Monday, September 22, 2014

Classic from the textbook–Election 2014

There is a substantial body of work in the areas of behavioural economics and cognitive psychology which shows that the record of political pundits, political forecasters and the media is appalling in forecasting all but the most obvious of outcomes (a comprehensive review is to be found in Nate Silver’s The Signal and the Noise).

Kahneman shows that we tend to have a systematic bias which involves:

  1. Consistently underestimating extremes – National won and by a significantly greater margin than was forecast while the left suffered greater than expected decimation; and,
  2. that we are very poor at estimating ranges or spreads – much the same thing here too. The gaps between the factions were not well estimated as forecasters cramped things around the averages.

Did forecasters do better than a random pick would have? Yes – very likely they did. Even the Peters NZ First result has become systematically bizarre to the point where it is unlikely to be the result of a coin toss (the main guy who ran the coin toss – Hone Hariwera dipped severely).

Did the pundits and forecasters do any better than an amateur would have done? Very unlikely – and they got the extremes wrong – just as the amateur would have.

Moral? Don't pay forecasters much attention and certainly not money unless you are wanting to buy moral support – and even that is a dubious purchase – just ask the left.

No comments:

Post a Comment