Saturday, November 30, 2024

Polling Heuristrics

There is no fool proof methodology for looking at Polls but in the US elections I find that the following heuristics, that I employed successfully for Election 2024, has utility.

  1. Use poll averages such as the Real Clear Politics (RCP) Average as opposed to a single poll to make a judgement call. Add 2% to the RCP average for GOP candidates as the polls routinely under sample Republicans. Polls also appear to over sample urban voters.
  2. Ignore polls taken a few days before an election. Many of these are push polls with questionable metrics.
  3. Presidential approval and disapproval polls are as a rule better indicators than polls that ask one to choose a specific candidate.
  4. Economic metrics are the single best predictor of election results. However don’t fall for the trap of the official metrics. The question here that most MATTERS is… What does the average person feel/think about the economy?
  5. Polling questions about the direction that a country is moving in (good or bad?) are also great indicators of election results.
  6. Don’t fall for talking heads who tell you that the election is done (one way or another) after a single event. In times they routinely walk these back.
  7. Polls taken after a debate generally don’t hold up in the long run. These are particularly driven by media spin. The same is true of polls taken after party conventions.
  8. Talking to likely voters as opposed to all voter samples will provide better results.

No comments: