split the calls about 50-50 between to voices to see if respondents chose different candidates based on which voice was asking them for their opinion. Here’s what we found:

  • In September, Hirono led 56-38 among those who heard the “local” voice versus 55-40 among those who heard the “mainland” voice. In October, Hirono led 54-40 among those who heard the “local” voice and 55-40 among those who heard the “mainland” voice. That’s essentially no difference.
  • In September, Hanabusa led 54-40 among those who heard the “local” voice versus a 49-43 lead for Djou among those who heard the “mainland voice.” In October, Hanabusa led 49-45 among those who heard the “local” voice and 57-41 among those who heard the “mainland” voice. Both gaps are sizable, but they run in opposite directions, making it impossible to draw any meaningful conclusions.
  • In September, Cayetano led 50-45 among those who heard the “local” voice and 52-38 among those who heard the “mainland” voice. In October, Cayetano led 53-41 among those who heard the “local” voice and Caldwell led 49-47 among those who heard the “mainland” voice. Again, the gaps run in opposite directions, making it impossible to draw any meaningful conclusions.
  • In September, respondents to the “local” voice were opposed to rail at a 50-40 mark and respondents to the “mainland” voice were opposed to rail at a 56-35 mark. In October, “local” voice poll respondents opposed rail 54-37 and “mainland” voice poll respondents opposed rail 49-45. Yet again, the gaps run in opposite directions.
  • In both surveys, respondents to the “mainland” voice were slightly more likely to say they were politically moderate than respondents to the “local” voice.

What does it all mean? The differences between the voices, both within each survey and between the two surveys, most likely reflect the type of random fluctuation one expects in survey research, particularly when dividing a survey sample into smaller subsamples.

The full crosstabs are searchable here:

About the Author