Editor's Note: The following information is from a forthcoming report on statewide election recounts. For more information, please contact Michelle C. Whittaker at firstname.lastname@example.org or 301-270-4616.
The ability to handle a recount of votes to ensure fair, accurate and genuinely democratic outcomes is broadly recognized as a critical component of effective election administration. Even though errors by humans and machines typically mean that any large count of ballots will result in slightly different vote totals from a previous count, the great majority of elections in the United States result in clear winners that are not disputed. However, particularly close elections may necessitate recounts. Losing candidates may challenge the outcome on the grounds of fraud or administrative error.
Trust in elections requires trust in a jurisdiction’s recount process – it is an ongoing process in learning how to conduct recounts well and determining when victory margins and data from post-elections audits should trigger a recount. In this report we quantify various aspects of statewide recounts in the United States between 2000 and 2015, including how often they occur, how often they change outcomes, how much vote totals change and how these figures vary with the size of the electorate. We conclude that:
- Recounts rarely occur: From 2000 to 2015, there were 27 recounts out of 4,687 statewide general elections – an average under two per year. Of these recounts, twelve were not consequential, meaning the original margin of difference between the top two candidates was 0.15% or more of the vote share. While recounts are generally rare, occurring on average once every 173 statewide elections, consequential recounts were even more unlikely, and took place in one out of every 312 statewide elections.
- Recounts rarely impact the margins: The mean average change in the vote margin in the 27 statewide recounts in 2000–2015 was 0.0191% of the vote (or 282 votes). The largest margin change occurred in Vermont in 2006, where initial errors in hand-counting resulted in an 0.11% shift in the recount margin. The next largest shift in the margin among the remaining 17 recounts was 0.076%. Although recounts with original margins of over 0.15% resulted in larger margin shifts (in terms of ballots cast) relative to recounts with closer margins, this margin shift on average typically widened the gap between the winning and the losing candidate instead of decreasing it.
- The election outcome was changed in 11.1% of all statewide recounts and 20.0% of all consequential recounts, representing one out of 1,562 statewide elections: Recounts altered the outcome three times in the 27 statewide recounts during the 2000–2015 period – that is to say, a recount changed the statewide election outcome one out of every 1,562 statewide elections. Recounts reversed the outcome, with 18.7% of all consequential recounts decided initially by less than 0.20% resulting in a change in the outcome.
As the number of voters increased, the shift in vote margins declined: Our analysis of statewide elections founds that random errors did not increase proportionally to the number of ballots cast. This means the impact of a single recounted vote on the margin of victory should decrease as the number of votes in an election increase. For example, correcting a single miscounted vote in an election with ten votes cast would change the margin by 10%, but a single error in an election with 1,000 votes would change the margin by only 0.1% percent. In the 27 statewide recounts in 2000–2015, a recount’s effect on victory margin indeed declined as the number of votes cast in the race increased. Consider that:
- For elections with combined vote totals under one million (ten cases), the margin swing (meaning the percentage in the margin changed by the recount) was on average 0.0389% of total votes cast.
- When the total number of votes cast was in the range of one to two million (ten cases), the margin shift was on average 0.0188% of total votes cast.
- When the total number of votes cast was above two million (seven cases), the margin shift on average was 0.0160% of total votes cast.
- No recounts were required for any election where more than six million total votes were cast for the two leading contenders.