RCP’s Thoroughly Unserious Forecast
by
Evan Scrimshaw | Aug 3, 2020 |
Uncategorized |
2 comments
I would genuinely love to know what RCP thinks it is doing with their election forecast, because I have spent weeks now trying to figure out what they’re doing and I’ve got nothing. Their whole purpose to the political landscape is their polling averages, and yet this map doesn’t work within the confines of their own averages, but it also doesn’t match any form of forecasting, with such a broad range of Tossups from Missouri to Minnesota that it can’t really be explained by state partisanship either.
In places, it seems to be a fairly simple average of polling, with Indiana and Missouri close on account of the fact that what limited polling we have in both states points to a massive swing to Joe Biden, and therefore, on that basis they should be competitive. In other places, they’re using the fact that a state was close in 2016 to justify their rating, like Minnesota, where the RCP average shows Biden up 9% but Hillary only won by less than 2%, so again, on that basis it’s an arguable Tossup. And then, in some other places, they’re using down ballot data from 2018 to make the case that states will be competitive, with Connecticut as Lean D being the clarion call for that one – I mean, I’m assuming it is that, because I got nothing to justify it otherwise. But RCP’s main problem is that none of these things will happen together.
Ask anyone who does US election projection, and they’ll tell you that there is no universe in which Missouri is a Tossup while Connecticut is close. This isn’t a partisan point, or a statement of any brilliance emanating from LeanTossup. Ask anybody else – the Nates Silver and Cohn, G Elliott Morris, Jack Kersting, or any of the people working at the rating agencies – and they’ll all agree with this. If Connecticut is close, then the Democrats have an avalanche of problems, and states that Trump won narrowly are going to go red, let alone a state where Trump won by 19% last time. The only answer that seems even remotely possible is that they have a few different forecasts, and they’re taking the results that get spit out of each of them, taking the closest to Tossup one of them, and making that the map. Or maybe they’re just really, really bad at this.
Either way, this isn’t a forecast as much as a Don’t Kick Me sign is a public health warning. This is an attempt to not be too wrong, because it is better to be stuck in the middle and have a decisive result than believe your own averages and be burned by a massive, unthinkably huge polling error. It is laughable that RCP thinks this is an acceptable way of building a forecast, letting fear of being wrong stop you from being right. This forecast is a claim that there is no way to accurately forecast an election, in which case, don’t bother trying. But when you offer up a forecast that is as unserious as it is laughable, you have to recognize it for what it is – the comedic relief. There’s immense value in contrary opinions and making yourself aware of other possibilities. This isn’t that – this is a wholly unserious forecast designed to avoid being too wrong, and it should be treated as such.