June 29, 2017

A Final Election Prediction (based on polling averages, demographics, the economy, early voting, and primary turnout)

This is my final prediction for how the election will go:

0AgWn

These predictions are based on a few different considerations.

First, the real clear politics polling average and the trendline of real clear politics polls. The trend is defined as the average change per day since the last change in direction of the polls.

Second, the polls plus 538 predicted voter share. This model takes the polls and corrects for pro republican/democrat bias that a pollster has historically exhibited, how accurate the poll has been in past elections, and the poll’s sample size and recency. This model also includes data on changes in the economy such that an improving economy improves the chances of the incumbent party candidate (Clinton) and demographics so that if a state has had a demographic change relative to the last election in a demographic group which tends to vote in a particular way their forecast will take that into account.

538 is generally a better picture than RCP of the polls. However, while 538 does correct for trends seen by looking at multiple polls issued by the same pollster, it does not offer an easy way to calculate trends over the last few days. Because the polls in many swing states have changed trends within the last few days, and because many swing states are very, very, very, close in the polls, very short term trends are important to look at.

538 does not include data on early voting or primary turnout because they incorrectly think that these variables are not predictive of election results. There are other articles on this website explaining why this is wrong.

The third variable I looked at is early voting. Specifically, I looked at the ratio of democrats to republicans in early voting in 2012, compared this to the final voting ratio of democrats to republicans in 2012, and in light of the disparity considered the ratio of democrats to republicans in 2016 early voting.

There is one specifical consideration to keep in mind when looking at early voting. This year, Clinton dedicated a ton of resources towards early voting and early voting turnout this year was a record high. It was high for both parties though so it is not clear how much Clinton’s efforts worked. However, to the degree that they did work she convinced some people who otherwise would not have voted to vote early and other people who would have voted in election day to vote early. To the degree that Clinton’s efforts have done the latter, early voting will overpredict democrat voting even more than it did in 2012.

Some states don’t have early voting or their early voting is very limited so all the states did not have numbers on this.

Lastly, I considered how primary turnout in each state changed for each part in 2016 relative to 2008.

I’ve also posted the results of each state in the 2012 and 2008 election. These numbers are important when interpreting the early voting and primary data.

Unless another hyperlink is offered, early voting data for 2012 comes from here. Data on changes in primary turnout comes from here. Real Clear Politics data can be found here. 538 data can be found here.

Below I’ve included data on all of RCP’s swing states except for GA and NM, which are not real swing states. When the data is mixed I’ve also included a little bit about interpretation.

 

Florida

Real Clear Politics Polling Average: Clinton + 0.2

Real Clear Politics Polling Trend: Clinton -0.5 for 2 days

Five Thirty Eight projected vote share: Clinton +0.3

2012 Election Outcome: Obama +0.9

2008 Election Outcome: Obama +2.8

Early Voting dem/rep 2012: 1.10

Final Voting dem/rep: 1.02

Early Voting dem/rep 2016: 1.04

Change in primary turnout (since 2008): -15.1% dem, +5.3% rep

The polling averages and primary turnout suggest that Florida is basically 50/50. With respect to early voting, we have some extra knowledge about Florida: Trump does nearly 20 points better among Florida voters who say that will be voting on election day when compared to those who say they are voting early. Given this, the polling trends and early voting suggest that Trump will win Florida. None of these variables strongly predict one outcome or another, but the ones favoring Trump are stronger than those not favoring Trump, so I predict Trump will win Florida.

 

Pennsylvania

Real Clear Politics Polling Average: Clinton +2.1

Real Clear Politics Polling Trend:  Clinton -0.65 points per day for 6 days

Five-Thirty-Eight Adjusted Polling Average: Clinton + 3.4

2012 Election Outcome: Obama +5.4

2008 Election Outcome: Obama +10.3

Change in primary turnout (since 2008): -30.7% dem, +87.9% rep

I’ve written in detail here about why I think that the polls in Pennsylvania are extremely problematic. Because of this, as well as the pro Trump trend line and the pro Trump primary turnout numbers, I think that Trump will more likely than not win PA.

