so, please, people. heed my analysis.
polling aggregates do not make any sense.
you want snapshots. the newest, best poll. you take that one new poll, and you throw the rest away.
the reason is that people swing dramatically out of the blue, on sporadically defined events. a poll on tuesday consequently completely negates and discards all polling done on monday. all an aggregate can ever do is pollute your snapshot polling.
so, why do these people do this? we had an election in canada in october, and i brought this point up repeatedly. i kept outperforming the models. and, they kept scratching their heads, right. so, my push back is not out of the blue. i've been pushing this point for months.
this is what they're doing: they're measuring advertising. they're interpreting the situation through some kind of vulgar marxist spin over manufacturing consent. the idea is that voters are only able to react to what campaigns throw at them. so, democracy is reduced to the voter reaction to marketing campaigns.
and, it does follow that an aggregate makes sense when you're measuring brand recognition, because it's a lot more random - and it's based on persistent exposure.
so, what a polling aggregate does is not just a problem in terms of the way it approaches the data. it's a problem in the way it approaches the voter. it really reduces the voter to a vessel to be filled with information - a robot that can be programmed and has no mind of it's own.
but, don't jump to conclusions. he's still down quite a bit in all of the polling i've seen in other states.
--
don't throw away the polls, either.
i read the polls right, didn't i? so, what's the problem here? the polls or the analysts?