I was over at my parents’ for dinner, just buzzing by the computer for a quick results check, didn’t even bother sitting down—and I remember seeing that result, sitting down heavily, and thinking, “Welp. The U.K. just left the E.U.”

The rest of the night went the same. Areas that were expected to vote for Brexit by a narrow margin voted for it by a large margin; areas that were expected to oppose it by narrow margins ended up supporting it by narrow margins. There were a few places where the anti-Brexit “Bremain” vote did better than expected, as there always are… but not many. Networks finally called the race hours after the result was clear.

This happened again in the 2016 presidential election. I told my family at 8:15 PM that Trump was very probably the President, because I could see that all the states Trump needed were getting called for him (or were on the cusp of it), and all the states Clinton needed were tossups… most of them leaning red.

I don’t expect that to happen again. President Trump starts today in a much worse position than he was in last time around. Last time, I told you that all Trump needed was a small polling error in his favor to win, that those kinds of polls happened all the time,and that the pundits were wildly underestimating Trump’s chances. Today, if anything, I think pundits (always fighting the last war!) are overestimating Trump’s chances. Last time, the polls were within the margin of historically normal error. That wouldn’t be nearly enough this time. But Trump could still pull it off. Totally plausible. Meanwhile, Republicans have a real shot at holding the Senate, or losing it only narrowly. So let’s take a look at three different ways we can track the race as it unfolds:

Congress-Level

Every seat in the House of Representatives is up for election this year, as usual. Congressional races are probably going to be the earliest (easily-accessible) indicator we have of how the night is going.

Here’s my Newcastle-upon-Tyne for Election 2020, my start-of-the-night bellwether:

Kentucky’s 6th Congressional District race between Rep. Andy Barr (R) and challenger Josh Hicks (D). Here is the Wikipedia article on the district, and here is the FiveThirtyEight race projection.

This is a pretty Republican district (Cook PVI R+9), but it’s a nice urban/rural/suburban mix around Lexington and Frankfort. The electorate resembles the voters both campaigns need most: Midwesterners and Pennsylvanians. (Sorry, Pennsylvania, you’re not in the Midwest.)

KY-6 is considered a battleground district, and polls there close early (5 PM Central), in a state where MOST of the vote is expected to be counted fairly quickly. There’s a lot of justified national concern about “red shifts” and “blue shifts” in the vote totals as the week goes on and absentee votes get counted, but there isn’t a lot of evidence that that will happen in this race.

So how can we map the results of this race onto the national outcomes?

On a “normal” night for Republicans, where the election forecasts are basically accurate, Barr wins by 10 points, dispatching Hicks fairly easily. Then Trump goes on to lose the presidential election and the Democrats end up with 52 Senate seats, because normal is not enough for Republicans who are this far down in the polls. Call this 20% likely to happen.

On a “good” night for Republicans, Barr wins by 15 points. Trump likely still loses the White House, but the Senate is competitive, and the GOP gains seats in the House, though not many. Also ~20% likely.

On a “great” night for Republicans, Barr wins by 20 points, the GOP holds the Senate and gains more House seats, and Trump is a photo finish for the White House. Expect to wait several days for a final result in that race. This is only ~15% likely.

On a “shocking” night for Republicans, Barr wins by 25, the GOP holds the Senate, makes major gains in the House (conceivably even winning it back), and we know fairly early that Trump has not only been re-elected, but has likely won the national popular vote. This is only 5% likely or so.

Now we’ll go the other direction. The probabilities are roughly the same.

On a “bad” night for Republicans, Barr wins by 5 points, Biden wins, Democrats end up with 54 Senate seats and substantial gains in the House.

On a “terrible” night for Republicans, Barr’s race is too close to call, Biden wins, and Democrats end up with 55-56 Senate seats.

On a “horrifying” night for Republicans, Barr loses by 5 or more; Biden wins Texas, South Carolina, Montana, and Missouri; Democrats end up with 57 Senate seats and 260+ House seats.

This is a lot of prognostication to base on the outcome of a single House race in Kentucky, so take it with some big grains of salt. In 2018, what looked like a “bad” night for Democrats at first eventually turned into a “normal” night for Democrats, so there’s definitely room for this to slide around as the night goes on.

This year, there’s a lot more reason to worry about that than usual, for all kinds of reasons (but mostly absentee voting). And there are always a few bizarre flukes in election outcomes, especially lightly-polled House races. KY-6 could be one of them, and that would make this whole analysis worthless.

