Political Junkie's Guide To The Elections - Week Ending October 1, 2016

Various Polls | October 2, 2016 | Political Junkie Too

[Political Junkie Too]

Previous Editions:

September 24, 2016
September 17, 2016
September 10, 2016 (Premier Edition)

Purpose of the Model

This was asked of me earlier this week. Here is my reply.

The purpose of my model over the years has never been to predict the outcome, as I am not a pollster gathering my own data. The purpose of my model has always been to reflect back the aggregate implications of the various polls that do exist, and leave it to the reader to draw insights and implications. Polls from different states with different margins of error are hard to conceptualize at an aggregate national level without a model like this to normalize the results and apply the Electoral College weightings.

I do not unskew or otherwise adjust poll results based on what I think. I use the data as presented in order for the reader to draw conclusions. You can reject it out of hand as garbage-in garbage-out, or you can isolate the garbage vs. your own perceptions and publicly call out the garbage and shame pollsters into improving their accuracy, or you can accept the results as true and take the appropriate mitigations.

Model Changes Since Last Week

• None.

The Race for the White House

This week we had the first debate, where Trump appeared unprepared and Clinton didn't collapse. Post-debate polling was brutal for Trump. Some suspect the polling is over-sampling Clinton-favorable demographic groups in order to maintain the Clinton narrative, because state polls are opposite to some national daily tracking polls.

A website called statespoll.com reviews polls against demographic trends and attempts to adjust the outcomes based on sample variances from known demographics. Where available and material, I am including the results of statespoll.com adjustments for comparison.

States reversed by statespoll.com:

• Colorado - Tilts Clinton to Tilts Trump
• Florida - Leans Clinton to Tilts Trump
• Nevada - Tilts Clinton to Tilts Trump

We shall see how this shakes out after the next debate.

Summary Results

Given the state of polling as it is, Donald Trump has an expected Electoral College Vote count of 239 versus Hillary Clinton's 300. Trump's probability of winning collapses to 28% from last week's 43%, still down -4% from last week.

With statespoll.com adjustments, Donald Trump would have an expected Electoral College Vote count of 254 versus Hillary Clinton's 284. Trump's probability of winning would be 39%.

State Polling Updates

Below are state polls since the last report, followed by the state ranking and history of Trump's chance of winning the state since Labor Day.

In California (55 EV), refreshed poll from KCBS/SurveyUSA of 732 likely voters (9/27 to 9/28) confirms Clinton's overwhelmingly lead 59%-33% in a 4-way race. California remains Safe Clinton (0.0%, 0.0%, 0.0%, 0.0%, 0.0%).

In Colorado (9 EV), a poll from CBS News/YouGov of 991 likely voters (9/21 to 9/23) has Clinton holding the lead at 40%-39% in a 4-way race (4.4% MOE). However, a poll from Gravis of 779 registered(?) voters (9/22 to 9/23) has Trump leading 41%-37% in a 4-way race (3.5% MOE), and a poll from CNN/ORC of 784 likely voters (9/20 to 9/25) has Trump leading 42%-41% in a 4-way race (3.5% MOE). A post-debate poll from PPP of 694 likely voters (9/27 to 9/28) has Clinton pulling bouncing to 46%-40% in a 4-way race (3.7% MOE). Polls from August were retired. Colorado remains a Toss-up but now tilts towards Clinton (4.2%, 60.0%, 57.6%, 46.3%).

statespoll.com adjusts Colorado to Tilting Trump (53.4%).

In Florida (29 EV), a post-debate poll from PPP of 826 likely voters (9/27 to 9/28) shows Clinton leading 45%-43% in a 4-way race (MOE 3.4%). A post-debate poll from Mason-Dixon of 820 likely voters (9/27 to 9/29) shows Clinton leading 46%-42% in a 4-way race (MOE 3.5%). A post-debate poll from Fox News/Opinion Survey of 619 likely voters (9/28 to 9/29) shows Clinton leading 47%-46% in a 4-way race (MOE 3.4%). Florida moves from Tilting Trump to Leans Clinton (40.9%, 58.1%, 68.1%, 52.8%, 30.9%).

statespoll.com adjusts Florida to Tilting Trump (51.0%).

