The following article by Ruy Teixeira, author of The Optimistic Leftist and other works of political analysis, is cross-posted from his Facebook page:
I’m not sure how many people have even absorbed the fact that there is now a competitor to the National Election Pool-sponsored exit poll, the AP VoteCast survey–used by Fox and AP affiliates–which has a very different methodology than the standard exits and purports to replace it.
Is it any better? Probably not. Both polls perform reasonably well in helping predict the outcome of races on election night, though some misses are inevitable. The real problem here is that people rely on the exit polls to give them a portrait of the demographic and attitudinal voting patterns of the electorate and therefore the underlying dynamics of the election. But the exit polls have been shown time and again to be inaccurate on this score and it is unlikely that the AP VoteCast survey is much better.
So which data source should you believe on voting patterns? Neither of them. The preferred approach is to look at surveys and carefully modeled results incorporating actual election returns that are released after the fact, like the Catalist results and the Cooperative Congressional Election Survey. From these sources, you can piece together a fairly accurate picture of the composition of the electorate and demographic variation in voting patterns. Harder but better.
This article by Steven Shepard in Politico goes over the current controversies between the exit polls and VoteCast, though as I indicate above, he misses the main point.