Thursday, November 17, 2016

The Pollcast: Were the U.S. polls right, wrong - or somewhere in-between?


The results of the U.S. presidential vote came as a surprise to many. Almost every national poll gave Hillary Clinton a lead over Donald Trump and the polls also suggested she held the lead in enough states to win the election.

Instead, though Trump lost the popular vote, he won the electoral college and will be America's next president.

As mail-in votes continue to be counted in Democratic-friendly states, Clinton's edge over Trump in the popular vote will likely grow — and the national error in the polls will shrink. Still, at the state-level the error was significant enough to up-end expectations.

What did the polls miss? Was there a problem with how polls — and the uncertainty intrinsic to the science — were interpreted? And where does the media and the polling industry go from here?

Joining me on this week's episode of The Pollcast is pollster David Coletto, CEO of Abacus Data.

You can listen to the podcast heresubscribe to future episodes here, and listen to past episodes here.

7 comments:

  1. What the polls ignored was the Electoral College and it's Republican lean

    ReplyDelete
    Replies
    1. The electoral college certainly leans or over-represents small rural or agricultural states. Since, each state has a minimum of three electoral votes it gives Republicans a lead of somewhere between 20-60 electoral votes from the start. It's about states not national polls and at the end of the day too much emphasis was placed on Clinton's national lead and ignored the situation on the ground. The last group of state polls had Trump surging with leads in unlikely places. Looking back conducting national polls for this type of election appears to be almost a complete waste of time.

      Delete
    2. To put it another way: When Democrats win it is because they win the 25 largest state. When Republicans win it is because they win more states.

      Delete
    3. The polls actually captured this alright. It's just pundits who ignored it that got it wrong.

      Delete
    4. The pundits were absolutely terrible. I re-watched the election night coverage from ABC where the pundits were still planning a Hilary victory even after it became apparent Pennsylvania was going Trump. Early in the night a field reporter from Ohio mentioned Black voting was down 16% in the state and there was no mention what a bad omen this was for Hilary's prospects in Pennsylvania and North Carolina. Much like our prime minister it appears American political pundits are chosen for their style not substance.

      I have never been one to believe the mainstream media is biased, at least not on purpose. While I still think most media coverage is unbiased. The US election demonstrates that it is probable an inherent or systemic leftish bias does inhabit the fourth estate at least down South. My conclusion is that if so institutions like the C.B.C. become more important as they can be publicly held accountable for their professional or unprofessional (perhaps biased) coverage. I trust however, that in an organisation such as the C.B.C. the philosophy of showcasing both sides of a story is still of the utmost priority.

      Delete
  2. Hi Eric,
    I was wondering how big an effect of sampling bias there might have been on the polls. Anecdotally, it sounds like many people who didn't vote previously cast their vote this year for Trump. Do pollsters take into effect this group (i.e., didn't vote in previous elections and turned out) and how big an effect might this have on the predictions?

    ReplyDelete
    Replies
    1. This is probably one source of error, as some pollsters judge likelihood of voting on whether someone voted in the past.

      A recent article on 538, though, posits that the miss is largely attributable to not contacting enough white non-college graduates. A lot of these people might have been these lapsed voters.

      Delete

COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.