Don’t believe the polls on third-party contenders

July 24, 2012

Could a third-party candidate sway the 2012 election? Joe Trippi, for one, thinks that former New Mexico governor Gary Johnson could play spoiler this year. The Libertarian candidate for president is getting 5.3 percent of the national vote in the latest Zogby poll.

But it's difficult to predict how third-party candidates will do. Historically, their success has never correlated all that well with normal electoral predictors such as economic growth or war fatalities. The only even halfway decent predictor appears to be presidential approval, and even presidents with approval below 50 percent, such as Gerald Ford or George W. Bush are occasionally spared a strong third-party challenge:










The correlation is only 0.56 — real, but not particularly impressive, especially given that we're only dealing with the 11 open races since 1948. The correlation is even lower if you include open races. Despite the very low approval ratings of Harry Truman in 1952 and George W. Bush in 2008 — a low approval which rubbed off on the candidate nominated by the incumbent's party — strong third-party challengers didn't emerge in those years.

One thing is for sure, though — we can't rely on polling, especially at this stage, but even later on, to predict how well a third-party candidate will perform. To see why, look at the Zogby tracking poll's estimates of Ralph Nader and Libertarian Bob Barr's vote shares in 2008. Nader averaged 1.75 percentage points in August; he ended up with 0.56, less than a third of that. Barr averaged 3.75  percent in August, with one poll giving him 6 percent, but he ended up with 0.4, even less than Nader.

The same story played out in 2004. The Zogby poll consistently estimated that Nader would get 1 to 2 percent of the vote, and he ended up with 0.38 percent. The point isn't that Zogby is especially incompetent, but that polls are bad at estimating small percentages like this.

Years with more prominent third-party runs than 2004 and 2008 don't see more effective polling either. Take 1992, when Ross Perot got 18.91 percent of the vote. In June 1992, Perot was beating both George H.W. Bush and Bill Clinton 39 to 31 to 25, respectively. In May, he and Bush were tied at 35 to Clinton's 25. Toward the end, the polling started to match results, but for much of the early summer, Perot was beating both candidates in the Gallup polls, and until he temporarily withdrew from the race in early July 1992, he was beating Clinton:








Same story in 1980. Moderate GOP Congressman John Anderson got 7 percent of the vote — a lot, by modern standards. But early June polls actually had him at 24 percent:








It's difficult to know for sure whether Gary Johnson will end up swaying the election in November. Something dramatic could change between now and then and make Johnson a real factor in the race. But be very, very suspicious of early polling.

Show Comments
Most Read Business
Next Story
Brad Plumer · July 24, 2012