Las Vegas Sun

March 28, 2024

POLITICAL MEMO:

How much does poll position matter?

Experts in polls lend advice on how much to read into their findings

Given the dramatic Democratic primary campaign between two strong, equally matched candidates, the population of politics junkies has soared, likely comparable to what happened to the population of actual heroin junkies after the release of the film “Trainspotting.”

Politics became glamorous and fun, attracting a whole new set of obsessives.

And for the politics junky, there’s nothing like the quick fix of a new poll.

Given the prospect of information overload, we’ve developed a poll-reading guide with the assistance of Mark Blumenthal, a former Democratic pollster who now publishes a godsend for junkies, pollster.com, and Los Angeles Times poll director Susan Pincus.

Blumenthal:

1. The margin of error applies to each candidate, not the difference between them. So take a recent Gallup poll, which had Obama with 47 percent and Sen. John McCain with 42 percent, with a margin of error of plus-or-minus 2 percent. This means Obama could be as high as 49, and McCain as low as 40 — or Obama could be as low as 45 percent, McCain as high as 44. There’s a 95 percent probability of attitudes being in this total range, with the highest probability in the middle of the range.

Some variance is assumed because pollsters draw from a random sample rather than trying to poll everyone. If you can draw people at random, such that any voter in the state or country has an equal probability of getting selected, you can interview 500 or 1,000 people and determine a margin of error. But other factors that can make the sample non-random can potentially further skew poll results even beyond the calculated margin of error: People who refuse to talk to pollsters, for instance, or who don’t have a landline.

2. Look at a lot of polls. Pollster.com aggregates polls and plots them on a graph, which gives a nice visual representation of trends and the range of variance. If you look at 10 polls and all 10 have the same candidate ahead by a point or two, that’s fairly reliable. If 10 polls are moving in one direction in the same week, then you can infer something.

3. Remember they’re snapshots of attitudes today. In competitive races there can be big shifts from July to November.

4. Who is a likely voter? Hard to say. Some polls include registered voters, some likely voters.

As for likely voters, some say they’ll vote, offering the same commitment as a New Year’s resolution to go to the gym. Gallup is famous for its likely-voter model. Its pollsters ask: Do you intend to vote? Did you vote before? Will you vote in the same precinct? And more like that.

They take the answers, score everyone and create a scale. Then they come up with a number — say, 60 percent — who estimate will vote in a presidential election, and they keep that numnber as likely voters. In 2004 they recognized that people were more engaged and increased their turnout model.

Look for big debates about whether Gallup is unfairly excluding young voters, drawn to Obama this cycle, from its likely voter pool.

Pincus tips:

1. Know the methodology. Is it a random sample?

2. Look at the entire poll: Are questions unbiased? (A poll should be viewed skeptically if it asks a biased question, such as, “Would Obama’s unpatriotic attitudes affect your decision about voting for him?” Are questions rotated? Are names within questions rotated? Is the horse-race question first, so the answer isn’t biased by information gleaned from the poll? (Respondents’ answers also can be altered by order of questions and the order of choices.)

The word “crosstabs” is all the rage in politics. This refers to the poll’s subgroups: Who do Hispanic married woman prefer? And so on. Often the poll’s best nuggets are contained in the crosstabs.

Finally, know this: Some polls are better than others.

Join the Discussion:

Check this out for a full explanation of our conversion to the LiveFyre commenting system and instructions on how to sign up for an account.

Full comments policy