Ref's Performance Analysis for 2016

National Popular Vote Data Summary

I will analyze my final projections here. I kept a running tally of polls and a battleground index for the last few months. You can see an example of that here. This tally produced a battleground index which is meant to compare to the national polls to determine the level of relevancy of the national polls in determining a likely winner this election cycle. A clear pattern emerged where Clinton held a larger national poll lead than she did in the battleground states, a tell tale sign that they could not be trusted. It is this long pattern that led me to conclude that we would have a popular vote/electoral vote split, which we did.

The national poll projection can be found at the bottom of this page and permanently linked at That page was uploaded on Tuesday morning at 7:51AM, as indicated by the screenshot below from my server. I have since noticed that I neglected to update the date range of the polls, but I'm leaving it that way so I don't change the upload date on the server. I will not update this page on my server anymore so this upload date will be preserved as evidence of the time of my popular vote projection.


Below is text from the final projection on Tuesday morning at 7:51AM.

A combination of poll sample bias, shy trump voter bias, and late breaking voters (incumbency rule) will keep the popular vote within one percentage point. There is substantial evidence for poll sample bias and the shy trump voter bias. I am using 47% as the threshold for the incumbency rule because there is a strong third party contingent and Hillary is a de facto incumbent.

Evidence suggests that African-American and millennial turnout will be below the expected levels. In this national projection I assume this will drop Democratic turnout from +6 above the GOP we saw in 2012 to +5, a conservative assumption, resulting in a 1 point bump for Trump. Evidence also suggests a substantial portion, as much as 2.4%, of Trump's support is shy about it, not admitting it to pollsters. I make a conservative assumption, cutting that number by more than half, and give Trump a 1 point bump here.

So after adjusting, today's result would be Clinton 45.3 to Trump 44.3. Then we have the late breaking vote which almost always goes to the challenger (incumbency rule). To be conservative I assume it splits 52/48 for Trump. Assuming Johnson and Stein carry 5% nationally, as some polls have shown them trending up, that leaves 5.2% remaining.

The final popular vote projection is Clinton 47.9., Trump 47.1.

A split between the national popular vote and the electoral vote will occur because of a surge in the Hispanic vote nationwide, but centered mostly in just a few states like California, where it will not result in an increase in Clinton's electoral vote total. Trump will also win traditionally red states like Texas with a smaller than average margin for a Republican, but he will win them nonetheless, which results in a reduced popular vote total for him but not a reduced electoral vote total. We have seen evidence for a popular/electoral vote split throughout this election cycle. The state polls show a clear path to 270 electoral votes if one assumes that the reduced turnout among African-Americans and millennials, which we have evidence for in the early vote, in contrast to high enthusiasm among Trump-backing white working class voters in Pennsylvania and/or Michigan, will blow up conventional wisdom.

Clinton 45.2, Trump 42.3
2012 Peformance
Weight Applied
Ref's Avg 10/31-11/6 - SOURCE: 538/NYT - - 45.3 42.3
11/4-11/7 1026LV Rank 1, Avg Error 0.9 - 3.1 43 45
Economist/YouGov 11/4-11/7 3677LV Rank 13, Avg Error 2.6 - 1.5 (est) 45 41
Google Consumer Surveys 11/1--11/7 26574LV Rank 2, Avg Error 1.6 - 0.65 38.0 35.7
11/1-11/7 2972LV Rank 4, Avg. Error 1.8 - 4.5 46.8 43.6
11/1-11/7 2354LV Rank 8, Avg. Error 2.0 - 3.0 48.8 46.2
Bloomberg 11/4-11/6 799LV ***Rank 1, Avg. Error 0.7 - 3.5 46 43
Fox News 11/3-11/6 1295LV ***Rank 19, Avg. Error 3.1 - 3.0 48 44
Gravis 11/3-11/6 16639RV Rank 16, Avg. Error 2.7 - 1.0 47 43
Monmouth 11/3-11/6 748LV ***Rank 18 (source) - 3.6 50 44
Reuter/Ipsos 11/2-11/6 2196LV Rank 6, Avg Error 1.9 - 2.2 42 39
CBS News 11/2-11/6 1426LV Unavailable - 3.0 45 41
11/2-11/3, 11/6 1500LV Rank 20, Avg Error 4.2 - 2.5 45 43
NBCWSJ 11/3-11/5 1282LV ***Rank 10 , Avg Error 2.1 - 2.7 44 40
11/3-11/6 2220LV Rank 18, Avg. Error 2.8 - 2.5 47 43
McClatchy/Marist 11/1-11/3 940LV ***Rank 11, Avg Error 2.5 - 3.2 44 43
*** indicates that the pollster did fewer than five polls in the last three weeks of the election, so the rating is less certain. Also, some estimation was involved to determine the pollster's rank.
I try to keep all data within a one week range when close to an election as late deciders move fast. Polls that substantially diverge from the other national polls and the bellwether state polls will be weighted down. See an example. Only one poll per media organization is included. This average mixes head-to-head polls with four-way polls. Four-way polls are better because they match reality, but some head-to-head polls, like the LA Times/Daybreak (4th most accurate 2012) and UPI/CVoter (8th most accurate 2012) polls, offer reliable data. Not all head-to-head polls are useful, in fact most are not because there will be a large third-party vote this year. Often news organizations or particular shows, like the Kelly File, will report only those head-to-head numbers because they artificially inflate Hillary's lead. Perhaps because the LA Times/Daybreak and UPI/CVoter are Internet polls, this bias seems less pronounced rendering those polls worth inclusion.


My Twitter account is just getting started. Follow me here!

More Analysis from the Ref

The unprecedented overwhelming media bias has distorted the polls, hiding Trump's strength

Nate Silver and other "data journalists" have to face reality . . . it's not science

The reasons many polls are understating Trump's support

We are about to see Trump fire on all cylinders and likely win the election

Hillary Clinton is a very weak incumbent

WATCH MY DOCUMENTARY: Consent of the Governed

Imagine MSM reaction if Trump revealed top secret tactical nuclear information in front of 70 million people

James O'Keefe promises devastating releases next week. He discussed it Friday on

Rasmussen Reports: TRUMP HOLDS THE LEAD

Ref's Ruling: The NBC/WSJ's small sample and timing led to an absurd result

Why the post-debate polls for CNN and YouGov are wrong

Reuters/Ipsos Battlegrounds: Still a close race