Can any of you statisticians out there explain in words a lawyer can understand how the Pew Charitable Trust poll shows a dead heat between JFK and GWB while Gallup shows GWB's lead in double digits?
In rank order: 1. Bias (known or unknown) 2. Sampling Error 3. Ethical Misconduct These are the classic 3 taught in any marketing research class (I may use this example in tonight's lecture!!)
One poll listed the alternatives as JFK vs. George W. Bush, while the other poll listed them as John F. Kerry vs. GWB? (Hey, even I'd vote for JFK, but is he available?) More seriously, I suppose that one possible explanation might be the margin of error in the polls. If in reality Bush is leading Kerry by a few percentage points, the polls might show them as tied if the errors for each candidate converge, while they may seem to be farther apart than they should be if the errors diverge. (But error that can make a ten point difference seems like an awful lot.) Disclaimer: I have never studied statistics, so what do I know?
Another factor is that the accepted opinion among media groups, is that Americans tend to tune out in blow out elections. If Americans feel that Kerry is out of the race, then they stop watching the various news programs. Less viewers means lower ratings, lower ratings mean lower revenue. This is another reason in addition to the 3 reasons stated by Han. Our media loves disasters, controversy, conspiracies and anything else that boosts ratings. Bush is pulling away, and unless Kerry creates a coherent message, he will be trounced. Bush has run a marginal campaign at best, but kerry is in the running for the worst campaign since Dukakkis or Bill SImon in California.
This site seems to be pretty accurate. This one will look at the electoral college according to latest polls. Interestingly enough, Bush was above 300 in the electoral college yesterday, with Kerry down for the count. Today, Kerry seems to have come within 10 electoral votes. Florida and PA are both extremely important in this race, and have gone back and forth quite a bit in the polls.
I saw a program a few nights ago on this very topic (MSNBC?). Appantly some polls are based on the overall national popular vote and some are based on a per state basis (electors). It was also pointed out that responses depend on the format of the poll: some polls ask up front questions like "Are we winning the war on terrism" or "Are we outsourcing too many jobs abroad" then ask, in a subsequent question, who is your choice for president. Depending on the early questions the choice for President varied. This points out that one needs to read the survey (any survey) in some depth to see the details.
