About
Learn More about our progressive online community for the Granite State.

Create an account today (it's free and easy) and get started!
Menu

Make a New Account

Username:

Password:



Forget your username or password?


Search




Advanced Search


The Masthead
Managing Editor
Mike Hoefer

Editors
elwood
susanthe
William Tucker
The Roll, Etc.
Prog Blogs, Orgs & Alumni
Bank Slate
Betsy Devine
birch paper
Democracy for NH
Granite State Progress
Mike Caulfield
Miscellany Blue
Pickup Patriots
Re-BlueNH
Still No Going Back
Susan the Bruce
New Hampshire Labor News
Chaz Proulx: Right Wing Watch

Politicos & Punditry
The Burt Cohen Show
John Gregg
Landrigan
Pindell
Primary Monitor
Scala
Schoenberg
Spiliotes

Campaigns, Et Alia.
NH-Gov
- Maggie Hassan
NH-01
- Andrew Hosmer
- Carol Shea-Porter
- Joanne Dowdell
NH-02
- Ann McLane Kuster

ActBlue Hampshire
NHDP
DCCC
DSCC
DNC

National
Balloon Juice
billmon
Congress Matters
DailyKos
Digby
Hold Fast
Eschaton
FiveThirtyEight
MyDD
Open Left
Senate Guru
Swing State Project
Talking Points Memo

50 State Blog Network
Alabama
Arizona
California
Colorado
Connecticut
Delaware
Florida
Idaho
Illinois
Indiana
Iowa
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Missouri
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Rhode Island
Tennessee
Texas
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin

Forecasting the NH State House Election - Part 1

by: William Tucker

Sat Oct 09, 2010 at 10:49:25 AM EDT


To better understand Granite State voting behavior, Judy Stadtman and I have compiled town-level New Hampshire voting data from each statewide election since 2000. One of the exercises in that analysis was to calculate and document the Partisan Voting Index for each of the state's 103 State House districts. That data has now become the basis for a model to forecast the results of the upcoming State House of Representatives election.

At a high level, the model is based on a strong correlation between the national partisan vote and results in New Hampshire House elections. I begin with the projected national two-party vote, adjust it for each House district based on the district PVI, and determine the probability of a partisan outcome for each seat. I then run 100,000 simulations of the outcome for each seat to project the total number of seats for each party.

In this post, I'll detail the methodology (warning: geek alert!) and test the model by retroactively applying it to the 2004, 2006 and 2008 House elections. Part 2 will follow with my fearless forecast.

William Tucker :: Forecasting the NH State House Election - Part 1

Methodology

I begin by comparing the average national two-party vote from 2004 and 2008 Presidential elections (Democrats: 51.2%) with the projected national two-party vote for the upcoming election to determine a projected partisan index for each district. For example, if polls predict a 52.2% Democratic vote, I adjust the House District PVI by +1.0.

I then calculate the probability of a partisan outcome in each district. I sort the districts based on PVI and then group them together in 20 groups of approximately 20 representatives per group. Next, I calculate the average percentage of partisan seats in each group for 2004 and 2008 (the years included in the PVI calculations) and plot those points that have partisan percentages between zero and 100% for both years.

I then factor each seat’s projected partisan index by the result to determine the probability of a partisan result. For example, a seat in a district with projected partisan index of +5 has approximately a 65% probability of electing a Democrat. I then run 100,000 simulations for each of the 400 seats to complete the forecast.

Validating the Model

To test the model, I simulated the State House elections of 2004, 2006 and 2008 (the period covered by the PVI calculations). For 2004 and 2008, I use the national two-party Presidential vote (in lieu of the projected national two-party vote), in 2006 I used the national two-party Congressional vote. In each case, the projected results are within five seats of the actual results.

Forecasting the 2010 Election

Looking backwards, of course, I have the advantage of being able to use actual rather than projected national voting results. To forecast the 2010 results, I will use the results from generic Congressional ballot surveys of likely voters -- more on that and the 2010 forecast in Part 2.

Cross-posted to Miscellany Blue

Tags: , , , , (All Tags)
Print Friendly View Send As Email
So here's my question... (4.00 / 1)
I run in Cheshire 2, which according to your table was a D+9 in 2008.  Does that calculation take into account that in 2008, only one R ran in a three-seat district?  Certainly each of the three Ds who ran would have gotten lower percentages of the vote had there been a full slate on the other side.

The object of life is not to be on the side of the majority, but to escape finding oneself in the ranks of the insane. --Marcus Aurelius, courtesy of Paul Berch

Methodology (4.00 / 1)
The model doesn't look at votes cast in a district, only the results. It takes into account the fact that Democrats won 2 of 3 seats in 2004 and 3 of 3 in 2008. BTW, the model calculates each seat in a +9 district had a 84% probability of electing a Democrat during that time. That is consistent with the Cheshire-2 results (5/6 = 83%).

"Politics ain't beanbag" - Finley Peter Dunne

[ Parent ]
I think I've got it (0.00 / 0)
The column on percentage of votes doesn't figure into the calculation, is that it?

The object of life is not to be on the side of the majority, but to escape finding oneself in the ranks of the insane. --Marcus Aurelius, courtesy of Paul Berch

[ Parent ]
The percentage... (0.00 / 0)
is % of seats, not % of vote. For example, approximately 60% of the seats from D+4 districts have been won by Democrats.

