A new automated poll in Kentucky makes the Democrats happy and gives the pollster a chance to explain how they sample cell phone-only voters. Gallup puts Texas Latinos under a survey microscope. And Politico and political scientists get into a spat about a forecast that's all too (wait for it)...predictable. This is HuffPollster for Friday, February 7, 2014.
SURVEYUSA GIVES GRIMES NARROW EDGE IN KY - Kentucky's U.S. Senate race is close, according to new SurveyUSA automated telephone poll for several media outlets in Kentucky, one of the first nonpartisan media polls on the contest. Democrat Alison Lundergan Grimes holds a narrow 46 to 42 percent advantage over Republican Senate Minority Leader Mitch McConnell among all registered voters. Sam Youngman: "[The] data reveals that her relatively strong starting position is more about voters' disapproval of McConnell than solid support for Grimes. 'This new poll confirms that Kentucky's U.S. Senate seat is up for grabs, and that the contest here will be one of the most exciting in the country,' said Stephen Voss, a political science professor at the University of Kentucky...In an especially troubling sign for McConnell, his job-approval rating was two points below the approval rating poll respondents gave President Barack Obama, who remains deeply unpopular in the state. Only 32 percent approved of McConnell, compared with 34 percent for Obama. Both men received a disapproval rating from 60 percent of voters.
Confirms other surveys showing close race - Eight polls on the race have been publicly released since October, six of which have shown the two candidates either exactly tied or separated by just a single percentage point. The five most recent used automated telephone methodologies. The Pollster poll tracking model, based on all of the publicly released polls to date, shows a virtual tie (Grimes 43.3 percent, McConnell 43.2 percent). [Pollster]
How do they account for cell-only households? - SurveyUSA, like other pollsters that rely on an automated, recorded voice method to interview voters , is barred by federal law from using its autodialers to call cell phones. A pollster that relies only on landlines will miss the roughly one third of adults who now live in households without a landline phone. As such, most of the firms that conduct automated telephone polls have been experimenting with alternative methods to sample and interview voters in cell-only households. SurveyUSA, like most of the others, interviews cell-only respondents drawn from a pre-recruited panel of respondents who volunteer to participate in online surveys, usually for a nominal financial incentive. The description of their procedure in the Kentucky release is a little vague: "Respondents not reachable on a home telephone (23% of registered voters), were shown a questionnaire on their smartphone, tablet, or other electronic device." [Courier-Journal]
How did SurveyUSA sample cell only respondents? - Back in 2012, HuffPollster asked founder Jay Leve for more detail, and he explained that respondents for each poll are "pre-recruited by [online panel vendor] United Sample" and "given a link to an HTML SurveyUSA questionnaire." They then completed the survey in a web browser on either a computer or a mobile device. SurveyUSA also asks a question that allows them to classify all respondents into 4 categories: "1: Cell Phone Only. 2: Cell Phone Mainly. 3: Cell Phone Supplemental. And 4: No Cell Phone." We checked in again with Leve today, and he confirmed that their method remains the same, adding additional detail:
· In the case of the KY poll you reference, we started with a 2:1 ratio of landline IVR [automated] calls to internet completions. (approximately 800 landline completions to 400 internet completions = 1,200 total adult completions).
· Of the 400 internet completions we received and analyzed, we determined that 111 of these 400 were in fact better characterized as “cell phone supplemental” or “no cell phone” respondents, and as such, those completions were added to the landline side of the equation.
· When we finished moving the 111 internet landline respondents over to the landline side of the respondent pool, we were left with, and what we reported out, were 911 landline completions and 289 cellphone completions. The 289 cell phone completions were among those judged to “cell phone only” or “cellphone mainly,” (or: 76% landline, 24% cell), in SurveyUSA’s final report, among adults.)
Leve also explained that respondents received the survey link either via email or, in some cases, by a dashboard app used by USamp panelists.
