07/16/2014 05:59 pm ET

HUFFPOLLSTER: Do Robopolls Get It Wrong?

photo division via Getty Images

A campaign pollster says it's time for the media to ignore surveys that don't call cell phones. Three new polls tackle immigration. And while Scott Brown trails in New Hampshire, the Iowa Senate race looks close. This is HuffPollster for Wednesday, July 16, 2014.

SHOULD THE MEDIA IGNORE POLLS WITHOUT CELL PHONE SAMPLES? - Yes, writes pollster Mark Mellman (D): "It’s long past time the press stopped publishing polls that do not sample cellphones. If an enterprising pollster offered a survey that, say, excluded everyone ages 30-44 or those over age 65, no one would pay any attention to it — and rightly so. Yet news outlets regularly publish polls that arbitrarily ignore a similar proportion of the population. A recent study found 39 percent of adults live in households with cellphones only, and no landline….None of this would matter much if wireless-only (or mostly) voters behaved like landliners. They don’t. In 2008, President Obama won wireless-only voters by 23 points, while tying John McCain among those with both landline and mobile service….If one arbitrarily excludes 20 percent or 30 percent of the population, the principle of randomness is violated and there is no reason to believe the survey will prove accurate. That is not to say it will be wrong, but it does mean there is no scientific reason to believe it will be right." [The Hill]

Internet panel 'fix'? - More from Mellman: "Many [robo pollsters] are now adding Internet interviews from (non-random) panels to their sample. The problem with this approach is simple: there seems to be no evidence that Internet panels represent in any way the wireless-only population the polls are excluding. Indeed, what evidence there is suggests they don’t. For example, Internet panelists tend to be higher income, while cell-only households are poorer. Cell-only voters are more likely to be minority, while online panels are whiter. More relevant, there is no evidence that the political attitudes of online respondents reflect those of cell-only voters. Indeed, David Johnson of Discovery Research Group reported “online access panel users were found to be more like landline respondents than cell phone ... respondents.” [ibid, Discovery Research report]

Reactions - HuffPollster asked for reactions from pollsters who now supplement their automated calls to landlines with online interviews drawn from opt-in panels. Here are their comments, via email:

-Jay Leve, SurveyUSA: "SurveyUSA uses live operators to interview cell-phone respondents on all surveys in which we use voter-list sample. We use internet panelists on those surveys in which we use RDD sample. In all cases, as has been reported in this space several times going back to 2010, SurveyUSA asks respondents which of 4 categories they fall into: cell phone only, cell phone mainly, cell phone supplemental, or no cell phone. Based on a respondent’s answer to this question, a respondent is characterized by SurveyUSA as either 'reachable on a home phone,' or 'not reachable on a home phone.' Landline respondents look nothing like internet cell-phone respondents, contrary to Mellman’s assertion. Mellman’s penultimate paragraph is pitiful and misleading."

-Mike Boniello, Rasmussen Reports: - "We disagree that an online panel coupled with phone surveys will not reach a representative sample of voters. There are ways to use online panels via quotas and multiple panels to get a representative sampling of cell phone users."

-Tom Jensen, Public Policy Polling (D): "The interviews we include from our online panels are cell only voters. This is not just reaching people with landlines over the internet instead of their phones. We added the online component to our polls to help us reach more younger voters and minorities, and that has definitely proven to be the case. For instance on our recent Louisiana poll, 52% of online respondents were black (compared to 28% of the overall sample) and 39% were under 30%, 31% were 30 to 45, 28% were 46 to 65, and only 2% were seniors. On our recent Michigan poll 15% of online respondents were black (compared to 13% of the overall sample) and 34% were under 30%, 34% were 30 to 45, 30% with 46 to 65, and again only 2% were seniors. These numbers are pretty typical of who we’re reaching with our online interviews- we don’t track respondents by income but the claim that online panels are whiter than the overall population is completely at odds with what we’re getting from our cell only online interviews."

