TIA-IF NO FRAUD, Dems will win a 238-200 House majority- a 99% probability

[Melissa G]

Well, I'm Still in intense yoga meditation thru next Sunday (though it is a bit lighter time-wise mid-week)
Anyway, an Elaboration from the the Previous TIA Channeling that I did has come through...
As always, I hope those Fans of TIA will keep this K&R'd for greater viewing pleasure!

IF NO FRAUD, Dems will win a 238-200 House majority- a 99% probability!


A Monte Carlo Forecast Simulation

How many of the 58-contested GOP House seats can the Democrats expect to win, assuming a FRAUD-FREE election? Corollary: How many elections will the GOP need to steal to maintain control?

This is an update to a prior analysis of 31 competitive GOP-held seats. The data base is bigger; it includes the latest polls for 58 seats. The Democrats need to capture 15 of them to gain control of the House. The analysis ASSUMES ZERO FRAUD. It is based strictly on the latest poll shares, undecided voter allocation assumption and the margin of error. The analysis will be run again in the coming weeks in order to update the model with new polling data.

The goal of the simulation is to calculate a range of probabilities of the Democrats winning a specified number of the 58 elections, over a range of undecided voter allocation assumptions. The simulation produces a VERY ROBUST ESTIMATE of the number of elections that would need to be stolen in order for the GOP to retain control of the House.

SIMULATION OBJECTIVE:
Calculate a 25-scenario probability matrix based on
a) 5 scenarios of undecided voters allocated to the Democrats and
b) 5 scenarios of the number of captured GOP seats

In a published study of over 150 elections, the challenger won the undecided vote 82% of the time. Even with the VERY CONSERVATIVE ASSUMPTION that the Democrats will win just 50% of the undecided, it's a virtual 100% probability that they will WIN AT LEAST 25 SEATS, TEN more than they need for a House MAJORITY, assuming NO FRAUD.

In the MOST LIKELY SCENARIO, in which it is assumed that the Democrats will win 60% of the undecided vote, they can expect to capture 32 of the 58 GOP-held seats. There is a 99% probability that they will win 30 or more. Therefore, in order to retain the House, the GOP will HAVE TO STEAL A MINIMUM of 15 elections.

This is a polling bar chart with the win probabilities:


Sensitivity Analysis
--------------------

UVA: Undecided voter % allocated to Democrats
UVA 50 55 60 67 75
Dems win ----------Probability%--------
25 seats 100 100 100 100 100
30 seats 35 85 99 100 100
35 seats 0 3 30 90 100
40 seats 0 0 0 5 70
45 seats 0 0 0 0 1

Polling Detail
--------------

Adjusted: Dem & GOP polls
(60% undecided to Democrat)
Prob: probability of Democratic win

Num District Pollster Poll MoE Poll Dem GOP Adjusted Prob
Code Average Sample 3.80 Date 45.4 45.9 50.6 49.4 57.8%

1 AZ 1 RT Strategies 983 LV 3.09 10/10 50 46 52.4 47.6 99.9%
2 AZ 5 SurveyUSA 509 LV 4.40 10/15 45 48 49.2 50.8 23.8%
3 AZ 8 Zogby 500 LV 4.50 10/2 45 37 55.8 44.2 100.0%
4 CA 4 RT Strategies 997 LV 3.09 10/10 44 52 46.4 53.6 0.0%
5 CA 11 Greenberg 413 LV 4.90 9/26 48 46 51.6 48.4 90.0%

6 CA 50 SurveyUSA 540 LV 4.30 9/12 40 54 43.6 56.4 0.0%
7 CO 4 Mason-Dixon 400 LV 5.00 10/7 36 46 46.8 53.2 0.6%
8 CO 5 Mason-Dixon 400 LV 5.00 10/7 37 37 52.6 47.4 97.9%
9 CO 7 RT Strategies 991 LV 3.09 10/10 47 47 50.6 49.4 77.7%
10 CT 2 Zogby 500 LV 4.50 10/2 41 44 50.0 50.0 50.0%

11 CT 4 Zogby 500 LV 4.50 10/2 46 41 53.8 46.2 100.0%
12 CT 5 RT Strategies 996 LV 3.09 10/10 46 52 47.2 52.8 0.0%
13 FL 13 RT Strategies 1024 LV 3.07 10/10 47 44 52.4 47.6 99.9%
14 FL 16 RT Strategies 1001 LV 3.09 10/1 50 43 54.2 45.8 100.0%
15 FL 22 RT Strategies 1022 LV 3.10 8/29 44 52 46.4 53.6 0.0%