TEXAS HISPANICS LEAN BLUE, BUT NOT BY ENOUGH TO FLIP THE STATE - Andrew Dugan: "Texas Hispanics are decidedly Democratic in their political party preferences, 46% to 27%, but that 19-percentage-point Democratic advantage is much smaller in Texas than the average 30-point gap Democrats enjoy among the Hispanic population in the other 49 states. And white Texas residents are decidedly more Republican (61%) than the average among whites residing in other states (48%), complicating whether Texas will turn into a 'blue' Democratic state in future elections….The biggest challenge for Democrats hoping to turn Texas blue may be in registering and turning out minority voters in that state. But the Democratic Party's relatively poor standing with white Texans will continue to impede its ability to compete on a statewide basis for the foreseeable future." [Gallup]
COULD GOP AMERICAN IDOL FANS HELP CLAY AIKEN'S HOUSE BID? - Josh Green: "Clay Aiken didn’t win American Idol, and he’s going to have an even tougher time getting elected to Congress. As I wrote this week, the incumbent Republican he’s hoping to knock off in North Carolina’s 2nd District, Representative Renee Ellmers, seems like a pretty good fit for the district. ...So Aiken is going to need to attract a lot of Republican voters to have any shot at victory. Can he do it? I’d say probably not. But if you dig deep enough into the demographics of who his likely supporters might be, it turns out Aiken may have a hidden source of strength: American Idol viewers….Prime-time television audiences generally skew Democratic, but a number of reality shows actually skew Republican. Viewers of American Idol are more Republican than audiences of many other reality shows—and as the size of the bubble indicates, there are a lot of them. Idol viewers also have a fairly high propensity to vote, unlike, for instance, viewers of Extreme Makeover: Weight Loss Edition, who seem to stay home on Election Day." [Businessweek]
The chart, via Businessweek and National Media
THE 'MODEL THING,' 2014 EDITION - Politico's Mike Allen: "BEN HIGHTON, a political scientist at University of California, Davis, elaborating in a guest post on 'Monkey Cage': 'Our model suggests … if Republicans win the Senate elections in only four states — Alaska, Louisiana, Iowa, and Montana — then they are very likely to take control of the chamber.'...NO DUH, dude. You don’t need a MODEL to tell you that. But that doesn’t translate to a current LIKELIHOOD of getting the majority. (Before primaries are done! So good.)" [Politico]
The Monkey Cage's John Sides responds: There is an odd tension here. On the one hand, the model’s conclusions are so obvious that they merit a 'no duh.' But on the other hand, the model is pointless because it’s too early and the primaries aren’t even over. So let’s review what the model is doing. The model draws on data from decades of previous elections. It includes baseline fundamentals that everyone — including Playbook’s Mike Allen, I am sure — knows influence congressional elections, such as how popular the president is and the partisanship in individual districts and states. The model then presents a forecast for every state and district assuming that conditions are what they are today. We conduct some simulations from the model to deal with the inherent uncertainty in the forecast. That’s what gives us the estimate of the likelihood the Republicans will win enough seats to take the Senate." [WaPost]
-Sasha Issenberg: "Are we really going to go through this 'what is a forecast' thing again every election year?" [@victorylab]
-Jonathan Bernstein: "Answering @victorylab Yes, we'll have to go through 'what is a forecast every election cycle....However, each time, what 'everybody knows' and therefore why we don't need these pointy-head numbers will get a little bit better....Similar to stathead-bashing sportswriters who now bash WAR or whatever - why do we need stathead formulas when we have OPS?" [@jbplainblog here, here and here]
Rasmussen Reports – Help Wanted. Rasmussen has a new position open for a senior data scientist/statistician to help lead us into the future of public opinion polling in a way that relies less on land lines and more on a combination of the various media channels available today. If your background is in statistics, mathematics and programming and you have a vision for the next wave in public opinion, please apply. Love of politics and policy a plus. Job listing and more details here.
HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday via email! Just enter your email address in the box on the upper right corner of this page, and click "sign up." That's all there is to it (and you can unsubscribe anytime).
FRIDAY'S 'OUTLIERS' - Links to the best of news at the intersection of polling, politics and political data:
-Partisans sharply divide on the use of executive orders. [YouGov]
-59 percent of women don't believe the GOP understands their concerns. [CNN]
-Drew DeSilver looks at who smokes in America. [Pew Research]
-Jonathan Bernstein explains how issues affect vote choice. [Bloomberg]
-Greg Sargent notes that Republicans are more likely to report being "hurt" by Obamacare than independents and Democrats. [WaPost's Plum Line]
-James Delorey (D) takes a critical look at the attempts to explain presidential executive orders "in just one chart." [GSG]
-Paul Donaldson (R) finds Americans ready to end the war in Afghanistan. [POS]
-David Wasserman unveils the Cook Political Report's Partisan Voter Index (PVI) for 2014. [Cook Political]
-How not to use random digit dialing. [XKCD]