What Went Wrong? Blue Overconfidence, Losing Jobs Voters, Rural Fears, Bad Polls
Political Scientists, Party Officials, and Pollsters will be studying the election results for quite a while. Here are my immodest first impressions.
How could Democrats take Wisconsin for granted? Republican Gov. Scott Walker is far right, has destroyed unions in the state, and survived a rare recall in the state. He was considered an attractive conservative Presidential candidate.
How could Democrats take Michigan for granted? Detroit is like an abandoned city. Yes, Obama did save the US auto industry. Michael Moore warned just a week ago that we were underestimating the suffering in the area, as he has been doing for years, and telling us of his fear of a Trump sweep as being real, as in BRexit. Trump has been harping on Ford taking its small car plant to Mexico. Ford has maintained that no jobs would be lost, since they were switching the plant to manufacture the more profitable large SUVs. Still, that wasn’t being added into Trump’s rallies. Even Trump has been warning us of a big BRexit.
How could the polls be so bad? There are usually two types of errors on any measurements. The first is the statistical error, or the width of a bell curve distributions of measurements. This is reduced by the larger number of people that you sample. For 1,000 people being sampled, the statistical error is about 3%.
The other set of errors are called systematic errors. They remain from known or unknown uncertainties. You try to measure them as well as possible, and reduce their errors equal to or below the statistical error. In the polls, as that of California, Clinton had 54%, Trump 30%, no answer 10%, Johnson 4%, and Stein at 3%.
There are many things the no answer can mean. People are wont to disclose their choice, since both candidates were highly unpopular, and they don’t want an expression of disapproval from the pollster. Some people still have not made up their minds. Some of those may have been ignoring the race, or they may be well informed and still debating in their own minds. Some may have been lifetime followers of their party, and very hesitant to cross to the opposite party, or skip the vote, or vote for an unknown third party candidate. Some of the people who sounded committed may change their mind. This could have occurred by a blockbuster October surprise, as the FBI finding Anthony Wiener’s computer, and unleashing a flood of innuendo, which Trump took full advantage of. Data show that rural America turned out 10% more than what was predicted. All of these are sources of systematic error that must be added to the statistical error. Only detailed studies will be able to assess the sizes of each of these systematic errors. Evaluating the systematic errors in addition to the standard questions in the polls can help estimate their size.
It seems clear that the systematic errors are on the order of 5% or larger, in general. A more careful analysis by the pollsters will give much more detail on this.