But the fun thing is… you can run this analysis for any remotely competitive House race. Just take any race, look at the FiveThirtyEight projection and draw the same lines on it.

For example, in Minnesota’s 3rd District, start from the left and count the first five balls. Kendall Qualls is in a photo finish in most of them, so that’s what would likely happen on a “shocking” night for Republicans. In the next 15 balls, Qualls loses by around ~10 points, so that’s what would happen on a “great” night for Republicans. If MN-3 announces that Dean Phillips won by 27 points, you know Republicans are having a “terrible” night.

Do this on lots of different House races as they get called—the more competitive the race, the better—and you’ll have a pretty clear picture of how things are going before the West Coast even starts counting.

UPDATE: It took a few hours to get the vote count fully in, and it started out blue because of the early vote in Kentucky getting tabulated first — but Barr won by just about 18 points, and Republicans went on to have a great night.

State-Level

States are going to be tough tonight. Different states always count at different paces (and California always takes weeks to get it all done), but tonight is going to have a lot more disparity than usual. The most important state in the election by far, Pennsylvania, is unlikely to have meaningful results for several days. Worse: there’s a strong divide between Democrats and Republicans this year on absentee voting, which has never happened before. Worse still: because different states count absentee ballots at different times, some states are likely to report heavily Republican results early, then more Democrats results later on, resulting in a “blue shift” toward the Democrats. Other states will have the opposite happen: a “red shift.”

In short, it stands to be a very confusing night for the decision desks at the various news networks. You, the viewer at home, will be well-advised to wait patiently for network calls at the state level, unless you’re some kind of precinct-level vote parsing machine. (See below.)

That doesn’t mean we can’t still watch the states, though. FiveThirtyEight has a very good Election Night Viewer’s Guide that tells us when each state expects to report results and what “shift” is expected (if any).

I will be following DecisionDeskHQ‘s calls all night. DDHQ is much more transparent and aggressive about its calls than the AP and most news networks, but it also has a track record of being right, because it follows the data.

FiveThirtyEight then has a really wonderful real-time projection calculator. I will be keeping this open all night, updating it as official state calls come in.

If New Jersey goes for Biden early, it doesn’t have much effect on the race… but, if Georgia goes for Biden early, the race is over.

Trump’s must-win battleground states are: Georgia, Florida, North Carolina, and Ohio. He trails in all of them right now. If he loses any of them, this is over.

Biden’s must-win battleground states are: Colorado plus two or more of Michigan, Minnesota, and Wisconsin. If Biden can’t hold his must-wins, he still (unlike Trump) has some chance to rebound (that’s the advantage of a big polling lead), but his path to victory becomes extremely narrow.

The tipping-point state: Pennsylvania. If Trump wins Pennsylvania and holds his must-win states, Biden’s only hope of winning runs through Arizona on what has turned out to be a pretty red night. By contrast, if Biden wins Pennsylvania and holds his must-wins (which he almost certainly will if he wins PA), then Trump has no path to victory. While it may not technically end up the tipping-point state when all’s said and done, it’s hard to overstate Pennsylvania’s importance if the election is close.

Unfortunately, PA is unlikely to turn in anything like complete results for days—weeks if there’s lawsuits. Plus, because Pennsylvania is run by deeply partisan institutions that hate each other and act like idiots, under an election law that might charitably be described as “dysfunctional”… Pennsylvania is ground zero for a Florida 2000-style recount disaster.

See my examination of Pennsylvanian/American election dysfunction, and how it could plausibly (but improbably) spiral into an American civil war, in my recent post “And The War Came.” Then pray for a landslide.

County-Level and Precinct-Level

If you are a bit of a data wizard, this is the best and fastest way to track the election. Once an entire precinct or entire county has reported results, you can compare the results to benchmarks from previous elections and/or benchmarks set by detailed polling and/or benchmarks set by demographic profiles of the electorate. (Ideally, you’d do all three.) Then you can run the same basic analysis I ran through for Congressional races and, boom, you have hyper-detailed data on how the race is going for certain demographics or in certain areas. (Is Trump beating his polls in non-Cuban Hispanic areas? Good news for his Arizona results! Is Biden racking up points in suburban precincts? Good news for his everything results!) Do this a few hundred times, and it aggregates to a hyper-accurate projection of how the race is going.

I can’t manage this this year. My benchmark site from 2016 has gone away, and I’m going to have too much difficulty sorting out the meaning of precinct-level results based on the state’s particular policy on how absentee ballots are reported.

