Canaries in the Coal Mine

Although we still do not know the winner of the electoral college, it is clear that the polls were systematically wrong in the swing states and several others.  Here is a recap, with links, of what we knew in advance but few dared discuss openly.

 

Election forecasting became an economics-free zone

 

I wrote about this last week:

 

(1) Better information-aggregation methods were showing less Biden lead and sometimes a Trump lead.

 

(2) The incentives for individuals to truthfully participate in polls were skewed relative to a neutral situation, and relative to 2016.  Incentives matter!

 

(3) The costs to individuals of participating in the election are different during a pandemic.  During this pandemic, perceptions of the new costs were correlated with party affiliation.

 

The forecasters were dismissing these economic insights.  That they used foul language and straw men rather than real argument and measurement was a telltale sign to somebody (me) who is an expert on economics but not on election forecasting.  There were exceptions such as Trafalgar – they cited some of the effects above and made better forecasts – but of course they were maligned for being different from the crowd.

 

 

Forecast errors were correlated over time, questioning their rationality

 

The June 2016 Brexit vote was an early notable case where educated people in big cities had no clue what their fellow citizens were thinking and made little effort to learn about it.  The British polls had been closed for hours before currency markets absorbed what happened.

 

The Obama Administration and most other Democrats were very slow to understand what happened around Brexit.   I tell the story in my book of “a conversation between a Laotian woman and Ben Rhodes (Obama’s speechwriter). Citing the British populist victory, the woman expects that candidate Trump will win. Mr. Rhodes confidently assures her that Mrs. Clinton will be the new president.”  People living on the other side of the planet knew better what was happening in the 2016 election than the “30 people setting the direction for the entire U.S. government.”

 

Back to British populism, polls indicated that there was no way that their leader Boris Johnson could form a majority coalition in the December 2019 election.  In reality, Mr. Johnson was getting most of the votes from areas of Britain that had long been solidly with his opponents.  No electoral victory in Britain since Thatcher had been as overwhelming.

 

Still in 2020 the conventional wisdom was that the populists would not surprise us again.  Any surprises would be purely random.

 

 

Companies with enough data to know acted like small things mattered

 

Twitter and Facebook have products that have overwhelmingly passed the market test, and they are free of charge!  Except that we pay with our personal data.  They know a lot about us as individuals and, especially, as a group.

 

In the weeks leading up to the election, these two companies tried to suppress the viewpoints of those on the side of our populist president, ranging from a Stanford professor to one of the biggest newspapers.  Why would Twitter particularly drive millions of consumers to its competitor?  Likely the company thought that the election was close enough that even small news stories might matter.

 

 

The Rallies

 

Trump was holding up to five massive rallies per day.  Biden rarely came out of the house and then to tiny crowds.  This difference was a signal from both sides of the market.  On the voter side, this was a more stark difference than in 2016.  On the campaign side, given that with 20-years-ago technology the Bush campaign knew that Florida was close and to have their candidate there at the end, we can assume that the Trump campaign knew that it was close in the places where Trump visited at the end (AZ, FL, MI, MN, NC, PA, WI).