New Senate Polls in Florida, Georgia, and Alabama

New Google Surveys polls show a trio of Republican incumbents in the Southeast well-positioned to be re-elected in November.

In Alabama, Richard Shelby leads Democrat Ron Crumpton 71–26. In Georgia, Johnny Isakson leads Democrat Jim Barksdale 58–36. In Florida, Marco Rubio leads Democrat Patrick Murphy 57–38.


Nobody expected this race to be competitive, and it is indeed not competitive. Richard Shelby has a massive 45 point lead in his race to win a sixth term in the US Senate, 71–26 (N=474). This is slightly ahead of the roughly 65% he has won in his past few election cycles.

For more information, view the raw data at Google Surveys.


While Georgia was talked about as a possible Democratic pickup earlier in the cycle, that seems unlikely at this point. Johnny Isakson leads by 22 points over Democrat Jim Barksdale (N=439).

Isakson is the clear favorite to win a third term this November. He has been endorsed by one of the state’s Democratic congressmen, while Barksdale appears to be waging a weak campaign.

For more information, view the raw data at Google Surveys.


Despite his failed presidential run, Marco Rubio appears to be successfully navigating his Senate campaign. He leads Patrick Murphy by 19 points in this poll (N=500).

While this was originally seen as a top opportunity for a Democratic pickup, Democrats have been significantly less optimistic recently. The DSCC has triaged the race and canceled its remaining ad buys, and Marco Rubio has led every public poll of the race in the past month.

I am personally skeptical that Rubio might win by 19 points (the highest lead he has on any 538 poll this month is 8 points), but this poll does suggest that Rubio is in a very strong position in the race.

For more information, view the raw data at Google Surveys.

A Note on House Effects

Google Surveys has shown house effects on a regional level, The GCS tracking poll has consistently given stronger results for Democrats in the Midwest and Republicans in the Southeast. A recent 538 article discusses per-state internet polls further.

We’ve generally found that 15–20% of respondents to these polls do not vote in the state which geo-targeting claims that they are in. While we filter them out of these results, it does create a non-response bias; those 15% of respondents will not be reached by polls of the state they do live or vote in.

Also, the distribution of computer use among the public may not be equal across various demographics. This is similar to the problems faced by “land-line only” polls. We expect this bias will be regional rather than on a per-state level; and in general neighboring states will have similar biases.


All three polls were in the field from Tuesday, October 18 to Thursday, October 20, using Google Surveys. In the past week, Google has rebranded “Google Consumer Surveys” as simply “Google Surveys”. Please read this methodology statement for more information.