Tuesday, August 3, 2010

"R" - Is It for "Republican," or "Rasmussen"?

Actually, shame on me, it’s a trick question. But, to quote the pragmatist philosopher and psychologist William James, “A difference that makes no difference is no difference.”

Such political news as there has been this last week, has been dominated by a Rasmussen “poll,” purporting to show Republican nominee Robert Bentley with a 20% lead over Democrat Ron Sparks. Not that we are surprised, but many in the media in Alabama immediately pounced on the story to show the irresistible momentum of the Bentley colossus. An example, from the Mobile Press-Register, is at the right, but The Birmingham News and others have been following the same lead, for weeks, from the same polling shop. Whether the effect is intentional or inadvertent, you can almost see their editors circling the block around their newspaper offices, beating a bass drum and shouting “It’s inevitable! Bentley will win! Join the parade while you can!” For this reason, journalists should be careful about how they handle polls, and we, as Democrats, should know the truth behind polls, their biases, and their issues, so we can intelligently respond to media stories such as the one from the Press-Register. Otherwise, such polls can become a self-fulfilling prophecy.

It’s a movie we’ve watched before. On November 7, 2002, The Birmingham News found itself with some explaining to do:

The Birmingham News, the Mobile Register and The Huntsville Times on Sunday published a survey of 600 likely voters that showed Riley leading Siegelman 47 percent to 39 percent, with 11 percent undecided and 3 percent favoring Libertarian John Sophocleus.

Powell, associate professor of communication studies at University of Alabama at Birmingham, conducted the poll Oct. 28-31 for the three newspapers and WHNT-TV in Huntsville. In Sunday’s article about the poll, Powell said, “Riley has pulled significantly ahead. He is poised to win the election.” But Tuesday’s election turned into a cliffhanger, with just a few thousand votes - a few tenths of 1 percent of the more than 1.3 million cast - separating Riley and Siegelman.

“The poll was wrong,” said Siegelman spokesman Rip Andrews, noting that Siegelman protested that it was wrong when the poll was released. “We never had any evidence that was credible to suggest anything but a very, very close race,” Andrews said. ‘Which is what we kept saying to some deaf ears in the media.”

At best, it’s a remake of a movie we’ve watched before, and as with most remakes, the cast isn’t even as good as it was in the original. But one recurring theme of the Alabama media saga is that it frequently has the same scriptwriter - Rasmussen Reports. “Rasmussen” is a name that recurs in news reports with remarkable regularity, especially in the columns of the Newhouse outlets in Alabama, and it’s well past time for them to come under a little scrutiny. (To be fair to Rasmussen, the 2002 botched poll was conducted, as noted, by UAB marketing professor Larry Powell.)

Let’s start with looking at Rasmussen’s record in Alabama. In a poll conducted only eight weeks before Bentley’s strong second-place showing in the June 1 Republican primary, Rasmussen didn’t even query respondents about Bentley! That’s not a promising start.

The Rasmussen shop states on its website that it “does not do polls-for-hire.” However, records indexed by the Center for Public Integrity show that Rasmussen was paid over $100,000 by the Bush campaign in 2004. Whatever explanation Rasmussen has for that discrepancy, I haven’t been able to find. Rasmussen’s website says that, because it has frequently been sought out to conduct polls for hire:

To meet this demand, Pulse Opinion Research was launched as a separate company several years ago to provide field work (interviews and processing) for surveys. Pulse licenses methodology developed by Scott Rasmussen and provides the field work for all Rasmussen Reports surveys. It provides all customers with the same quality field work that we rely on every day.

This creates obvious problems. If Pulse and Rasmussen are closely affiliated - say by interlocking ownership - then the “independence” of Rasmussen is a sham, about which they are proactively deceptive by affirmatively touting it. Rasmussen states that it contracts out its “field work” - the actual making of calls - to Pulse. If the two are truly independent of each other, this raises obvious questions about quality control and oversight.

Nate Silver, at fivethirtyeight.com, while kinder in some other aspects to Rasmussen, had the following to say about his methods:

I’m not saying that Rasmussen’s question wording is always biased. It isn’t. And I’m sure you could find a couple of cases where the wording tends to portray the liberal argument more favorably. But cases like these happen consistently enough with Rasmussen that I’d say it’s a concern. And when they do use unorthodox question wording, nine times out of ten it favors the conservative argument.

