Because I turned 18 in 1999, I had a chance to vote for the first time in the 2000 general election. A life-long Illinois resident, I registered in Wisconsin, where I was attending school. On Election Day 2000 I stood in line with a friend at the Memorial Union at University of Wisconsin-Madison and cast my first ballot. Two years later, in my last year of school, I almost skipped the mid-term elections, but a last-second sense of guilt overcame me and I jumped on my roommate’s scooter, making it to the polls right before closing time.
I’ve since moved to New York City and despite the appeal of filling out an absentee ballot and mailing it to a state that actually matters, I’ve registered in New York. I’ll be voting, of course — politics is my livelihood.
Thus the shock Monday when I took the seven-question test Gallup uses to determine who’s a “likely voter” and who isn’t.
I flunked.
There you have it, a Campaign Desk reporter with a bachelor’s degree in history and political science and a perfect track record in previous elections doesn’t have a history consistent with those who show up on polling day. More to the point, if Gallup were to poll me I would not be included in the likely voter horserace number that often makes headlines in USA Today and on CNN broadcasts.
Every polling outfit has its own unique way of determining likely voters. There are various schools of thought on this issue, but all focus on past voting behavior or self-identified intent to vote. Some think past voting behavior is the best indicator for a likely voter, others think self-identified intent to vote provides the best results, and still others use a combination. Some pollsters, like the American Research Group’s Dick Bennett, think you can determine a likely voter in three questions. Others, like Gallup and Pew Research, think it takes seven questions.
And as the Harris Poll recently showed, different likely voter models can alter the results. On October 20 Harris released its latest round of numbers applying two likely voter models. One model looked for respondents who are registered to vote, who are “absolutely certain” they will vote, who were old enough to vote in 2000, and who did so. That group gave Bush an eight-point lead, 51 percent to 43 percent. But when Harris applied a model that did not factor in whether the respondent had voted in 2000, Bush’s lead shrank to 2 percent, 48 percent to 46 percent.
Gallup uses a seven-question model that is weighted heavily toward past behavior. A reasonable facsimile thereof is available for every political junkie to sample via USAToday.com. (Note: The USA Today reproduction is not entirely accurate and should not be taken at face value.)
So, let’s examine why I am not a likely voter, question by question. Respondents receive one point for each answer that indicates he or she is likely to vote; in order to be considered a likely voter a respondent must score a six or seven.
Question One: How much thought have you given to the upcoming election for president? Quite a lot or only a little?
Thomas Lang: As it’s my job to report on and critique campaign coverage, I answered “quite a lot.”
Question Two: Do you happen to know where people who live in your neighborhood go to vote? Yes or no?
TL: No. I have a sheet of paper sent to me by the New York City Board of Elections confirming that I am registered to vote which I know lists my voting place, but I can’t tell you off the top of my head.
Question Three: Have you ever voted in your precinct or election district? Yes or no?
TL: Nope. I just moved to New York City in January.
Question Four: How often would you say you vote?
TL: I would say always. (Gallup assigns a point for those that answer “always” and “nearly always.”)
Question Five: Do you plan to vote in the presidential election in November?
TL: Yes.
Question Six: Did you vote in 2000?
TL: Yes.
Question Seven: If “one” represents someone who definitely will not vote, and “ten” represents someone who definitely will vote, where on this scale would you place yourself?
TL: Ten. (In the past, Gallup assigned points for eight, nine, and ten. More recently Gallup is only assigning points for nine and ten.)
Gallup would award me a score of five for these answers, docking me a point each for not knowing my polling place and for not having voted in my precinct in past elections. Essentially, I’m not a likely voter because I moved since the last election.
In Gallup’s defense, its conclusions about transient voters are not fabricated out of thin air. The Census Bureau’s report (PDF) on voting and registration in the 2000 election found that “Individuals with more established residences, as measured by home ownership and duration of residence, were more likely to vote than those who rented housing or recently moved into their homes.” Gallup’s Senior Editor, David Moore, said that my situation was unique and represented a very small portion of people that will vote but are not included in the likely voter model.
However, those who move are not the only ones who often find themselves excluded from the Gallup likely voter model. Since the Gallup model is based primarily on past behavior, it’s difficult for those planning to vote for the first time to make the cut as a likely voter.
