Press Box

Projectile Voting

Do election-night predictions reduce voter turnout?

Is it too early for me to start publishing the exit-poll data from the 2008 presidential election?

Slate triggered a ruckus—and made a few enemies—in 2000 by printing embargoed but leaked exit-poll numbers on the days of presidential primary elections, well before the polls closed. Good-government types denounced us for contaminating the electoral process, saying that our acts would deter people from voting. Voter News Service—the election research consortium run by CBS, NBC, ABC, CNN, Fox, and the Associated Press that produced the numbers—threatened Slate with a lawsuit.

Slate agreed to VNS’s demands that we stop running the numbers because we’d made our editorial point about network hypocrisy. Our gripe went like this: All the TV anchors are privy to this exit-poll data, but in their on-air reporting either 1) pretend not to know who the data favor, or 2) routinely violate their own embargo by hinting at the substance of the data with leading comments. For example, while the Michigan presidential primary polls were still open in 2000, ABC News anchor Peter Jennings posted this on the network’s Web site: “[W]e are seeing the first wave of exit polls from the Michigan Republican primary. … Right now, John McCain and George W. Bush are statistically dead even—not a deep breath between them.”

In 2000, we also didn’t want to fight a lawsuit. But we started publishing exit-poll projections again during the 2004 campaign because our legal nemesis, VNS, had been replaced by a new consortium rendering our old agreement null and void. We had another reason for running the numbers: The networks had become even more hypocritical about exit polls. In covering California’s 2003 special gubernatorial election, practically every network gave away the exit-poll goods about Arnold Schwarzenegger’s excellent showing while the precincts were still open—without overtly acknowledging they were doing so.

Alas, it is too early to publish 2008 data. But it’s not too early to revisit the common claim that early information about election results erodes voter turnout. Thanks to a 2005 book that had escaped my notice until now, I think we have the answer to the question. It’s “no.”

Election Night News & Voter Turnout: Solving the Projection Puzzle, by William C. Adams, a George Washington University professor of public policy and public administration, is a compact (149 pages) and expensive ($49.95) volume. The author finds little or no evidence that significant numbers of West Coast voters avoided voting after hearing the projected winner on television or radio.

Adams analyzes voter behavior from the last dozen presidential elections. In eight of these, the networks projected a winner before all polls closed. From focus-group research with registered voters in Western states, he learns that projections make them feel like they’re second-class citizens whose votes the networks publicly devalue. Yet their outrage is manufactured. None of Adams’ subjects were tempted to skip the ballot box after hearing the projections, and none had personal knowledge of anybody doing so, leading him to write that they were outraged by “damage supposedly done to other people.”

Adams tests the thesis that early projections deter turnout with what he calls a “natural experiment.” He looks at voter data produced by two adjoining counties in Oregon, Grant and Malheur, with similar demographics. The counties share the same TV market, and in the test year of 1984, their ballots were remarkably similar, neither having a hot local contest. Even the weather was the same on Election Day.

Polls close in Oregon at 8 p.m., but because Grant County is in the Pacific Time Zone and Malheur County is in the Mountain Time Zone, Grant precincts open one hour later and close one hour later than Malheur precincts. The upshot of this time-zone disparity is that Malheur voters experience one less hour of exposure to network projections than Grant voters. If projections had any effect on turnout, you’d expect to see a lower Grant turnout.

Ronald Reagan smothered Walter Mondale in the 1984 election, and every network projected Reagan the winner by 5:30 p.m. PST. It was the earliest call the networks had ever made as a group. What happened to the turnout in western Oregon? Adams writes:

[The] citizens of Grant had both an hour more of darkness and an hour more to hear TV projections, but they still turned out to vote at slightly higher rates at day’s end than did their cohorts in Malheur.

In the 1988 presidential contest between George H.W. Bush and Michael Dukakis, two of the three networks called it for Bush by 6:21 p.m. PST, and the third by 7:30 p.m. PST. Once again, Grant voters turned out at slightly higher rates. Other time-zone studies of voter turnout that matched counties across state lines echoed the Oregon results.

When Adams inspected the hourly voter-turnout records for 16 elections in Los Angeles County, he found post-projection turnout “slightly higher in every instance than turnout during the corresponding period of the unprojected elections.” He found no evidence that Republicans in the West didn’t vote after the networks called Florida for Al Gore in the 2000 election, nor any hint that more Western Democrats did. And in an election-night survey conducted in the Portland, Ore., area, Adams found only 0.2 percent of registered voters who blamed projections for their decision not to go to the polls.

Obviously in a country as large as ours, some registered voters who intended to vote bailed after hearing a projection. Just because Adams didn’t find them doesn’t mean they don’t exist. But it’s just as possible that an equal number of registered voters in the West who hadn’t planned to vote have at some time rushed to cast their ballot for a Democratic senatorial candidate after leaked exit-poll projections told them a Republican would take the White House.

Adams’ fine book won’t end the argument over projections. But as the 2008 campaign commences, it will become the place where the argument begins.


I’ve voted in every presidential election since 1972 and never voted for a winner. Can anybody out there beat that record? Send dismal voting results and other bad news to (E-mail may be quoted by name unless the writer stipulates otherwise. EarthLink folks: Turn off your spam filters if you want me to write back.)