Elections Analytics - Forecasting
Overview
This corner examines ways to model elections and forecast the outcome based on factors other than polling information.
For example, it is a standard theory that the state of the economy affects whether the incumbent is re-elected or
defeated.
Modeling Elections Outcomes
Building a model typically involves choosing a variable to model, and then choosing a set of factors to model that
variable against. Generally, the model is then fit using some form of
Regression.
In the United States, there is a two party system. One party is currently holds any particular seat, and the other
party challenges the incumbent. As a rule, it is thought that the incumbent enjoys an advantage over the
challenger, particularly if the economy is doing well.
- Incumbent wins or Loses the Election - This is modeled using a 1 for a win and a 0 for a loss. Then the model is
fit to the explanatory factors using logistic regression.
-
Percentage of Votes Won by the Incumbent - in this case, a win is represented by a value of 50 or higher.
The percentage vote model would likely be fit using ordinary least squares regression
Factors Affecting Elections Outcomes
After defining the variable to be modeled and the methodology to fit the model (see above), one needs to select a set of factors
to use as predictors. As mentioned, a common place to draw the predictors from is
Macro-Economics.
Common variables include