In Georgia (16 EV), a poll from WSB-TV/Landmark of 600 likely voters (9/21 to 9/22) shows Trump leading 47%-43% in a 3-way race (MOE 4.0%). Georgia remains Strong Trump (62.2%, 62.2%-, 9.2%, 90.6%, 90.0%).

In Louisiana (8 EV), a poll from JMC Analytics of 905 likely voters (9/22 to 9/24) shows Trump leading 45%-35% in a 4-way race (MOE 3.3%). Louisiana remains Safe Trump (100.0%, 100.0%, 100.0%, 100.0%, 100.0%).

In Maine At-Large (2 EV), a new poll from Press Herald/UNH of 513 likely voters (9/15 to 9/20) has Clinton ahead at 40%-37% in a 4-person race (4.3% MOE). Maine's at-large votes remain Leans Clinton (36.3%, 29.7%).

In Maine CD-1 (1 EV), a new poll from Press Herald/UNH of 266 likely voters (9/15 to 9/20) is has Clinton winning 50%-28% in a 4-person race (6.0% MOE). Maine's CD-1 vote is Safe Clinton (0.0%, 0.0%).

In Maine CD-2 (1 EV), a new poll from Press Herald/UNH of 231 likely voters (9/15 to 9/20) is has Trump winning 48%-34% in a 4-person race (6.0% MOE). Maine's CD-2 vote is Safe Trump (100.0%, 100.0%).

In Michigan (16 EV), an updated poll from Fox 2 Detroit/Mitchell of 1,956 likely voters (9/27 to 9/27) closes 5-point Clinton lead (down from +6% to 46% to 41%) in a 4-way race (2.24.0% MOE). A new poll from MIRS-GSCI of 600 likely voters (9/18 to 9/24) also has a 5-point Clinton lead of 46%-41% in a 4-way race (2.2% MOE). A new poll from Detroit News of 600 likely voters (9/27 to 9/28) has a 7-point Clinton lead of 42%-35% in a 4-way race (2.2% MOE). Michigan remains Strong Clinton (2.3%, 2.3%, 22.0%, 22.0%, 5.0%).

In Minnesota (10 EV), a poll from KSTP/SurveyUSA of 625 likely voters (9/16 to 9/20) shows Clinton leading 46%-39% in a 4-way race (MOE 4.0%). Polls from July and older were retired. Minnesota remains Strong Clinton (14.7%, 14.7%, 8.4%, 7.2%, 1.5%).

In Missouri (10 EV), a poll from CBS News/YouGov of 1,087 likely voters (9/21 to 9/21) holds Trump's lead at 46%-37% in a 4-person race (3.9% MOE). A refreshed poll from Remington Research of 1,279 likely voters (9/26 to 9/27) also shows Trump leading at 49%-39% in a 4-person race (3.0% MOE). Polls from August were retired. Missouri moves from Strong Trump to Safe Trump (64.6%, 79.4%, 92.4%, 99.3%, 100.0%).

In Nevada (6 EV), a refreshed poll from Suffolk of 500 likely voters (9/27 to 9/29) has Clinton beating Trump 44%-38% in a 4-way race (MOE 4.4%). A poll from the Las Vegas Journal Review of 800 likely voters (9/27 to 9/29) has Clinton beating Trump 45%-44% in a 4-way race (MOE 4.0%). Nevada moves from Leans Trump to Toss-Up tilting Clinton (32.1%, 32.1%, 61.4%, 74.1%, 48.1%).

statespoll.com adjusts Nevada to Tilting Trump (50.3%).

In New Hampshire (4 EV), a poll from WBUR/MassINC of 502 likely voters (9/27 to 9/29) shows Clinton leading 42%-35% in a 3-way race (MOE 4.4%). New Hampshire remains Strong Clinton (0.1%, 4.4%, 9.5%, 6.4%, 4.6%).

In New Mexico (4 EV), a refreshed poll from Ipsos/Reuters of online likely voters (9/2 to 9/22) totally flips to Clinton leading 47-38% in a 2-way race (MOE unknown, assume 4.5%). A poll from Albuquerque Journal of 501 likely voters (9/27 to 9/29) shows Clinton leading 35%-31% in a 4-way race (MOE 4.4%). New Mexico switches from Strong Trump to Strong Clinton (0.0%, 0.0%, 85.1%, 85.1%, 9.2%).

