More Things
These numbers are unlike Excel numbers (blog.contegix.com).
by Tony Petrangelo
Jul 8, 2013, 10:00 AM

hPVI 2013, a prelude

With the 2012 election in the books and the legislative session now over, it’s time to roll out the new edition of hPVI. Not to be confused with the actual New Edition.

If you are unfamiliar with hPVI, this post goes into some detail about the metric itself, but here’s a short version.

Partisan Voting Index (PVI) is a metric that assigns a partisan score to a congressional district based on the results of the last two Presidential elections in that district compared to the results of those elections in the nation as a whole.

hybrid Partisan Voting Index (hPVI) is a metric that assigns a partisan score to a Minnesota legislative district based on the results of the last Presidential and the last Gubernatorial election in that district compared to the results of those elections in the state as a whole.

From the above descriptions it should be clear that hPVI is a direct descendant of PVI, its creation influenced heavily by its predecessor.

There’s always been this thing about hPVI though, a sort of bug that resulted in a roughly four point Republican bias. The following comes from a post I wrote detailing this precise problem:

The above scatter plot contains a formula for the regression line that runs through it. If we work that formula backwards, solving for a 0% margin of victory, we get R+3.8 (or -3.8) as the answer.

What this means is that we would expect a district with an R+3.8 lean to result in a tied election and anything greater than that would be a GOP victory, anything less, a DFL victory. Meaning a pure toss-up district is between R+3 and R+4.

Ideally this wouldn’t be the case. Ideally we would expect a district with an Even hPVI to result in a tied election. And since the point of hPVI is to be calibrated for use in Minnesota, not to be true to its predecessor, I felt a change was in order.

So a change has been made.

Instead of subtracting the average of the Democratic share of the two-party vote in a district from the average share of the Democratic vote statewide (which tends to run a couple points north of 50%), instead I’m subtracting the average of the Democratic share of the two-party vote in a district from 50%.

This won’t change the relationship of individual districts to each other, rather, it will result in all districts shifting a couple of points in the liberal direction.

This is a minor change, all the same data is being used, that data is just being compared to a different baseline. Doing the calculation this way for the data set cited in the paragraph above would result in an R+1.1 district being the break-even point as opposed to the previous R+3.8.

As the title indicates, this post is merely a prelude to the posts that will contain the actual hPVIs. Those posts will follow shortly.

Thanks for your feedback. If we like what you have to say, it may appear in a future post of reader reactions.