Saturday, November 10, 2012

Which Polls Were the Best?

Nate Silver has a nice analysis of which polls did well and which polls did not so well in predicting the 2012 presidential election.  Well worth the read.  You'll see a more comprehensive article in Public Opinion Quarterly some time in the future, if they do as they normally do and do an election synopsis of poll accuracy.

Read his article.  Basically, live polls far outperform robo-dial polls.  No surprise there despite the pathetic efforts of some to defend such polling.  Of the bottom five polls in accuracy, three were robo-bullshit-polls (full list, any polls conducted).  The problem with the robo polls, other than being annoying as hell, is they cannot legally call cell (mobile) phones, which gives them among other things an age bias.  That makes them only slightly more accurate in predicting an election outcome as cutting open an animal and studying its innards.

Of the "big boy" polling firms, the ones who performed poorly (i.e., sucked) were Gallup, InsiderAdvantage, Mason-Dixon,  American Research Group, and Rasmussen (this of shops that conduced at least five polls).

Google's poll did well, leading Silver to suggest:
Perhaps it won’t be long before Google, not Gallup, is the most trusted name in polling.






No comments: