Nov 9, 2020 - Economy & Business

P.S. Economic polls are bad, too

Illustration of binoculars with broken lenses
Illustration: Aïda Amer/Axios

This week America witnessed a forecasting failure of almost unprecedented magnitude. October's unemployment rate came in at 6.9%, after dozens of the best-paid and most experienced economic forecasters in the world had predicted the number would come in at 7.7%.

Why it matters: Those forecasters had literally millions of data points of information to go on, and have had ample experience with unemployment releases, which come out like clockwork on the first Friday of every month. But there will be no great post-mortem about why they got the number so wrong.

  • Forecasts, by their nature, are often wrong, even with reliable inputs and well-calibrated models. Wall Street understands this and has long learned to live with it.
  • In the case of presidential election polling, the inputs are much less reliable and the models can only be roughly recalibrated once every four years. That makes electoral forecasts extremely unreliable — an uncomfortable fact that almost no one in the media has really grappled with.

The bottom line: Americans will always latch onto whatever polls they can find, desperate to find a narrative during the long months of campaigning. But we'd be better off ignoring them entirely.

Go deeper