-Brent Seaborn, Vox Populi Polling (R): "Mobile research reaches more cellphone only voters than cell phone interviews. We have tested online research as a supplement to both live operator and automated calls, we prefer using online mobile research - surveys taken on a smart phone app. We find this supplements our research methods best. Mobile research tends to be younger, more racially diverse, and lower income. Part of Vox Pop's mission is to get more information on public opinion into the dialogue and make good quality research available to many organizations who could not afford the traditional opinion research. We recognize there are always cost quality tradeoffs but we have worked very hard to maximize quality and make public opinion research available to many."

-Douglas Rivers, YouGov - "Unfortunately, Mark has fallen for the ecological fallacy. Even if the average internet panelist has higher income than the average cell-only voter, this doesn’t imply that the average cell-only internet panelist has higher income than the average cell-only voter. Similarly, even if internet panels over-represent whites (which they do), this doesn’t mean that cell-only panelists over-represent whites. Reputable surveys from internet panels don’t randomly select respondents from full set of panelists and hope everything works out. Rather, they are selected to be representative of population. The study Mark cites is based on a very small sample (200 people) from an online panel and, according to the report, used "an open form of sampling with no quota stops.” This is asking for trouble and I'd agree with Mark that this isn't adequate. If I was sampling cell-only households, I would select a sample of cell-only respondents to match the demographics of those in the National Health Interview Survey—our best source of data about this population." [Note: YouGov, which partners with the Huffington Post on a daily national survey, does not conduct telephone surveys, but rather online research using a sample selected from YouGov's opt-in online panel.]

How are the supplemental online respondents selected and weighted? - Those trying to evaluate the newer online sampling methods used by the automated pollsters confront several critical issues: What kinds of respondents are sampled (anyone in the online panel or only those in the "cell only" population?) and how are online respondents selected or weighted (is there an effort to match the demographics of the "cell only" population?) Unfortunately, disclosure of the details of these practices has been minimal, so it's hard to assess how many automated pollsters are, to paraphrase Rivers, randomly selecting online panelists and hoping everything works out.

NEARLY HALF OF AMERICANS SAY MIGRANT KIDS SHOULD BE DEPORTED - Rachel Lienesch: "A near-majority of Americans want the undocumented immigrant children currently being held at the border deported as soon as possible, even though only about one-third of adults think these children have someplace safe to return to, a newHuffPost/YouGov poll finds. Forty-seven percent of Americans said deporting children to their home country as soon as possible should be America's priority in dealing with the current border crisis, while 38 percent said the priority should be letting immigrant children stay in America until it's certain they have a safe place to return….But Americans aren't certain if undocumented children will actually have a safe place to return if they are deported. Asked to choose between two statements about immigrant children crossing the border illegally, only 36 percent of respondents said they think 'most of them have a safe place to return to but would prefer to live in the United States,' while 39 percent said they believe 'most of them are fleeing unsafe places and don’t have somewhere safe to return.' Twenty-five percent said they weren't sure." [HuffPost]

Few approve of president's handling of the issue - Pew Research: "President Obama gets very low ratings for his handling of the issue. Just 28% of the public approves of the way he is handling the surge of children from Central America, while twice as many (56%) disapprove. That is one of the lowest ratings for his handling of any issue since he became president….And as was the case in January, neither party has a significant edge when it comes to dealing with immigration; 42% say the Republican Party could do a better job on the issue while 40% say the Democratic Party." [Pew]

Concerns about immigration issues rise - Lydia Saad: "With thousands of undocumented immigrant minors crossing the nation's southern border in recent months, the percentage of Americans citing immigration as the top problem has surged to 17% this month, up from 5% in June, and the highest seen since 2006. As a result, immigration now virtually ties 'dissatisfaction with government,' at 16%, as the primary issue Americans think of when asked to name the country's top problem. This is not the first time that immigration has spiked in the public's consciousness. Most recently, Gallup found the issue increasing to 10% in 2010, at a time when a new immigration law in Arizona was making news. And prior to that, it increased twice in 2006 to 15% or higher, amid congressional debate over immigration reform." [Gallup]

