SurveyUSA is out with a new poll for KGTV and UT San Diego. This poll has Democratic congressman Scott Peters up 1% over Republican Carl DeMaio in the CA-52 congressional race. This is a contrast to their June poll, which had DeMaio up 7, and the June primary. The GOP won there by 15%.
I don't know what the current spread is, but SurveyUSA has once again shown they don't understand how to survey California. They have a policy. If someone says they're likely to vote, they consider the person a likely voter and include their answers. In California we know how many of the last five elections someone has voted in and use this data to predict whether the person will turn out. SurveyUSA has again overestimated the electorate. Let's set the way back machine to last February. SUSA put out a survey right before the San Diego mayoral race in February that predicted an amazing 91.3% turnout that had Republican Kevin Faulconer winning by 1%. Turnout was 43.5% and Faulconer won by 6%. Turnout was half what they predicted, although it was about what most observers expected. SurveyUSA caught a lot of non-voters and that's likely one of the reasons they were off by 5%.
We'll go further back to predict the turn-out this year. In a similar district, turn-out was 64.8% of registered voters in 2010. Turn-out in this district in 2012 was 74.5%. When SUSA did their June poll they determined that 73.9% of the registered voters they talked to were likely to vote. That was too high, but not awful. In this survey they have 82.8% of registered voters voting. So they've gone in the other direction and added more non-voters. That makes their results suspect.