We all have various theories for what is drives elections. For example: left v.s. right, race, social issues, economic issues, skilled campaigning, endorsements, slander, self-interest, get-out-the-vote, voter-suppression, candidate height, etc. etc. In thinking about that it is critical to draw a distinction between what garners votes for a candidate, say his height, and how he votes once elected, say for environmental regulations.
As a practical matter if you want your guy elected it’s important to set aside the issues and focus entirely on the question of what will get him votes.
Let’s say you accept that. So what do you do? Well, you go find the literature and experts and extract from them the secret formula. Wait. Why would that formula be secret? Certainly discovering that formula would be a central theme in political science? So, I don’t think it is a secret. But as usual it is a bit hard to find the formula in the face of so much noise. Some of that noise comes from the disputation nature of scientific enquiry; the political scientists are always disputing the question and puzzling out who’s clueful and who’s not is often difficult. Much of the noise comes from confusing the first distinction – lots and lots of people assume that you get elected because of where you stand on the issues. Maybe that’s true (the formula would tell us), but I presume it’s not and so all explanations about elections based on issues are just noise.
This formula might look something like this. P = a*x + b*y +c; where P is the percentage of the vote the incumbent gets. And the various terms represent the answer to our original question; what are the drivers that garner votes. Of course there could be dozens of terms, and it could be that the constants, a, b, c; vary substantially from one election and venue to another. But, maybe not.
One school of thought about what drives elections is that it’s all about the economy. In that case you give the incumbent some constant c; say 50% of the vote, you then select a measure x, say unemployment, multiply it times a, add them up and you’ve got your prediction. If this model has high predictive power then you’d better work on x. Everything else you work on is about issues and policy, it’s not about getting reelected.
Models like that are sometimes called “economic determinism.” Andrew Gelman wrote a short post on economic determinism back in 2008 providing pointers to the political scientists who have worked on the question.
The first paper (pdf) I read is simple enough. Bob Erikson fitted a somewhat more complex model P = a*x + b*n + c. The economic metric x was how much per-person income had changed, weighting the recent past more than the more distant past.
Bob’s model fit his small sample (10 presidential elections) really really well. He concluded that economic term was very potent. A 1% change in income drove a 2.7% change in the vote. But so was second term b*n.
The metric he selected for n is called NCE, or net candidate advantage. That’s based on a voter survey (the National Election Study) of how much the voters like/dislike the candidate.
Based on the table in his paper and my calculations on average 12% of the votes the incumbent gets are driven by economic issues, 6.2% was driven by the NCE, and only 2.5% was left unexplained. So 12% is economics and 8.8% is other stuff. While economics is not entirely deterministic it certainly sets up a powerful current your either swimming with or against.
I’ve not followed any of the other bread crumbs that Andrew tossed out in his posting. I have some strong doubts about Bob’s model, in particular the last election mentioned is Reagan’s 2nd; and that puts the entire data set before the last three decades of Republican party shift to a highly disciplined far right party. Further economic growth was largely positive thru that entire period, while since then it’s been flat for voters.
FYI – the US economy is in a recession. The entire economy is running at about 80% of it’s potential. You can see that 20% short fall in lots of statistics, labor participation, railroad freight, it’s all over the place. Not a happy time for incumbents.
X = x + …
Please explain further.
Good. You have corrected it.
Now I’ll start asking you questions on the model…