Skip to content

What March polls really show

Those who cannot remember history are doomed to be confused.

The polls numbers aren't as bad as they look. Quite. stability.ai

Table of Contents

The latest poll results being reported are concerning, but not quite so much when put in context.

When you look at the results of polls conducted in the seven swing states in March, taking into account the results of the 2020 election as a starting point, two adjustments are necessary.

Under the 2020 Census reallocation of electoral votes, North Carolina, which Trump won, gains an additional vote so that it now has 16, rather than 15. Pennsylvania drops a vote, to 19 and Michigan drops a vote to 15.

Apples and apple seeds. The 2020 election results may never be accepted by some, but they are hard numbers about the actual votes of millions of voters. The polls conducted in March are soft numbers about the hypothetical votes of hundreds of voters in each swing state. The two sets of numbers don't deserve equal weight for that reason, even though the polls are fresher.

A statistical approach called Bayesian analysis allows a model that adjusts previous results with new information. Applied to the polling:

  • Movement away from Trump isn't apparent in North Carolina.
  • Movement away from Biden is apparent in Georgia and Arizona, both of which are definite toss-ups with outcomes more uncertain than even taking into account simple random sampling margins of error suggest.
  • Movement away from Biden is not apparent in Pennsylvania, Michigan, Wisconsin or Nevada.

Implication: Biden needs to hold Pennsylvania, Michigan, Wisconsin and Nevada if he cannot hold Georgia and Arizona and cannot flip North Carolina.

More detail can be found at SwingWatch, which will soon be updated with information from polls conducted in April. SwingWatch is a totally commerce-free source of election news and data analysis.

Latest