As of this writing, with state officials still counting some votes, Joe Biden appears likely to have been elected president, Republicans to have narrowly held the Senate (pending runoff elections in Georgia) and Democrats to have narrowly retained control of the House.

Here are several takeaways for our understanding of U.S. national elections.

Historically high voter turnout

Over two-thirds of eligible voters — nearly 159 million Americans — cast ballots in this election. That’s proportionally the highest voter turnout since at least 1900. Turnout was highest in states that allowed early voting, vote-by-mail, automatic voter registration or election-day registration.

Such high turnout will have long-term effects on voting patterns, even if this level is particular to the 2020 election. Why? Because voting is habitual: Once someone votes, they are more likely to do so in the future.

Time to rethink election ‘fundamentals’?

Political science considers a few factors to be “fundamental” predictors of a U.S. election’s outcome. And indeed, the poor economy and the covid-19 pandemic probably diminished President Trump’s vote. But these factors are just part of the electoral landscape, and much of the research estimating their impact was done during a far less partisan era. Over the past few decades, and especially during the past eight years, partisanship has increasingly become part of our social identities. As a result, the effects of partisanship may be starting to overwhelm even these long-standing fundamentals.

Candidates can matter even in nationalized elections

As political scientist Dan Hopkins and others have shown, increased partisan polarization and the decline of local news has increasingly “nationalized” elections: Most people vote now for the same party up and down the ticket.

But this doesn’t mean that candidate attributes don’t matter at all: Ideological moderation and strong personal brands can still net candidates a few percentage points on margin. For example, Maine’s Republican Sen. Susan Collins outperformed the president’s vote by some 20 percentage points. And in Kansas, moderate Democratic state Sen. Barbara Bollier fared roughly 4 percentage points better than Biden. Moderate Republicans won gubernatorial elections in Vermont and New Hampshire.

This election will have long-term consequences

Every 10 years, state governments must redraw boundaries for congressional and state legislative districts based on the latest census data. Electoral results suggest that Republicans will control more state governments than will Democrats — allowing GOP-led states to draw maps that favor Republicans while concentrating Democratic voters into fewer districts. These maps will give Republicans an advantage in congressional and state legislative elections for the next decade.

Electoral cycles matter

The vast majority of presidents win reelection to a second term. This incumbency advantage probably helped Trump this year, despite his relatively low approval ratings. Further, the political party that controls the White House nearly always loses congressional seats in the next midterm. For instance, Democrats lost seats in 2010 and 2014 after Barack Obama’s victories in the previous elections, and Republicans lost seats in 2018 after Trump’s victory in 2016.

If Biden wins the presidency, Democrats are likely to lose congressional seats in the 2022 elections. This means Republicans could gain control of the House and keep control of the Senate. (That assumes that Republicans will win both Georgia Senate runoff elections in January 2021.)

However, historically the midterm slump is larger when one party controls the presidency and both houses of Congress than it is when control is divided between the parties. That’s probably because unified control of government enables the party in power to pass unpopular legislation, and the public punishes them in the next election. If divided government makes it more difficult for a President Biden and the Democrats to implement unpopular policies, they might avoid at least some of this midterm penalty.

Don’t trash polls yet

On average, the polls appear to have been off by about 2.5 to 3 percentage points this year. The absolute error in polls is not necessarily larger than in previous years. But, crucially, both public and private polls continued to overestimate Democratic vote share in Florida and in Midwestern states. These inaccuracies persisted despite many methodological improvements in state polls since 2016, including weighting by education, sampling based on the voter file and fielding more surveys closer to Election Day to guard against late shifts in the electorate.

However, it would be premature to abandon survey research generally. Election polls are intended to do a single, highly specific task: predict vote share. In contrast, most survey-based social science research has the goal of better understanding relationships among demographic, behavioral and attitudinal factors, including new work that examines how partisanship influences who follows public health recommendations about covid-19, like wearing a mask.

And in many cases, survey evidence about the public’s position on a particular issue is corroborated by voting patterns. For example, not only do most Americans support raising the minimum wage across a range of surveys, majorities also consistently vote in favor ballot initiatives designed to do this.

Chris Warshaw is an associate professor of political science at George Washington University. He’s on Twitter at @cwarshaw.

Emily Thorson is an assistant professor of political science at Syracuse University. She’s on Twitter at @emilythorson.