Showing posts from November, 2016

The Quantum Effect of Polls

If you ask 10,000 people who they are going to vote for, and 6000 say Blue, and 4000 say Red, then within a very small margin of error, you can predict that out of a million voters, around 600,000 will vote Blue, and 400,000 will vote Red. Donald Trump's unexpected victory is the fourth time we have recently seen the pollsters and political commentators get it absolutely wrong. The Brexit shock, the British National Elections Tory trouncing of Labour, and  Netanyahu's victory in the last Israeli election, were similarly against every poll prediction and the pontifications of every credible political commentator. What is apparently at stake is not just the credibility of polls, or of political commentators, but the very basis of statistics. If political polls are worthless, then perhaps their mathematical basis is also bunk? Or, as Churchill would have it - "there's lies, damned lies, and statistics". However, these same statistical theories used by th