What is the relationship between primary turnout and general election outcomes?

We've often heard the argument that because Democratic primary participation is disproportionately higher in a particular state, Democrats stand a chance of performing better in that state in the general election. After Alex Thurston made this argument about the upcoming Oklahoma Senate race, I decided to find out if this claim is actually true.

To validate this argument, we want to compare the percentage Democratic primary turnout (out of the total primary turnout) to the performance of the Democratic presidential candidate in the general election for each state. To account for crossover voters, it makes sense to only consider state primaries that were competitive on both sides. For simplicity's sake I decided to look at the 2000 election, but the 1988 election would work as well. Data and full details on how I selected the states here.

Here are the states with competitive primaries:

Clearly, in 2000 there was a correlation between partisan turnout in competitive primaries and the general election outcome in those states. A linear regression with 2000 Gore performance as the dependent variable has an R-Squared value of .829 and a p-value less than 0.001, which indicates that there is a strong linear relationship.

Given that, this year's primary turnout in states that were competitive on both sides may be a useful input when determining which states are going to be "swing states" in the general election.

Another argument, that Democrats will do better in the upcoming general election because Democratic primary turnout is higher nationally this year than in previous years, is harder to validate. As Matt Stoller pointed out, this is because there have been too few competitive primaries on both sides since the McGovern reforms to begin making that judgment. Once we know the 2008 general election results, we can start to think about this by examining the turnout in competitive primaries in 1988, 2000, and 2008.