Fortunately for us all, Nate Cohn is still running the wonderful Upshot blog at the New York Times (a spin-off of FiveThirtyEight from its days as an NYT property, because all good election forecasting websites are run by 538 alumni; see also Harry Enten at CNN). And the Upshot is known, above all else, for one thing:

The Upshot is bringing back The Needle! (But only for three states.)

The Needle was the first major news tool to realize that Donald Trump was going to win the 2016 election (permanently traumatizing some), and it did that because The Needle does all this county and precinct tracking that I’m talking about, brilliantly, automatically, in real time.

Due to the same problems I mentioned, The Needle is limited to three states this year: Georgia, Florida, and North Carolina. Oh well. Those are all Trump must-win states.

So keep an eye on The Needle. If Trump wins all three—especially if he wins them early—he’s a tossup to win the election, and it will probably all come down to Pennsylvania. (There are other possibilities, but Pennsylvania looms large.) On the other hand, if Biden wins any of the states The Needle is tracking, or keeps them close for a lot of the night, it’s going to be very hard for Trump to make a comeback. The Needle will tell you how it’s looking a long, long time before the networks call it.

In fact, the network decision desks all basically have their own internal versions of The Needle, and the usual rule, post-2000, is that a network calls a race only once their internal Needles announce a >99% likelihood of a candidate winning it. So watching the Upshot’s Needle is a lot like looking in at a network decision desk in real time, hours before the call gets made.

(Assuming, that is, that The Needle doesn’t break, which is exactly what it did in 2018, to our great dismay.)

That’s it!

One asymmetry worth noting before I sign off: if Biden wins, we are likely to know it by Wednesday morning. If Trump wins, we very probably won’t know for days or weeks. That’s both because Trump is likely to win only in a very close race (which may hinge on a recount), and because most of Trump’s paths to victory go through Pennsylvania, which counts slow.

It’s also possible that a close race takes days to resolve and ultimately goes to Biden, but this is not especially likely. So Trump supporters should be cheering for a long, long night tonight (and tomorrow, and the day after).

For more viewing-guide fun, check out the Upshot’s hour-by-hour guide to the results, and follow FiveThirtyEight’s election liveblog!

Some passages in this post are openly plagiarized from a similar post I wrote in 2016.

This, in the end, was the story of the entire Brexit results night. Areas that were expected to vote for Brexit by a narrow margin voted for it by a large margin; areas that were expected to oppose it by narrow margins ended up supporting it by narrow margins. There were a few places where the anti-Brexit “Bremain” vote did better than expected, as there always are… but not many. By the time the results were tabulated in England’s version of  “battleground states,” the result was fairly clear on the strength of the vote totals in the “safe” constituencies alone.

Lesson for Americans: you can infer a great deal about the state of the presidential race from even early results in small geographic areas.

We talked this morning about the fact that Donald Trump needs to beat his polls, Brexit-style, in order to win the election, while Hillary Clinton only needs to match her polls to beat Trump. That’s why Clinton is a 2-to-1 (nearly 3-to-1) favorite to win. Let’s take a look at three different ways we can find out quickly which story is the true one:

State-Level

FiveThirtyEight has a very good Election Night Viewer’s Guide that goes through each battleground state, hour-by-hour, and explains the significance of each. I’ll focus just on a couple of very early states (poll closing times in each state in parentheses, all times EST):

Trump Must-Wins: Georgia (7:00), Florida (8:00)

Clinton Must-Wins: Virginia (7:00) , Pennsylvania (8:00)

Bellwether: New Hamsphire (8:00)

Early States: Kentucky, Vermont (both 7:00)

One of the interesting things about this election is that the candidates are competitive in so many places that there are very few true “must-wins”; there is almost always a plausible path for either candidate to recover if they lose one of their key states.

But the four states I’ve listed here really are “must-wins” — without them, there is almost no visible path to victory. Trump has a narrow lead in Georgia and is narrowly behind in Florida. Clinton is comfortably ahead in both Virginia and Pennsylvania. You might say, “That’s not fair to Trump! His must-wins are harder than hers!” Well, that’s because Trump is currently losing the election.

Results in these early states will give us a good idea of how things are going to go. If Florida goes early for Clinton, then she is beating her polls and the race is pretty much over. But if Virginia or Pennsylvania are looking like close races, then Trump is beating his polls, and there’s good reason for Clinton fans to take a few drinks and settle back for a long night.

