Despite a recent projection in the gubernatorial race for Democrat Josh Stein, the presidential race in North Carolina remains too close to project. According to 538's final forecast, Trump is slightly favored to win -- 59 out of 100 simulations -- but Democrats have long eyed the state as a battleground as Republicans' margins have shrunk in recent years. In 2016, for instance, Trump won the state by about 3 percentage points compared with his 1-point win in 2020. The last time the Tar Heel State voted for a Democratic presidential candidate was in 2008, when Barack Obama narrowly defeated John McCain. Four counties in the state are fully reporting: Graham, Pender, Polk and Swain. In three of the four, Trump increased his vote share from the 2020 race. But in one of the counties -- Pender -- Trump's support decreased slightly, though he still has a significant edge over Harris here.

In polling this cycle, Democratic Senate candidates have often overperformed their party's presidential candidate. For example, 538's final polling average in Arizona has Democratic Senate candidate Ruben Gallego up by 4 percentage points over Republican Kari Lake, while Trump is up by 2 percentage points in the state. Trump is also leading slightly in Nevada's state polling average by 0.3 percentage points, while the Democratic Sen. Jacky Rosen polled ahead of her challenger by 6. With smaller gaps, the same is true of each competitive Senate race in key swing states, with Democratic Senate candidates doing better than Harris in their respective races in Wisconsin, Pennsylvania and Michigan. At a time when elections are highly nationalized and split-ticket voting is down, can we expect these gaps to persist come Election Day?

To answer that question, I looked at October and November surveys leading up to Election Day in the 538 databases for competitive 2016 and 2020 Senate races (those where the final two-party margin was 10 points or less) that also included questions about that year's presidential race. I then averaged how each Republican and Democratic candidate polled and compared those averages with the election results.

For example, in Florida in 2016, Republican Senate candidate Marco Rubio led Democrat Patrick Murphy by an average of 5.4 percentage points in the polls, while presidential polling in the state gave the Democrats a smaller 1.9 percentage point advantage. That meant the polls had the GOP Senate candidate outrunning the GOP presidential candidate by 7.2 percentage points. Come Election Day, both Republicans prevailed, but Rubio outpaced Trump by 6.5 percentage points. So while the polls were off, they were just about accurate as to Rubio's overperformance relative to Trump.

By comparing those differences across states, we can estimate how well polls do at estimating over- or under-performance by Senate candidates. It wasn't just Rubio's race -- across the seven races we examined in 2016, the polls generally predicted which Senate candidates overperformed and which underperformed the top of the ticket, and the magnitude is typically in the right ballpark, too. The average difference in the polling gap and results gap between Republican candidates and Trump was an absolute difference of 0.9 points, with the biggest miss coming in New Hampshire, where Democratic Senate candidate Maggie Hassan and Hillary Clinton actually finished much closer than polls predicted.

The story is similar in 2020, as the chart below shows. Of the 10 states we analyzed, the polls predicted the direction of the overperformance in all but one. The exception was North Carolina, where the polls indicated that the Democrats' Senate candidate Cal Cunningham would narrowly outperform Biden. In fact, both candidates did a bit worse than polling indicated, though Cunningham was dogged by a late-breaking scandal and ended up underperforming Biden by under half a point.

Certainly, the polls didn't get the magnitude of the overperformance right in all cases in 2020. Take Maine, where the polls showed both incumbent Republican Sen. Susan Collins and Trump both losing, but Collins overperforming Trump by 7.8 percentage points. In fact, the gap was more than double that, with Collins going on to win easily while Trump lost, overperforming him by a whopping 17.7 points. Still, though the polls understated GOP support in both 2016 and 2020, they were much more accurate in identifying when a Senate candidate is likely to do better than the party's standard bearer.

This year, swing-state Democrats may be hoping to take a page from Collins' book, playing up their bipartisan, moderate credentials and previous work with Trump in an effort to win over split-ticket voters. Recent history suggests that their current overperformance in the polls is a good sign for them, though certainly not a sure thing.

In the last two presidential elections, there have been some major surprises. Surprise was, of course, the theme for 2016, when Trump won a decisive Electoral College victory with the crumbling of the "blue wall" of Wisconsin, Michigan and Pennsylvania. With Clinton ahead in the polls that year, this a significant upset. In 2020, Biden's victories in Arizona and Georgia came as a surprise to many observers -- neither had gone to the Democratic presidential candidate since the 1990s.

Is it possible for us to be surprised this year? Most race predictions allowed for a wide range of outcomes, with either candidate winning the popular vote, the Electoral College or one but not the other. The major competitive states still seem like anyone's guess. A Harris victory in North Carolina, where votes are still being counted, might be a slight surprise -- Trump won there in both 2016 and 2020 -- but so far, we haven't seen anything unexpected, and it's not clear that we will.

ABC News projects that Trump will win the presidential contest in Arkansas, gaining the state's six electoral votes. He won with 61% of the state's votes against Clinton in 2016 and 62% against Biden in 2020. His former press secretary, Sarah Huckabee Sanders, is the state's current governor.

Back in 2020, we anticipated that some states would see especially pronounced "red shifts" and "blue shifts" in their vote tallies as returns came in through the night. Here in 2024, we can expect at least some of the same, although the picture will differ in some respects.

That is, some batches of votes that are more heavily Republican-leaning may be reported at similar times, and then batches of more Democratic-leaning votes. This has to do with the tendency for Republicans to prefer voting in-person on Election Day, while Democrats are far more likely to vote by mail; the timeline for when states and their counties process ballots of different types can lead to major gains for each party as they're added to the statewide tallies.

Now, the good news is that we aren't holding the 2024 election in the midst of a global pandemic, and many states have adjusted their electoral rules in recent years to better handle the greater preference for voters to use mail ballots to cast their votes. For instance, Georgia counties must now report their early in-person and mail-in votes within an hour of polls closing, which may lead a large majority of the state's votes to come in pretty quickly and reduce the chances of dramatic partisan shifts as Election Day votes are tallied. Meanwhile, Michigan is allowing localities to pre-process mail ballots ahead of Election Day such that the state will quickly have counts for those types of votes to add in with Election Day votes, again potentially diminishing red or blue shifts. Plus, evidence from early and mail voting suggests we may not see as large a partisan split in preference for voting method, further reducing potential partisan shifts in the election night count.

That being said, some states like Pennsylvania and Wisconsin have not dramatically altered their rules for processing mail ballots, which could mean we once again see some shifts in those places based on the types of ballots that have been tallied. So we still have to be careful to monitor not just where freshly-tallied votes are coming from, but also how they were cast.