Silver also notes that “So far in the 2010 cycle, their polling has consistently and predictably shown better results for Republican candidates than other polling firms have. (Emphasis in original.) Finally, Silver notes that “they have a knack for issuing polls at times which tend to dovetail with conservative media narratives.” That observation - made in January - really rings true with respect to the “20% Bentley lead” poll. From the June 1 primary until the runoff two weeks before the Rasmussen “poll,” Bentley was actively advertising on television, making daily appearances, had canvassers and phone banks working, and was the subject of regular news coverage, including favorable “Bentley Wins!” headlines on July 14. Sparks, during this same interlude, was not running paid TV ads, was not frequently covered in the news, and was working on resting and assembling a field operation, not deploying one. If I were to pick the optimal time to poll for Bentley, the two weeks before the Rasmussen poll would have been the time. Presuming Rasmussen is knowledgeable about polling, he knows this, too, and the timing of his poll cannot have been accidental.

Silver is not the only analyst critical of Rasmussen’s techniques and slants. As one political science professor was quoted about Rasmussen on Politico:

“He polls less favorably for Democrats, and that’s why he’s become a lightning rod,” said Charles Franklin, a University of Wisconsin political scientist who studies polling. “It’s clear that his results are typically more Republican than the other person’s results.”

Rasmussen uses automated calling, often called a “robopoll,” a practice so prone to error that many major media organizations, including The Washington Post, have an editorial preference to not publish their results. While polls like Rasmussen like to point out that they have the advantage of giving every respondent the same pronunciation and inflection in the questions, that advantage is more than lost in the shortcomings of a robopoll. Any robopoll is going to have to limit itself in the number of questions it asks. People are just not going to listen to the robot as long as they will to the person who asked “Hi, how are you doing?” at the beginning of the call. This is critical because it limits the “screening” questions that can be asked about the age, location, race etc. of the person answering the call. Has the computer picked too many white households to call? Too many from Mobile, compared to its share of the vote? How many times has the respondent actually voted in the last few elections? A serious, and accurate, poll needs to know a lot of this information to determine if its raw data needs to be mathematically adjusted so that the published result is a fair cross-section of the actual electorate. When you are dealing with the limited attention span of someone listening to a recording, you have to limit those questions severely. A related issue involves what former Secretary of Defense Donald Rumsfeld called the “unknown unknown” (it pains me to quote him, but his memory is good Democratic motivation) of people who simply hang up when they hear the computer talking. We simply have no way of knowing how those people differ from those who merrily sit there and punch “2” for Bentley. Another “unknown unknown” in any robopoll is how many of the “respondents” were actually six-year-olds who thought it would be good grownup fun to punch a number when the nice computer asked them to. A live caller will catch this one.

Another robo-calling polling outfit is Public Policy Polling out of North Carolina. They released a poll a week before the June 1 primary, showing Artur Davis with a 10% lead over Ron Sparks. We all remember how accurate that turned out to be. The same shop came out with at least one poll showing Troy King leading Luther Strange in the GOP primary for Attorney General - which Strange eventually won, 60.2%-39.8%.

So who is the personality behind Rasmussen Reports? Scott Rasmussen serves as the President of the Ocean Grove Camp Meeting Association in New Jersey. Under New Jersey law, Orange Grove is treated as a municipality. In 2007, Rasmussen, as President, terminated the public’s right to rent the Association’s beachside pavilion, rather than be forced to allow homosexual couples, including longtime Ocean Grove residents, to use it for civil union ceremonies under New Jersey law. The pavilion had been leased for weddings for decades prior to this action. If this sounds familiar to Alabamians, it should. The City of Montgomery, faced with a Federal court desegregation order in 1959, closed all the city’s parks, swimming pools, athletic fields and playgrounds, and kept them closed until 1965, in the famous case of Gilmore v. City of Montgomery, 417 U.S. 556 (1974). While the pavilion episode does not directly pertain to polling, it does give a pretty valid perspective of Rasmussen’s political point of view, and it’s certainly not liberal or moderate.

There are two lessons to take away from all of this. First, for journalists, it’s important to take that extra five minutes and look into the background and methodology of the pollster whose “results” you’re about to put on page one. Especially if the pollster’s name starts with the same letter as “Republican.” You run the risk of repeating the mistake of 2002, when false reports of a Riley lead created the bandwagon that pulled him close enough to Siegelman for the “After Midnight” recount to do the rest. For Democrats involved in a campaign, at any level, it’s important to be informed about the weaknesses and biases of polls, and to push back when a pollster like Rasmussen is trying to declare an election over. This is true whether you’re the candidate or campaign spokesperson in Montgomery talking to the reporter, or just the local Democrat at the morning coffee table at the local meat-and-three cafe. With only one or two exceptions in Alabama, reporters will balance their story on a poll if we are ready and willing to give them the information they need.

5 comments:

  1. Lies, damned lies, and statistics. It's a numbers game.

    ReplyDelete