Debunking the national polls that COMPLETELY CONTRADICT state polling (AP, ABC, USA Today)

If you include enough Democrats in your sample, Hillary will always win, and why 2012 is not the new floor for Democratic turnout

We've got some bad national polls this time around

First, it is true that the party ID question in polling is not an actual report of the real party distribution in a voting group. It represents what people currently want to associate with, not how they are registered. The reason it's a good way to analyze the credibility of a poll, however, is that exit polls use the same method to determine party distribution. I am not comparing apples to oranges by looking at the party distribution in phone/Internet polls and comparing them to exit polls.

The exit polls represent the only data of party distribution we ever receive. The exit polls are not actual hard voter data; they are just polls like the ones we are doing here except that they derive from people who actually turn out and vote, but they are, nonetheless, just polls. We want the party ID in a phone/Internet poll to match the exit polls because that means it will be more accurate.

In truth, Democrats do outnumber Republicans in this country and they will probably outnumber them in the election. Of course the real question is by how much? The AP poll assumed a +12 Democratic voter turnout. If this happened, no Republican would win any national election. So let us dispense with that absurd poll.

The ABC partisan assumption of a +9 Democrat vote is also too high. Again, if Democrats outnumber Republicans by almost ten percent, that leaves the GOP with precious little hope of winning any national election, and that's just not the makeup of our nation right now. With enthusiasm for each candidate at least being a draw and Democrats having less of a tendency to turn out in general, that +9 Democrat advantage will likely not materialize.

These kind of misleading polls really strike at their credibility as pollsters. They seem to want the big Hillary lead, probably to throw off the poll averages and suppress Republican voter turnout. I write this only because there are good methodological reasons to use different voter turnout models. It's just bad poll modeling. I believe they are actually sacrificing credibility when everyone is looking to game the system, and yes, I know this is a bold thing to assert. They are transparently embracing bad polling methods and reach an incredible result, so you tell me why.

But the real question, will it work? I have my doubts. I think this kind of thing actually can hurt Democrats because Republicans and Republican leaners get their news from alternative media. Democrats are more likely to hear it because they listen to mainstream media sources. Hearing a poll like this could actually give a voter who wants Hillary to win, but may not actually want to go vote, an excuse to go to the movies rather than vote.

Hillary Clinton: Ignore the polls "because we’ve got to turn people out”

So now for the USA Today/Suffolk distribution of 38% Democrat, 32% Republican and 24% Independent. That number looks familiar . . . oh right, it matches the 2012 partisan distribution exactly. I can see that assumption possibly making sense. If Democrats were filling arenas and minorities and young people were going to turn out in historic numbers like they have never turned out before, then I could buy it. That's what happened in 2012. Pollsters missed that in 2012 and they are making up for that mistake by assuming it almost across the board now. That is except for the pollsters that didn't miss it, namely the IBD/TIPP and LA Times polls, both of which got the distribution right in 2012. They are showing a tie race now, so perhaps we ought to take a lesson, but don't tell that to Larry Sabato who just loves the assumption that the 2012 turnout model will repeat itself.

Fox News also assumed this turnout model, assuming that Democrats would have a +7 advantage. Hillary's lead was only three percent in the Fox poll, however, so we can assume the margins of error are kicking in. The best practice here would be to average the two results with a similar turnout model, so +3 Clinton and +9 Clinton turns into +6 Clinton. That makes more sense to me. Let's assume for the sake of this article that Fox and USA Today both came to a +6 for Clinton, because essentially they should have but the margins of error threw them in different directions.

To arrive at a +6 for Clinton you assume a +6 advantage in turnout for Clinton. Funny how that works. The problem is that assuming a +6 advantage for Democrats requires you to assume that Hillary can duplicate the mega turnout that Obama brought out in 2012. On the contrary, I think we should assume that that Democrat turnout will be down around its minimum because people are bummed out by this election and many are tuning it out. That's a real bad cocktail for Democrats to swallow because their voters are prone to staying home unless they are super enthusiastic.

This assumption ignores Trump's strategy of driving down Hilary's turnout. He is intentionally firebombing her to get the soft voters to quit going to Facebook for a while to avoid the politics and instead start some Netflix marathons. There is evidence this is working.

Left-wing media is dangerously undermining democracy in 2016

Reuters just pointed out that young voters are doing just that, tuning out the election because of its negative tone.

Looking at Party ID in the sample is enough to prove a bad poll

There are far more ways than bad party ID assumptions to reach a desired result in any poll, but particularly a national one. Pollsters report the regions they call to, but not the particular types of locations, rural verse urban. If they make more calls to urban areas, the Independents and Republicans will more likely lean Democrat.

Does this happen? I suspect that it does. Much was made in conservative media of an email from the 2008 Clinton campaign that appeared to request oversampling for the purpose of maximizing impact on the "media polls." While Politifact labeled this just a request for an oversample of minority groups for the purpose of getting a clear picture of what is going on within a demographic group that is inadequately represented in the normal sample, it could have been more than that. Politifact didn't account for the impacting media polls part of the email. But that's beside the point really. One need not prove this kind of poll rigging to undermine a poll's credibility.

It's quite tempting to speculate that pollsters might yield to their partisan leanings from time to time and call more urban areas in their national polling to find more liberal Republicans and Independents. Of course that can and likely has happened. But we don't have evidence for that, so we can't really conclude that it happened here and we don't need to.

It's enough for Trump to know that pollsters, including the Fox poll that showed a +3 Clinton lead, are probably assuming too big of an advantage for Hillary when they assume 2012 turnout will repeat. Romney did not attempt to drive down turnout among Democrats like Trump is doing. Hillary has enough problems on her own generating enthusiasm. When Trump slams her like it's his job every day when her negatives are already astronomical, it stands to reason that he is probably succeeding in driving her turnout down.

At the same time, Trump world is populated by "centipedes" who possess resilient, in fact invincible, enthusiasm for the Donald. While centipedes are a feature of the_donald subreddit, the same high energy level of support present there extends to all reaches of Trump world. Enthusiasm is in fact infectious. While it may not yield the same level of enthusiasm in a passionate Trump supporter's wife, children or friend, it will probably result in the not so passionate potential Trump voter showing up at the polls because their Trump supporter dad, husband or friend has asked them about it eleven times.

So this dynamic I just described may end up in a Democratic turnout advantage of 2% or 3%. How does Donald make up the rest? This is where the new and unaccounted for voter makes the difference. The biggest reason the IBD/TIPP, LA Times and Rasmussen polls are showing Trump tied while the more traditional polls are showing a Hillary lead is because the traditional polls make it harder to reach likely voter status. The IBD/TIPP poll and LA Times/Daybreak poll essentially just take your word for it that you are a likely voter. They have successfully predicted the last two elections better than any other pollster, so to argue with this assumption, Chris Stirewalt, Larry Sabato, Nate Silver, you have to discount real election results.

If in fact the more traditional pollsters are reaching Trump's new voter base but excluding them because they did not vote in previous elections, a typical screening question in the older parts of the industry, their polls would be biased toward Clinton. This voter turnout assumption is used widely in most of the polls, which is why you can't trust the poll averages this time around. If you average a bunch of garbage stats, you are going to get a garbage average.

In summary, the old polling is showing Clinton up six when they assume a 2012 redux voter turnout model. In truth, I think that number should be at most Dem +2 or +3 assuming that Trump is not drawing new voters, which he almost certainly is. The final result will come down to how much Trump can drive down Hillary's turnout and bring out his army of brand new voters. I think the odds of this finishing within three points nationally are a virtual certainty. Who wins, nobody knows yet and that's the truth.

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