In 2009, after working with Vermont Freedom to Marry, I created the Vermont District Index in an attempt to measure the electoral risk that legislators had taken in breaking with their party on the big vote. The original model used Presidential, Gubernatorial, and State Legislative results to come to its conclusions - the same variables that remain at the core of the LDI project today.
For a given state, the following elections are used to compute the ratings:
A district's rating is calculated as a weighted average of the margin of victory in the above six races. But why those six?
1) Using three presidential elections ensures that the best indicator of political preference (given that it has the highest turnout) is not unbalanced because of one figure. For example, Al Gore, John Kerry, and Barack Obama have boosted Democratic strength in very different areas of the country - by including the past three elections, instead of the past two, the index does a better job accounting for the differences between national-level candidates. Additionally, by decreasing the weight of the results over time, as opposed to the simple average used by a method like Cook PVI, the rating system can better keep up with changing dynamics.
2) Presidential results, however, aren't everything - there are plenty of parts of the country where the state parties have outperformed their national affiliates - Democratic performance in Arkansas and Kentucky is a perfect example. This component of the district balances these national preferences with state-based preferences, as ultimately we seek to measure the strength of a party's brand within a district - not the popularity of the party's national politicians.
3) Finally, the most recent lower chamber election for state legislature is used in the index. This additionally serves as a counterbalance against conventions about voters' preferences. Given the much more intimate nature of state legislative seats, this 15% helps to account for places that may never vote for one party for major office, but have no problem supporting local candidates of the opposing party - something that frequently shows up in legislatures with low constituent-to-representative ratios.
After the weighted average has been taken from each district, all the districts are average together to find the average result for the state. Deviation from this average by an individual district is reported as R/D+x, where x is a regular integer and D and R correspond to whether the Democratic margin of victory is greater or lower than the average result for the state. For example, in a state where the average district is 50-50, a district that is D+10 produces an average Democratic margin 10 points larger, which would be 55-45.
