poll tracks; Just How Likely Are You?

American Demographics, Dec, 1999 by G. Evans Witt

Diamond mining is a dirty, complicated, and expensive process that involves blasting tons of rock, crushing the boulders between huge metal rollers, and then looking for the gems hidden within.

America's pollsters have begun their biennial search for diamonds, but the task is even harder than looking for those magic crystals of pure carbon. The pollsters, you see, are looking for likely voters - members of that rare species who are registered to vote, are going to vote, and will tell the pollster that they are going to pull the lever in their state's presidential primaries.

That may not sound too hard, in theory. But it turns out to be quite difficult in practice. The main reason is sheer numbers. Take a look at Iowa, which is scheduled to hold its first-in-the-nation presidential caucuses on January 24, 2000. In 1996, 96,451 Republicans gathered for the GOP caucuses on a cold Midwestern night in February. That's not much more than half of the 162,393 Republicans who would vote in the GOP gubernatorial primary in 1998. It's only 7 percent of the 1.4 million Iowans who cast presidential ballots in 1996, and about 5 percent of the 1.8 million registered Iowans that year. And it's an even smaller share - 4.5 percent - of the 2.1 million Iowans who were 18 years of age and older and eligible to vote.

Of course, Iowa's caucuses are, well, a little strange. Technically, they are not elections at all, but meetings, and the rules are a bit weird. There was no count of how many Democrats went to the caucuses in '96, for example: President Bill Clinton was running for reelection, his campaign didn't want any counting, and there wasn't any counting.

So what about New Hampshire? The Granite State holds the first-in-the-nation primary, a real election, and has since 1952. In the 1996 Republican presidential primary, 209,938 New Hampshire citizens cast their votes. In the general election, 499,175 voters went to the polls - 66 percent of the 754,771 registered voters in the state and 57 percent of the 871,000 residents who were 18 years of age and older.

So here is the task that faces the pollsters: In Iowa, you have to find the one out of 22 adult residents who will go to the Republican caucuses. In New Hampshire, the task is slightly easier: The pollsters simply have to identify the one out of four residents who will vote in the presidential primaries. To put this another way, after talking to 1,000 Iowans, the pollster would only have found 45 people who are going to the GOP caucuses. In New Hampshire, talking to 1,000 residents would net the pollster 241 likely GOP primary voters.

Pollsters have developed various methods for identifying likely voters that work pretty well for general elections. This involves asking if the person is registered to vote, if they know where their voting precinct is, how interested they are in the election, and how often they have voted in the past. The combinations of these questions (and often several others) winnows down the public into a group of citizens who are fairly safe bets to actually cast their ballots. In a general election. Where half or more of the people vote.

But all these questions don't work as well in presidential primaries, or any primaries for that matter, where at most one in four citizens in the state may be casting a ballot. Asking if a person voted in the last primary doesn't work as well as asking the similar question about a general election. The habit of voting in general elections is just much stronger than that of voting in primaries, which is often motivated by attachment to or extreme dislike of a particular candidate. But those factors are harder to gauge.

The numbers that make finding likely voters difficult is also the reason that pre-primary opinion polls of all adults in a state or of just registered voters can be wildly misleading. After all, if only one in eight registered voters is actually going to cast a ballot in the primary, why should one put much weight on the opinions of all the registered voters in the state? You are actually interested in the opinions of only one-eighth of that sample and results based on all registered voters don't give you any help in determining who they are.

The pre-primary polls are now pouring out in advance of the 2000 presidential primary contests. Each of these polls should be weighed carefully. Pre-primary polls are difficult to do well. After all, the pollsters have to go out and find those rare diamonds: the likely primary voters.