2020 Election Forecast
Fewer than two months remain until the first Tuesday after the first Monday in November—Election Day. Polling changes weekly, and occasionally daily, with media outlets, statisticians, and thinktanks throwing their resources into the effort. The number of polls and specificity of predictions has never been greater, and while the refrain of “ignore the polls” has become something of a battle cry for both Democrats and Republicans since 2016, polling remains central to political strategy.
This year presents a different race, with a different opponent for a now-incumbent Donald Trump, but is nonetheless informed by the patterns seen four years ago. 2016 remains a watershed moment not only in the political history of the United States, but in the world of empirical polling and statistical modeling. No one has forgotten the final results of the 2016 election, and many are aware of the groups and states which moved the needle toward Trump’s electoral victory. In The National Watch’s first prediction of the 2020 race, we look to the past to recount what happened, and what it may mean for the next 50 days.
2016: How Wrong Were the Polls?
In this article, polling averages from popular aggregators RealClearPolitics (RCP), FiveThirtyEight, and 270toWin are referenced. There is some overlap, as all three sites compile their averages using some of the same polls. However, enough variation exists to result in appreciably different averages for each site, both across states and nationally. For the purpose of this analysis, a battleground state is defined as being decided by 3 percentage points or fewer. Margin of error can vary greatly from poll to poll, depending on the size and characteristics of the group sampled, but 3 percent is chosen here to identify the races expected to be the very tightest.
The tables below, compiled from final polling prior to the 2016 election, summarize 1) the states predicted by each site to be battlegrounds 2) states which turned out to be battlegrounds, despite not being projected to be. State names highlighted in red are those projected to be battlegrounds which ultimately were not. The first two columns give the projected margin in each state and the actual margin observed. Negative/red numbers correspond to a margin in Donald Trump’s favor, while positive/blue margins favor Hillary Clinton. The third column shows the difference between predicted and actual margins, with a positive/blue number indicating an overprediction of Clinton’s margin, and a negative/red number indicating an overprediction of Trump’s margin. For example, RCP’s prediction of a Trump win by 1.0 point in North Carolina actually overpredicted Clinton’s strength by 2.8 points, with Trump carrying the state by 3.8.
Out of the three aggregators, RCP proved the most prescient, predicting two-thirds of the nine eventual battleground states in the election. Overall, among the 13 projected and/or actual battleground states above (AZ, CO, FL, IA, ME, MI, MN, NE, NH, NC, OH, PA, and WI), the average margin for the three sites was Clinton +0.5 (RCP), Clinton +2.3 (FiveThirtyEight), and Clinton +2.2 (270toWin). Compared with the true average margin, Trump +1.5, pre-election polling was, on average, within a reasonable margin of error (2.0-3.8%) in battleground states. In short, the polling wasn’t wrong at all, from an overall perspective. Support for each candidate was well-estimated, but the specific states where that support lived was a different story.
Neither FiveThirtyEight nor 270toWin hinted at the closeness of the races in Pennsylvania and Wisconsin, with FiveThirtyEight overestimating Clinton’s support by 4.9 points in Pennsylvania and 6.3 in the Badger State. 270 to Win missed the mark by 5.2 and 8.0 points in the same states. Clinton’s massive underperformance in the Midwest caused Pennsylvania and Wisconsin, along with Michigan and Minnesota, to turn into major swing states. Worse yet, her struggles caused expected battlegrounds Iowa and Ohio to instead be won in blowout fashion by Trump.
Certainly, polling overestimated Clinton’s strength on average, but it did not do so to a historically unusual extent, and in most cases fell within the accepted margins of error. Perhaps it was the unconventional nature of Donald Trump, or perhaps it was hindsight born of shock, but something has perpetuated a misconception in the public imagination. The 2016 election pitted two polarizing figures running campaigns against the backdrop of a polarized nation, and that polarization was borne out in the pre-election polling.
Ultimately, the election was decided by fewer than 80,000 voters in Pennsylvania, Wisconsin, and Michigan. In 2013, the University of Michigan’s football stadium hosted over 115,000 fans during a single game. Three years later, the President was chosen by a group of voters small enough to fit into Michigan Stadium with 37,000 seats to spare.