Imagine that you have never voted before but, for whatever reason, you feel compelled to vote this time around. Now take the test that I failed to pass. You have thought about the election “a lot” (one point). You know where to go to vote (one point). No, you’ve never voted in your precinct (zero points). You’d say that you seldom vote because you’ve never voted before (zero points). Yes, you plan to vote in this upcoming election (one point). No again to the next question because you did not vote in 2000 (zero points). Finally, you respond that on a scale from zero to ten you give yourself a nine because you are pretty sure nothing is going to keep you from going to the polls on November 2 (one point). Total up your score: You get a four. In Gallup’s likely voter world, you’re a dead man. In most cases only those people with scores of six and seven are counted in the likely voter model. Therefore, you, Mr. or Ms. New Voter, aren’t even close to being a likely voter as defined by Gallup.
Gallup’s Moore confirmed that a hypothetical “35-year-old who has never voted but absolutely intends to vote is not included as a likely voter.”
As always, there’s a catch. According to Moore, Gallup cuts voters from the ages 18 to 21 a break. Since respondents in this group were ineligible to vote in 2000, the past voting behavior questions are essentially discounted, and 18-to-21-year-olds make the grade as likely voters if they indicate they are interested in the election and intend to vote. For this group, Moore says, “it’s mostly about desire” to vote.
Since it’s almost impossible for older first-time voters to show up as likely voters and easier for younger first-time voters to show up, 18-to-21-year-olds make up 85 percent of first-time likely voters. First-time voters, according to Moore, account for six percent of all likely voters in Gallup’s survey.
Yet just under 12 percent of all adults indicate they’ll be first-time presidential voters this year. Other numbers indicate that Gallup’s six percent estimate might be low. Election 2000 exit poll results found that four years ago, self-identified new voters made up nine percent of the electorate. As for this election, other polls, such as the ABC/Washington Post poll have reported the number of first-time likely voters at ten percent.
If Gallup’s six percent number for new voters is unrealistically low, how does that affect the overall horserace number?
First off, as some Democrats including Kerry’s pollster Mark Mellman, have noted, first-time voters trend Democratic. A 2000 election exit poll found that first-time voters preferred Gore over Bush by nine percentage points, 52 percent to 43 percent. In comparison, return voters were split down the middle on Gore and Bush, 48 percent to 48 percent.
Gallup’s poll itself confirms that preference for 2004. First-time likely voters, Gallup found in its most recent poll, prefer Kerry to Bush by a 14-point margin, 56 percent to 42 percent. Repeat voters — the other 94 percent of likely voters — prefer Bush to Kerry, by 52 percent to 45 percent. (It’s worth noting that margin of error for the Gallup poll’s likely voter results is plus or minus three percentage points and is accurate 19 times out of 20. The error is greater for sub-groups. The margin of error for the Harris poll is plus or minus 3 points and is accurate 19 times out of 20. All polls are also subject to other errors including question wording and order.)
Even weighting first-time voters so that they comprise 12 percent of the likely voters moves the final numbers by only about one percentage point, says Moore, taking away about half a point from Bush and adding half a point to Kerry.
Our own decidedly primitive arithmetic supports Moore’s assertion.
Without a doubt, one can raise questions about Gallup’s likely voter model. Moore admitted that the likely voter model is “probably going to exclude some people who will actually vote and is probably going to include some people who will not vote.” The key, he says, is to find a balance that provides a group of likely voters that will represent the opinions of those that will vote on election day — not necessarily identify those who will actually go and vote.
While Gallup has voluntarily opened up its process to scrutiny, many pollsters play it close to the vest when it comes to likely voter models. Mark Blumenthal, a Democratic pollster and author of the Mystery Pollster blog, which has explored the likely voter conundrum, told me that pollsters “treat their likely voter model as a secret sauce” feeling that it would put them at a “competitive disadvantage” to release their exact formula. Sure enough, Zogby International spokesperson Shawnta Watson Walcott was only authorized to tell me that they ask six to eight questions, some of which focus on past voter behavior, some of which do not. (A Slate story offers some insight into which pollsters release information about their likely voter models.)
As for my own fate as an unlikely voter, I’ll keep in mind that if I’m going to move before 2008 I had better do it before the mid-term elections. Otherwise, even if I signed on as staffer in a political campaign or ran for office myself — I was thinking about dogcatcher on the island of Maui, actually — I could be doomed to another four years in the unlikely voter ghetto.
Thomas Lang was a writer at CJR Daily.