Internet Electoral Data Demigod
Patrick Ruffini has just posted his
2008 GOP straw poll. He provides lots of interesting ways to break down the poll data (by state, region, even by referring blog!) that make data nuts like me want to
hug him and squeeze him and call him George. Go vote, and then take a look at the overwhelming popularity of one "fantasy" candidate.
An Internet poll is only different from a traditional party straw poll in being much larger and more geographically dispersed. Traditional straw polls have exactly the same selection bias characteristics as this one does.
ReplyDeleteBefore you can say whether a poll is "useful" you have to know what its ostensible use is. If its purpose is to divine a sensus fidelum of candidate preference among a set of people normally thought of as "the party base" or "highly motivated voters" you could do a lot, lot worse than an Internet straw poll. In fact, a standard stratified random phone poll would be much inferior.
Kathy is right. Ruffini is a party operative in touch with people who are very active in the party. By taking a straw poll, he can find out who the most energized party activists support. Now, will that give you the eventual nominee? Not necessarily. It might give you the next Howard Dean.
ReplyDeleteI voted. Sam Brownback for round one. Condi Rice as the fantasy candidate.
Tlaloc, there is no point in blathering about parametric statistics when the purpose of the poll is not to produce parametric estimates. Ruffini runs his polls for the purpose of finding out what Internet-aware party activists and highly motivated and interested conservative bloggers and blogreaders are thinking. You're still stuck in Gallupland, assuming that all polls are designed to estimate total population fractions from a stratified random sample. This poll gets the sample it wants by using the very characteristics you claim make it invalid.
ReplyDeleteMultiple votes might be a problem, but its my sense that most people do not overvote. Ruffini does track votes by referring site, so would be aware of a coordinated pollbomb effort.
I think Kathy's kind of right on this one. The selection bias, in this case, is really what Ruffini is going for: it's how he gets his convenience sample. So, in terms of selection, there's no problem for him there. He doesn't want random people's votes, he wants an idea of what a GOP primary might look like if it was held today.
ReplyDeleteJust admit it, T, you just want to disagree with anything stated in a post on this forum at any time. Kathy and I have both had graduate stat. methods courses, so we're well aware of all the stuff you are complaining about. Ruffini's poll is clearly going to yield better info than a magic eight ball, but not as good as a gallup poll. Get over yourself.
ReplyDeleteI didn't admit you were right all along. Your position was always "no value." Our position was always "some value."
ReplyDeleteI didn't admit you were right all along. Your position was always "no value." Our position was always "some value."
ReplyDeleteAnd my position is more extreme than Hunter's: Ruffini's poll does not just have "some value"; in this case, given the target population, the timing, and the subject, it has more value than a traditional Gallup-style poll. It is more than three years until the next presidential election and about 2.5 until the first primaries. There is zero value in polling average Americans about their presidental candidate choices right now. They are not engaged, and fewer than one in 20 could even minimally identify all the names on Ruffini's poll. The preferences that are important right now are the preferences of the people who are going to put effort and money into propelling candidates into the primary cycle. The same people that Ruffini attracts.
ReplyDeleteThat's a good point. Plus, I haven't seen any drives to pack the Ruffini ballot box. Nor has any one else I know. I think people are honoring the straw poll.
ReplyDelete