Some phenomena are beyond the reach of polling.
2020: The More Things Change…
So, where does this leave us in the home stretch of the 2020 election? As in 2016, FiveThirtyEight projects six states (as of this writing) to be within 3 points on Election Day. 270toWin points to the same six states (Florida, Georgia, Iowa, North Carolina, Ohio, and Texas), while RCP swaps Michigan for Texas. All three sites lean on traditional swing states Florida, Ohio, and Iowa, while alluding to the increased purpling of Texas, a state no Democrat has carried since Jimmy Carter in 1976, and Georgia, only claimed by Democratic candidates once since 1980. Read on for The National Watch's September 2020 predictions for the victor in each state. Note that margins shown for each state reflect, in order, current RealClearPolitics, FiveThirtyEight, and 270toWin polling averages.
Florida (Biden +1.8/Biden +2.8/Biden +1.8)
Florida has tightened significantly in the past month. In the summer and fall of 2016, polls swung seemingly monthly between a 4-point Clinton edge and a dead heat, and 2020 is telling a similar story. Between mid-April and late-July, Biden built an 8-plus point lead, per RCP; since then, that advantage has all but vanished. To be fair, Clinton never approached the type of polling margin Biden held this summer, but this should be little consolation for Democrats. The trend is the same, and Trump’s resurgence suggests he is erasing an even larger deficit at a faster velocity than 2016. Moreover, while Biden does well with seniors, his support among Latino voters faltered during the Democratic primaries. As The Hill reported this week, signs of weakness continue to rear their heads in Florida. Polling inaccuracies? Easing Covid concerns? Whatever the cause, the Sunshine State is far from decided.
The National Watch Prediction: Trump
Georgia (Trump +1.3/Trump +1.5/Trump +0.2)
An uphill battle at best, a longshot at worst, Georgia has been more of a pseudo-swing state for the Biden campaign. Typically only competitive when a southern Democrat is on the ticket, Republicans have won the state by five points or more in the last five general elections. However, margins have slimmed as demographics have shifted; as the non-white population increased from 34.9 to 40.3 percent between the 2000 and 2010 Census, double-digit Republican victories in 2000 and 2004 were followed up with a substantially smaller, 5.2-point win in 2008. This wasn’t simply an Obama bump, either; Trump’s 5.2-point victory in 2016 only matched McCain’s 2008 margin and fell well short of Romney’s 2012 edge. The Census Bureau estimated that the percentage of white, non-Hispanic residents in the state had dropped further, to just 52.0%, by mid-2019.
The Biden campaign was revived by a Clyburn endorsement and a leap year triumph in the neighboring Palmetto State. With a rapidly shifting electorate and an energetic push for voting access, the Peach State may indeed prove ripe for a Democratic breakthrough.
The National Watch Prediction: Biden
Iowa (Trump +1.7/Trump +1.7/Trump +1.0)
The demographics of Iowa are shifting as well; in 2000, 93.9% of Iowans identified as white only. In 2019, that number had fallen to…85.0%. Jokes aside, though it’s a far less diverse state than Florida and Georgia, Iowa is the swingiest of all: it has been evenly split, 7-7, between Republicans and Democrats in the previous 14 elections (dating back to 1964). Though that includes just two Republican victories in the eight races since 1988, one of those two came in 2016. Starting out the new millennium with tight races in 2000 and 2004, Iowa gave sizable victories to Obama in 2008 and 2012, only to famously give Trump a nearly double-digit victory in 2016. Trump seems to be on track to repeat that victory in 2020, and though polling shows slim margins, Biden’s fourth-place (very nearly fifth-place) finish in the Iowa Democratic Caucus casts doubt on his strength in the state. While a competitive finish is possible, don’t expect the trailing candidate to pour resources into Iowa down the stretch; due to lagging population growth, Iowa, which once represented 13 electoral votes, will be worth only 6 in 2020.