In North Carolina (15 EV), a new poll from High Point of 404 likely voters (9/17 to 9/22) has Clinton narrowly leading 43%-42% in a 4-way race (MOE 4.9%). A post-debate poll from PPP of 861 likely voters (9/27 to 9/28) has Clinton leading 44%-42% in a 4-way race (MOE 3.3%). North Carolina falls back from Leans Trump Toss-Up tilting towards Trump (39.2%, 44.7%, 42.5%, 66.0%, 50.5%).

statespoll.com adjusts North Carolina to Leans Trump (74.4%).

In Ohio (18 EV), a poll from Gravis of 850 registered(?) voters (9/22 to 9/23) shows Trump leading 43%-42% in a 4-way race (MOE 3.4%). Ohio remains Leans Trump (26.8%, 65.6%, 62.4%, 69.7%, 68.5%).

In Pennsylvania (20 EV), a poll from Morning Call of 486 likely voters (9/19 to 9/23) has Clinton's lead dropping back to 2% with 40%-38% in a 4-way race (MOE 5.0%). A poll from CNN/Opinion Research of 771 likely voters (9/20 to 9/25) has Clinton's lead shrinking to 1% with 45%-44% in a 4-way race (MOE 3.5%). A post-debate poll from PPP of 886 likely voters (9/27 to 9/28) has Clinton's lead bouncing back to 45%-39% in a 4-way race (MOE 3.3%). Polls from August were retired. Pennsylvania moves from Strong Clinton to Leans Clinton and back to Strong Clinton (2.6%, 3.9%, 16.0%, 12.2%, 21.6%).

statespoll.com adjusts Pennsylvania to Leans Clinton (31.5%).

In South Carolina (9 EV), a poll from Winthrop of 475 likely voters (9/18 to 9/26) has Trump leading 42%-38% in a 4-way race (MOE 4.5%). South Carolina remains Strong Trump (89.5%, 89.5%, 98.4%, 98.4%, 89.4%).

In Utah (6 EV), a poll from the Salt Lake Tribune of 820 likely voters (9/12 to 9/19) has Trump leading 34%-25% in a 4-way race (MOE 3.4%). Utah remains Safe Trump (97.3%, 97.3%, 100.0%, 100.0%, 100.0%).

In Virginia (13 EV), a poll by CBS News/YouGov of 1,237 likely voters (9/21 to 9/23) shows Clinton's lead continuing with 45%-37% in a 4-way race (MOE 3.3%). A poll by Christopher Newport University of 1,003 likely voters (9/15 to 9/23) shows Clinton's lead at 39%-33% in a 4-way race (MOE 3.9%). A post-debate poll by PPP of 811 likely voters (9/27 to 9/28) has Clinton's lead at 46%-40% in a 4-way race (MOE 3.4%), no spread change from their prior poll. Virginia remains Strong Clinton (30.9%, 30.9%, 14.6%, 2.3%, 1.3%).

In Washington (12 EV), a poll from Emerson of 700 likely voters (9/25 to 9/26) has Clinton leading 44%-38% in a 4-way race (MOE 3.6%). Polls from August and earlier were retired. Washington moves from Safe Clinton to the outer fringes of Strong Clinton. (0.0%, 0.0%, 0.0%, 0.0%, 1.3%).

Must-Win States

This is a chart of the states that were won when Trump reached 270 or more Electoral College votes. The numbers indicate the percent of the time the state was won when Trump won the election.

States in the 90+% range must be won to reach 270 electoral votes.

Elecctoral College Vote History

Using an average of polls from Real Clear Politics, the results of simulated elections are listed in the table below. The definitions of the columns are:

• "GOP Electoral Votes P10” – the value where the probability of the actual number being lower is 10%.
• ”GOP Electoral Vote EV” – the expected value (mean) or probability-weighted average of the simulation.
• ”GOP Electoral Vote P90” – the value where the probability of the actual number being lower is 90%. Or, it is the value where the probability of the actual number being higher is 10%.
• ”Probability of 270” – The probability associated with value 270 (the Electoral College majority goal).