IOWA RACE DEADLOCKED, SCOTT BROWN TRAILS IN NEW HAMPSHIRE - HuffPollster: "The U.S. Senate race in Iowa is shaping up to be deeply competitive, a new poll finds, while former Sen. Scott Brown (R-Mass.) looks like less of a threat in his effort to unseat Sen. Jeanne Shaheen (D-N.H.). Rep. Bruce Braley (D-Iowa) and state Sen. Joni Ernst (R) are deadlocked at 43 percent in the race to replace retiring Sen. Tom Harkin (D-Iowa), according to an NBC/Marist poll released Wednesday….HuffPost Pollster's model, which incorporates all publicly available polling, finds the race tied…..The Senate race in New Hampshire has been more stable, with Shaheen leading Brown in virtually every survey taken. NBC/Marist finds Shaheen ahead 50 percent to 42 percent, with just 6 percent undecided. She's helped by a 52 percent favorability rating among voters, with 39 percent viewing her unfavorably, while views of Brown are split at 40 percent favorable and 39 percent unfavorable. HuffPost Pollster's average gives Shaheen a 9-point lead over Brown." [HuffPost]

Candidate image matters - Chuck Todd, Mark Murray and Carrie Dann: "[T]here’s an explanation why our new NBC/Marist polls show Iowa’s Senate contest all tied up (with Democrat Bruce Braley and Republican Joni Ernst at 43% each), while Sen. Jeanne Shaheen (D-NH) leads likely GOP challenger Scott Brown by eight points, 50%-42%, in New Hampshire. The answer: Shaheen is popular while Brown isn’t, and both Braley and Ernst are polarizing in the Hawkeye State. In New Hampshire, Shaheen has a 52%-39% fav/unfav rating (+13) in the poll, but Brown -- the former senator from next-door Massachusetts -- has just a 40%-39% score (+1). And in Iowa, Braley’s fav/unfav rating is 36%-32% (+4), while Ernst’s is 38%-33% (+5)." [NBC]

Will Brown's image improve? - As NBC's team notes, one challenge for Brown: he and Shaheen are both known quantities to New Hampshire voters, leaving him little room to grow. The UNH/WMUR Granite State poll, which has tracked the race since January, puts his name recognition higher than any of his Republican rivals for the nomination, but shows his favorability rating remaining basically flat. [UNH]

OBAMA SEEN AS MORE EMPATHETIC, LESS EFFECTIVE THAN BUSH - Alec Tyson: "A new Pew Research Center survey finds that President Barack Obama’s overall approval rating has held steady at 44%, even as he receives low marks for his handling of the surge of undocumented child immigrants at the U.S. border. While Obama’s job rating has been below 50% for the past year, it stands eight points higher than that of his predecessor, George W. Bush, at a comparable point eight years ago. In July 2006, 36% approved of Bush’s job performance. Obama gets much higher marks than his predecessor for empathy and honesty. But his ratings on leadership and his ability to get things done are about the same as Bush’s at about this point in his second term." [Pew Research]

HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday via email! Just click here, enter your email address, and and click "sign up." That's all there is to it (and you can unsubscribe anytime).

WEDNESDAY'S 'OUTLIERS' - Links to the best of news at the intersection of polling, politics and political data:

-Quinnipiac finds the Colorado gubernatorial race tied between John Hickenlooper (D) and Bob Beauprez (R). [Quinnipiac]

-Rich Morin reviews the demographics of gun ownership. [Pew Research]

-Jews, Catholics and Evangelical Christians are among America's best-liked religious groups. [Pew]

-Nate Cohn ponders the electoral implications of splitting California into six states. [NYT]

-Tom Edsall sees a generational split coming among Democrats around economic redistribution. [NY Times]

-Forget Nate Silver, Andy Biao says, here's how to flawlessly predict anything on the Internet. [Medium]