Short term update: Polling results should never be such a sure thing; and sloppiness ensues the poor job polling forecasters have done through the primaries. Nate Silver engages in more eccentric probability "calculations", again resulting in untamed and misleading election night thoughts, such as: "According to our final polls-plus forecast, Hillary Clinton has a greater than 99% chance of winning the Michigan primary." May also like this newer article about Trump's wealth versus S&P.
This is a quick look at the current polling of the 2016 general election match-up between Donald Trump and Hillary Clinton. Of course anything can happen between now and Election Day (a state holiday in many parts of the U.S.) to shift these polls further. And it is possible that either Ted Cruz or Bernie Sanders, or both, may instead win their respective primaries. There are a handful of regular national canvassers who we looked at here: NBC/Wall Street Journal, MSNBC, USA Today, CNN, Fox News, Rasmussen, Quinnipiac, and PPP. Due to the high resource constraints, statistical polling doesn't take place every day, but in recent months we have reasonable data and can make out some fundamental current patterns!
We see that Clinton typically has a 5 percentage point advantage over Trump. Generally speaking the results are 45% preferring Clinton, while 40% prefer Trump. That's not bad for the latter given that until recently he was only a business celebrity, while the former has been scheming a White House rule for decades. But how do we take such seemingly humdrum information to instead ascertain what are the odds for either candidate? Looking at the data a bit more closely, it is easy to notice that even as Trump gains in the Republican primaries, there has been a very recent (past several weeks) scaling back of his favoritism versus Clinton, in the general election polls.
Meanwhile Americans are increasingly uncertain who they would vote for regardless. The undecideds' have drawn from earlier leanings towards both Clinton and Trump, and acutely in the most recent weeks that the portion of voters who are "undecided" has been exceptionally volatile. Still, with the critical standard error about the general election spread shown (in blue), we would expect that there is still a 25%-30% chance that Trump would be elected POTUS. Despite last summer Huffington Post making a sham of one's presumption of unbiased news-"reporting" by subjugating Trump's "side show" coverage to their ludicrous Entertainment section, the probability of his winning is certainly not zero. And in fact it climbs to only a 1 in 3 chance even if most of the surveyed undecideds (in red) eventually vote for him. Despite serious stakes, the outcome is similar to rolling a die with 2 faces painted red (for conservative) and 4 faces pained blue (for liberal): despite the bias, both colors will eventually have a chance to be landed on!
Unrelated updates: there have been many recent exciting news. For example, an entertaining share by the excellent Vinod Khosla (billionaire co-founder of Sun Microsystems) of our viral Strategists: full of bull article.
This article also continues to have shares including by The Irish Times, Andrew Busch (long-time contributor with Maria Bartiromo on both Fox news, and CNBC's well-recognized Closing Bell show), and George Papadoplouos (Wall Street Journal contributor, alongside high-profile experts sampled below).
Our An escalator of optimism article shared by RealClearMarkets, Lance Roberts of Houston's KSEV 700 AM radio, Chart of the Day in Abnormal Returns, and Zero Hedge. And our 2-month slide v. August nose dive article shared by Bloomberg and Washington Post columnist Ritholtz.
Our Flipping volatility regimes was both a cover story and a top news on Zero Hedge.
Our Kevin Bacon's 4.5 degrees of separation article shared by another Wall Street Journal contributor: distinguished technology executive Henry Elkus, and later our conclusions were replicated by Microsoft (who incidentally is guilty of having their own kooky age-"prediction" game, with statistics just as dismal).
Our "And the winner goes to ..." article liked on social media by William (Bill) Watts, the markets editor for Wall Street Journal's MarketWatch, and most viewed and shared for a week on KDnuggets (and by the University of Nebraska department of quantitative and data sciences).
Last, will be at a reception and named on plaque at HQ as a charter member of a new science and philanthropic society for the American Statistical Association. This donation came from all of our Statistics Topics book sales (here, here, here). Also the book will be promoted for $1, this week only (tell your friends!)
This is a quick look at the current polling of the 2016 general election match-up between Donald Trump and Hillary Clinton. Of course anything can happen between now and Election Day (a state holiday in many parts of the U.S.) to shift these polls further. And it is possible that either Ted Cruz or Bernie Sanders, or both, may instead win their respective primaries. There are a handful of regular national canvassers who we looked at here: NBC/Wall Street Journal, MSNBC, USA Today, CNN, Fox News, Rasmussen, Quinnipiac, and PPP. Due to the high resource constraints, statistical polling doesn't take place every day, but in recent months we have reasonable data and can make out some fundamental current patterns!
We see that Clinton typically has a 5 percentage point advantage over Trump. Generally speaking the results are 45% preferring Clinton, while 40% prefer Trump. That's not bad for the latter given that until recently he was only a business celebrity, while the former has been scheming a White House rule for decades. But how do we take such seemingly humdrum information to instead ascertain what are the odds for either candidate? Looking at the data a bit more closely, it is easy to notice that even as Trump gains in the Republican primaries, there has been a very recent (past several weeks) scaling back of his favoritism versus Clinton, in the general election polls.