Week	GOP Electoral College P10	GOP Electoral College EV	GOP Electoral College P90	Probability of 270
03-Sep-16	148	214	295	21%
10-Sep-16	152	231	298	29%
17-Sep-16	181	258	343	40%
24-Sep-16	192	256	318	43%
01-Oct-16	180	239	302	28%

Probability Curve

State Rankings Definitions

• Safe - 100% probability of winning
• Strong - 75% to 99.99% probability of winning
• Lean - 62.5% to 74.99% probability of winning
• Toss-Up - 50% to 62.49% probability of winning

24-Sep-16 State Rankings

Clinton - 292						Toss-Up - 30				Trump - 216
Safe		Strong		Leaning		Tilt		Tilt		Leaning		Strong		Safe
EV	State	EV	State	EV	State	EV	State	EV	State	EV	State	EV	State	EV	State
55	California	7	Connecticut	29	Florida	9	Colorado	15	North Carolina	18	Ohio	11	Arizona	9	Alabama
3	District of Columbia	20	Illinois	2	Maine	6	Nevada					16	Georgia	3	Alaska
3	Delaware	16	Michigan									11	Indiana	6	Arkansas
4	Hawaii	10	Minnesota									6	Iowa	4	Idaho
10	Maryland	4	New Hampshire									9	South Carolina	6	Kansas
11	Massachusetts	14	New Jersey									38	Texas	8	Kentucky
29	New York	5	New Mexico											8	Louisiana
3	Vermont	7	Oregon											6	Mississippi
1	Maine CD-1	20	Pennsylvania											10	Missouri
		4	Rhode Island											3	Montana
		13	Virginia											5	Nebraska
		12	Washington											3	North Dakota
		10	Wisconsin											7	Oklahoma
														3	South Dakota
														11	Tennessee
														6	Utah
														5	West Virginia
														3	Wyoming
														1	Maine CD-2
119		142		31		15		15		18		91		107

Electoral Map

[Political Junkie Too]

There is no way around concluding that Donald Trump had a very bad polling week.

Florida was the biggest loss for Trump this week. He needs to reestablish his position there, falling from a 52.8% probability to 30.9% chance of winning the state.

Trump also took a big hit in Colorado, falling from a 57.6% chance of winning to 46.3%.

Nevada is another state where Trump's solid lead (74.1% last week) fell to 48.1%.

North Carolina is also still a Toss-Up, having pushed towards Trump last week but falling back again this week.

New Mexico flipped the most to Clinton, but polling here was always suspect. Trump's lead until now may have been a mirage due to lack of credible polling in the past.

And finally, Pennsylvania is a state where people have high hopes, but so far, polling is not bearing it out. It's true that Trump's chances are improving here from 12.2% to 21.6%, but that's too far away at this point in the race to be counting on a complete reversal unless something drastic happens in the race or in polling rigor.

-PJ

[Political Junkie Too]

Pinging to the latest report.

Not a good week for Trump. His work is cut out for him. He really needs to rethink his approach to the next debate a week from today.

As always, comments/feedback is always welcome.

Let me know if you want on or off the ping list.

-PJ

[LS]

I have Trump solidifying 272. Pretty damn good week. He got FL and OH this week.

[NYC-RepublicanCT]

Agreed.. not a good week, but there’s still time.. that is, if he stops with the “Romney nice guy” performance and blankets the airwaves with ads ALREADY!!!!

[bankwalker]

At this stage of the game the polls are still just a mind manipulation tool. They will become more realistic about a week before the election to help preserve the pollsters' credibility.

[vooch]

good number crunching

my question is ‘if Crooked is doing so poorly in must-win demographics how can she be ahead ?’

for example:
Blacks, she must win 90-95% and is barely getting 75%

under 30s, she must win 70-75% and is barely getting 55%

[NYC-RepublicanCT]

Please add me. Thanks

[Political Junkie Too]

I hope so, but I am dealing with the facts of the polls as they are published.

I ask you... how long can the pollsters go reporting polls opposite to what the on-the-ground observations suggest?

I agree with you that Ohio looks solid for Trump, and the polls bear that out, but I cannot say the same thing for Florida based on the polls. Even statespoll.com adjustments (not comprehensive of Florida polls) has it at a 51%-49% chance for Trump.

If I ONLY include the statespoll.com adjusted polls (and ignore the polls not looked at), then Florida is 74%-26% probability for Trump. But I don't want to fall into an omission bias, either.

-PJ

[LS]

I put no stick in polls. I know, 2012.

