NOTE: The report referenced in this article has been replaced. For updated information on recounts, see the new report: A Survey and Analysis of Statewide Election Recounts, 2000-2009
- Recounts are very rare: Between 1980 and 2006, there has been less than one statewide recount per year, and that includes all statewide elections from races for governor and senator to judgeships and ballot initiatives. That means that out of over 7000 elections, only 23 have resulted in recounts, either requested or automatic.
- The changes are insignificant: When recounts do happen, the margins of change tend to be a middling couple hundredths of a percent, and only twice has a recount resulted in a change of outcome. That’s right, just two out of over 7000 elections have flipped winners in the past 26 years. In fact, those two elections occurred within the past three years (the 2004 governor’s race in Washington and a 2006 auditor race in Vermont), so between 1980 and 2003 the number of reversed outcomes due to recounts was exactly zero.
- Bigger is better: As the number of voters increases, the need for a recount decreases. A larger pool of voters makes it less likely that a margin of victory would be small enough to warrant rechecking the results, so as the stakes for an election rise, the chances that the election will need disputing drop. That's a relief, isn't it?
Average number of statewide elections per year, 1980 - 2006 (est.): 283
Number of those elections resulting in recounts per year: less than 1
Average change in victory margin following a recount: 0.012%
Sparked by the disputed 2000 presidential election in Florida, Americans have grown to live in fear of recounts; the endless legal battles, bureaucrats huddled in a room staring through magnifying glasses, passionate charges of fraud and corruption, an electorate unsure of whom they might have put into office, and straight-faced discussion of dimpled chads.
No one wants to relive that scenario at any level, but the increased polarization of today's politics heightens the fear for many that a slew of tiny victory margins are leading to recount after recount.
The truth about recounts, however, is reassuring, and is an important reason why losers of very close elections so rarely call for a full recount, as evidenced most recently in the special election for Congress in Georgia where the winner made the runoff by under 200 votes and won the runoff itself by only about 400 votes.
While these statistics mean that Americans can breathe a little easier about the specter of contested elections, they also tell us something about the prospects for a national popular vote for president. The aforementioned Florida recount battle in 2000 is often cited as a prime example of what can go wrong in something as important as a presidential election. Who would want to see it repeated at the national level?
The good news is that the probability of a theoretical national popular vote for president resulting in a recount is extremely low. As we have seen, recounts that change the result of an election are even less likely, so even if a national popular vote were held that was close enough to warrant a recount, it would be extremely unlikely to result in a change in victors.
Most importantly, though, is the fact that the more votes cast in an election, the less likely it is that the margin of victory will be small enough to contest. Without outright corruption of the electoral process (concerns for which demand our constant vigilance) we estimate that in an election with 100 million votes, it would take a margin of 12,000 votes to trigger a recount where the challenger might have a meaningful chance -- albeit not a good chance -- to change the outcome. The closest presidential election in the past century, Kennedy vs. Nixon in 1960, was won by a margin of about 120,000. Even the famously "razor thin" 2000 election had Gore winning the popular vote by more than 500,000 votes. If these elections had been decided by a national popular vote, they would have come nowhere near a recount-inducing margin of victory. The probability of a disputed outcome sharply increases with state-by-state races for presidential electors, because the pool of voters in each race is smaller.
The National Popular Vote interstate compact is making its way through statehouses all over the country to do away with all of the problems posed by the Electoral College's state-by-state races, and it has already been signed into law in Maryland. As direct election of the president comes closer to being realized, some detractors fret about the horrors of a national recount. They can rest assured that, conservatively estimated, a national popular vote for president would only necessitate a recount once every 16,000 years. The odds, it would seem, are in everyone’s favor.
And as for fears of outright vote stealing, well, that's another critical issue for another day.
For more on this subject, take a look at FairVote’s report, A Survey and Analysis of Statewide Election Recounts, 1980-2006.
For more information from FairVote, contact Paul Fidalgo at (301) 270-4616, or by e-mail at firstname.lastname@example.org.