Election Notes, Labor Day Edition
The election that would never end is suddenly speeding toward its conclusion. For so many months, all our speculation was purely theoretical — and all commentary ended with disclaimers about how much things might change. Once Labor Day comes, however, the presidential picture begins to firm up. Over the last 15 cycles, no candidate with a polling lead at Labor Day has ever lost the popular vote. Two have lost the presidency — including Hillary Clinton in 2016 — but polls are no longer meaningless. So as we enter the home stretch, let’s consider where the election stands and what to expect.
What the Polls Say
While not without caveats, Joe Biden enjoys a large lead in national polls, and a comfortable advantage in state polls. According to 538 (more about that site below), Biden currently holds a 7% lead in national polling averages (50.4% to 43%). That’s a substantial lead. Polling errors happen, but rarely at that scale — particularly in races that are as stable as this one, with two very well-known candidates. Keep in mind that errors happen in both directions, so if an 8% error is possible in Trump’s favor, it’s also possible in Biden’s. Finally, Biden has consistently been polling above 50%, an important benchmark. If accurate, that means voters who have already decided to vote for Biden would have to switch to Trump. There just aren’t many gettable voters for Trump to pick up.
Of course, the electoral college changes the picture, so here’s where we are at the state level. Biden leads state polling averages in six Trump 2016 states. Trump trails in all three Clinton-won states he has targeted to flip.
Wisconsin (Biden +7.4%), Michigan (+6.7%), Arizona (+4.7%), Pennsylvania (4.5%), Florida (2.8%), North Carolina (1.8%)
Minnesota (Trump -6.3%), Nevada (6.5%), New Hampshire (-8.2%), Maine (-9.9%)
These show a closer race, but Biden holds a clear advantage. One way to think about them is imagining different polling errors. Even if state polls are off by 4% in Trump’s favor, Biden will win the electoral college 289–249. If the polls are accurate, he’ll win 333–205. If there’s a two-point polling error in Biden’s favor, Biden will win 411–127 (picking up GA, IA, OH, and TX). In other words, as polls stand now in a very stable race, Biden is doing very well.
Now, about those caveats.
National Elections are State Elections
Trump “won” the 2016 election by garnering three million fewer votes than Clinton. He doesn’t need to beat Biden to win the election, he needs to beat him in the right states. Here Dems are rightly concerned that Trump will steal the election via voter suppression, mail-ballot shenanigans, and other efforts. And he is certainly trying. Yet elections in the US are overseen at the state rather than federal level, and so these efforts must happen piecemeal, with participation by state elections officials.
However, in Michigan, Pennsylvania, Arizona, and Wisconsin, the Secretaries of State overseeing elections are Democrats. This is important. Tactics that suppress votes require active efforts by states: limiting polling locations (always, of course, in BIPOC neighborhoods), limiting early voting, dumping people from voting rolls, and, in 2020, limiting mail voting. It’s a lot harder to rig elections in states where officials prioritize ballot access, as Democratic states will do this year. (And it’s worth noting that not every Republican is so committed to a Trump win they’re willing to help him cheat. Just last week, Oregon Secretary of State Bev Clarno, a Republican, ruled against the state GOP on an issue involving the explanatory pamphlet mailed to all registered voters).
I don’t want to sound Pollyanna-ish. Many states with Republican leadership will do these things, and their efforts will earn Trump an additional few points. It’s why wins in Georgia and Texas are much less likely than polls suggest. Florida, which historically runs some of the most corrupt elections, will be an interesting test case. The polls there show Biden with a lead of about 3% — a deficit local Republicans might be able to bridge by suppressive efforts. Winning there would be a strong sign of strength.
Biden may yet win a landslide, but he’ll need to widen the gap in Republican-led states that will attempt to suppress his voters.
Whom to Trust
Election seasons feature a lot of very hot takes. Typically the biggest news story of the year, news orgs trumpet an avalanche of breathless reporting and opinion. One of the biggest criticisms is that their relentless focus on the horse race leads to baroque “theater criticism” — style and form over facts and substance. That’s all valid and true, but what it misses is just how bad the coverage is. Incentives are misaligned; pundits are focused on attracting eyeballs, not getting things right. After an election, the worst analysts suffer zero consequences and, if they drove traffic, are rewarded by leading the charge four years later.
A small group are held accountable: data analysts like Nate Silver at 538. Because they quantify their punditry, they are crucified for even small misses. In 2008 and 2012, Princeton neuroscientist Sam Wang moonlighted as a Nate Silver alternate. He was so confident Hillary would win, he promised to eat a bug if Trump prevailed. That ended his career as a polling data guy, and he turned his attention to gerrymandering.
Nate Silver isn’t the only one standing, but no one has an identity so exclusively tied to elections. He is under incredible pressure to get this right. As a result, he spent weeks fine-tuning the model that got 2016 right, and obsessed over ever factor that might throw it off. Although he’s a lefty, he isn’t even slightly motivated to shade his prediction to Biden; getting the election wrong would be far more damaging to his career.
His analysis, revealed most fully on Twitter, is incredibly subtle. If you want to understand the election, read his feed. It will give you a deep sense of the factors that decide elections (and ones, like the hoary old “fundamentals” pundits cite, that don’t), and what to watch. Silver also understands probabilities in a way no reporter does, and that will inform your thinking the next time a pundit starts babbling about “key factors.”
Is the Cake Already Baked?
Looking forward, everyone worries about events that might scramble the race. Debates, October surprises, economic changes, the pandemic — might these not rewrite the election and allow Trump to rally like he did in 2016?
No one would be fool enough to rule it out, but it’s remarkable how stable the race is and how little has changed it. In head-to-head polls dating back nearly a year, Trump has never fallen below 40% against Biden, and never risen higher than 45%. Biden has ranged from 47 to 53%. During that time, Trump was impeached, 190,000 people died from a pandemic, and 30 million people lost their jobs. No matter — nothing pushed Trump below 40% nor Biden above 53%. It’s possible Biden’s floor is lower, yet his voters tell pollsters there’s almost no chance (5%, similar to Trump’s) they’ll switch their vote. In 2016, undecided voters broke 2–1 for Trump at election day, but this year there are very few to break.
Labor Day is always an important milestone because it signals the moment casual voters start tuning into the election. It’s why the numbers become less volatile. There’s every reason to think that’s especially true this year, with a 48-year political veteran and former Vice President running against an incumbent. It’s hard to imagine a surprise large enough to change anything substantially. Perhaps Biden will become so lost during a debate voters will fear profound senility, but beyond that?
There’s little doubt Trump will become more desperate and more chaotic in the final weeks of the campaign. It’s hard to imagine him conceding defeat or admitting loss. Much about 2020 will create a sense of doubt about the election — on both sides. Those are matters for a different post. The basic contours of the election, however, are coming into focus, and they favor Biden.