Most interesting will be New Hampshire, which has been running very close to the exact center of national opinion. An early call for either candidate there is a very good sign for that candidate (though it doesn’t mean the end of the race, since New Hampshire only has 4 electoral votes).

Keep your eyes on the early states, too. Both Kentucky and Vermont should be called immediately for Trump and Clinton, respectively, but keep an eye on the margins in each state: if Trump is going to win the election, he “should” win Kentucky by about T+22 points, and, if Clinton is going to win the election, she “should” win Vermont by about C+57 C+27 points. (UPDATE: +57 was a pretty serious typo! Sorry!) Of course, you should ignore the early counts, when only a few precincts are in, but, by the time each state gets to 60% precincts reporting or so, we should have some idea whether either candidate is having a “Newcastle-upon-Tyne” moment.

North Carolina is a really important state in this election (Trump probably has to win it to win the election), but the vote counting there was a little erratic in 2014. Early N.C. returns showed a near-certain win for Democratic Sen. Kay Hagan based on early voting, which is what the polls had said, but Thom Thillis dominated Election Day voting, and ended up pulling out a squeaker. So just ignore North Carolina until the networks officially call it. (A Clinton win in North Carolina is a very bad sign for Trump, but, because North Carolina relies so much on early voting, it’s not necessarily fatal: Trump could recover with wins in Pennsylvania or Michigan, neither of which have much early voting.)

The problem with state-based forecasting is you have to wait ages for entire states to report in. Let’s drill down to a more granular level.

Congress-Level

Every seat in the House of Representatives is up for election this year, just like every year. Republicans are expected to lose seats but keep their majority, but the number of lost seats depends largely on how many people turn out to vote for Trump and Clinton. (The course of Speaker of the House Paul Ryan’s future may well be determined by how large his Congressional majority is, since the GOP caucus is already so deeply divided.) If Trump supporters turn out, iffy Republican seats will be saved, and iffy Democratic seats will be lost. If Clinton supporters show up, the opposite will happen.

So here’s another chart from FiveThirtyEight (the whole article is worth reading), which gives us a really good way of tracking which side showed up:

My take:

If Democrats have a “good” or “very good” night by the standards of this chart, Clinton will win the presidency.

If it’s even (with maybe one or two surprises), Clinton is very likely to win the presidency, but it’s not quite a closed book.

If Republicans are having a “good” night, then the presidential race is a tossup.

If Republicans are having a “very good” night, then Trump will win the presidency.

That may seem unfair: the Democrats only need to run even in the House to have a strong chance at winning the presidency, while the Republicans have to have a “very good” night to have the same chance. That’s because Trump is currently losing in the polls, and requires a strong Republican showing across the board just to pull even in his race for the White House.

And, of course, as in every election, there will be upsets — possibly even multiple upsets on both sides of this chart. When you’re looking at small districts rather than huge states, things go against the trend a little more frequently just because of the greater variability.

So if the race in the 9th Indiana Congressional District (Trey Hollingsworth vs. Shelli Yoder) looks close, don’t immediately jump to the conclusion that it’s all over and Clinton won. But if Democrats go on to win all four seats in Florida… then yeah, it’s probably all over and Clinton won.

But there’s still one more way we can track the trends in this presidential election: county-by-county.

County-Level

This method is simple: find a county where 100% of precincts have reported results. Let’s say rural Boone County, Kentucky, at the very tippy-top northern part of the state. It’s an extremely safe county for Trump, and their polls close at 6 PM Eastern — less than an hour after this post is published — so I expect them to report fairly quickly. (Kentucky’s polls overall don’t close until 7:00, so don’t expect results to be released until after that.)

You can find detailed county-by-county results in each state in many places around the Internet. I like to use Politico’s user-friendly results report. (Click the state you’re interested in, then click “detailed results” to see a county-by-county map.)

Remember that you’re looking for counties with 100% or very nearly 100% reporting. Figure out who is winning and how much they’re winning by. Great. That’s step 1.

Step 2: go to ShareBlue Benchmarks. Find that same county. Hover your mouse over it. You’ll now see a (decent) estimate of who will win the county and by how much, based on statewide polls and demographic information about the county.  (Boone County is Trump +38%.)

If Trump is beating his ShareBlue benchmarks in a lot of counties in several states, he’s got a really good shot at winning.  If not, or if Clinton is beating her benchmarks, then he’s pretty much going to lose.

