Dem
51
GOP
49
The professional association for pollsters, the American Association for Public Opinion Research, held its 79th annual conference recently in Atlanta, with dozens of sessions, workshops, panels, short courses, and idea groups on many topics related to public opinion research. Many of the sessions were on specific topics, such as "Integrating Text-to-Web into Election and Political Surveys," "Issues that Divide: Attitudes Toward Policing and Immigration," "Messaging That Matters, Contact Strategies and Materials," "Gender and Sexual Identities: Trends and Measurement," "Innovations in Question and Scale Development," and "We Can Do It: Sampling Rare and Unique Populations." Sessions ran 90 minutes and had 4 papers. Hundreds of papers were presented. The president of AAPOR is Jennifer Agiesta, who runs polling for CNN. We don't have the attendance figures for 2024, but in 2023, 1,411 people attended.
One of the topics that came up over and over was how to increase response rates. Also, how to make sure all subpopulations are adequately sampled, especially those whose views may differ from the majority's. The conclusion is that no one contact mode works anymore. In the old days, say 30 years ago, random-digit dialing worked because everyone had one phone at home and most people answered it when it rang. Those days are not coming back.
SSRS discussed an experiment that recruited respondents different ways, including phone, SMS text message, and postcard. Potential respondents were given six ways to respond: webpage URL, QR code to scan, text message, e-mail, a phone number to call (inbound dialing) and SSRS calling them (outbound dialing). Text messaging is becoming increasingly popular, especially text messages that contain a link to a web page where the questions are. There is also an economic benefit to this mode: Sending a text message is cheaper for the pollster than calling the respondent and talking to him or her for 15 minutes. One problem with text-to-web, though, is that many people have been told by multiple sources never to click on links in text messages.
One result presented is that younger respondents liked interactive voice response calls while older ones preferred talking to human beings rather than to a computer.
Another finding is that sending people postcards increased participation by Republicans. Apparently, postcards are a medium they understand and trust. Web responders tended to be Democrats. Several papers noted that reaching voters of color and people with lower educational attainment worked best when an actual human being called them on the phone.
One problem that has plagued pollsters is the shy-Trump voter effect. Ipsos reported that by adding a sentence to the pitch saying that they were especially interested in hearing from people from underrepresented groups boosted the number of Republicans who took the survey. The University of Pennsylvania tried putting an American flag in the invitations to try to get Patriotic Americans to sign up. It did give a more balanced sample—but it also reduced the total number of responses.
Given that the pollsters know that using multimode methods to contact respondents is not going to give a representative sample at all, weighting becomes extremely important—in fact, critical. SSRS found that rather than asking people for their partisanship, asking them specifically who they voted for in the past gave a better result, even though some people don't remember who they voted for in the past. These are typically the swing voters who decided at the last minute and who are crucial. Another problem is that each presidential cycle, about a quarter of the voters didn't vote last time, either due to aging in or being a marginal voter who votes only when the spirit moves him or her.
Will any of these new strategies help? We don't know and they don't know. (V)