Opinion polling, especially in swing states, is a critical element of the electoral process. Market research firms select a sample statistically representative of the overall population of an area and question them on their opinions of candidates and issues.
There are several methods that these firms employ, but the most popular by far is automated polling conducted via the use of IVR systems and live outbound telephone calls conducted by surveyors.
An article in the New Republic by Nate Cohn explores the differences in the data collected from each of these methods, and why discrepancies in the results exist. Data from three “swing” states (Ohio, Virginia and Florida) was collected via both live and automated polls that questioned respondents on their presidential voting preferences. There was a significant divergence in data collected via live cell phone calls versus that collected via IVR-powered polls.
In the live polls, which were conducted by survey administrators telephonically, Obama was shown to have a substantial lead. In automated polls, also administered telephonically but via an IVR system, the results were much closer with Obama either leading by less significant margins or Romney leading. So what’s at issue, and why the inconsistency?
Primarily, the issue is not with the polling itself but rather with the rules and regulations that govern the automated versus live polling. IVR polls cannot be used to contact those using cell phones; they are only able to get in touch with households via landlines. Per Cohn, the “problem is that cell phones have seriously complicated the argument on behalf of IVR polling over the last few years. Automated pollsters are unable to contact voters who only possess a cell phone.”
There has been a large uptick in the percentage of the population exclusively using cell phones as their primary and only phone. Pollsters first need to come to a common agreement on what Cohn calls the “science of cell phone polling” and decide what percentage of a total sample size cell phone users should represent.
In addition, there needs to be diversification in polling methods. Live polling is costly but yields more accurate results because live pollsters can contact users via cell phones. Automated technology is cost effective and very advantageous as a supplemental component of a polling strategy, but it is at a disadvantage due to polling law dictating operations. A truly accurate poll will have pulled data from a combination of sources and will be representative of the overall population.
For more information on polling rules and regulations on a state-by-state basis, visit the Floodlight blog.