Uncategorized

The “unskewed polls” guy is back to faking polls

Dean ChambersDean Chambers bought himself his five minutes of fame in late 2012 by fundamentally misunderstanding how public polling works. He insisted that all mainstream media polls were biased towards Democrats either by intentionally including more Democrats than Republicans, or by a process called weighting, where you take a number and adjust it up or down based on known factors, and created a website to “unskew” all these biased polls.

Pollsters do weight results by age, sex, race, ethnicity, and some other factors because we know with great accuracy what the American population looks like from the decennial census, and they must do this because polling samples are random. If your poll of 1000 people (random phone numbers) is made up of 51% females, then the results without modification will underrepresent women, which make up 55% of our population. Weighting the results to artificially increase (or decrease) the representation of known groups is not only scientifically legitimate, it’s necessary to make the poll results accurate.

What Chambers didn’t understand is that pollsters don’t weight by party affiliation/ID. (Actually, Rasmussen, one of the most pro-Republican biased pollsters in existence, does weight by party affiliation, which is probably why they are so inaccurate.) Party affiliation is considered an opinion, and there’s no hard scientific data (like the census) on party ID. There really can’t be, because pollsters have observed that a person’s party identification can change many times over their life and turn on a dime, so to speak. Many life-long Republicans started calling themselves Independents (while not changing their voting behavior at all) with the emergence of the Tea Party movement in 2009, for example.

Because of that fundamental error, Chambers tried to “unskew” what he thought were skewed polls by re-weighting them to artificially increase Republican representation to what he personally thought it should be. Initially, he re-weighted accurate mainstream polls to reflect the already skewed Rasmussen model (in effect, skewing them twice). Then he began taking extremely inaccurate polls on his website and began skewing mainstream polls based on those results.

Rather than correcting skewed mainstream polls to be more accurate, what Chambers ended up doing was taking mostly accurate mainstream polls and skewing them until they were no longer accurate, but “looked good”.

The results were ugly and as inaccurate as you’d expect. Chambers “unskewed” polls showing Barack Obama leading in most swing states that he went on to win rather easily so that they showed Mitt Romney leading instead, and based on that, he predicted Mitt Romney would win the electoral college 275-263, when Obama went on to win 332-206. Chambers missed the result by by 114 electoral votes.

Not to toot my own horn, but my projection relied almost entirely on the supposedly “skewed” mainstream polls and I only missed the result by 29 votes and 1 state (Florida).

Chambers realized what most Republicans calling for a Mitt Romney landslide did after the election, that Democratic turnout in 2008 was not a fluke. He wrote this mea culpa on November 10th:

Most of the polls I “unskewed” were based on samples that generally included about five or six or seven percent more Democrats than Republicans, and I doubted and questioned the results of those polls, and then “unskewed” them based on my belief that a nearly equal percentage of Democrats and Republicans would turn out in the actual election this year. I was wrong on that assumption and those who predicted a turnout model of five or six percent in favor of Democrats were right. Likewise, the polling numbers they produced going on that assumption turned out to be right and my “unskewed” numbers were off the mark.

What Chambers still didn’t understand is that polls don’t use “turnout models”. If there were five to six percent more Democrats in virtually all polls, that means one of two things:

1. There are five to six percent more Democrats in America.

2. All pollsters were using the same flawed methodology (which is not the same thing as being skewed or being biased.)

Polls like Gallup’s that do track party ID (but don’t use it for weighting) have shown that #1 is the case for many years. The reason that Democrats don’t always win Presidential elections vary, but if Democrats ever managed full efficacy (100% turnout) in every Presidential election, the executive branch would be a permanent Democratic institution. That’s just the math of it, regardless of whether you think that’d be a good thing.

It was heartening to hear people who had sworn up and down that non-partisan analysis like mine was skewed/delusional/blah admit that they were wrong after the election. In fact, it was their own analysis that was skewed. (Hardly surprising when you go back and look at the data.) Rush Limbaugh admitted the day after the election at the top of his show that mainstream polls were right, there are more Democrats in America (although he didn’t use the word Democrat, he said “them”, because that’s more scary than “Americans who legitimately disagree with me”.)

Chambers did the same thing, and said that he was getting out of the poll unskewing business:

That leaves some asking where do I go from where and what about UnSkewedPolls.com. I remain an independent journalist and columnist and will continue to publish articles and commentaries for the readers who read my articles on Examiner.com and my web sites. UnSkewedPolls.com is just one web site and one project I did, and depending on your point of view it was proven wrong, it has run its course or it will fade away.

The “non-apology apology” version of “I was wrong” should have been a warning sign that Dean Chambers was being disingenuous. It doesn’t depend on anyone’s point of view, the facts are clear: mainstream media polls were skewed slightly in favor of Republicans, not Democrats, but were generally accurate and correctly predicted the outcome of the election. Chambers didn’t understand what he was doing and was proven beyond a shadow of a doubt to have been hysterically wrong about nearly everything he did and said on that site.

So I guess it shouldn’t surprise anyone Chambers has pulled a complete 360 and is right back “unskewing” accurate and reliable polls because he doesn’t personally like the results. The same arrogance and ignorance that led him to a humiliating error with the outcome of the 2012 election has led him to begin intentionally skewing opinion polls of Barack Obama’s job approval rating.

I don’t subscribe to terms like “Bush/Obama Derangement Syndrome” but not hard to understand why they exist. If believing all polls were skewed and biased towards Democrats, leading you to predict a Mitt Romney victory, only to have them turn out to be quite accurate when they showed an Obama win isn’t enough to trigger introspection, what the fuck is?

It looked for a while that Chambers and a large chunk of the political right understood what they had done wrong in 2012 that led them to make faith-based predictions that blew up in their face, while data-based projections were anything from generally correct to unimpeachable. Perhaps their mistake was honest, and their reflection sincerely.

Turns out… not so much.

This stupidity had infected the Romney campaign right down to its core. Mitt Romney’s internal “unskewed” polls looked a lot like those that Chamber had faked, and Romney himself believed what they were doing was right and smart. The Republican Party needs to understand what it did wrong to avoid repeating that mistake in 2016, because it’s a part of what cost Romney the 2012 election. Romney wasted time campaigning in Pennsylvania, which Obama won by over three points, while Florida remained in the blue column by less than a single point.

Florida flipping wouldn’t have changed the election winner, but it would have made the election much more winnable, from 332-206 to 303-235. And Romney had a much better chance of winning Florida than he ever had at winning Pennsylvania.

If Chamber’s sudden reversal is also playing out in the rest of the professional right, the Republican Party has virtually no chance to win either the 2016 or 2020 presidential elections, which are going to be tough enough as it is without trying to campaign based on faulty information. Only Chris Christie poses a credible challenge to Hillary Clinton (if she runs), but why would he want to run when a big chunk of his base turned on him after Hurricane Sandy?

It looks like we’re in for another four years — perhaps an eventual eternity — of Republicans denying polls, which means another four years of Republicans campaigning on flawed data, which is significantly worse than campaigning blindly on no data at all. It looks like Republicans may be lost in the woods for longer than we thought.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.