The Key House Races Pickup Prediction Model

KeyHouseRaces.com ^ | 14 October 2010 | Interceptpoint

[InterceptPoint]

Political Junkie Too, the resident statistical wiz at Freerepublic.com has taken the average KeyHouseRaces ratings by our 6 "Experts" for each of the districts on our Master List and built a computer model to generate an estimate of the outcome of the 2010 election for the House of Representatives. The model uses a technique known as a Monte Carlo simulation that utilizes probability functions for each of the KeyHouseRaces ratings categories - Leans D, Toss Up, etc and then runs a simulated election based on these probabilities.

As we started to get some decent poll data PJ Too decided to build a second model that utilized poll data less than 60 days old and, for districts where we had no poll data, to utilize the basic "Expert Rating" version of the simulation. The chart and the plot below show the calendarized results of this "hybrid" poll and expert-based simulation of the November election.

The results shown below are the average of 32,000 runs of the hybrid computer model each week since July 31st. Since the polls tend to lead the "Experts" it is not surprising that the poll-based simulation predicts a larger number of GOP seats than the Experts based model. This is simply due to the fact that Republican prospects have been improving weekly and the expert ratings updates tend to lag the polling data.

You can see the original Political Junkie Too Expert Base simulation results HERE.

Poll Based Model

The following chart summarizes the results of a 32,000 run Monte Carlo simulations of the outcome of the 2010 election for the House of Representatives based on polls that are available for the Key House Races Master List.  Where polls are not available, the "Expert" ratings are used in the same manner as is done in the  "Expert Ratings Model" simulation. The expected number of GOP seats shown is the rounded expected (mean) value of the GOP seats for the 32,000 runs.

Hybrid Expert - Poll Based Model
Week	House Seat Projection		Probability of GOP Control	Projected GOP Pickup
	GOP	Dem
31-Jul-10	212	223	8.43%	32
7-Aug-10	211	224	4.51%	31
14-Aug-10	211	224	4.44%	31
21-Aug-10	213	222	11.83%	33
28-Aug-10	217	218	42.58%	37
4-Sep-10	218	217	50.40%	38
11-Sep-10	219	216	66.39%	40
18-Sep-10	219	216	62.30%	39
25-Sep-10	218	217	61.17%	39
2-Oct-10	221	214	83.86%	41
9-Oct-10	223	212	96.15%	44

 

 

 

Poll Based Model

The data from the above table is plotted below.

The Dashed Red Line shows the probability that the GOP will re-take the House and it crossed over the 50% point on September 11th.

The Dark Black Line shows the trendline for House GOP seats that has steadily risen since early August.

That Purple Line at the top of the plot represents an estimate of the maximum number of seats that the GOP could pickup. Only 10% of the Monte Runs exceeded these value.

Likewise, the Green Line provides an estimate of the lower limit of the GOP seats - only 10% of the Monte Carlo runs fell below these values.

The bottom line: The trend is up and we need to keep it going up right to November 2nd.

[RockinRight]

I like those odds.

[InterceptPoint]

[wideawake]

So in 20 days the probability of a GOP takeover went from 11% (August 21) to 66% (September 11)?

[wideawake]

So in 20 days the probability of a GOP takeover went from 11% (August 21) to 66% (September 11)?

[Michael.SF.]

BTTT

[InterceptPoint]

So in 20 days the probability of a GOP takeover went from 11% (August 21) to 66% (September 11)?

Sure. As you get closer to the magic 218 seats and then exceed that number the probability of the GOP taking back the House rises rapidly and approaches 100%. It never makes it but it gets arbitrarily close.

[1035rep]

Great news. Thanks for all the hard work.

[GonzoGOP]

So in 20 days the probability of a GOP takeover went from 11% (August 21) to 66% (September 11)?

Unlike political junkies like most of us here, most people didn't start paying attention to politics until September first. There was a critical 10 seat swing between the dates mentioned as a large number of undecided voters in the tossup races broke GOP.

[MtnClimber]

I hope it keeps moving in this direction!!!

[bIlluminati]

This model does come close to a good number. Compare to intrade.com, which predicts Republicans gain between 55-60 seats, and yet only 85% chance of Republicans taking the House. Nice overlay here.

Also, read the fine print on DemonRat control of the Senate on Intrade. Intrade doesn’t count Lieberman and Sanders, and won’t count Crist or Murkowski, for either party. So, at about 10%, neither party controlling the Senate is low, and Demons at 50% is too high.

[jazusamo]

Thanks! This is looking great, let’s hope the seats keep increasing for the Republicans.

[GonzoGOP]

Interestingly of the 100 seats that they show as being in play only seven are currently held by Republicans. And only three of those are in danger of flipping. This shows that the dems are playing defense big time. There are also 22 seats where the dem is up by less than one percentage point. If the turnout model favors the GOP as much as it appears likely to we may pick up more than the 44 predicted.

[AuH2ORepublican]

Given that most of the pundits are still behind the curve, I expect that GOP pickups will be far, far north of 44 seats. For the sake of argument, you should move every pundit rating one spot towards the GOP (likely Dem to leans Dem, leans Dem to tossup, tossup to leans GOP, leand GOP to likely GOP, etc.) and run the numbers again to see what happens.

[saganite]

Historically, what is the largest number of seats to swap during an election for either party? Just interested to see if this election is approaching that record.

[1035rep]

“There are also 22 seats where the dem is up by less than one percentage point.”

I think the tide will bring those in as well.

[saganite]

This is also on your page and it’s interesting to note the different projections. There’s quite a disparity based on the different models but all but one project a Republican takeover.

Republican Pickup Projections

Dick Morris 73
Fred Barnes 60
Unlikely Voter 60
RCP 51
Freedom’s Lighthouse 50
Sabato 47
NYT-Silver 47
Cook 50+
KHR 44
EP 47
Rothenberg 37-45
CQ Politics 36

We Need 39 to Win the House

[Liz]

Nice work--thanks for the headsup.

Heh--this is a good omen. LV paper reporting panic in Reid's camp (Intrade: Angle 55, Reid 44) DSSC now pouring money into Nevada.

When Rand Paul called his opponent a "Democrat, Rand's numbers shot to a 15 point lead.

So how bad is it? "Angle's running the worst campaign in the country, and she could still win." Carville's comments about Sharron Angle could be said of almost any Repub candidate---with a good or bad campaign going. NOTE THAT Not One Democrat is campaigning on their legislative victories, O'care, bailouts, Tarp, and "summer of recovery."

STUCK ON STUPID No matter what progressives do, it backfires against them. Obama is between a rock and a hard place---he's damned if he does and damned if he doesn't. Not even Ohaha giving Rahm Emanuel the bum's rush out of the WH can turn the tide of the Republican tsunami forming offshore---bringing the deluge of Nov 2, 2010.

[InterceptPoint]

1894. The pickup was something like 130 seats. Not likely this year but 50 to 70 is looking possible. I’m betting that the PJ Too model will keep rising right up to election day.

[WOSG]

“Compare to intrade.com, which predicts Republicans gain between 55-60 seats, and yet only 85% chance of Republicans taking the House.”

This makes sense, because an outlyer “october surprise” might happen, so there is that 10-15% chance of a shift that denies GOP a House majority, but almost certainty that we are on track to a big GOP victory.