The National Watch Prediction: Trump
Michigan (Biden +2.6/Biden +6.7/Biden +5.5)
Michigan was decided by fewer than 11,000 votes in 2016. Home to GM, Ford, and a disenchanted working class, Michigan, like Iowa, made an immense red swing after a decisive Democratic win in 2012. If 2016 is any indication, jobs and stability will decide this race. Even prior to the coronavirus pandemic, Michigan, and the nation as a whole, fared better economically under Obama and Biden. Now, in the midst of a recession which has brought the most rapid job and GDP losses in American history, working-class voters will again be voting for a post-recession president. In 2008, they chose a Democrat. They didn’t have to wait long to see the results in their state. Biden’s role as one of the architects of the 2009 auto industry bailout will undoubtedly have an impact on the votes of struggling workers, even 11 years later. The question is, how much? The track record is on Biden’s side, but Trump won for a reason. Voter perception is animated by personal experience, not necessarily macroscopic data. No one can predict how the people of Michigan will perceive the depth of the recession in November. Only time, and perhaps a few more weekly jobs reports, will tell.
The National Watch Prediction: Biden
North Carolina (Biden +0.6/Biden +1.8/Biden +1.4)
As Iowa slows, North Carolina grows. The Tar Heel State has become the ninth-most populous in the nation, representing a 15-vote electoral prize. Like Georgia, Democrats don’t often find success in North Carolina, but also like Georgia, North Carolina is changing. The state is home to rapidly growing city centers such as Charlotte and Raleigh, becoming progressively younger and more diverse as a whole. Trump has the decided advantages of incumbency, the historical redness of the state, and typically low turnout among younger voters. However, North Carolina has the potential to be a major crack in the “red wall”, just as Michigan shattered the blue façade for Trump four years ago. It also shouldn’t be underestimated that here, in a state that prides itself on its military heritage and service, the President’s denigrating rhetoric, past and present, could further narrow his support.
The National Watch Prediction: Trump
Ohio (Biden +2.3/Trump +1.9/Trump +2.5)
Though split on the victor, all three referenced polling sites expect Ohio to remain a battleground in 2020. Ohio, with 18 electoral votes, could be the most important midwestern state in the race for the White House. With roots in manufacturing and a population that is around 80 percent white, superficial factors would seem to favor another Trump victory in the state. Biden has seen stronger support than Clinton in 2012 among seniors in the state, which is slightly older than the country as a whole. This helps to explain the relative tightness of the race, but the failure of Trump to protect manufacturing jobs (a campaign promise), including those at the Lordstown GM plant which closed in 2019, truly hit home for Ohio voters. Trump’s margin over Clinton was unquestionably impressive, but he is by no means safe in this economic climate.
The National Watch Prediction: Trump
Texas (Trump +3.5/Trump +1.0/Trump +2.3)
Did someone say purpling? In 2018, Beto O’Rourke became an insurgent candidate, coming within three points of unseating Senator Ted Cruz in the closest US Senate race in Texas in four decades. More than 55% of Texans are non-white, making Texas one of four so-called “majority-minority” states in the US. The turnout and priorities of these groups will tell us the direction of this race. Will the efforts of the Trump administration to dismantle the DACA program (an Obama-Biden construct), enforce child separation policies, and divert defense funding for border wall construction evoke an en masse electoral rebuke? Will white voters, not a majority but still a plurality, make what could be a final stand against a blue Texas? Ultimately, until Biden shows the ability not just to court, but to enthuse Latino voters, the Lone Star State is Trump’s to lose.
The National Watch Prediction: Trump
Electoral Prediction
Among other states identified by RCP as battlegrounds, but not polling quite so tightly, Biden looks poised to win Arizona, Wisconsin, Virginia, New Hampshire, New Mexico, Minnesota, and Colorado. Maine’s 4 electoral votes will likely be split between the candidates (Biden 3, Trump 1), and the remaining three states, Pennsylvania, Arizona, and Nevada, remain close (4-5 points). Given Biden’s unsteady support among Latino voters, Trump could very well erase deficits in Arizona and Nevada, capturing both states. Our electoral prediction map gives these to Trump, while calling Pennsylvania for Biden. The closest battleground states are shown with gold borders.
Despite polling staying remarkably stable throughout the campaign, this race is far from over. We may have Biden ahead for now, but debates and October surprises are right around the corner. In the meantime, stay tuned for an updated prediction in the weeks to come, and whoever you vote for, vote!