This time I think they go to the mat because they ALL are done when Trump wins.

Again I don’t look at polls, I look at absentees and early votes. Those are 180 degrees opposite the polls. She is already behind in FL 47-33. Not a lot of votes but real votes, not polls. She’s behind 145,000 in absentee requests with ONE QUARTER of all GOP ballots already. No, FL is gone.

Now, NV gone. I will post later on this.

[Political Junkie Too]

Some thoughts...

Blacks, she must win 90-95% and is barely getting 75%

African-Americans are about 16% of the total population. This is concentrated in the major cities. As long as those cities are in traditionally Democrat states, it's possible that there are still enough votes in those dense Democrat states to sustain winning the state.

under 30s, she must win 70-75% and is barely getting 55%

Are they filtering out as not "likely voters?" Are they not sufficiently represented in polling due to predominantly cell-phone use? If these were Sanders supporters, they may choose to stay home rather than vote for Trump.

-PJ

[Political Junkie Too]

If nothing else, my model is a documentation of the history of the polling throughout the election, starting from a baseline way out all the way through final voting.

In 2012, Romney peaked at 45% probability of winning in the final week, and that was using my old techniques.

In 2008, my earliest model had McCain peaking at 62.5% chance of winning in mid-September with 275 EV, but that turned out to be a "Palin bounce" coming out of the convention, something that McCain could not sustain.

Bottom line... I rely on readers like you to point out where observations conflict with polling. I won't change the results based on personal observation, but will point out where polling seems at odds with a preponderance of personal observations, or where polling seems stuck on a narrative at odds with personal observations.

-PJ

[gdzla]

I trust that come Nov. 8, 2016 the clinton caliphate will be defeted.

[nathanbedford]

I wish I had been wrong but this result was evident from halfway through the debate.

[Political Junkie Too]

Did I just take sides in a feud? Care to include me in a discussion within my thread?

-PJ

[nathanbedford]

No feud as far as I'm concerned rather a frank discussion of the effect of the debate on the polls and, by extension, on the election itself. I Cited your thread because of the valuable service you perform in so evenhandedly delivering valuable data.

[SharpRightTurn]

“Trump ... really needs to rethink his approach to the next debate a week from today.”

I concur. Ignore the war on women, tax, and personal questions. Hammer Madame Mao on trade deals, immigration, loss of jobs, foreign policy failures, and corruption. Non-stop.

[Art in Idaho]

Still think, with the voter fraud, this will be very close. The voter fraud will affect PA for sure and several other states, possibly FL, NC, NV, (Reid - remember Sharron Angle) and CO.

My current guess:

My 270 to Win puts Trump at 270 and Hillary at 268. Way too close for comfort. If Trump takes FL and OH, then it's CO, IA, NC, NV and NM deciding in my estimation. Man o' man. . . The whole future of the nation, whether our Sacred Republic lives or dies, decided by 2 or 3 states. . Sounds like the Supreme Court.

Every vote counts. Get out the Vote!

[Political Junkie Too]

Thanks.

-PJ

[mrsmith]

A link to statespoll that works: http://statespoll.com/

About that PPP poll...
They asked whom the respondents voted for in 2012.
Here’s the ‘overage’ of the poll results above the actual vote for Obama:

Virginia +3%
Penn +4.5%
Colo +1%
Florida +4%
NC +4%
Three of of those are outside the professed MOE of the poll!

So I REALLY discount it.

I don’t think Trump lost votes in the debate, but he may jave lost some momentum. No way to tell.

[mrsmith]

You guys might enjoy this:
http://www.buenosairesherald.com/article/222172/colombians-expected-to-say-%E2%80%98yes%E2%80%99-to-peace
“ According to two opinion polls released yesterday, Colombian voters will approve the government’s peace deal with Marxist FARC rebels in Sunday’s plebiscite, ratifying an end to more than 52 years of war.

One survey, conducted by polling firm Ipsos on behalf of local media RCN Radio, RCN Television, La FM radio and Semana magazine, showed 66 percent of those voting would approve the deal, while 34 percent would vote against it. Another poll by Datexco, published by newspaper El Tiempo and W Radio, estimated that 55 percent of Colombians will vote “yes” in the referendum and 36.6 percent will vote “no.” “

“NO” won LOL!
There’s something serious wrong with polling these days.