My first read of this newsletter and very enjoyable. After observing elections at the national level since 1968, my conclusion is that there are way too many variables out there for anyone, even the most astute, to be able to detect any other than narrow amplitude trends. It's best to figuratively shrug ones shoulders and say "who knows?" The situation improves as the elections narrow down to state and best, local.. there predictions become more accurate. I'm not saying one can't make descriptors of the current political situation at time t° but the accuracy rapidly falls off as time increased even with our best Bayesian methodology. Case in point, Trump in 2016.
Thanks, Michael! I appreciate that. I heartily agree with you on this: it's very difficult to know how an election is going to go. And yet we try. I think it's driven a couple related factors. One is people's loathing of uncertainty and deep desire to be rid of it. When pundits confidently declare what the future holds they're mostly responding to demand (and the experience that being wrong carries little cost). Another factor is the large number of media outlets who need to product enormous amounts of elections-related content, as well as the fact that predictions/horse-race content performs better than other kinds. So even though the answer to "what will happen in the election" is generally "it's not possible to know," we will probably continue to get a ton of confident predictions in the lead-up to elections.
You know Ethan, what's mysterious to me is the unreasonably tight nature of our national presidential races. We are at an improbably even split. This, to me, is a central mystery of our politics. Realistically, we should be somewhere on the slopes of a bell curve rather than teetering at the top. Why not more 60/40 and 70/30 outcomes? Something is exerting a balancing influence. But what?
My first read of this newsletter and very enjoyable. After observing elections at the national level since 1968, my conclusion is that there are way too many variables out there for anyone, even the most astute, to be able to detect any other than narrow amplitude trends. It's best to figuratively shrug ones shoulders and say "who knows?" The situation improves as the elections narrow down to state and best, local.. there predictions become more accurate. I'm not saying one can't make descriptors of the current political situation at time t° but the accuracy rapidly falls off as time increased even with our best Bayesian methodology. Case in point, Trump in 2016.
Thanks, Michael! I appreciate that. I heartily agree with you on this: it's very difficult to know how an election is going to go. And yet we try. I think it's driven a couple related factors. One is people's loathing of uncertainty and deep desire to be rid of it. When pundits confidently declare what the future holds they're mostly responding to demand (and the experience that being wrong carries little cost). Another factor is the large number of media outlets who need to product enormous amounts of elections-related content, as well as the fact that predictions/horse-race content performs better than other kinds. So even though the answer to "what will happen in the election" is generally "it's not possible to know," we will probably continue to get a ton of confident predictions in the lead-up to elections.
You know Ethan, what's mysterious to me is the unreasonably tight nature of our national presidential races. We are at an improbably even split. This, to me, is a central mystery of our politics. Realistically, we should be somewhere on the slopes of a bell curve rather than teetering at the top. Why not more 60/40 and 70/30 outcomes? Something is exerting a balancing influence. But what?