Tuesday, October 25, 2016

Comparing Election Models: FiveThirtyEight, The Upshot, and Princeton Election Consortium

Today on The Upshot I found a nice table that compares all of the major election models on a state-by-state basis. The results look a lot like what I'm seeing in my own model.

Using three of the major models—FiveThirtyEight, The Upshot, and Princeton Election Consortium (Sam Wang)—I created my final list of battleground states. Here it is:


Battleground (EV) D Participation D Win Prob. R Participation R Win Prob.
Florida (29) 82%–86% 0.82–0.86 99%–100% 0.14–0.18
Pennsylvania (20) 95%–97% 0.95–0.97 39%–55% 0.03
Michigan (16) 96%–97% 0.96–0.97 23%–44% 0.04
Wisconsin (10) 94%–96% 0.94–0.96 12%–31% 0.04–0.06
Minnesota (10) 95%–99% 0.95–0.99 4%–12% 0.01–0.05
Colorado (9) 92%–94% 0.92–0.94 18%–26% 0.06–0.08
Nevada (6)89% 0.89 20%–26% 0.11
New Hampshire (4) 98% 0.98 4%–8% 0.02
Tue 25 Oct 2016

In many of these states, there are competitive Senate races going on:


To make the list of battleground states, I simulated the three models separately and identified battleground states according to each model. The states above are those that emerged as battleground states in at least two out of the three models.

(To repeat the "battleground" notion that operates here: a state counts as a priority for candidate X if the state goes for candidate X in a relatively high fraction of the candidate's wins, even as the probability of candidate X winning the state is relatively low. Intuitively, priorities are states you need but don't yet have. Battleground states are those that appear on both candidates' priority lists. Note that choosing priorities this way involves an element of arbitrariness, or at any rate, judgment: what is a "high" fraction of wins? what is a "low" probability?)

The fact is, the battleground analysis paints a terrible picture for the R candidate. From what I remember, it looks nothing like the race in 2008. One way to see the D-R asymmetry at a glance is to look at a scatter plot of the win probability against the participation rate. Here is the D plot:



In this plot, each dot is a state. (Or an electoral-vote granting entity such as the District of Columbia or Maine Congressional District 2.) The horizontal axis is the probability of D winning the state. The vertical axis is the percent of D victories in the simulation for which D wins the state. The plot above represents a reasonable situation to be in: the more you need a state, the more you have it.

Compare to the R plot:


There are many states you need (up high on the plot) that you don't really have (left half of the plot).

So on paper, this is looking like a slaughter. But as the sportscasters like to say after every upset victory, That's why they don't play these games on paper. Get out and vote! This is especially important because of down-ballot elections.

In fact, if neither of the presidential candidates turns you on (I think I threw up a little in my mouth just now), leave the top spot blank. Don't vote for President. Just go the polling place and vote for whatever else looks like it makes sense.

I have some friends and relatives who never vote. The main reason they give is that they don't know enough about the candidates or the issues to make an informed choice. I can sympathize with that, because I often feel that way myself. Back when I lived in small-town Vermont, I remember one election when I only voted on a single race from the entire ballot, because I just didn't know anything. (I knew one of the people running and generally admired her judgment, so I voted for her.)

And in the most recent primary election in New York state, I had no idea which of the people running for Congress I should vote for. You know what saved me? Ballotpedia. I spent 15 minutes skimming Ballotpedia and I basically had an idea of which candidate seemed best. If you haven't voted recently, give it a try.

Another reason I've heard for not voting is that "The candidates are all the same." This—forgive me—is madness. Had John McCain been elected President in 2008, we wouldn't have Obamacare today, and we wouldn't have joined a range of new climate treaties. You may hate these things, you may cheer these things, but either way, they are big decisions for a country to make, and they directly reflect the choices that Americans collectively made at the ballot box in 2008 and 2012.

A related reason people don't vote is that "My vote doesn't make a difference." This feeling is understandable; even economists struggle to explain voting within a framework of rational self-interest. But when all else fails—when the demands of work and parenting and everyday life make voting seem decidedly optional—I do it anyway. My parents taught me that I have a duty to vote.

No comments: