Examining the RNC's Polling Recommendations
The Republican National Committee's self-described Campaign 2012 "autopsy," released on Monday, contains five separate recommendations for revamping the party's polling and survey research apparatus, including some of the same fixes being discussed by pollsters reportedly participating in the National Republican Congressional Committee's review of their practices.
The section of the report dealing with the party's polling begins with this data point: Seven-in-ten Republican pollsters said that Democratic polling was better than the GOP's work in 2012, with 22 percent saying Democrats "did 'much better' than the Republicans when it came to accuracy and reliability."
The report raises many of the known challenges facing pollsters, including the increasing use of cell phones, bilingual interviews, declining response rates and "flawed turnout models." But the report also makes some specific recommendations for addressing some of these issues.
One recommendation states: "The RNC should strongly encourage that all Republican surveys include no less [sic] than 25 percent cell phone subsamples and recognize that in certain states and districts a higher number may be required." That recommendation "does not apply to surveys done using auto dialers," which are not legally permitted to call cell phones. But implicit in that recommendation is the idea that polls that do not call enough cell phones (or do not call them at all) are less reliable.
Another encourages the RNC to conduct a test comparing two sampling techniques: random-digit dialing, in which a computer randomly dials a phone number within a specified area code or exchange, and registration-based sampling, in which voters are randomly selected according to their voter file. "The Obama campaign and the Democratic Party are moving to listed voter samples and away from random digit dialing," the report states. Many GOP pollsters already use voter files for sampling.
The report also calls into question the ways in which Republican polling firms determine the likely electorate. Many believe that one reason for the GOP's inaccurate polling last year was that pollsters were excluding voters who were determined to be less likely to vote but voted anyway -- mostly for Democrats. These low-frequency voters are more likely to be younger and non-white; in other words, they're more likely to vote Democratic. "Special attention needs to be given to this question to ensure that we are not screening out casual interest voters who nevertheless show up on Election Day," the report states. "Screeners that are too robust, particularly during presidential cycles, have the potential to skew results to the favor of our candidates because they exclude too many young and minority voters."
Similar to efforts spearheaded by the NRCC, first reported last week by Politico, the RNC is also proposing a "quarterly summit" of GOP pollsters "to discuss the current political environment and debate assumptions, sampling, screening, and weighting of samples in an attempt to generate more accurate and consistent data across multiple committees and platforms."
The RNC is looking to take advantage of the spirit of cooperation that has emerged following last year's elections, which saw Democrats win another term in the White House, expand their majority in the Senate and chip away at the GOP's majority in the House. But these polling firms that would participate in these summits are competitors, and that creates a natural disincentive to cooperate. Additionally, since many of the proposals would involve increased costs -- like calling more cell phones or conducting interviews in languages other than English -- some firms may opt against implementing some of these recommendations to offer cheaper prices to candidates.