A cautionary tale from 2010 for pollsters and pundits as we near Election Day 2016. As Senate Majority Leader Harry Reid (D-NV) awoke on Election Day 2010, he was considered “a goner,” so said Tom Kludt of Talking Points Memo. As Kludt noted,
“In the dying embers of the 2010 midterms, the Senate majority leader appeared to be on the ropes. Polls from a variety of well-known outlets showed tea party champion Sharron Angle leading him in the final weeks of the U.S. Senate race in Nevada. No less of an authority than Nate Silver, who correctly predicted 34 of the 36 Senate races that year, pegged Angle as the favorite in his final forecast. But on Election Day, the Nevada polls were proven wrong. Reid defeated Angle by nearly 6 points…”
So what happened? According to Senator Reid himself, “I would not be the majority leader in the United States Senate today, but for the Hispanics in Nevada.” An impressive turnout operation, powered by Culinary 226, combined with inaccurate Latino polling, badly underestimated Reid’s strength.
How did the pollsters and pundits miss the fact that Reid was poised for a resounding victory? Here’s a recent analysis by Gabriel Sanchez, a principal of Latino Decisions:
16 pre-election polls of the Reid/Angle race had Angle up by 3 percentage points total, on average, while estimating that Latino support for Senator Reid was between 55%-70%.
According to 2010 election eve polling of Latino voters in Nevada conducted by Latino Decisions, Reid won Latino voters by a whopping 90%-8% margin.
Had the pre-election polls accurately estimated a 90-10 split among Latinos, they would have found Reid leading Angle by 3 to 5 percentage points overall (close to the final, actual tally).
The Latino Decisions’ election eve poll result of a 90%-8% margin was subsequently corroborated by regression analysis of official precinct level vote choice.
Unfortunately, it seems the mistakes of 2010 are being replicated in 2016.
Let’s start in Nevada. A NBC/WSJ poll released yesterday found that Donald Trump and Hillary Clinton are tied overall in the state at 43% and that GOP Rep. Joe Heck is leading Democratic nominee Catherine Cortez Masto by a 49%-42% margin. However, a look inside at the Latino internals raises serious doubts. According to the NBC/WSJ poll: 36% of Latino voters in Nevada are voting for Donald Trump – a whopping fifteen percentage point swing among Latinos in favor of Trump over the past month. In addition, the poll found Heck leads over Cortez Masto among Latinos by a 43%-42% margin.
Really? Trump on the move with Latinos in Nevada? Heck out-polling a candidate who would be the first Latina ever elected to the U.S. Senate?
As Nevada’s leading political journalist Jon Ralston said on MSNBC in response to the Latino findings in the poll, the findings don’t “make a lot of sense to me.” Later, on Twitter, Ralston called the Latino findings in the poll “highly unlikely.” In contrast to the NBC/WSJ poll, a recent Latino Decisions poll of Latino voters in Nevada – a poll conducted by an experienced firm with bilingual callers who are able to speak to voters in the language they prefer – found that with Latino voters Clinton leads Trump 72-17% and that Cortez Masto leads Heck by a 54-24% margin.
Another example from yesterday is the Bloomberg Politics poll of Florida. Bloomberg found Clinton with a lead over Trump with Latino voters of 51-36%. Latino Decisions’ most recent Florida poll found Clinton leading by a 63-23% margin. Latino Decisions has not been shy about pointing out potential methodological flaws in the Bloomberg poll here and here. In addition, Florida political observer and expert Steve Schale criticized the poll’s overall design, writing, “They have the electorate at +3 Republican. 42-39. I don’t even know even the most optimistic GOP operative who agrees with this model.”
As to why all of this matters, Frank Wilkinson of Bloomberg yesterday examined the stakes and implications of getting Latino polling right.
“Surveys of Hispanic voters reflect vast disparities … In an analysis of earlier poll disparities, Emory University political scientist Alan Abramowitz said that a 28-point difference in the Hispanic vote would translate into a 2 to 3 point difference in the national vote. The difference between the October Economist/YouGov and Latino Decisions polls is 43 points, or perhaps 4 points in a national election, larger than Obama’s margin of victory in 2012.
Abramowitz, in an e-mail, pointed out that Latino Decisions has an ‘excellent’ track record from 2012 and that its large margin for Clinton has been supported by other polls that also used large samples of Hispanics, conducted interviews in both Spanish and English and employed other techniques designed to obtain more representative voter samples.
…A presentation by Latino Decisions pollster Gabriel Sanchez pointed out that 2010 Senate polls in Nevada and Colorado, both of which have large Hispanic populations, generally underestimated Democratic support. (Democrats won both contests.) Many polls similarly undercounted Obama’s Hispanic support in 2012.
…The trajectory of the fast-growing Hispanic vote will help determine not just party power but party dynamics for years … If the Latino Decisions end of the polling spectrum is correct, however, the 2017 Republican autopsy promises to be a doozy.”
According to Frank Sharry, Executive Director of America’s Voice, “According to the Latino vote experts at NALEO, 13.1 million Latinos are expected to vote this year, a 17% increase in turnout and a nearly 9% increase in the Latino share of the vote. Latinos will be a huge factor in swing states such as Florida, Nevada, Colorado, Arizona, North Carolina, Virginia and, yes, Texas. It is high time that the better pollsters in the country stop treating this community as an afterthought, with shoddy methodology and inaccurate findings. It’s time to make adjustments and get the Latino vote right in order to accurately represent a fast-growing community and accurately represent the state of the overall race.”
Check out a recent panel discussion from Latino Decisions, NCLR, Catalist and Professor Alan Abramowitz on what constitutes a good vs. bad Latino poll and why both many pre-election polls and the national exit polls miss the mark when it comes to Latino voters.