Wednesday, July 8, 2015

Forecasting the 2016 election economy: is it the economy, stupid?


 - by New Deal democrat

Is it "the economy, stupid?" Or is the economy just one contributing factor to presidential election results? Either way, what economic metrics best correlate with the election outcome?  Can we forecast those metrics reliabily enough to usefully forecast the election result itself even a year in advance?

With the advent of poll aggregators such as Nate Silver and Prof. Sam Wang of the Princeton Election Consortium, forecasts of election results have improved markedly, especially as we get down to the wire.

But can we usefully predict where those poll results might be?  Well, that's an experiment I am going to undertake over the next year.

There are two components to this undertaking.  The first is to identify those economic metrics which most correlate with election results.  The second is to identify useful leading indicators for those metrics.

Fortunately, political scientists and statisticians have pored over economic data and proposed a number of metrics -- for example, real GDP or real income -- that they believe have had predictive value for election results.  Typically these focus on the 2nd and/or 3rd quarter of the election year.

So my first goal is to examine those metrics, and select from among them perhaps three or four that appear to most closely match the outcome of the election.

Once I have identified those metrics, then I will try to forecast them, generally by variations on existing long and short leading indicators for the economy.

In the best case scenario, hopefully by the end of this year I will be able to forecast with confidence the likely value, or range of values, of several critical economic variables, which in turn will shape the results of the 2016 elections.  Even if the exercize doesn't pan out, it should be useful and hopefully fun (in the nerdy sense).

In 2011, Nate Silver listed over 20 economic variables which best correlated with election results.  In my next post, I will examine the most accurate of those variables in order to select a few to be tracked most closely.