I don’t often cite PowerLine, but here’s a statistic that wants a little discussion.
It wasn’t easy for the Times/Siena poll locate its sample of 800 likely voters. The poll made 51,983 calls in search of those 800 likely voters.
This is what The New York Times said about the sample—absent the hype:
Can Democrats turn Texas blue? We made 51,983 calls, and 800 people spoke to us.
Hyped or blandly reported, this is a statistic without context, and so it’s meaningless.
Is the list called actually a randomly drawn sample? That’s the only way the 800 can be considered random and so representative. The NYT, well down in their description, admits their sample is not random.
People who respond to surveys are almost always too old, too white, too educated and too politically engaged to accurately represent everyone.
Pollsters compensate by giving more weight to respondents from under-represented groups.
But weighting works only if you weight by the right categories and you know what the composition of the electorate will be.
There’s a hint in that last, too.
Is getting 800 respondents out of 52,000 tries typical? What’s the trend over the last, say, dozen or so elections, both Presidential and mid-terms considered separately as well as together? NYT chose to be silent on that.
Is telephoning really a good way to get a random sample? Sure, things have changed since the Harry Truman polling debacle, but…. We’re out and about a lot more today, too, and it’s easier to just ignore incoming calls from numbers we don’t recognize.
Aside: what’s not being talked about in the NLMSM (I wonder why…) are these two statistics from the same Times/Siena poll, carefully ghosted so as to be made less noticeable: Beto O’Rourke’s Approval/Disapproval rating
41% favorable rating; 44% unfavorable; 15% don’t know
compared with Ted Cruz’ Approval/Disapproval rating:
51% favorable rating; 42% unfavorable; 7% don’t know