A Safe rating implies a very high confidence level, approaching 95%, so unless it's an upset of Weltian proportions, the chances of another outcome are not good. A Tilt rating is still likely to go in that direction, but there may be a red flag or two that is cause for hedging. A Toss-up means just that, it's a wide open race.
Quick Overview
First off, let's quickly go over the bottom line numbers before getting into the minutia of how the ratings were made and the actual ratings.
If these ratings are close than control of the house is in play this election, but all Democrats need to do is win a third the toss-ups to retain control and since most of the toss-ups are currently held by Democrats it's likely that they will. And if 32 seems like a lot of toss-ups, it's less then 25% of the total seats. I'll do an update to these when we're closer to the election that will likely push some of the Toss-ups out to Tilt status.
As you can see the majority of the toss-ups are Democratic held seats and even some of the Tilt GOP seats are currently held by Democrats, so it certainly seems likely that there will be gains made by the Republicans in November, the big question is how big these gains will be.
The Process
Putting something like this together with no polling information to go on presents some challenges. To figure out how to go about actually rating the races I looked at some recent election history. I divided all of the state house and senate elections into two groups; incumbents running for re-election and open seat races, I didn't include special elections, only general election contests from 2004 and on.
Why 2004? Because the 2002 redistricting makes it very difficult to compare results from that year and after, so I just looked at house elections in 2004, 2006 and 2008, and senate elections in 2006.
Incumbents running for re-election
Looking at the results of races where an incumbent ran for re-election and lost the one thing that immediately stands out is that in the overwhelming majority of these instances the incumbent who lost received less than 54.5% of the vote in the previous election with the median being just over 52%. So that's our first incumbent red flag, % of vote in the previous election.
There are exceptions however, with the most obvious being the 2004 Bill Kuisle - Andy Welti race, where the incumbent Kuisle had received just over 70% of the vote the previous year. Of all the incumbents in my database who lost re-election only one other had gotten over 60% of the vote the previous year and that person got 61%, so this was quite the upset.
Again, since there is no public polling of these races there aren't that many options to help identify which races are more competitive than might be expected. The best tool I found for this purpose is the amount of money a candidate raised from individual contributions.
Minnesota campaign finance information is reported in three waves; pre-primary report, a pre-general election report and a post-election report. Obviously the post-election report won't help us for these purposes and the pre-general election report isn't due until the end of October, so the only thing we can look at right now is the pre-primary report.
As it turns out the pre-primary information is more helpful than the pre-general election report for pointing out incumbents in trouble. In my database of incumbents who lost election the median amount of money raised by their challenger, as a % of the total raised between the two candidates, was 58%. Whereas the median challengers share of money raised between the pre-primary reporting period and the pre-general election reporting period was only 45%.
In the aforementioned Kuisle - Welti race, Andy Welti received 75% of the individual donation money, so in the absence of polling looking at how people are voting with their checkbooks can be helpful. This is our second incumbent red flag, if their challenger is raising more money than them in individual donations.
I also found it to be helpful in predicting the results to add in a correction for the districts hPVI, although this tends to work more as a bonus for an incumbent in friendly territory, then as a penalty for one in a hostile district.
One thing I found that had no correlation with predicting outcomes was how a candidate did in relation to either the presidential or gubernatorial candidate of their party. The performance of the candidate for higher office on it's own was helpful, but this is part of hPVI. Weather the candidate for house or senate either overperformed or underperformed the candidate for higher office however had almost no correlation with future outcomes.
So to get at the rating for an incumbent seeking re-election I'm looking at three factors; their previous share of the vote, the amount of money they raised in individual contributions compared to their opponent, and the districts hPVI. The same three factors go into rating open seat races, but there are some differences.
Open seats
Open seat races are a bit more tricky as without an incumbent in the race the uncertainty factor goes up. With open seat races, like rating incumbents, the vote share of the previous incumbent is a telling indicator, but not nearly to the same degree. In fact, on average you can expect the candidate of the party of the previous incumbent to do about 4% worse, but the standard deviation between the two is quite high, unlike with incumbents.
It was somewhat helpful to look at how much pre-primary money the challenger raised, but with open seat races it was actually more helpful to look at the amount raised between the primary and the general elections. I suspect this has to do with primary fights being more prevalent in open seat races, where the winner gets a surge of fundraising as the party coalesces.
hPVI on the other hand was more helpful in open seat races than in those with an incumbent, this might have something to do with district's regressing to the mean in open seat races, as incumbents can sometimes defy gravity and stick for a while in districts that seem either to liberal or conservative for them.
What I then did was combine these factors into two different scores, an incumbent retention score and an open seat score, I used these scores in determining the race ratings. Now these ratings are by no means perfect, nowhere in the ratings is their any accounting for subjective factors like volunteer enthusiasm, candidate quality, individual district quirks, and on and on. Instead this is just an analysis of the three factors mentioned above for each race, if you think I've got one called wrong let me know in the comments.
With that let's get to the ratings!
The Ratings
The column on the left is the house district, red means it's currently held by the GOP, blue the DFL. The next column is the Democratic candidate, then the Republican candidate. If a name is in bold that means the candidate is an incumbent, if a row is shaded grey that means it's an open seat race.
No comments:
Post a Comment