B Y J A M E S C A R V I L L E
POLL-UTION Why some voter surveys when my side is winning, a presidential election can't get too boring for me. But I recognize that in this, as in most things, I'm a bit odd. After last week's less-than-scintillating debate, most of you all are sitting there with your eyes glazed over. Even some Democrats are probably praying for some sort of gaffe that might make this race interesting — like Al Gore visiting a spelling bee and telling an eight-year old, "No, it's spelled R-A-D-I-C-C-H-I-O-E." There's so little suspense in this race that the Republican rats are already jumping ship. The Republican chairman in Florida, a usually reliable Republican state Dole absolutely has to win, said last week that "it will be nothing short of a political miracle" if Dole carries Florida. Tommy Thompson, the Republican governor of Wisconsin who was on Dole's V.P. short list, told Don Imus and his millions of listeners, "I thought George Bush's campaign was probably the poorest run presidential campaign — and I think this is a close second." Now, obviously I'm no fan of the Dole campaign. They fired my wife, for God's sake. But this kind of disloyalty is incredible. Where I'm from, you wait for something to die before you bury it. These folks are putting Jack Kevorkian's efforts to shame. As long as the Republicans are already declaring this election a done deal, I thought I might use this opportunity to step outside of my usual rapid-response punditry and get a little professorial. When I'm out on the road, which is just about every day now, people ask me questions about the mechanics of running a political race. A lot of the questions have to do with polling: Why are all the polls so different? Why should one poll give Clinton a 15-point lead and another give him 23 points? And why do some polls jump around like EKGs while others hold much more steady? Very good questions. And the answers say some interesting things about the way the media distort our perceptions of political campaigns. The most reliable way to track a race would be to average all of the national polls together into one big sampling. Since that's not easy to do, your best bet is to follow the Wall Street Journal/ NBC poll, which is the work of Republican Bob Teeter and Democrat Peter Hart. It's the gold standard, and I'll tell you why. First, it is based on a healthy sample of at least 1,000 people. Second, the information comes from phone calls over the course of three or more nights. That means there is time to call back a household that does not answer the phone the first time, keeping the sample as random as possible. Third, Teeter and Hart spend a lot of time screening out respondents who are registered but not likely to vote. Many pollsters say their surveys are based on nothing but likely voters, when in fact they're including hundreds of people who won't get anywhere near a voting booth on election day. Perhaps the best example of a reputable survey that is much less reliable is the CNN/ USA Today poll. It is what is known as a "rolling" or "tracking" poll. For the CNN/ USA Today survey, pollsters make enough calls in one day to get about 250 respondents. Since the time frame is just one day, the pollsters make few call-backs to people who were not home on the first try. Let's say the pollsters collect their 250-person sample on a Monday. To get their final poll result, they take that Monday sampling and average it in with other 250-person samplings they took on Saturday and Sunday. When CNN and USA Today take another sampling on Tuesday, they average that new data in with the results from Monday and Sunday, and throw out the information they collected on Saturday. In other words, the poll keeps rolling along, always factoring in new one-day samplings and tossing out the least-recent results. Hence the name "rolling" poll. Sounds pretty solid, doesn't it? Well, it's not. Even though the overall sample includes data gathered on three separate days, each day's sample is so small it often produces screwy results. And one day of screwy results can easily throw off the three-day average. I'll give you an example that makes for some good cocktail-party trivia. Watch the CNN/ USA Today poll on Mondays. You'll see it often registers an upward bounce for Democrats on that day. Why? Because their sample is skewed in favor of those who tend to be at home on weekends, which is to say those who are older and those who are less well-to-do. Both of those groups tend to vote Democratic. By now, you're surely wondering why so many news organizations rely on these lousy rolling polls. It's simple: they're trying to do their polling on the cheap. Small sample sizes are obviously less expensive, especially if you're not doing a lot of call-backs. There's another reason news organizations are not eager to switch to better methods: Volatile polls are not necessarily a bad thing. In fact, they're a great way of adding artificial excitement to a boring race. Hell, why rely on a scrupulous, uneventful poll when you can commission a crappy poll that shows a big overnight swing? A big swing makes big news. So the next time you see a television commentator or op-ed writer pontificate about this nation's fickle voters, take their words with a grain of salt. Many of the fluctuations we see are more artifact than real. Many of them are the media's own self-serving creations. Is the media too addicted to political polls? Join Carville in Table Talk.
Bookmark: http://www.salon1999.com/weekly/carville.html |