"Politics ain't beanbag" - Finley Peter Dunne

[ Parent ]
Numbers, and then there are the other factors (0.00 / 0)
I am looking at your State House Data Sheet 1.  There is a column headed Vote.  In that column, the number after Dan Eaton's name is 84%, Tara Sad is 81% and I am 79%.  Waht does that signify?  

I took it to be an indicator that we had only one opponent, who was relatively unknown.

I asked the questions because I'm trying to understand the data, but the real point behind the questions is this:  

Races for the NH House are, more than any other race, dependent on name recognition and character than they are on political party.  When one observes recounts, many ballots have votes for Reps whose votes are, for the most part, going to cancel each other out in the House.   But platform is less important to people on this most local level than is knowing who they are voting for.

The object of life is not to be on the side of the majority, but to escape finding oneself in the ranks of the insane. --Marcus Aurelius, courtesy of Paul Berch


[ Parent ]
Oh, I see (4.00 / 1)
My mistake. The colunn you're referring to was an attempt to show the relative strength of the candidates. But, as you point out, it's doesn't mean much. The model doesn't use it.

I agree completely with your last point. It would be silly to think we could predict any given House race just by looking at the partisan makeup of the district. But I do think it's important to understand it.

"Politics ain't beanbag" - Finley Peter Dunne


[ Parent ]
So very grateful for William Tucker's analysis (4.00 / 1)
Thanks for the enormous amount of effort that you put into this.  Very few people offer any analysis whatsoever of House districts, so the information you have compiled is invaluable--so long as it does not make anyone complacent.

The object of life is not to be on the side of the majority, but to escape finding oneself in the ranks of the insane. --Marcus Aurelius, courtesy of Paul Berch

[ Parent ]
Thanks, Lucy n/t (0.00 / 0)


"Politics ain't beanbag" - Finley Peter Dunne

[ Parent ]
Can this method be used to establish the unfairness of large multi-member districts? (4.00 / 1)

ie do the results significantly skew in heavily gerrymandered districts? If so it might be a useful tool in establishing fairness in redistricting.

"But, in the unlikely story that is America, there has never been anything false about hope." Si se puede. Yes we can.  

Thank you (0.00 / 0)
Interesting, as I run in Cheshire-01, which , according to PVI, SHOULD be a Dem seat, but is not.

Here I am, awaiting Part 2, instead of kocking on doors......


That is the danger... (0.00 / 0)
models and just that. The only real way for us to maintain our majority is if you go door to door.

Obama/Biden '08

[ Parent ]
actually, (0.00 / 0)
I did spend the morning at the Westmoreland Transfer Station a/k/a The Dump.  

[ Parent ]
And Cheshire 1 is a case in point. (4.00 / 1)
It went D in 2004 and 2006, and then had a candidate who had way too many other distractions in 2008.

We had an easy ride in 2008 in Cheshire 2 with only one not very well-known opponent against the three of us.  This year, there are three, all well known, and they are working hard.  Tara knocked on doors today, I made phone calls.

The object of life is not to be on the side of the majority, but to escape finding oneself in the ranks of the insane. --Marcus Aurelius, courtesy of Paul Berch


[ Parent ]
Just to be clear (0.00 / 0)
This model attempts to forecast overall results based on historical data. I don't claim that it is predictive of individual races. It doesn't take into account the unique characteristics of each race or the ground game (canvassing, phone banks, etc). I've made a couple of thousand phone calls this election cycle and I wouldn't do it if I didn't believe it makes a difference.

"Politics ain't beanbag" - Finley Peter Dunne

[ Parent ]
Predictability (4.00 / 1)
As I understand this:
  1. The data show a strong correlation between national election results and state House district results
  2. If the national results favor one party by 5 points and a particular district historically favors the other party by 3 points, that district results will tend to give the nationally-favored party a 2 point margin
  3. The missing link is the jump from "local races are correlated with national results" to "local races can be predicted from national polls."
  4. The apparent gaps in that link include: which national poll[s] to use; their predictive ability of what they attempt to measure; and their predictive ability of the Congressional results. (Getting from "generic ballot" numbers to individual US House race numbers isn't beanbag.)

It would be interesting to see a check of historical national poll results, rather than electoral results, as correlated to local NH races...


Good questions (0.00 / 0)
1. Yes.

2. If I understand the question... PVI measures a district's vote against the average national vote. 2008 PVI is based on a 51.2% Democratic vote. A D+3 district had an avg 54.2% D vote. If the national vote favors the other party by 5% (47.5%-52.5%) you would expect the district to favor the Democrat by 1 point (50.5%-49.5%). But this model uses the 3.7 point swing in its calculations.

3. Agreed. Even if there is a perfect correlation with the national vote, the model is only as accurate as the your national vote projection.

4. There are a couple of obvious options. The last Gallup Generic Congressional Ballot poll has been an accurate predictor of the eventual vote in midterm elections. Or there's the Nate Silver approach of aggregating all polls and weighing by sample size and age. Given the recent fluctuation in the Gallup polls, I'm leaning toward the latter. But if you've got a better suggestion, I'm all ears.

"Politics ain't beanbag" - Finley Peter Dunne


[ Parent ]

Connect with BH
     
Blue Hampshire Blog on Facebook
Powered by: SoapBlox