Princeton neuroscience professor Sam Wang has determined that Hillary has a 99%+ likelihood of becoming our next president. Princeton Election Consortium Economist and professional sports statistician Nate Silver just stated: fivethirtyeight.com/...
All of this data is nevertheless consistent with Clinton being an Electoral College favorite. She has a 64 percent chance of winning the Electoral College in our polls-only model and 65 percent in polls-plus, putting her somewhere in the range of being a 2-1 favorite.
Both Sam Wang and Nate Silver aggregate a state’s polls then apply statistical methods to calculate a probability of winning that state. These methods should produce similar results but they don’t. That discrepancy indicates there is probably something wrong with one of the methodologies. Careful examination of polling and early voting data finds 5 reasons that Sam Wang is probably right and that Nate Silver’s methodology has 2 major flaws.
1. Professor Sam Wang is more scientific and has greater past forecasting skill than Nate Silver
Sam Wang, PhD, has demonstrated expertise in statistics and science that Nate Silver lacks. Quite simply Sam is both smarter and better trained than Nate. And Sam has a scientific reputation that is at stake if he makes an unscientific mistake in his methodology. Nate has a monetary interest in getting as many clicks as possible. This point would be an example of an appeal to authority and confirmation bias on my part but for one thing. Sam has a stellar track record in presidential election years. Nate has a great record as well but made an election day adjustment to bolster his success rate. Thus Sam’s success rate is better than Nate’s.
2. Nate is “unskewing” the polls
A number of experts have pointed out an obvious flaw in Nate’s methodology. The disagreement between Nate and the experts was written up by one of Huffington Post’s most credible reporters, Ryan Grim.
The short version is that Silver is changing the results of polls to fit where he thinks the polls truly are, rather than simply entering the poll numbers into his model and crunching them.
Silver calls this unskewing a “trend line adjustment.” He compares a poll to previous polls conducted by the same polling firm, makes a series of assumptions, runs a regression analysis, and gets a new poll number. That’s the number he sticks in his model ― not the original number. …
Guess who benefits from the unskewing?
By the time he’s done adjusting the “trend line,” Clinton has lost 0.2 points and Trump has gained 1.7 points. An adjustment of below 2 points may not seem like much, but it’s enough to throw off his entire forecast, taking a comfortable 4.6 point Clinton lead and making it look like a nail-biter.
Nate Silver responded via Twitter that Grim's piece was “idiotic and irresponsible” Then he made a claim that opens up his methodology to scientific dissection.
The reason we adjust polls for the national trend is because **that's what works best emperically**. It's not a subjective assumption.
If we examine the ABC news/Washington Post poll above we see almost no long-term change in the national poll numbers. There is very little, if any long-term trend since the nominations were settled. What we see is a lot of short-term variability that directly responds to news events. But, because Comey’s unprecedented meddling in the race happened ten days from election day, the final news blip pushes the trends in Nate’s methodology towards Trump.
I will make a confession. I suck at statistics. (Note that Sam Wang is brilliant at statistics.) However, I am a PhD geochemist who has developed and managed nuclear waste safety research. I have studied the possible effects of climate change on nuclear waste transport in great detail. I have looked at mountains of data, figures and analyses. And it is obvious that Nate has made an elementary error. His methodology overweights the effects of Comey’s announcement because it came at the end of the polling cycle. This is scientifically similar to the sophomoric cherry picking done by climate deniers when they used the hot El Niño year of 1997 as a starting point in their analyses to claim that global warming had stopped.
3. Nate allows new Republican polling operations like “Remington” to bias his data set.
The top hit for Remington on Google says:
remingtonresearchgroup.com/Remington Research Group voted polling automated telephone surveys. ... Axiom Strategies is the largest Republican political consulting firm in the country.
Nate made “house bias” corrections, based on past performance for well-established polling firms but Nate has no way to correct for “house bias” for new polling operations. Remington, a Republican political operation out of Kansas, just polled a number of swing states for the first time. Remington's numbers appeared to have a large Republican bias compared to other polls taken about the same time, but the sample size is too small to determine a correction factor for house bias. Nate should have not allowed Remington’s apparently biased polls into the overall sample because they may have created a false trend towards Trump just before the election. It appears that Remington may have gamed Nate’s methodology to help Trump. Remington certainly had the motivation to do so.
4. Nate has discounted early voting results
Nate claims his methodology is empirical, but he ignores actual voting results that show his trend adjustment is wrong. Record smashing democratic early voter turn out in the Latiño district of Las Vegas has created an insurmountable lead for Hillary Clinton in Nevada, yet Nate has stuck with his 2 point trend adjustment favoring Trump to Nevada’s polls. Thus Nate predicts that Trump is likely to win Nevada despite the unprecedented turn out of Hispanics who have been repeatedly vilified by Trump. Moreover, Nate simply ignores the analyses of John Ralston, Las Vegas’ most respected local political reporter. This is political and scientific malpractice by Nate.
5. Nate Ignores Hillary’s GOTV machine & Trump’s lack of GOTV
In recent presidential elections Democrats and Republicans have both had good GOTV. Thus they cancelled each other out. This year Hillary has a state-of-the art GOTV machine while the Republican party is relying on Trump’s free publicity machine and the anger of Trump’s white male base. However, Trump has mobilized women and Hispanics to vote Democratic like never before by his hateful speeches and vile behavior. Voters screened out by “likely voter screens” may turn out like never before to stop Trump. So, the combination of Hillary’s much superior GOTV plus evidence from early voting that “unlikely voters” are turning out in record numbers to stop Trump indicates that the late surge is towards Hillary, not Trump. The ABC/Washington Post poll above may be hinting at a last minute surge towards Hillary, but it certainly isn’t showing movement towards Trump.
Conclusion:
Nate Silver’s methodology has 2 major flaws this year. He claims his methods are “empirical” but ignores evidence that shows his methods are flawed. Sam Wang’s methods are more scientific than Nate Silver’s.
