Everyone remembers the 2000 presidential election with its excruciating post-election vote counting controversy in Florida and the Supreme Court decision that awarded the presidency to George W. Bush. But how many people remember the Gallup 2000 election tracking poll? I'm sure the Gallup folks would like us to forget it, but in light of the controversy over Gallup's recent poll for USA Today that showed a big discrepancy between registered and likely voters, it's worth recalling just how well Gallup's likely voter screen performed in the weeks leading up to the 2000 election.
As Robert Erikson of Columbia University has demonstrated, Gallup's likely voter screen had the effect of producing tremendous volatility in the results of their daily tracking poll. The volatility was due less to variation in the preferences of registered voters than to differences in which voters were identified as likely to vote.
Just how volatile where Gallup's results? Here are a few examples:
On August 8 Gallup had Bush 2 points ahead of Gore. Six days later, on August 13 they had Bush 16 points ahead. But one week later, on August 20, they had Gore back ahead by one.
The craziness didn't end with the nominating conventions either. On September 21 Gallup had Bush ahead of Gore by 10 points. Four days later, on September 25, Gore was back in the lead by 3. Then on October 5 Gallup had Bush ahead by 11. The next day, on October 6, his lead was down to one, and one day later, on October 7, Gore was ahead by 7.
Here's my favorite. On October 24 Gallup had Gore ahead of Bush by one point. Three days later, on October 27, they had Bush ahead by 13.
In their final pre-election poll Gallup gave Bush a 2 point lead--not too bad since Gore won the popular vote by less than a point. But two days earlier, Gallup had Bush ahead by 5.
No other poll during the 2000 campaign showed anything like the volatility of the Gallup tracking poll and so far Gallup's 2008 tracking poll has shown nothing like the volatility of their 2000 tracking poll. That's probably because they haven't started to apply their likely voter screen to the tracking poll. So here's some friendly advice to the folks at Gallup: Don't, at least not until the last few days before the election when it might actually work.