How Good Is the Data…
The image of the just re-elected Harry Truman holding the front page of the Chicago Tribune is one of the most recognizable photos and one of the most spectacularly inaccurate headlines in American political history.
All the respected pollsters of the day – it was 1948 – got the outcome of that election wrong. Most had quit polling a week before the election certain that New York Gov. Thomas E. Dewey had a lock on the White House. It turned out to be an historic mistake that the Truman-hating Tribune was all too eager to believe as the early results trickled in on election night. Truman, of course, staged what it now considered one of the biggest comebacks in political history, likely turning the tide in the last few days of his feisty campaign.
Polling has come a long, long way since Truman’s day and even farther since the Literary Digest famously predicted Alfred Landon beating Franklin Roosevelt in 1936. In that case, the Digest relied for its self selecting sample on folks with a telephone or an automobile registration. Many Americans didn’t have either in 1936.
Yesterday I offered up some observations on the state of the current Otter – Allred race for Idaho governor that were prompted by a recent Rasmussen poll. Rasmussen’s Idaho research – he bills himself as a strictly independent pollster – was one of two or three polls on various political races that he releases every week.
In a nutshell: Otter appears to have a comfortable lead and Allred is still introducing himself as a first-time candidate unknown to most Idahoans. In my view, Allred may also come to regret missing a strategic opportunity during the legislative session to cast himself as the anti-Otter. While the press and many in the public, including education supporters, human rights activists, park users and those who think texting while operating a motor vehicle is a bad idea, were focused on the daily actions of the legislature and the governor, Allred hardly got up on the stage to offer a counterpoint. We’ll see if that turns out to have been a big mistake or not.
The Rasmussen poll shows Otter up 60-28.
How Good Is Rasmussen’s Research?
That is a good question regarding Rasmussen’s poll, or for that matter any poll that finds its way into the public arena. Just how good is the data?
The first thing to say about Scott Rasmussen is that he nailed the 2004 and 2008 presidential races and has a respectable record in state level contests.
Left-leaning bloggers don’t like him and accuse him of partisanship. The respected polling analyst Nate Silver noted in January that Rasmussen’s numbers “tend to be more favorable to Republican candidates and causes than most other polling outfits.” Silver is quick not to accuse Rasmussen of bias. It could be, he says, an issue with the methodology of Rasmussen surveys; he screens for “likely voters” when other pollsters don’t and Rasmussen uses automated data collection techniques that some folks question. And Silver notes, Rasmussen, who did polling work for Republicans and George W. Bush in 2003 and ’04, could be right on with his numbers even as some question his methods.
In any event, Silver’s analysis of Rasmussen’s work and methodology is worth reading, as is a piece by Mark Blumenthal in the National Journal who asks and tries to answer the question of when a poll is “partisan.” The conclusion: it is getting harder and harder to tell.
Thoughts to keep in mind as you read about polls
- Most reporting on surveys is less than adequate. Even the big news organizations like the Washington Post and CNN never seem to provide enough context as to how the survey was conducted and what was going on that might influence the results. Idaho reporters are at an even greater disadvantage in reporting on polls since they are often writing about something for which they have no first hand knowledge. An Idaho reporter gets what looks like interesting information from a Rasmussen – or soon you can bet from a more Democratic-leaning pollster – and about all they do is report the findings and add the comments from the opposing camps. When it comes to polls we need more context. We need explanation of how the surveys were conducted. What and how many questions were asked? We need more detail. We need more reporting.
- The real value in polls is contained in the “internals.” We all love the horse race question: “if the election were held today”…and those results typically get the headlines. The really valuable strategic information is always buried deeper in a good survey. How are the demographics of age, religion and gender sliced? Do Idahoans feel the state is heading in the right direction? What issues make one candidate or the other vulnerable? The horse race is fun and it tells us something, but it is far from the complete picture. I’d love to see such a survey, but that information is going to be held very close to the vest by both campaigns and pollsters like Rasmussen don’t do that kind of sampling, at least not that they make public.
- Idaho news organizations would do all political junkies and the election process a real favor if they were to develop their own research capabilities. Good research costs money, but perhaps a collection of news organizations could pool the resources – The New York Times/CBS News model – and provide the context and “internals” that would provide real value to voters and policy makers.
Lessons From Distant Campaigns
I have been deeply involved in two statewide races for governor – 1986 and 1990 – and have watched every race since from the back row. One of the surprising findings from our research in 1986 – Cecil Andrus was mounting his comeback that year after having been off an Idaho ballot for a dozen years – was that fully a third of the probable voters didn’t know the former governor and Secretary of the Interior from a bale of hay. He just didn’t register with those voters who had come of voting age or moved to the state since he had been governor in the 1970’s. In other words, the candidate needed to be introduced to these voters.
The lesson: most candidates underestimate the level of public understanding of who they are and what they stand for. This is a particular problem for first time candidates and it often proves fatal.
[A footnote: Andrus relied on Jimmy Carter’s controversial and outspoken pollster Pat Caddell for research in 1986 with mixed results. Doug Schoen, who polled for Bill Clinton, did the job splendidly in 1990. In the small world category, Scott Rasmussen touts endorsements from both Caddell and Schoen on his website. Both Caddell and Schoen often provide contrarian views at odds with national Democratic talking points and both provide commentary for FOX News. Each pollster has predicted massive Democratic loses at the polls this fall as a result of health care legislation.]
Another polling lesson for me comes from 1992 when four-term Democratic Rep. Richard Stallings, who represented southern and eastern Idaho, ran for the United States Senate, I’m going to bet Stallings’ name ID north of the Salmon River never got above 60%. I can’t prove that notion, but the election outcome demonstrated that Stallings was not able to connect with voters in that region of the state. For a guy from Rexburg, the territory north of Riggins might as well have been in another state.
The lesson: being well known in Boise doesn’t mean much in St. Anthony or Sandpoint. Statewide name recognition is a long, hard and expensive slog. You earn it with time or with money or both. It is but the absolutely first step to a successful political campaign. There is an old, old fomulation in politics that holds that every candidate must travel a cycle. First the name must be established, then who they are as a person can be developed, and finally comes the message. But it all starts with name recognition. You don’t have that you don’t have much of a political campaign.
Here’s my guess: Otter is not in quite as good a shape as the Rasmussen poll indicates and Allred is not in quite as bad a shape. Such polls measure name ID and party affiliation and not a lot more. Having said that, and with the acknowledgement that it is early in the cycle, campaigns do develop a certain rhythm and pace – call it the narrative – and this one is starting to firm up. Today it is very much Otter’s race to lose and his name ID, his long record of familiarity with Idaho voters, Idaho’s strong R tendencies and this being a GOP year all put him in solid shape to be re-elected.
A further guess: Otter will run a very traditional, tried and true Republican campaign based on presenting a united GOP front and emphasizing the party’s anti-tax stand. Couple that message – we’re Republicans and you can trust us on taxes – with a strong ground game to turn out voters and that has been enough for a GOP gubernatorial candidate to win every time over the last 20 years.
If Allred is to have any chance of pulling the big upset, he had better start running soon with the political equivalent of football’s “wishbone offense.” He needs something to revolutionize the game. He has to shake up the race in a very significant way, change the developing narrative and move the polls or he’ll find himself on the wrong side of “Dewey beats Truman.”