North Carolina

Real Clear Politics Polling Average: Trump +1.4

Five Thirty Eight Adjusted Polling Average: Clinton +0.6

2012 Election Outcome: Romney +2

2008 Election Outcome:  Obama + 0.3

Early Voting dem/rep 2012: 1.51

Final voting dem/rep 2012: 0.96

Early Voting dem/rep 2016: 1.31

Change in primary turnout (since 2008): -35.2% dem, +98.3% rep

Everything points to Trump winning NC.

Nevada

Real Clear Politics Polling Average: Trump +1.5

Five Thirty Eight Adjusted Polling Average: Clinton +1.2

2012 Election Outcome: Obama +6.7

2008 Election Outcome: Obama +12.5

Early Voting dem/rep 2012: 1.19

Final voting dem/rep 2012: 1.15

Early Voting dem/rep 2016: 1.13

Change in primary turnout (since 2008): -38.4% dem, +46.3% rep

Real Clear Politics doesn’t offer a trend line for Nevada. 538 and early voting both suggest Clinton will win. Primary turnout favors Trump but we haven’t seen much of a bunch in republican voting turnout relative to democrat early voting turnout. Given this, Clinton will probably win Nevada.

Colorado

Real Clear Politics Polling Average: Clinton +2.9

Five Thirty Eight Adjusted Polling Average: Clinton +3.8

2012 Election Outcome: Obama +5.4

2008 Election Outcome: Obama +9

Early Voting dem/rep 2012: 0.95

Final voting dem/rep 2012: 1.12

Early Voting dem/rep 2016: 0.99

Change in primary turnout (since 2008): -10.6% dem, N/A rep

Ohio

Real Clear Politics Polling Average: Trump +3.5

Real Clear Politics Voting Trend: Trump + 0.7 per day for 1 day

Five Thirty Eight Adjusted Polling Average: Trump +1.7

2012 Election Outcome: Obama +3.0

2008 Election Outcome: Obama +4.6

Change in primary turnout (since 2008): -47.9% dem, +78.8% rep

Trump will obviously win Ohio.

Arizona

Real Clear Politics Polling Average: Trump +4

Real clear trend: +1.15 for 4 days

Five Thirty Eight Adjusted Polling Average: Trump +3.0

2012 Election Outcome: Romney +9.1

2008 Election Outcome: McCain +8.5

Early Voting dem/rep 2012: .8196

Early Voting dem/rep 2016: .85

Change in primary turnout (since 2008): -9.1% dem, +2.3% rep

Everything points to Trump winning Arizona.

New Hampshire

Real Clear Politics Polling Average: Trump +1.167 or Clinton +0.6

Real clear trend: none

Five Thirty Eight Adjusted Polling Average: Clinton +3.3

2012 Election Outcome: Obama +5.6

2008 election outcome: Obama +9.6

Change in primary turnout (since 2008): -15.4% dem, +14.5% rep

Polling averages in New Hampshire are significantly impacted by a single poll which found Clinton winning with an 11 points margin. This poll is about 2.5 standard deviations from the mean. With the poll, RCP predicts Clinton willing by 0.6 points. If you remove this poll, Trump comes out winning by 1.2 points.

538 included this poll in their analysis. Given this, I think we have to take it with a big grain of salt. Primary turnout is in Trump’s favor, as are the polls once you remove the outlier poll. Given this, I think that NH will go to Trump.

Michigan

Real Clear Politics Polling Average: Clinton +4.7

Real clear trend: Clinton +.7 per day, 1 day

Five Thirty Eight Adjusted Polling Average: Clinton +4.2

2012 Election Outcome: Obama +9.5

2008 Election Outcome: Obama +16.4

Change in primary turnout (since 2008): +99.7% dem, +49.9% rep

Clinton will win Michigan.

Virginia

Real Clear Politics Polling Average: Clinton +5.0

Real clear trend: Clinton -0.2 for 1 day

Five Thirty Eight Adjusted Polling Average: Clinton +5.4

2012 Election Outcome: Obama +3.9

2008 Election Outcome: Obama +6.3

Change in primary turnout (since 2008): -27.5% dem, +91% rep

Clinton will win Virginia.

