> Tom, I suspect that it has to do with the likely voter definition. Many
> college students did not vote 4 years ago, so they are not counted as
> likely voters.
I wouldn't say "counted as"; it's a weighting equation and college students are certainly a politically active population, so I wouldn't be surprised if their poll responses are given a certain consideration.
College students (to refer to an earlier post) also tend to have household, rather than personal, incomes higher than 20k - otherwise known as the bank of mom and dad. A significant part are independent or self-financing, but most, I believe, are still dependents and so a demographic analysis won't necessarily pick them up as having low incomes.
> Question: Don't the polls seem to underestimate how well the right will
> do?
IIRC, the unweighted polls usually favour the Democrats substantially, by a wide margin, and so they underestiate how well the right will do. IOW, you're entirely correct.
Also IIRC, the (weighted) polls tend to underestimate how well the left will do, but by a small margin, and it depends on how big the turnout is. Small turnouts favour republicans, or so goes the conventional wisdom.
The point of weighting (ie, the "likely voter" rather than the "registered voter") is to correct the estimator bias - that is, correct the raw sample to account for people that favour one candidate and answer accordingly but won't actually go out and vote. They also control for fluctuations in the sample itself (too many men, women, minorities, Democrats, Republicans ...) just to keep things in hand and the error (and cost) low.
Maybe I should quit my job and go back to study polling :).
Marco
,--------------------------------------------------------------------------.
> Marco Anglesio | Chance favours <
> mpa at the-wire.com | the prepared mind <
> http://www.the-wire.com/~mpa | --Louis Pasteur <
`--------------------------------------------------------------------------'