Predicted Electorate Random Sampling (PERS) is a technique for preparing a random sample of voters that utilizes voter history from registration files. PERS makes political polls less expensive to conduct and the results are more accurate than polls based on Random Digit Dialing (RDD). PERS represents an advancement from pollsters’ basic use of voter history in sampling such as selecting among past voters, etc. PERS is based on the RBS methods developed by a group of researchers headed by Professors Donald Green and Alan Gerber at the Institution for Social and Policy Studies at Yale University.

RBS saves money because it creates a sample that reflects the composition of the likely electorate without costly wasted calls to unlikely voters. Studies show that applying likely voter screens does not improve the accuracy of RBS polls, so pollsters do not have to pay for calls they can’t use. Accuracy is improved because RBS creates samples more representative of the actual electorate than other sampling techniques for political polls. Studies show that weighting the results of RBS polls does not improve accuracy, so pollsters eliminate the possibility of bias from weighting the results. With the recent demise of reliable exit polling, the absence of weighting is an important feature of RBS.

The advantages of RBS increase as the expected turnout shrinks. RBS eliminates the large number of wasted calls and the difficulty of identifying likely voters in polling for primary, off-year, and municipal elections.

Studies by Professors Green and Gerber and their colleagues at Yale have shown that when RBS samples are prepared using their methods the poll is less expensive to conduct and the results are more accurate than polls based on Random Digit Dialing (RDD). See more articles describing these studies.

The essence of the RBS method is to divide registered voters into several groups, called “strata”, based on their expected rate of turnout. The standard set of strata are:

a) Individuals who voted in the last two similar elections⁽¹⁾
b) Individuals who voted in the last similar election but not the similar election four years before
c) Individuals who did not vote in the last similar election but did vote in the election four years before
d) Individuals who were registered in time to vote in the last similar election but did not vote in either of the last two similar elections
e) Individuals who registered to vote after the last similar election

The creation of an RBS sample requires extensive voter history to complete two stages of calculations for each strata:

1) The rate of voter turnout in each strata is calculated for the similar election four years prior to the current election. The electorate in this historical election is divided into the same strata using voter history from six and eight years ago.
2) These voter turnout rates then determine the skip rate in each strata for the current electorate so that the proportion of voters selected in each strata matches expected voter turnout.

The PERS variation used by VCS contains two important improvements suggested by Professor Robert Hughes of the University of Colorado at Colorado Springs; Christopher Mann, a PhD Candidate at Yale; and several members of the VCS staff. First, studies by Professor Hughes and others have shown that people who registered within 5 months of the registration cut-off date for the current election turnout at a significantly higher rate than those who register in the first 19 months of the election cycle. Therefore, new registrants are divided into two strata: 1) individuals who registered in the first 19 months of the two year election period and 2) individuals who registered in the final 5 months prior to the cut-off date.

Second, PERS uses the historical turnout rate in each strata to determine the current skip frequencies. The original RBS technique was more static with the proportion of the poll sample from each strata determined by the proportions in the historical election four years prior to the current election. This improvement in PERS means that the sample for the current election will automatically adjust for changes in the composition of the electorate between elections (e.g. a large influx of new registrants).

For more information contact any VCS Representative.