Meanwhile Americans are increasingly uncertain who they would vote for regardless. The undecideds' have drawn from earlier leanings towards both Clinton and Trump, and acutely in the most recent weeks that the portion of voters who are "undecided" has been exceptionally volatile. Still, with the critical standard error about the general election spread shown (in blue), we would expect that there is still a 25%-30% chance that Trump would be elected POTUS. Despite last summer Huffington Post making a sham of one's presumption of unbiased news-"reporting" by subjugating Trump's "side show" coverage to their ludicrous Entertainment section, the probability of his winning is certainly not zero. And in fact it climbs to only a 1 in 3 chance even if most of the surveyed undecideds (in red) eventually vote for him. Despite serious stakes, the outcome is similar to rolling a die with 2 faces painted red (for conservative) and 4 faces pained blue (for liberal): despite the bias, both colors will eventually have a chance to be landed on!
Unrelated updates: there have been many recent exciting news. For example, an entertaining share by the excellent Vinod Khosla (billionaire co-founder of Sun Microsystems) of our viral Strategists: full of bull article.
In an engaging article concerning BlackRock. Our An escalator of optimism article shared by RealClearMarkets, Lance Roberts of Houston's KSEV 700 AM radio, Chart of the Day in Abnormal Returns, and Zero Hedge. And our 2-month slide v. August nose dive article shared by Bloomberg and Washington Post columnist Ritholtz.
Our Flipping volatility regimes was both a cover story and a top news on Zero Hedge.
Our Kevin Bacon's 4.5 degrees of separation article shared by another Wall Street Journal contributor: distinguished technology executive Henry Elkus, and later our conclusions were replicated by Microsoft (who incidentally is guilty of having their own kooky age-"prediction" game, with statistics just as dismal).
Our "And the winner goes to ..." article liked on social media by William (Bill) Watts, the markets editor for Wall Street Journal's MarketWatch, and most viewed and shared for a week on KDnuggets (and by the University of Nebraska department of quantitative and data sciences).
Interviewed by the terrific Robin Powell of U.K.'s Ember Television and Regis Media: "I’ve been speaking to statistician, author and blogger Salil Mehta who has studied Warren Buffett's returns in detail." Key takeaway: "It’s nearly impossible to beat the market by a statistically significant (or risk-adjusted) amount over the long run. Warren Buffett used to do it, but even he stopped doing that a while back." Article also had many shares and likes including by great writers at both Money (US) and La Tribune (France).
Last, will be at a reception and named on plaque at HQ as a charter member of a new science and philanthropic society for the American Statistical Association. This donation came from all of our Statistics Topics book sales (here, here, here). Also the book will be promoted for $1, this week only (tell your friends!)
Free email subscription enjoyed by thousands (note you must complete process by accept the immediate email confirmation):
Article fails for not considering the fact that this current election cycle is culmination of about a decade of dramatic social and political changes causing political polling to decline in the reliability of results due to old perceptions and methods not suitable for this new world.
ReplyDeletePolitically major change is "flipping" of political axis from traditional Left-Right to Progressive-Libertarian one. Thus population-wide re-calibration we still do not understand and hence, do not poll for. (Ex. H.Clinton and Jeb Bush attract the same group of voters in this new world versus B.Sanders and D.Trump group). Political candidates being "flippant" and inert to old Left-Right system, try to be attractive to their "old" constituents too. Hence, for now political axis are actually "swirling" around. Laying ground for very, very hard polling, great errors in it ...
Socially major change is in the Internet social media/social portals/networks,... and general cultural change toward polarization and instability. This also causes swirling of the political axis mentioned above, just by another cause. Also, quick impact of this media. Results taken day apart may show dramatic differences due to it.
We still do not know how to deal with either of these two major problems in polling/predicting political outcomes. We do not understand fundamentals underlying them enough to convert that understanding to some meaningful numbers. Finally, obsolete polling methods and new people mindset about them make collected data much more error prone than generaL, old practice implies.
We do not understand what we are currently measuring (be it that many are misguided thinking that they know based on the past that does not apply now), we do not understand what and how to measure nor how to interpret that proper measurement. It might take couple of Presidential election cycles 'till polling starts to be Scientific again. For this election, all current polling results are so error ridden on all levels that they do not predict anything.
Thanks Dusan, I disagree that statistics are not helpful, only because you might see it as "this time is different". Please see other related articles here on the elections, and Trump. http://statisticalideas.blogspot.com/2016/05/the-flip-flopping-pollster.html http://statisticalideas.blogspot.com/2016/03/mercurial-pollster-results.html http://statisticalideas.blogspot.com/2016/04/trump-v-s-salilstatistics.html
DeleteNo doubt your blog have a nice and great value in public good work and keep posting.
ReplyDeletePresidential Fishing Debate
I search many blog about the US Political News but did not get many specific detail about the this topic but your blog have allot of information about my topic which I always searching.
ReplyDeletePolitical News Articles