September 12, 2012
As originally published in the New Hampshire Union Leader
What you think you know is often wrong. We are bombarded in election season with facts that aren’t facts and conclusions that are so convenient to our worldview that checking them against the facts would be inconvenient.
More important to most people is starting with the conclusion we have and then accepting the theory that confirms what we already believe. After all, if it agrees with me it must be true because I can’t possibly be wrong.
Today’s papers will be full of election news, so I had better start with election examples or you won’t bother reading. Nowhere does the truth take longer holidays than in election analyses. Pollsters who show our candidates winning are obviously correct, but if my guy is down then the pollster is obviously a biased charlatan.
This is very common right now in the presidential race. The President (of whom I am not a fan) is leading in most polls. The average of recent polls posted on RealClearPolitics.com shows the President ahead by 3.1 points. I have friends who are eager supporters of the other guy (Romney is his name, in case you haven’t been paying attention) eager to dismiss some polls or even every poll.
The theory is that the polls have a biased sample. “They oversample Democrats and get a biased result.” Respondents self-identify and in quite a few polls self-identified Democrats outnumber Republicans by 4, 5, or 6 percentage points. “That’s ridiculous,” I’m told. “2008 was D+7 but that was unusual. The sample should be roughly even. So if a Democrats-plus-5 poll shows Obama plus 3, it means Romney wins by 2.”
I’m not vouching for that logic, but it always annoys me that my whining friends can’t tell me what the historical average is even though they are certain it should be moved in their favor. So I looked it up.
According to exit polls, self-identified Democrats outnumbered self-identified Republicans by 7 points in 2008 (39-32). The annoying class is right that 2008 was above average for Democrats and that 2004 was an even split (37-37). However, both years are atypical. The four elections before 2004 had self-identified Democrats with either a 3 or 4 point edge. Interestingly, Independents remained steady nationally at 26 or 27 percent of the electorate except in 2008 when they bumped up to 29 percent.
So that crazy pollster with a slightly Democratic-leaning sample by 5 points is off the historical average by maybe 1.5 points. Hardly ridiculous. Nor need it be. If your guy is up or down by three points with eight weeks to go, the race could go either way.
The next polling myth is the incumbent rule. We all “know” that if an incumbent in a big race is below 50 percent he’s in trouble because “undecideds will break heavily against the incumbent” or “they’re not undecided about him.” This is well known and oft repeated. It’s also wrong.
Nate Silver writes a terrific blog called fivethirtyeight for the New York Times. I’m quite right-wing and he’s quite left-wing but (and here’s the lesson if you’ve been waiting for it) he runs numbers and does factual analysis that is very good even if you’re happy when he’s sad or sad when he’s happy.
So he looked at the incumbent rule to see if candidates under 50 percent with a month to go had troubles. They didn’t. On average, an 8.1-point lead turned into a 7.2-point victory. Undecideds broke only very slightly against them. However, if your guy is down 6 points to an incumbent, I’m sure the math will be different for you.
I use these two election samples not to rain on your parade, but to warn you to check the numbers before you believe generalizations in politics.
Ronald Reagan slashed spending. No. Spending under Reagan increased 6.7 percent each year. OK, but that was all defense spending. No, non-defense spending went up 6.1 percent each year. By contrast, spending over the next 10 years is projected to increase by 4.4 percent.
Supposedly the “Ryan budget” also slashes spending. Again, no. Spending would increase over 10 years from $3.5 trillion to $4.9 trillion under that supposedly radical alternative. The average annual increase would be 3.2 percent instead of the projected 4.4 percent.
This has been an election season message urging you to ignore the whining and check the data yourself. You might find that what you think is true isn’t.