2/ Now, near the end of my dissertation, the deviation between the two has become almost an ever-present fear. I'm still writing up the relevant discussion section, but here's how I would put it: Polls aren't polls; they're pushes.
-
Show this thread
-
3/ When we take polls and compute statistics on them, we think we're measuring what a person believes, albeit subject to sampling noise both in terms of who we picked *and what they happened to emit at the time* (think RAS).
1 reply 0 retweets 2 likesShow this thread -
4/ But, it's worse than that because it's generally hard to understand *how much* the context matters to them. You (the interviewer) pushed the context on the interviewee. They'll give you a response but it's not a natural expression. It's not a poll; it's a push.
1 reply 0 retweets 3 likesShow this thread -
5/ Again, this doesn't just introduce noise. A particular belief context integrates aspects of a particular person's information environment. This includes how useful that belief is as a social and asocial instrument.
1 reply 0 retweets 2 likesShow this thread -
6/ Socio/economic/demographic factors induce some regularities but the decomposition is hard to obtain. When you push a context, it's even harder. On the other hand, in a free-form interviews, proffered opinions on self-selected contexts contain more of what we want to measure.
1 reply 0 retweets 2 likesShow this thread -
7/ The problem is, again, interviews take skill and are way less convenient to perform. So, in a lot of ways, we just keep taking increasingly elaborate convenience samples. For want of a strong p-value, the kingdom was lost?
1 reply 0 retweets 7 likesShow this thread -
8/ P.S. This is still a hot-take.
1 reply 0 retweets 5 likesShow this thread -
Replying to @generativist
is this a hot-take? i wonder how many people doing public opinion research would disagree with your basic premise that surveys measure nudged-beliefs. zaller and feldman described similar concern like 25 years ago. i dunno, this seems reasonable, if not mainstream.
1 reply 0 retweets 1 like -
Replying to @ntdPhD
I alluded to them via "(RAS)" but I think there's a general "yea sure I agree okay but it's what we have so whatever show me the GSS/ANES/X/Y/Z survey data" position in practice.
1 reply 0 retweets 1 like -
Replying to @generativist
ah, skimmed and missed it. yeah, true. its default position to just use best-available and not spend the additional time probing around. artificial publics, artificial constraint. its why i dont get too jazzed about "is the public ideological?" if you're relying on survey data
2 replies 0 retweets 1 like
Yea, exactly. If you want to measure what someone will do in a situation — their behavior — opinions are a weak guide at best. If you want to measure a mixture of someone's tastes (a la Bourdieu) and social environment, go ahead —ask their opinion. But, beware your inferences.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.