There’s going to be a lot of random statistical noise down here in the weeds, so don’t invest too much faith in any one county. Remember that, in Newcastle-upon-Tyne, Brexit beat its polls by 10 points, but only won nationwide by 4 points.  That’s going to be true here, too: Trump will beat his benchmarks by wide margins in some areas, Clinton will do the same in others, and the question is whether Trump is able to do it consistently enough, in enough places, by large enough margins (2-3%) to come from behind and win the election.

Although I caution you against reading too much into this, I think it is also the fastest way to draw useful inferences about what’s happening in the presidential election, so it’s what I’m going to be focused on during the early part of the evening.  You can do a lot by comparing actual results to benchmarks.  I would much rather have results from 100% of precincts in rural Boone County than from 5% of precincts in the entire state of Florida. 5% of Florida is way more people than live in Boone County, but, because I don’t know where those votes are coming from or how they’re being mixed with the early vote, I can’t draw inferences from them.

UPDATE: The New York Times Upshot blog will be doing a lot of this automatically tonight, which will save us all a great deal of work. Tune in there for rapidly estimated early results, complete with statistically-reliable margins of error!

The Bottom Line

The odds are, this election will be over early. The networks won’t call it until Clinton officially has 270 electoral votes, but Clinton is going into this thing with a comfortable polling lead.  We should know after a couple of states whether the polls were basically on-target, and, if they were, Clinton is almost certainly going to win.  Trump needs to beat his polls, which is doable, but we should know fairly soon whether he actually is. And there are plenty of reasons to think he won’t.

Those of you following the election with this blog, then, have a very good chance of knowing the outcome of the election by the time polls close in Minnesota (9 PM Eastern), before — perhaps long before — the networks call it.  I’ll be staying up late to see who wins the Senate — which is a genuine nailbiter according to the current polls.

On the other hand, Trump could beat his polls, and, if he does, you will be among the first to know that it’s going to be a long night.

Either way, I’ll be watching Gods of Egypt (as promised) after the final results are known. Expect my review by the weekend.

I may or may not post further updates tonight, but I can tell you I’ll be following FiveThirtyEight’s LiveBlog, the Upshot’s projections, and I’ll probably be tweeting a bit as the spirit moves me, so tune in there if you want to know what’s up.

May God bless America. Vote SMOD 2016.

A few weeks ago, Trump supporters were talking about how their man was going to win even though all the polls showed him losing, “just like Brexit!”

Those of us who knew something about Brexit said this was stupid, and rightly so: the final Brexit polls showed Brexit losing by about 0.5%, and Brexit went on to win by 3.8% — a polling error of 4.3%. The global elites who had been handicapping the race had been predicting a loss for Brexit, but they had ignored the polls (which showed a close race). The elites looked like idiots after Brexit won, and rightly so, because the lesson of Brexit is you should always pay attention to the polls.

So when Trump supporters started saying they were going to win by beating their polls, “just like Brexit,” those of us with sense went and checked the polls.  At the time (right after the third debate), Trump was losing by an average of 7 points.  In other words, even if Trump beat his polls by 4%, “just like Brexit,” he’d still lose the election by 3%!

Brexit comparisons three weeks ago were silly.

Brexit comparisons today are not.

The polls have gotten much closer in the past couple of weeks. Clinton still has a clear polling lead, but that lead has shrunk to about 3%.  Worse for her: Trump has a clear advantage in the electoral college, because her best groups (minorities) tend to be concentrated in safe states, while Trump’s best groups (non-college whites) tend to be concentrated in swing states. Clinton’s lead in swing states is somewhere around 2%.

In other words, if Trump supporters show up in a big way and beat the polls “just like Brexit,” based on today‘s data… then Trump will win.  He’ll win the national popular vote by 1%, he’ll win the electoral college, and he’ll tie the vote in Michigan, with a map that looks like this (you didn’t know Colorado was so close, did you?).  He doesn’t even need a Brexit-sized polling error, at this point: a perfectly average error will do.

On the other hand, there’s no particular reason to think that Clinton couldn’t pull off a Brexit surprise herself.  Clinton could just as easily beat her polls by 3.8%, and then she would win by around 11% 8% — a landslide that brings her 358 electoral votes or so.

But Clinton doesn’t need to do that.  All she needs to do is make her polls to win (or miss them by a very small amount).  Trump is the one who has to beat his polls.  That’s why Trump still has a chance to win, but Clinton is still the 2-in-3 favorite.

There will be signs early in the evening on Election Night of what kind of a night it’s going to be. Brexit, once again, lights the path.

…but we’ll get to that in my next post. (Assuming I manage to post again before it’s all over.  I might not!)