Wednesday, August 4, 2010

A bit more on forecasting midterm elections

I offered a forecast for the 2010 midterm elections in a previous post.  I thought it might be helpful to provide a visualization, as well.  In the below graph, I have shown the predicted gains in midterm House seats by the president's party based on two main variables: the growth in per capita real disposable income (from the 3rd quarter in the year before the election to the 2nd quarter in the election year) and the president's Gallup approval rating on Labor Day.  I've provided three different levels of presidential approval: 35, 50, and 65.
As the graph makes pretty plain, it's rare for a president's party to gain seats at a midterm: only unusually popular presidents presiding over reasonably strong economic growth (think Clinton in '98 and Bush in '02) have much of a chance of adding to their margins.  And of course the current economic growth figure is 0.88, and Obama's approval rating is at around 45%, which is where the forecast of Democrats losing 40 seats comes from.

4 comments:

Eric Rubin said...

im boycotting your blog until you put my banner back up. WTF?

Esau said...

Does your model account for the proportion of seats that the presidents party controls? That seems like something that would relevant to any predictive models.

Seth said...

Esau, no, I didn't account for that. I didn't find the variable that helpful, although I've found some studies that utilize it.

Seth said...

Eric, it's not personal. It's strictly business.