At the risk of winning the General John Sedgwick award (General Sedgwick, the highest ranking Union general to die in the Civil War, told his troops at the Battle of Spotsylvania Court House in 1864, seconds before being felled by a Confederate bullet, "They couldn't hit an elephant at this distance"), I will make a prediction: despite a flood of media reports anticipating Florida-type recounts in state after state, it's unlikely that any of the key U.S. Senate and gubernatorial races will go to a meaningful recount with one exception: the U.S. Senate race in Alaska. In that race the wildcard is unlikely to be the initial victory margin, but instead disputes over the tens of thousands of write-in votes to be cast for potential winner Sen. Lisa Murkowski.
In a forthcoming FairVote report on recounts, we rigorously identified every statewide general election that took place from 2000 to 2009 and each of those races that went to a recount. We found that:
- A meaningful statewide recount took place once out of every 262 statewide elections in the 10 years of elections from 2000 to 2009: 11 out of 2.884 races. (That would be consistent with just one meaningful recount in U.S. Senate races taking place in this period.)
- Seven additional recounts occurred because of automatic recount laws or requested recounts. Of the 18 total statewide recounts in 2000-2009, the average change in victory margin was 0.027%, or 296 votes.
- Three of these 18 recounts (and three out of the 11 meaningful recounts) resulted in a change in outcome -- that's one out out of every 961 statewide general elections in this period.
We of course still need to make sure the results in close elections are right. We still should work for better post-election manual audits. But our findings (provided in more detail below) are important to keep in mind as we vote tomorrow and settle in to watch the returns tomorrow night.. (And a big thanks to our Democracy Fellow Emily Hellman for her hard work updating the report.)
What You Need to Know About Recounts:
An analysis of a decade of statewide elections, 2000-2009
The ability to resolve close elections is an important component of representative democracy. All states have chosen to hold statewide popular vote elections to elect their statewide elected offices, and many states have laws and plans to hold recounts in close elections Following is a summary of our research into every statewide recount that took place over the ten year period between 2000 and 2009:
- Recounts take place rarely: From 2000 to 2009, there were 18 recounts out of 2,884 statewide general elections. Of these recounts, at least seven were not meaningful recounts. (Their original margin was 0.15% or more, which is well beyond the largest margin shift to take place in any of the last decade's 18 recounts.) In sum, a recount occurred on average once every 160 statewide general elections over the last decade, and a meaningful recount (one where the original margin was less than 0.15%) took place only once out of every 262 such elections.
- Recounts change the margin by insignificant numbers: The mean change in the vote margin in these 18 recounts was 0.027%, or 296 votes. The largest change was in Vermont in 2006, where errors in hand tallies were key to an 0.11% shift in the margin. The next largest shift in the margin among the remaining 17 recounts was only 0.076%.
- Recounts changed the outcome in one out of every 961 statewide general elections since 2000: Recounts altered the outcome three times (16.7%) out of 18 recounts in statewide general elections during the 10 year span - that is to say, a recount changed the statewide election outcome one out of every 961 elections held from 2000 to 2009. Those elections were the Washington State governor's race in 2004, the Vermont auditor's race in 2006 and the U.S. Senate race in Minnesota in 2008.
- As the number of voters increases in an election, the shift in the margin of vote in a recount declines: The margin of victory in which there exists a credible opportunity for changed outcome via a recount theoretically would decrease in percentage terms based on the number of votes cast in the election (e.g., in an election with 10 votes cast, a single miscounted vote would change the margin by 10%, but a single error in an election with 1000 votes would change the margin by only 0.1% percent). In the 18 statewide recounts, as the number of votes cast increased, the percentage of the recount margin shift indeed did decrease: 1) or elections with combined vote totals under one million (8 cases), the margin swing - meaning amount which the margin changed between initial count and recount - was on average 0.0392% of total votes cast (less than 1 for every 2,500 votes cast). 2) When the total votes cast was in the range of one to two million (5 cases), the margin shift was on average 0.0188% of total votes cast (less than 1 for every 5,300 votes cast); 2) When the total votes cast was above two million (5 cases), the margin shift was on average 0.0156% of total votes cast (less than 1 for every 6,400 votes cast). Not a single recount was required for any election where more than six million total votes were cast for the two leading contenders.