Friday :: Sep 17, 2004

There's A Problem With The CBS News/NYT Poll As Well


by Steve

Iíve had several people send me notes tonight and in the comments section trying to undo my argument against the Gallup Poll sampling today by pointing to the just-released CBS News/New York Times poll out tonight, which shows that amongst registered voters Bush holds an 8-point lead over Kerry (50%-42%). The contrary argument from these folks apparently goes that if Gallup is wrong, then CBS News/NYT shows a similar large lead for Bush.

Not so fast. Check out the sampling used by CBS News/NYT in their poll, and again, then compare it to the recent party IDs in the last two presidential elections.

Total Respondents: 1287

Weighted GOP: 426 (34%)
Weighted Dem: 399 (31%)
Weighted Ind: 462 (36%) ?

Again, Iíll ask my question from this morning all over: If the Democratic Party ID from the last two presidential elections was 39%, and the GOP party ID was no higher than 35%, then:

-Why is CBS even using a sample that reflects a 3% GOP lead over Dems?

-Why is CBS using a sample that shows Indies higher than they have been in recent presidential elections?

This CBS poll shows Bush with a 50%-42% lead over Kerry, yet as Ruy Teixeira of Donkey Rising says far better than me, if this poll had been reweighted to reflect more closely the actual turnout of the 2000 election, Bushís 8-point lead evaporates into a 47%-46% tie.

This is not a complex issue. I donít pretend to be a polling expert, so I donít get into discussions about arcane things like when in the poll the pollster asks the respondent what their party ID is. To me, Iím still back to Square One: why trumpet a poll as being representative of registered voters when it clearly undersamples Democrats and in this case grossly oversamples independents?

As I said this morning, the bigger problem is that media outlets splash these polls across their front pages without questioning whether or not the samples used in these polls are reflective of a likely presidential turnout in two months. And the only way to deal with this is to write the paper in this case and ask how they can run a poll with a 3% Democratic deficit against the GOP and with Independents coming in at 36%.

Steve :: 11:07 PM :: Comments (18) :: Digg It!