For a change, SurveyUSA conducted a CA-52 congressional poll using Registration-Based Sampling (RBS) not Random digit dialing (RDD.) RBS is just what it sounds like. The pollster makes criteria on a voter registration list and then polls those people. RDD is when a pollster dials phone numbers at random and then asks if the person is a registered voter and if they intend to vote. If they say, they're included. Due to the plethora of voter information in California, most pollsters use RBS and prequalify respondents.
Carl DeMaio only leads by 1% in the poll, but it's a very curious poll. He leads 53%-42% with the 24% of voters who say they've already voted, but trails by 2% among those who haven't. The already voted percentage is believable based on the much heavier Republican VBM rate that I've mentioned in previous posts.
It's the remainder that I have issue with. The sample is made up of 35% Democrats/33% Republicans. In 2012 it was 37% Republicans/33% Democrats and this year the VBMs are coming out more Republican than they did in 2012. That's not surprising since a mid-term is expected to be more Republican. This poll is less Republican. They appear to have oversampled young voters and minorities and arrived at a fairly Democratic sample. Based on that, I believe DeMaio is up by more points.