Confirmation Bias for Fun and Profit
Quick note. I am not the only one who thinks poll aggregators are destroying their own raw materials:
One likely reason why there are fewer state polls in 2016 than in 2012 and 2008 is that aggregators get most of credit and media attention. https://t.co/pvaUdpESIW— Walter Shapiro (@MrWalterShapiro) October 30, 2016
posted by James | 11:17 AM
4 Comments:
I always thought this was a great point of yours.
How to "fix" the problem is also pretty interesting and tricky. For example, a natural solution would be for the aggregators to take the state polls in-house, so that the polls capture the surplus that's currently free-rider'd by the aggregators.
So but what if 538 (e.g.) takes over a bunch of state polls, and then other aggregators are able to free-ride on 538's state polls? Etc.
Making the polls private and exclusive to the aggregators would "fix" this free-riding issue, but that would result in losses -- maybe enough to make most state polls unsustainable. (There are part-way solutions here that could maybe work: release the polls to the public, but several days after they are used internally, etc.)
I guess one thing that should be said is that maybe the social value of accurate political polling is... not that high? Like it would save a lot of money on psychotherapy for Democrats to know that Clinton is really winning (Inshallah), but on the other hand wouldn't it just make Republicans more angry and upset? (And vice versa.) What is the net benefit that accurate polling gets us? Slightly better for business firms to be able to predict the future, but that seems marginal in the case of presidential polling.
You can give examples like, "If people had believed the Brexit polling they would have tried harder or even voted differently," but again it seems as though that's an example that is limited to its facts (you can imagine a case where you get a better result because people don't know the outcome ahead of time).
Interesting question. Accurate polling is quite recent, so it may be possible to derive some natural experiments from our own history. But it's tricky.
The clear lesson of history is that less accurate polling is linked to racism, sexism, homophobia, and continent wide war in Europe. Thank god for modern polling.
Post a Comment
<< Home