Skip to main content
ABC News
Recounts Rarely Reverse Election Results

The Wisconsin vote in the presidential election is undergoing a recount. Green Party candidate Jill Stein, who requested the Wisconsin recount, also has raised funds for a recount in Pennsylvania and is seeking more funds for one in Michigan and possibly other states. Could the recounts possibly change the outcome in any of the states? Not if they go anything like statewide recounts over the last 16 years.

Recounts typically don’t swing enough votes to change the winner. Out of 4,687 statewide general elections between 2000 and 2015, just 27 were followed by recounts, according to data compiled by FairVote, a nonpartisan group that researches elections and promotes electoral reform. Just three of those 27 recounts resulted in a change in the outcome, all leading to wins for Democrats: Al Franken’s win in Minnesota’s 2008 U.S. Senate race, Thomas M. Salmon’s win in Vermont’s 2006 auditor election and Christine Gregoire’s win in Washington’s 2004 gubernatorial race.

Recounts also typically don’t change the margin by an amount that would be large enough to affect the result of this year’s presidential election. The mean swing between the top two candidates in the 27 recounts was 282 votes, with a median of 219. The biggest swing came in Florida’s 2000 presidential election recount, when Al Gore cut 1,247 votes off George W. Bush’s lead, ultimately not enough to flip the state to his column. In each state Trump won or leads in, his advantage is more than 10,000 votes, according to counts to far. Some statewide races that have undergone recounts have far fewer votes than the closest states in the 2016 presidential race, but even in percentage terms, the average swing was 0.2 percentage points, which could be enough to flip Michigan but not any other states (and therefore not the Electoral College; even with Michigan, Clinton would be 22 electoral votes short of the 270 needed to win).

“I think there’s real value in post-election audits, in large part to catch something systemic — even more so for a big election like president, and with our Electoral College rules that can exaggerate the impact of results in one or two states,” said Rob Richie, executive director of FairVote and co-author of a forthcoming report on recounts — updating earlier FairVote reports — from which the data in the preceding paragraphs is drawn. “But as a general matter, the likelihood of a statewide recount having an impact on outcome is extremely small unless the original outcome is exceptionally close.”

Citing data from recent recounts, Marc Elias, general counsel for the Hillary Clinton campaign, wrote in a Medium post on Saturday that “the number of votes separating Donald Trump and Hillary Clinton in the closest of these states — Michigan — well exceeds the largest margin ever overcome in a recount.” The Clinton campaign nonetheless will monitor the recount process, as is typical of affected campaigns in recounts.

It’s possible that typical recounts won’t be useful precedents for recounts of this election. Computer scientist J. Alex Halderman, in a Medium post of his own last week advocating for recounts, argued that electronic voting machines could have been hacked. If they were, hackers could have modified vote totals by more than the usual amount that recounts changed vote totals. However, an analysis of county-level vote data doesn’t provide evidence for hacking: Whether a county used electronic voting machines doesn’t seem to be associated with Clinton’s performance relative to expectations. And even if the machines were hacked, a recount might not find evidence to prove that hacking took place.

Carl Bialik was FiveThirtyEight’s lead writer for news.