Election Modeling

I’ve been spending a lot of my time working on a class assignment in which we are asked to model the U.S. presidential election. The model is by necessity fairly rudimentary – I’m afraid I won’t be giving Nate Silver a run for his money any time soon – but it’s nonetheless been very interesting to think through the various steps and factors which influence how election results play out.

The basic approach is borrowed from the compartmental models of epidemiology. Essentially, you treat all people as statically equivalent and allow for transitions between discrete compartments of behavior.

Consider a simple model with the flu: you start with a large pool of susceptible people and a few infectious people. With some probability, a susceptible person will come in contact with an infectious person and become infected. At some average rate, and infectious person will recover. Thus, you can separate people into compartments, Susceptible, Infectious, Recovered, and with average transition rates can estimate the number of people in each compartment at each time step.

Of course, sophisticated epidemic models can be much more complicated then this, and trying to interpret the complexity of an electoral system through such a simple model has proven to be challenging.

First there’s the question of how to transfer this metaphor to electoral politics – what does it mean to be ‘susceptible’, ‘infectious’, or ‘recovered’ in this context?

But perhaps the piece I have found most interesting is trying to understand the system’s “initial conditions.” I am not an epidemiologist, but a simplified model of disease spreading where some people start susceptible and a few people start infected makes intuitive sense to me. We even worked out mathematically how moving people from “susceptible” to “recovered” via vaccination helps prevent a serious outbreak of a disease. (PSA: get your flu shot.)

But I’ve had a much harder time wrapping my head around what initial compartment a voter might belong in.

There’s an idealized version of politics in which all eligible voters start with a completely open mind – a clean slate ready to be filled with thoughtful judgements and reflections on the merits of each candidate’s policies.

But that’s not really how electoral politics works.

I, for example, have always been a staunch partisan, and while it perhaps would be better if I entered an election season as a clean slate – I always enter with a whole host of biases and preconceptions. The debates and TV ads were never going to change my mind.

So what has been most striking in the process is how little movement actually takes place – especially considering just how long this election has gone on.

When you take the partisan leaning of Independents into account, Pew estimates the current population of registered voters as 44% Republican/lean Republican, 48% Democrat/lean Democrat, and 8% no leaning/other party.

FiveThiryEight‘s weighted average of national polls shows some fluctuations over the last six months, but currently puts Clinton at 45.8%, Trump at 39.4%, and Johnson at 5.8%. That’s not a direct correlation to the raw partisan leaning, but it’s close enough to show that – in the epistemological framework – relatively few transitions are happening.

In fact, the earliest FiveThiryEight numbers, from June 8 of this year put Clinton at 42%, Trump at 38%, and Johnson just shy of 8%. So I guess this makes me wonder:

…Couldn’t we have held this election back in June and saved ourselves the trouble?

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.