RABA Research begins with a data file that runs deeper than those that exclude households without an active voting history. This is done to ensure that the beginning universe of potential respondents doesn’t shut out those who may have registered to vote more recently than 2014, and to not overstate representation of those with active voter histories who may be less enthusiastic about voting in 2016.
RABA then employs random digit dialing (RDD) to a universe of more than 20,000 potential voters. Those marked as cell phones are stripped from the sample. IVR surveys are sent to the remaining landlines.
The same then compiles a randomly-selected list of Cellphone Only Households (COHs) and surveys these respondents in two ways: first, by live-dialing to the numbers and patching through willing participants to the same recorded survey taken by those with landlines; second, by screening for COHs with an online panel featuring an opt-in universe of more than 3 million.
The likely voter screen is a combination of self-reported vote likelihood and RABA’s bipartisan consensus on participation of various demographic groups, and is informed by – but not limited to – voter scoring.
The resulting data is then weighted to race, gender, age, and party affiliation using a proprietary method that uses census data, exit polling, and internal voter data.
The weighting process is overseen by a pollster with more than a decade of experience in public opinion research and voter targeting. Our surveys typically have a sampling error of plus or minus 4 percentage points or smaller, at the 95 percent level of confidence.
Using the methods above, RABA Research became the only public polling firm to earlier this year correctly call the Missouri Democratic presidential primary within the margin of error.