Public Policy Polling nailed the 2008 results, but PPP’s polls have sometimes seemed a little strange so far in 2012. First, they leaned heavily toward Obama. Then they came back down to earth, but showed Obama struggling among non-white voters (an unlikely scenario). So what’s going on? It might be education.
Since their June 14-to-17 survey, PPP has released educational breakdowns for their DailyKos/SEIU national tracking poll. In every survey, PPP's respondents have been better educated than the 2008 electorate. PPP doesn’t weight for education, and less-educated voters are less likely to respond to pollsters, so this might not be especially surprising. But, it's possible that response rates among less-educated voters are even more lower for automated pollsters than live interviews.
Oversampling postgraduates and undersampling high school graduates could make a difference. Democrats excel among voters with a postgraduate degree—Obama won them by 18 points in 2008. In contrast, Obama did worse among high school-educated voters, only winning them by 6 percent in 2008 (slightly lower than his 7 point victory nationally). If recent polls showing Obama performing poorly among high school-educated voters are accurate, then the education gap could widen, making the educational composition of a poll more significant than it was in 2008.
Just how significant? Well, re-weighting PPP's results to reflect the exit polls would produce a consistent but slight shift in Romney’s direction—up to a net-2 points in some of their polls. However, this simple "re-weight" doesn't necessarily reflect a hypothetical poll properly weighted by education, since just re-weighting by education might throw other variables, like race, out of alignment.
Education might also help explain PPP's odd racial sub-samples. While most polling firms show Obama scoring well among minorities but struggling among white voters, PPP shows the opposite, with Obama performing near 2008-levels among whites, but down nearly 10 points among non-white voters. Education could help explain this phenomenon: If Obama is performing well among college-educated whites but struggling among those without a college degree, then oversampling well-educated voters would tend to inflate Obama’s standing among whites. The reverse is true for minorities: Obama does better among less-educated minorities and worse among those with a college degree.
Many pollsters don't weight for education, and PPP just happens to admirably releases comprehensive cross-tabs. But it's possible that this is a bigger issue for automated pollsters with low-response rates. Remember, though: These issues might fade as we approach the election, especially if less-educated low interest voters become more likely to respond to pollsters.