Thursday, April 23, 2015

Q&A - 23/4

Question

Here, u talk about prez elections, predictions. But this model is too simple!

Yes

The model is simple, but it is, in fact, deceptively simple. Let's see what the modeler did not include in his model: he did not include crime rates, the wheather, population increase, the price of oil or whether or not the presidential candidate loves his mama. What he did include however are these three variables, and these variables alone (we hear Abramovitz is also toying with another variable - polarization, which measures how much the electorate is divided along party lines, etc., but this is still one variables out of bazillion he could have included). He also did not have a seperate prediction for Reps another one or Dems, he only had one - a prediction for the incumbent party. By choosing the variables the way he did, he included his intuition about what really matters in an election in his model; up and down vote on the party in power. And the results are bang on.

Simple doesn't mean something is shoddy. Simple can be very effective.

And simple ain't easy. "Writing shorter takes longer" said the wise man, right?