Member Newsletter
For most of this election season so far, FiveThirtyEight hasn’t surfaced its own presidential race average. Compiling these averages in a sophisticated and honest way is actually kind of complicated. (They explain their methodology here.) We used to play in this space with PollTracker. So I know from experience. For me it’s not that I put so much weight on the specific number. It’s just a simpler way of visualizing the trend over time.
Well, now they’ve finally rolled theirs out. So I don’t have to be so reliant on RealClearPolitics, which has both a clumsy methodology and goes to comical length to add or not add polls to juice their favored candidate. All that said, the absolute averages as of today as remarkably similar. Trump up .4% on FiveThirtyEight and .3% on RCP. This reminds us of some of the ironies and … how can I say it, melancholy and ennui of complexity. The FiveThirtyEight model is HUGELY complex and sophisticated. It factors in all sorts of different variables. It incorporates findings from individual states into the national averages and vice versa. And yet for all this statistical leg work the averages as of this moment are essentially identical.
There’s probably some statistician’s book of aphorisms somewhere which has a chapter called “complexity is its own reward.” And I think that’s right. These are highly sophisticated models. But at the end of the day we’re dealing with foundational data that has a lot of haziness built into it. The models are best to look at for broad trends, as we discussed yesterday.
Another interesting nugget in the FiveThirtyEight methodological breakdown piece is a look at the swing states. The headline here is Trump and Biden tied at the national level and Trump holding what they call a “tenuous” lead in the swing states. But the breakdown of their averages shows something more specific. The two candidates also basically tied in the Blue Wall states of Michigan, Wisconsin and Pennsylvania. The precise averages are actually Trump +1.1 (Michigan), Biden +.1 (Wisconsin) and Trump +0.9 (Pennsylvania). But those datasets are still weighted toward GOP-leaning polls at the moment. In any case, those are basically ties and I’m fairly confident Biden wins those. It’s the Southern-tier states of Georgia, Nevada, North Carolina and Arizona which each have Trump leads of around 5 points, give or take.
Here’s the key. If Biden holds the Blue Wall states and wins that single electoral vote in Nebraska he gets to 270 votes. Literally the absolute minimum to win. That’s if he loses Arizona, Georgia, North Carolina and Nevada. So 270 to 268. If this seems odd to you it’s because those southern states have gained a few electoral votes at the expense of blue states in the northeast.
You can see why Republicans really, really want to switch the rules for electoral votes in Nebraska.
Now if Biden wins Nevada or Arizona suddenly he’s also 40 electoral points up. On the other hand, if they’re able to change the rules in Nebraska you actually have a real chance of a tie in which the House picks Trump as President. Note that even if Democrats take the House they are highly unlikely to control a majority of state delegations.
As we make clear about the national numbers and really all the polls, we’re six months out from the election. So we can’t put too much into these numbers. But they do tell us that there’s a good chance the election will be decided in Arizona or Nevada. Or at least that’s where Biden has a shot of putting it away.
A couple final points on FiveThirtyEight. What we’re discussing here isn’t a forecast. It’s a very sophisticated average or perhaps better to say composite of current polls. I assume they’ll release a forecast at some later point. But for now we have an average — not a prediction or predictive model. Also, you probably remember that FiveThirtyEight founder Nate Silver parted ways with FiveThirtyEight or year or so ago. His replacement is a guy named G. Elliot Morris. He’s good. I follow a lot of these guys — mostly guys who are kind of the new generation of political numbers crunchers. I have a feel for whose judgment and technical know-how I trust. And he’s good. As I said, technical sophistication and complexity mostly needs to be its own reward probably. But if you’re going to be nerding out about averages and models I think you’re in good hands with him.