Thursday :: Sep 30, 2004

Gallup's Latest State Polls Are Better, But Still Have GOP Bias


by Steve

Gallup released three battleground state polls late yesterday, and two of the three showed a close race amongst likely voters. After obtaining the breakdowns by party ID of all three of the likely voter samples used by Gallup in yesterdayís polls, it's clear that Gallupís latest likely voter samples include more Democrats than before. But in all three states, Gallup continues their GOP bias and advantage over Democratic respondents by anywhere from 1% (in Ohio) to 5% (in Florida). Again, if the election were held this week, how likely do you feel that such GOP advantages amongst those voting would materialize?

Ohio
Polled 9/25-9/28

Registered Voters
Kerry: 49%
Bush: 46%
Nader: 2%


664 Likely Voters
Bush: 49%
Kerry: 47%
Nader: 1%

Likely Voter Sample Composition
252 Republican (38%)
244 Democrat (37%)
168 Independent (26%)

For reference, the 2000 Ohio exit poll showed a 38% Democrat, 37% GOP, and 26% Independent split. According to Gallupís poll yesterday. Kerry has a 51%-39% margin over Bush with independents in Ohio. If he does, and independents make up 26% of the sample, then why is Kerry trailing by 2%? Notice under Gallupís likely voter model, which samples more GOP respondents that Democratic ones, Bush gains 3% while Kerry loses 2% in a state with greatly increased Democratic registration.

Pennsylvania
Polled 9/25-9/28

Registered Voters
Kerry: 49%
Bush: 45%
Nader: 3%

654 Likely Voters
Bush: 49%
Kerry: 47%
Nader: 1%

Likely Voter Sample Composition
282 Republican (44%)
268 Democrat (41%)
100 Independent (16%)

For reference, the 2000 exit poll in Pennsylvania showed a 40% GOP, 40% Democrat, 20% Independent split. According to Gallupís poll yesterday, Kerry has a 43% to 41% advantage over Bush with independents in Pennsylvania. But is a 16% composition of independents a valid composition of likely voters if the election were held this week? Also, note that under Gallupís likely voter model Bush gains 4% while Kerry drops 2% in another state which has seen increased Democratic registration.

Florida
Polled 9/24-9/27

Registered Voters
Bush: 49%
Kerry: 44%
Nader 2%

704 Likely Voters
Bush: 52%
Kerry: 43%
Nader: 1%

Likely Voter Sample Composition
294 Republican (42%)
256 Democrat (37%)
142 Independent (21%)

For reference, the 2000 exit poll for Florida showed a 40% Democrat, 38% GOP, 22% Independent split. So the Gallup model understates Democratic turnout from four years ago by 3% and overstates GOP turnout by 4%. Note that under Gallupís likely voter model, Bush gains 3% and Kerry loses 1% in another state with increased Democratic registration.


Gallup says they aren't predicting what will happen in November, but rather is reporting with a high degree of confidence what will happen amongst likely voters if the election were held this week. It is heartening to see that Gallup is using likely voter models in at least Ohio and Pennsylvania that more closely resemble what happened in 2000, notwithstanding concerns we would have about using likely voter results that will ignore recent registration gains and GOTV efforts. But Gallupís Florida likely voter model still has a questionable GOP bias not supported by what happened four years ago.

Steve :: 1:52 PM :: Comments (3) :: Digg It!