Pollsters survey too many non-voters

I’ve noticed that a couple of pollsters (SurveyUSA and CNN/ORC) get their likely voter pool by randomly calling registered voters and asking them if they’re likely to vote. Others may use the same criteria. Using this criteria SurveyUSA indicated over 90% of San Diego’s registered voters would vote in mayoral race. The actual number was less than half that. Asking people if they’re likely to vote includes people who will vote along with many who won’t. These non-voters showed the San Diego mayoral race to be much closer than it actually was.

What pollsters need to do is to use a percentage similar to those that’ll actually vote. The linked chart shows the percentage of registered voters who actually vote.

2008 average state: 71% turnout
2010 average state: 50% turnout
2012 average state: 69% turnout

Notes: I couldn’t find accurate registration numbers for New Hampshire or Rhode Island. Thus, I excluded them from the chart. The 2014 registration numbers reflect the most recent report the state has on their website. In some cases this could a year old. Wyoming allows same day registration and voting. Other states due also, but Wyoming doesn’t make them permanently registered voters in their number registered. So their turnout in Presidential years is above 100%. Many states separate active voters and inactive voters. Most of these include all these voters in their registered tallies. Colorado and Nevada do not. So their turnout rate is relatively high. It’s worth noting that Washington and Oregon, which have all VBM elections have a higher percentage of registered voters voting than other states.

As the chart shows, on average half of registered voters voted in the last mid-term. Some pollsters pre-qualify voters based on their voting history. Those that just choose registered voters randomly, regardless of whether they’ve voted or haven’t, need to find a way to exclude nearly half the voters who answer their surveys. That SUSA survey in CA-52 used 83% of registered voters. This SUSA Florida survey uses 86% of registered voters. In 2010, 49% of Florida’s registered voters voted. The inclusion of non-voters mean that you just can’t know how accurate the poll is.