Early today, I expressed reservations about a new Wilson Research Strategies poll showing Carly Fiorina leading Barbara Boxer by three points and Meg Whitman leading Jerry Brown by one.  The methodology seemed unusual, and without crosstabs to analyze, the poll looked more like an outlier.  After I posted that this morning, Wilson contacted me, providing me with the raw poll data (which they have also posted at their site), and asked me to take a second look.

Just to refresh everyone’s memory, these are the toplines of the survey:

Carly Fiorina currently leads by three points against Barbara Boxer in the race for United States Senate from California.

U.S. Senate Ballot
Total Fiorina 46%
Total Boxer 43%
Total Other Candidates 5%
Undecided 6%

Meg Whitman leads by one point in the race for Governor of California.

Governor’s Ballot
Total Whitman 45%
Total Brown 44%
Total Other Candidates 4%
Undecided 7%

In analyzing the turnout model used by Wilson, I compare it to the exit polling data from a media consortium in the 2008 presidential election in California.  The Wilson assumptions seem to mirror that exit polling pretty well:

• Partisan turnout: Wilson’s sample has a 12-point advantage for Democrats, 46/34/20, compared to 42/30/28 in 2008.  This actually looks as though it oversamples Democrats and Republicans at the expense of independents.  Since Fiorina and Whitman have an edge among independents, that may indicate that this poll understates the support for the two Republicans.  Bear in mind that this turnout in 2008 still resulted in a 24-point blowout for Obama, however.
• Also, bear in mind that the registration advantage for Democrats in California is 44.3%/30.9%, or 13.4 points.  The question will be whether the parties turn out close to their registration profile, the 2008 turnout profile from exit polling, or if Republicans surge in California.  If the GOP surges, then this poll will end up underestimating the strength of both Fiorina and Whitman.
• Gender: Wilson’s model has a 52/48 advantage for women; in 2008, the turnout was 54/46, and that might reduce Democratic support for both candidates in the poll — although Fiorina has a slight edge among women in this poll, 44/43.   Whitman trails Brown by two among women, 43/45.
• Race: Wilson has it 58% white, 17% Hispanic, 7% black, 9% Asian.  The 2008 exit polling showed 63% white, 18% Hispanic, 10% black, and 6% Asian.  This seems more like a wash.

The methodology isn’t quite as unique as I first assumed, either.  The Muhlenberg poll in Pennsylvania uses a similar technique, although their results seem rather difficult to accept in the Pennsylvania Senate race.  However, overall, it looks as though Wilson’s survey model closely approximates the California electorate and played it down the middle for the sake of this effort, and it could show that Republicans have started to gather some momentum in the two statewide races.  Both are still to close to call, but especially for Boxer, the low number seems to indicate an inability to best her challenger.

Update: The 24-point blowout was for Obama, of course, not McCain.