Improving election polls at a time of mistrust

Pollsters are jittery going into the midterm elections. In both presidential contests of 2016 and 2020 pollsters seriously underestimated support for Donald Trump and other Republicans. Pollsters recognize that the traditional approaches they have been using no longer accurately capture the mood of the electorate.

One reason that has been offered to explain their forecasting errors is that some Republicans, distrustful of the media and other elite institutions, were less likely to participate in polls. So, pollsters are now employing new techniques to improve their predictions of electoral outcomes. These techniques include inviting potential respondents to participate in surveys via text message, administering online surveys to panels of people who are paid for their participation, and adding voter turnout history to voter files.

But there is a critically important element missing in these efforts to improve accuracy in predicting elections.

Pollsters need to go out into the communities where voters live and talk to them face-to-face in natural settings (e.g., restaurants and bars, laundromats, stores, parks, etc.) and not just ask questions remotely using a standardized script.

This was the methodology pioneered by Samuel Lubell, who analyzed the reasons for Truman’s upset victory over Thomas Dewey in 1948, which was not predicted by any of the major pollsters at the time. In the 1950s and 60s, Lubell successfully predicted elections by sampling key precincts and talking to voters there informally in their homes and outdoor settings to understand their political views more deeply than his contemporaries in the business. He then added that voter sentiment to the quantitative data he analyzed.

Pollsters and qualitative researchers used a similar approach in the 1988 presidential election between George H.W. Bush and Michael Dukakis. Rather than holding focus group discussions in research facilities, voters were recruited and paid to meet in neighborhood bars and other familiar places where they regularly socialized. It was in these discussions where disaffected white male Democratic voters expressed, in a depth not heard before, their anger about feeling pushed aside by the Democratic party.

What do these two qualitative techniques suggest for how researchers can achieve greater accuracy in predicting electoral outcomes?

First, it may be important to hold discussions outside of focus group facilities so that the environment is more informal and participants feel a greater degree of  ownership over the discussion. When people meet to discuss topics they feel deeply about, they are more likely to speak candidly and build trust with the interviewer when it’s a bar or a diner they regularly go to in their own neighborhood. These are places they know, and the familiarity helps to diminish their belief that the interviewer is just using them to extract information for use against them.

Second, unlike traditional focus groups, which rely mostly on interviewing people who don’t know each other, voters may be more likely to feel comfortable expressing their views without self-censorship when they know each other. This approach builds on the technique of using friendship pairs in qualitative research in which focus group participants recommend a friend to also serve as a panelist. By adopting this approach, people are less likely to feel inhibited about expressing themselves honestly.

In our current political environment, where voters feel mistrustful of the media, politicians, and the Washington establishment, the interviewers themselves may be also perceived as part of this mix. The more electoral researchers can meet voters on their own turf, empower them by having them sit and talk with people they know, the less the interviewer will be identified with the environment voters mistrust.

In a familiar environment, voters can feel that they have the voice they want and deserve and be able to speak candidly. In turn, pollsters can develop greater insights, which can help them to make more accurate predictions. 

This result is not just a nicety that will make pollsters less jittery. Our democracy stays healthy when voters have input that accurately reflects their opinions and can be shared with fellow voters. 

On-target poll predictions will inject a dose of badly needed trust in our electoral system.

Peter Tuckel, Ph.D., is a professor emeritus at Hunter College, CUNY.

Barbara Kaplan-Tuckel, Ph.D., is former president of InVision, a qualitative research firm.