Don’t fall off the wagon
The other day I suggested that people take a pledge not to read, at least until the beginning of 2012, articles about presidential polls, and particularly to
ignore polls about President Obama’s approval rating, a subject on which the Republican media are constantly trying to jerk us into a triumphalist frenzy, notwithstanding the fact that Obama’s always “plummeting” approval rating never gets below 43-45 percent, about where President Reagan’s approval rating was for a long time during his first term.Here’s an example of what I mean, appearing in the Must Reads today at Lucianne.com:
Poll Numbers Continue Alarming Trend for ObamaWhen you see a link like that, as exciting as it may seem, don’t click on it, don’t even be tempted. Cast a cold eye on Obama’s “sinking” poll numbers. Web surfer, pass by.
Alexis Zarkov writes:
Opinion polls are largely useless in forecasting the outcome of a presidential election before and even during a campaign. Only until right before the election do we get useful predictive information. We know that political science research, apart from polling, can predict the popular vote very accurately months before the vote. Andrew Gelman discusses the disconnect between polls and predictions from political science fundamentals here. The article is fairly technical, but non-mathematical readers can still gain a lot from the narrative. In particular, look at Table 1 to see how much political science beats polling. If the federal government would release third quarter economic data in September 2012, we would have a very good idea who is going to win in November absent a very close election. The change in real disposable personal income per capita is one of the key predictor variables to watch. More so than the polls. In fact the wildly varying opinion polls seem to converge to the political science prediction as discussed by Gelman. The press and political strategists seem unaware of what’s common knowledge among quantitative political scientists. Not surprising. Dry economic data and complex statistics have little entertainment value as compared to personalities and polls. The idiots in the press seem to think that Michele Bachmann’s minor gaffes about irrelevant facts (was John Wayne born in Waterloo, Iowa or Winterset, Iowa) matter. They think that they can influence the election by focusing on such trivia. Presidential (unlike Congressional) elections are strongly dependent on factors outside their control.Jake F. writes:
I love the fact that you encouraged people to take the don’t-read-polls pledge, and are reinforcing the need to do so with a “lead me not into temptation”-style post about Lucianne.com. That, along with your persistent reminders about the “free fall” in Obama’s numbers — even though those numbers never seem to move — provide a welcome dash of realism to what would otherwise be a lot of wishful thinking.James R. writes:
A. Zarkov is right but I’m not sure those involved are unaware of what is common knowledge among political scientists, because that premise is based on a misunderstanding of what polls are for, especially at this stage. These polls aren’t a predictor of who will win a year from now. The purpose of these polls is to shape expectations and outcomes, to massage the electorate, and give an impression of who our “inevitable” candidates are (as opposed to who our “outside the mainstream no chance to win so you shouldn’t pay attention to them” candidates are); the pols are for positioning the “credible” candidates the electorate will chose from. As the esteemed T. Herman Zweibel used to say “Why help pick the monkey when you can own the organ grinder?”Paul K. writes:
I agree with your advice to avoid watching polls. At the Drudge site, alongside the dubious poll numbers, there is this foreboding headline: “Jewish Dems losing faith in Obama.”LA replies:
That’s a riot. So the article starts out talking about this big worry for Obama (“he’s losing the Jewish donors!”), which turns out not to exist.
Posted by Lawrence Auster at June 29, 2011 12:30 PM | Send