This weekend Survey USA released a new poll on the San Diego Mayor’s race. This new poll showed some changes, but not an overall shifting of the top three candidates. However, rather than focus on the findings, we at PDI have been looking at the demographics of those polled in the survey. Unlike polls that use PDI data and call voters with known partisanship and ethnicity, this survey is performed based on random digit dialing and counts on the respondent self-identification to help formulate the baselines.
A blog post regarding the last survey resulted in a call from the Union Tribune and an article looking into the Survey USA methodology. This provided a snapshot of the overall race based on the partisan breakdown of the respondents, then a second based on the partisan breakdown of PDI’s 13SDP3 universe.
However, the greater concern could be the ethnic breakdown used in the survey. In the first poll, the Latino share of the survey was 25%, while the white portion was 58%. This Latino figure is twice the historic share of voters in low-turnout elections. The new survey places White voters at 59% of the survey with Latinos at 20%.
Looking at historic turnout for a set of high and low turnout elections (using the 2004 – 2012 Presidential General and Primaries) shows the actual registration rates for Latinos and Whites along with their historic performance. As this shows, White registration has fallen in San Diego over the past 10 years, as Latino registration has grown. And, consistently white turnout surpasses their rates of registration, while Latinos consistently under-perform. 
The survey USA model of 20% Latino turnout and 59% white turnout, while much improved for this latest poll, still produces significant outliers. While the data from Presidential races isn’t perfectly translatable to a Mayoral Special Election, it does provide historic trends that cannot be easily dismissed. In fact, historically statewide Latino turnout drops even more in special elections – providing some evidence that Latinos will provide less than the 12.8% of the
electorate shown in the 13SDP3 universe.
How would this difference in the ethnic breakdown of each poll impact the results of the candidate horse race? The real impact would likely be small, maybe a couple percentage points either way. But, with a multi-million dollar campaign and a fight from all campaigns to make the runoff, these one or two percentage points could be very important.