There once was a shade-tree political consultant who got fired by a politician who accused him—falsely, I hope to heck—of just flat-dab making up a poll.
That was a couple of decades ago.
Now they all make them up.
Not from whole cloth, of course. They draw their samples. They make their calls. They ask their questions. They compile their numbers.
Then they say, “Oh, heck, this isn’t right. Women and old folks answer landlines a lot more than men and young people do. We’ve got to weight this thing to produce a true facsimile of what we think the demographics are going to be for the real electorate when people actually vote.”
Weighting to reflect the pollster’s projection of the demographic makeup of the coming electorate, applying a multiplier to actual data to increase the men and younger people, and to assume their preferences or extend the ratio of preferences as set by a sample deemed too small, or to do whatever it is the pollster thinks needs to be done to true-up his poll … that’s making it up.
You may be making it up very smartly. You may be a scientist. But you’re a bit of an artist too.
This all arises from two factors:
One is that people don’t use landlines so much any more, but cell phones. And they otherwise do their telecommunications via the Internet through email and Facebook and Twitter and the rest.
So the old reliable drawing of a random sample from the white pages and letting the data fall where it falls is no longer so reliable. Now we have pollsters calling not only cell phones, but building samples from Internet users.
The second is that campaigns now put out their own “internal” polls to shore up their money solicitation. And now the new media regurgitate the self-serving data.
Those kinds of polls have always been suspect owing to self-interest and the ways a pollster can push an answer. But nowadays they merely could be weighted.
So Mark Pryor says he’s ahead and Tom Cotton says he’s ahead.
The New York Times reported a poll a month ago saying Pryor was 10 points ahead and now reports one saying Cotton is four points ahead.
What happened? Did Pryor blow off the tomato festival for a secret getaway at a resort with the Koch brothers?
No. Here’s what happened: That first poll surely was unweighted or misweighted. This new one contains an online component, meaning a sample of Internet users.
But while online usage is nigh unto universal elsewhere, there remains a significant segment of the Arkansas electorate that is not online. And it’s a demographic inclined to vote Democratic.
See what I did there? I made the case for weighting those online numbers to true-up the sample to the peculiar Arkansas electorate. I made the case for making up numbers for Pryor.
After all, he’s going to get the lion’s share of people without computers. And the Democrats are going to push turnout of poor people. And then we need to factor in the gender gap in Pryor’s favor, because, I’m telling you, women are worked up against Cotton.
By the time I’m done, I may have Pryor out front.
The point is that political polls have proliferated; they are sometimes suspect on a partisan basis, and they otherwise are not generally as reliable as they once were. That’s because the way we use telecommunications in America has been wholly revolutionized.
The last time I checked my voice mail on the landline, I had 60 messages. Most were hang-ups, probably pollsters. So the white, aging male Southern moderate was under-represented in those surveys, which hardly seems possible.
The other thing I’m doing is getting this column on the record in case what happens in November bears remote resemblance to what many of the polls are saying now, which seems entirely possible.
John Brummett’s column appears regularly in the Arkansas Democrat-Gazette. Email him at email@example.com. Read his blog at brummett.arkansasonline.com, or his @johnbrummett Twitter feed.