Iowa

Real Clear Politics Polling Average: Trump +3.0

Real clear trend: +.8 Trump for 2 days

Five Thirty Eight Adjusted Polling Average: Trump +2.7

2012 Election Outcome: Obama +5.8

2008 Election Outcome: Obama +9.5

Early Voting dem/rep 2012: 1.31

Final voting dem/rep 2012: 1.13

Early Voting dem/rep 2016: 1.23

Change in primary turnout (since 2008): -34.2% dem, +51% rep

Trump will win Iowa.

Maine 2 CD

Real Clear Politics Polling Average: Trump +0.5

Five Thirty Eight Adjusted Polling Average: Trump by less than 0.1

2012 Election Outcome: Obama +8.5

2008 Election Outcome: Obama +11.2

The data here is very weak, but it slightly suggests a Trump victory.

Maine

Real Clear Politics Polling Average: Clinton +4.5

Real clear trend: Clinton -1.1 in 1 day

Five Thirty Eight Adjusted Polling Average: Clinton +7.0

2012 Election Outcome: Obama +15.3

2008 Election Outcome: Obama +17.3

Change in primary turnout (since 2008): 0.0% dem, +232.5% rep

The trend isn’t good for Clinton but it is not strong enough to put Trump over the top in a single day. 538 obviously suggests Clinton will win. Primary turnout looks very good for Trump here, but I doubt it will be enough.

Conclusion

There is a lot of uncertainty here. I could very easily be wrong on several states, but this is what I think the most probable outcome is. Even though this is what I think will most likely happen, there are lots of other plausible outcomes, and a majority of plausible outcomes have Clinton winning. Things can go wrong for Clinton in a few states and she can still win by performing strongly in others. Trump has to do well in every important state in order to win.

I’ve tried to reason through this carefully, but it is still worth noting that I am a Trump supporter and this map is giving me the outcome I want it to. If my bias is impacting me, it will probably have the strongest effect on PA, which is a fairly subjective call.

Facebook Comments
  • Hellno Kitty

    Trump is winning big league!

  • fukkinrokkin

    White Pill City !!

  • Max Gibs

    wait, you have in your summary that you don’t think trump will win PA but your map shows it going to trump?

    • ” I think that Trump will more likely than not win PA.”

      By that I mean that the probability of Trump winning is higher than the probability of Trump losing. Sorry if the wording is confusing.

      • Max Gibs

        oops, thanks for the clarification

    • JasonVoothees

      I think you read “I think that Trump will more likely than not win PA” as “I think Trump will more than likely not win PA”. That’s what I saw when I first read through it.

  • JasonVoothees

    Do you think there are any policies Trump could have supported which would give him an edge in Maine without costing him other states?

  • User Name

    Does Ryan also agree with this prediction? (BTW please please please Un-private the prediction video,as i am very curios to see how it will all turn out,i hope there will be a follow up video to it)

    • Ryan Faulk

      Sean and I are basically in full agreement. Sean was more pessimistic about Florida, now we agree. He was more optimistic about Colorado and Nevada, now we agree. I was more pessimistic about New Hampshire, now I’m more positive about it, in line with Sean.

      If you’re in a swing state, vote early and vote often.

  • Goybeans

    Not a big deal but your map has Trump winning ME but losing ME2. He should be at 284.

    • Ryan Faulk

      Yeah Sean sorta threw this article together in a few minutes based on things we had been talking about for about a week.

      Ironically, when we do the research for a particular article, the articles tend to be better. But when it’s just a quick write up of things we’ve talked about for a while, it’s sloppier and assumes the reader knows things they don’t necessarily know.

  • Kris Polk

    NICE JOB GUYS

  • Goybeans

    tfw a small alt-right analysis website BTFOs mainstream analysts by being the only ones to correctly call the election.

    You guys called the monster vote so hard. Only thing is you didn’t call it hard enough in MI and WI. Looks like NH will be blue but it was so close it could have gone either way.

    Congrats. Did Ryan keep himself to his plan of not checking election results until the morning?

  • kraj

    Pretty good analysis. I genuinely didn’t expect MI and WI to come through like that. And NH was only barely lost.