Skip to main content
ABC News
What 2021’s Biggest Upset Elections Tell Us About The Losing Parties

2021 was bookended by high-profile elections. In early January, traditionally GOP-leaning Georgia sent two Democrats to the U.S. Senate in two runoff elections, tipping control of the chamber into Democratic hands. Less than 10 months later, on Nov. 2, blue-leaning Virginia elected a Republican to statewide office for the first time since 2009.

To be sure, the races in those two states were fought in very different contexts — and had very different outcomes. But even so, they shared a key feature, as our analysis of precinct-level data makes clear: The losing party lost the most ground on its own home turf. In Virginia, it was heavily pro-Biden precincts that delivered the governor’s seat to the GOP, while in Georgia, it was heavily pro-Trump precincts that sent two Democrats to the Senate.

Why is that noteworthy? First, the results provide a valuable reminder that while voters tend to seek balance by backing the party out of power, that balancing isn’t guaranteed. Voters’ support for the party out of power depends on both parties’ positions, candidates and rhetoric. Also, at a time when commentators have emphasized geographic segregation by party, our joint analysis of Virginia and Georgia found that both election outcomes hinged on reduced geographic polarization, with each party losing ground in its respective stronghold to the other.

Let’s start with Virginia. To understand how now-Gov.-elect Glenn Youngkin defeated Terry McAuliffe in a state in which now-President Biden defeated former President Donald Trump by 10 percentage points in the 2020 presidential election, we compared McAuliffe’s vote share to Biden’s and Youngkin’s vote share to Trump’s in each election precinct. From left to right, we then plotted the change in votes from the 2020 presidential election to the 2021 gubernatorial election by how much a precinct supported Trump in 2020. Where the lines are flat in the chart, gubernatorial performance equals presidential; downward slopes mean that the gubernatorial candidates performed worse than their party’s presidential candidates, upward slopes better. The steeper the slope, the bigger the change those precincts provided to the outcome.

And as the chart makes clear, McAuliffe’s biggest underperformance relative to Biden came in heavily Biden-leaning areas. We see that in the blue line’s sharp drops on the left side of the chart — and that McAuliffe’s underperformance there was much larger than his losing margin of 64,000 votes.

Now, in a lower-turnout race, you might expect a party to lose most of its votes in the communities where its support was initially higher.26 percent lower in the Virginia gubernatorial election than in the 2020 presidential election, with both McAuliffe and Youngkin receiving fewer votes than Biden and Trump. Of course, we would expect a non-presidential election to have lower turnout than a presidential election.

">1 But the story here is more complicated because if we look at heavily Republican Virginia precincts — those that gave Trump at least 70 percent of their two-party votes, which are on the chart’s right side — Youngkin saw a net gain of nearly 3,000 votes. That’s striking: For all the coverage of Trump’s strength in certain communities, Youngkin actually outperformed Trump in Virginia’s most pro-Trump precincts in an off-year election.partisan divisions in vote method in 2020, that alone doesn’t explain the patterns we document here. ">2 

Still, it was in the pro-Biden precincts that this race was decided. In those precincts, the Democratic vote total dropped by hundreds of thousands, giving Youngkin the margin he needed to win.

As for the Georgia Senate runoff elections, our second case study, the Democrats actually won both races even though off-cycle elections tend to go poorly for the president’s party. Technically, Trump was still president and in his final days of office, but by the time of the Jan. 5 runoffs, it was clear Biden had won the presidency and would be soon assuming office. As such, GOP Sens. David Perdue and Kelly Loeffler should have been well positioned for the runoff. After all, in a similar situation in 2008, then-Republican Sen. Saxby Chambliss turned a 3-point lead in the first round in November into a win of almost 15 points in the December runoff.

Still, the period between November 2020 and January 2021 was hardly ordinary, given Trump persistently and falsely insisted that he had, in fact, won the presidential election. He also repeatedly attacked Georgia election officials, pressuring them to overturn the results of the election. And, of course, this all culminated in the attack on the Capitol on Jan. 6, the day after the Georgia runoffs.

Demonstrators in Washington, DC hold up signs that read “Black Voters Matter.”