16 ID 1 RT Strategies 998 LV 3.09 10/10 43 49 47.8 52.2 0.3%
17 IL 6 RT Strategies 997 LV 3.07 10/10 47 47 50.6 49.4 77.8%
18 IL 10 Mellman Group 400 LV 4.90 10/11 32 49 43.4 56.6 0.0%
19 IL 14 RT Strategies 1003 LV 3.08 10/10 42 52 45.6 54.4 0.0%
20 IL 19 RT Strategies 1023 LV 3.08 10/10 36 53 42.6 57.4 0.0%

21 IN 2 RT Strategies 989 LV 3.07 10/10 50 46 52.4 47.6 99.9%
22 IN 8 Indiana State 626 LV 3.90 10/12 55 32 62.5 37.5 100.0%
23 IN 9 SurveyUSA 512 LV 4.40 10/15 48 46 51.6 48.4 92.3%
24 IA 1 Bennett, Petts 400 RV 4.90 10/8 48 37 57.0 43.0 100.0%
25 IA 2 RT Strategies 1006 LV 3.09 10/10 48 47 51.0 49.0 89.8%

26 KY 3 RT Strategies 996 LV 3.09 10/10 48 48 50.4 49.6 69.4%
27 KY 4 RT Strategies 1000 LV 3.09 10/10 46 49 49.0 51.0 10.2%
28 MN 1 RT Strategies 1024 LV 3.08 10/10 47 48 50.0 50.0 50.0%
29 MN 2 SurveyUSA 519 LV 4.40 10/15 42 50 46.8 53.2 0.2%
30 MN 6 RT Strategies 995 LV 3.09 10/10 50 45 53.0 47.0 100.0%

31 NV 3 Mason-Dixon 400 RV 5.00 9/21 37 47 46.6 53.4 0.4%
32 NH 1 Research 2000 300 LV 6.00 9/14 31 56 38.8 61.2 0.0%
33 NH 2 Univ of NH 220 LV 6.20 9/24 36 46 46.8 53.2 2.2%
34 NJ 7 RT Strategies 1022 LV 3.10 10/10 46 48 49.6 50.4 30.6%
35 NM 1 RT Strategies 986 LV 3.09 10/10 52 44 54.4 45.6 100.0%

36 NY 3 RT Strategies 984 LV 3.09 10/10 46 48 49.6 50.4 30.6%
37 NY 19 Abacus Assoc 600 RV 4.00 9/15 44 49 48.2 51.8 3.9%
38 NY 20 Grove Insight 400 RV 4.90 10/13 41 42 51.2 48.8 83.1%
39 NY 24 RT Strategies 1029 LV 3.07 10/10 53 42 56.0 44.0 100.0%
40 NY 26 RT Strategies 1056 LV 3.07 10/10 56 40 58.4 41.6 100.0%

41 NY 29 Cooper&Secrest 503 LV 4.40 9/21 39 43 49.8 50.2 42.9%
42 NC 8 RT Strategies 1029 LV 3.08 10/10 51 44 54.0 46.0 100.0%
43 NC 11 RT Strategies 979 LV 3.09 10/10 51 43 54.6 45.4 100.0%
44 OH 1 Anzalone-Liszt 500 LV 4.40 8/01 45 45 51.0 49.0 81.4%
45 OH 2 RT Strategies 1003 LV 3.09 10/10 48 45 52.2 47.8 99.7%

46 OH 6 RT Strategies 982 LV 3.10 8/29 56 40 58.4 41.6 100.0%
47 OH 15 RT Strategies 1015 LV 3.09 10/10 53 41 56.6 43.4 100.0%
48 OH 18 Greenberg 400 RV 5.00 10/11 48 41 54.6 45.4 100.0%
49 OK 5 SurveyUSA 435 LV 4.70 10/10 33 62 36.0 64.0 0.0%
50 PA 6 RT Strategies 1023 LV 3.07 10/10 52 46 53.2 46.8 100.0%

51 PA 7 RT Strategies 1017 LV 3.08 10/10 52 44 54.4 45.6 100.0%
52 PA 8 Grove Insight 400 RV 4.90 10/15 44 40 53.6 46.4 99.8%
53 PA 10 Bennett,Petts 400 RV 4.90 10/9 51 37 58.2 41.8 100.0%
54 VA 2 RT Strategies 982 LV 3.10 10/10 46 48 49.6 50.4 30.6%
55 VA 5 SurveyUSA 502 LV 4.40 10/10 40 56 42.4 57.6 0.0%

56 VA 10 RT Strategies 1004 LV 3.07 10/10 42 47 48.6 51.4 3.7%
57 WA 8 RT Strategies 1015 LV 3.09 10/10 45 48 49.2 50.8 15.5%
58 WI 8 RT Strategies 983 LV 3.09 10/10 48 46 51.6 48.4 97.9%

[IndyOp]

1. IGNORE THE POLLS. GOTV. GOTV. GOTV. GOTV.

Thanks for posting this - let's keep our eyes on the prize (and ready our pitchforks for the after election fraud rumble).

