Overthinking It

The Washington Post is trying to figure out what went wrong with the British election polling.

There are theories about sample size and cell phones and response rates.

They glide over the obvious answer, which is that parliamentary elections can be tricky to gauge.  In a US Presidential or Senatorial race, you have a binary choice.  If you're voting in a Scottish constituency, you can vote Labour, Lib Dem, SNP, Conservative, Green, UKIP and probably a few other parties.  That allows for volatility in your choice.  Maybe you're a Conservative, but you know that a Tory can't win in your constituency.  So, you vote Lib Dem or UKIP.  Or maybe you're a Lib Dem, but you're fed up with the party over the austerity coalition, but can't quite make yourself vote for Labour.  Suddenly the SNP might look appealing.

Polling in Britain was slightly off, when it looked at the overall percentage of the vote each party would get nationally, but it wasn't disastrously off.

Polls are most useful for penetrating BS punditry.  Here is a piece ranking GOP presidential candidates.  Bush, Rubio, Walker, Cruz, Paul and Huckabee are all right there 1-6, and that roughly corresponds to the polling data.

But then Cillizza includes Kasich, Christie, Jindal and Perry to round out his top 10.  He notably leaves off Ben Carson and Rick Santorum, despite the fact that Carson in particular is more viable according to the poll numbers.

But Christie and Kasich are Creatures of the Beltway.  Republicans don't really like them and are very unlikely to vote for them.  And probably they won't vote for Carson either, because a lot of what he says is crazy.  But the GOP voting pool if full of crazy.