Related: How Democrats’ Failure To Pass A Voting Rights Bill Fits A Pattern Of Failing Voters Of Color Read more. »

This is important for understanding what happened in Georgia, because Trump’s rhetoric had the potential to undermine faith in voting and possibly dampen turnout among Trump’s supporters. Trump’s claim that he had won the election also made it harder for Perdue and Loeffler to convince voters that, if they returned to the Senate, they would serve as a check on the new Democratic president. 

At the same time, we should be careful not to focus on just one explanation. Trump’s approval rating had also fallen between November and January, and other events, including the Republicans’ and Democrats’ campaigns, could have proven influential, too. 

But once again, to understand how each party won in a state that usually backs the other, we compared their vote share to Biden’s and Trump’s in the 2020 presidential election across precincts. In this case, we just looked at Perdue’s race against Ossoff, as that contest proved closer (1.2 points) than Warnock’s against Loeffler’s (2 points).

Turnout declined less dramatically in Georgia from the presidential to the runoff than in Virginia from the presidential to the gubernatorial,3 but as in Virgina, we plotted the change in two-party votes from the 2020 presidential election by how much a precinct supported Trump in 2020. And as you can see in the chart below, both Ossoff and Perdue won fewer votes relative to their party’s presidential candidates in all types of precincts.

Still, the places where they lost votes weren’t even. There was pro-Democratic mobilization in heavily Black areas, but the votes cast did not quite keep pace with presidential votes from 2020 (lines slope slightly downward). Ossoff lost more ground relative to the November presidential race in heavily pro-Biden precincts, meaning that Perdue improved on Trump’s margin in those precincts. In fact, if Perdue had also received Trump’s margin over Biden elsewhere, he would have won. 

But that’s not what happened. Instead, Perdue ended up losing the election primarily because of underperformance in heavily pro-Trump precincts, as the sharp downward slope of the Republican line on the chart’s right side makes clear.

Simply put, Perdue lost the election in places where Trump did best. We cannot say for certain why he underperformed so substantially heavily pro-Trump precincts — whether the underperformance reflected turnout, persuasion or both. But because Ossoff did not see large gains relative to Biden in those precincts, it is reasonable to think Perdue’s problem was, at least in part, that many November Trump voters stayed home in January.

Right after the January runoffs in Georgia, it seemed quite plausible that in heavily pro-Trump precincts, Trump had a mobilization advantage over a more traditional Republican like Perdue. But the Virginia results showed that another businessman-turned-politician could outperform Trump, even in Trump’s strongholds and even in a governor’s race in which overall turnout was less than in the presidential election. 

When viewed alongside Virginia, decreased turnout in heavily Republican precincts in Georgia’s runoffs lends itself to another interpretation. Republicans don’t always face a disadvantage without Trump on the ballot — Virginia makes that clear. But they did face a disadvantage before Biden took office, when Trump’s false election claims and accusations against Georgia elections officials were at a fever pitch

But perhaps the biggest takeaway is that in both contests, Virginia in November and Georgia in January, the losing party faltered most on its own home field.

It’s still not time to pay attention to 2024 polls | FiveThirtyEight Politics Podcast


  1. Turnout was 26 percent lower in the Virginia gubernatorial election than in the 2020 presidential election, with both McAuliffe and Youngkin receiving fewer votes than Biden and Trump. Of course, we would expect a non-presidential election to have lower turnout than a presidential election.

  2. Some caution is warranted here because absentee ballots are only reported at the county level, and the 2020 election saw nearly double the rate of mail-in ballots as the 2021 election. That means some of the change in votes is likely due to 2020 absentee voters returning to in-person precinct voting in 2021, although given the extent of partisan divisions in vote method in 2020, that alone doesn’t explain the patterns we document here.

  3. Turnout in the January 2021 runoffs was down by 10.3 percent compared to the November presidential election.

Seth Hill is a professor of political science at the University of California San Diego. He studies American elections, representation, and political beliefs.

Dan Hopkins is a professor of political science at the University of Pennsylvania, and his research focuses on American elections and public opinion.