:kick:

[Pab Sungenis]

2. Can we count on 60% of the undecided vote?

Honestly? I'd like to think so, but I'm pessimistic. What do the numbers look like with 51% of the undecideds?

[Melissa G]

5. pabsungenis, Check your inbox...

[Melissa G]

6. From the Op at 50%...

"Even with the VERY CONSERVATIVE ASSUMPTION that the Democrats will win just 50% of the undecided, it's a virtual 100% probability that they will WIN AT LEAST 25 SEATS, TEN more than they need for a House MAJORITY, assuming NO FRAUD."

[bleever]

3. Thank you, MG.

Om shanti shanti, and


GOTV! GOTV! GOTV!
GOTV! GOTV! GOTV!
GOTV! GOTV! GOTV!
GOTV! GOTV! GOTV!
GOTV! GOTV! GOTV!

[Melissa G]

4. INDEED Bleever! Get Out the Effing Vote and monitor The Heck out

of those Effing Black Boxes and Tabulators!

[stellanoir]

7. ahhh

TIA's stats are a soothing balm for my weary eyes. Thanks for the refreshing post. Enjoy your yoga and meditation dear.

K & R'd

[Melissa G]

8. Thanks, stellanoir!

:hi:

[Kurovski]

9. K&R for the most limber member of ER.

Oh, and that whispering little birdy, too.

[Melissa G]

10. LOL..I don't know about Limber.. esp after 10.5 hours of yoga class

but I likely chant more than any other ER member.. and my husband does say I'm a wee bit twisted..:-)

[Kurovski]

11. But twisted in just the right direction...

:hug:

[foo_bar]

12. a 99% probability three weeks before the election? is this a joke?

Oh right, it's TIA.

[Awsi Dooger]

13. It's a 100% probability there are 435 House seats

Not 438.

[OnTheOtherHand]

15. FWIW it looks like TIA's number would be

that Dems have a 99% chance of picking up 30 seats, so, 233-202 or so. The 238-200 in the OP appears to be an editorial interpolation.

Of course, TIA never even thinks to list as one of his assumptions that the polls are accurate within sampling error -- all house effects are zero, etc. Here are some thoughts from Mark Blumenthal, whom TIA used to like until Blumenthal had the temerity to disagree with him: http://www.pollster.com/mystery_pollster/handicapping_the_house_part_i.php

Nothing here says that the polls are necessarily wrong on average, just that the uncertainties are a lot larger than TIA's "VERY ROBUST ESTIMATE" admits. Damn, I hate it when people hawk false certainty on progressive websites.

But if people are encouraged to go vote, I'm all for that.

[foo_bar]

16. does overconfidence historically inspire people to vote?

I think the key unstated assumption is "if the election were held tomorrow", except TIA's headline reads "99% PROBABILITY Dems will win over 30 GOP House seats - IF NO FRAUD!" (I can't link the original post since it's against DU rules, and I'm neither clever nor devious enough to invoke spiritual powers), so the assumption might be "if nothing at all happens in the next three weeks except fraud". Either way it's worth the comic relief, so hats off to TIA in his quest to be right just this one time, which must be as statistically inevitable as the truthfulness of a stopped clock.

[OnTheOtherHand]

17. heck if I know

To take your question seriously and acknowledge uncertainty -- which, of course, is sort of weird in a TIA thread -- it seems to me that in order to understand the effects of "overconfidence," we would need multiple empirical measures tapping alternative definitions. I think some forms of 'overconfidence' could indeed encourage turnout in some circumstances. People like to be part of a winning cause. If people like the way TIA's posts taste, maybe they will brush longer. Who knows?

Yes, the forecast assumes that nothing changes -- but the probability also has to assume that sampling error is the only source of error. On the other hand, if we only assume that the individual polls have larger -- but uncorrelated -- error variances, the forecast might not change very much. So you may well be right that nothing-changing is the single most important unstated assumption. (Of course it also assumes that the polls have no systematic error, which is not a very conservative assumption.)

Treated as a game, it's pretty fun.

[livvy]

14. Thanks for the post.

There's nothing quite like a TIA post to get me fired up. I don't always get all the details, but the gist is always plain!

[DeeDeeNY]

18. 99% sounds good, but

the percentage that is equally important to me is the the probability that the election will be fraud free.