So, having been disastrously wrong in my analysis of the dueling California polls, I owe SurveyUSA an apology and congratulations: they alone constructed a likely-voter screen that predicted the amazing, historically unprecedented surge in Latino turnout, which accounts for Hillary's margin of victory. The Field poll, which predicted Obama winning very narrowly and which I'd assumed would be close to correct, actually got more or less the right breakdown within each racial group (the white vote was split equally between the two).
What they (and I) missed, by a mile, was the racial composition of the electorate. They predicted 20 percent Latino/12 percent black. It was actually 29 percent Latino/6 percent black (a decline in relative black turnout since 2004, and a near doubling in relative Latino turnout), which could mark a major turning point in California Democratic politics. What this suggests is that Latinos aren't going for Hillary by default or thanks to name recognition: there appears to be a genuine and positive enthusiasm for Hillary in the Latino community. (She also ran up surprisingly big margins among Asians, 75-23.) He did slightly better among Latinos elsewhere--losing only 55-41 in Arizona, for example--but this is going to be a major hurdle for him in Texas on March 4. Judis was right.
Incidentally, SurveyUSA had a great night all around. I, for one, welcome our new robot overlords.
--Josh Patashnik