Lindholm Company Blog

The partisan registration difference alone can accurately predict 91% of the House contests from 2002 through 2008.

On its surface, this is not that interesting. Put it another way, and it really stands out: campaign quality differences, campaign expenditures differences, and candidate incumbency factors, etc., among other facts, are only needed to explain 9% of the outcomes.

This model holds during the great Republican year of 2002 and the “bad” Republican years of 2006 and 2008.

The analysis was performed using a logistic regression of partisan registration difference as the independent variable and winner of the contest as the dependent variable for all Oregon House District contests from 2002 through 2008.

Source of data: Oregon Secretary of State.