The Effect of Discussion Group Composition on Policy Preferences in the United States

Cynthia Farrar
Jennifer E. Green
David W. Nickerson
Steven Shewfelt
Fieldwork by:
United States of America
3,128 people in 330 discussion groups
2004 - 2005
Target group:
  • Men and boys
  • Rural population
  • Voters
  • Women and girls
Outcome of interest:
  • Attitudes and norms
Intervention type:
  • Information

Social scientists have long been interested in the effects of group discussions on politics on individual opinions. In a series of three evaluations throughout the United States, researchers evaluated the effect of group composition on individual participants’ political views. They found little evidence that the ideological and demographic complexion of the group influenced post-discussion opinions.

Policy issue

Social psychologists have long studied the effects of group deliberation on individual opinions. A person’s views might change during a group discussion to be more accepted by their peers or, on the other hand, they may also be convinced by someone else’s argument. Group dynamics have recently become a key research question for political scientists. Previous studies’ results suggest that individuals with minority views or profiles may face a double disadvantage: not only are their views unlikely to prevail in decisions made based on the majority opinion, but these individuals may also be pressured into changing their opinions to better match those of the majority.1 However, many of these group deliberation studies were conducted with small samples in lab settings, making their results hard to compare to real-world settings. The studies described here represent a large field evaluation, including many more discussion groups, topics, and variations than have been included in previous work.

Context of the evaluation

Researchers partnered with the University of California Berkeley Survey Research Center to recruit 3,128 participants across 25 states for three related studies. In all studies, the sample pool closely resembled the larger US population. However, the sample was comprised of a somewhat larger proportion of non-Hispanic whites than the general population due to the demographic profile of the sites where the evaluations were carried out, and participants were also much more likely to have graduated from college than the average American. This may make the results most relevant topopulations with above-average rates of college education.

All three studies took place from 2004-2005. The first study included 721 participants; the second study included 1,559 participants, and the third study included 848 participants.

Details of the intervention

Researchers conducted three randomized evaluations to test the impact of group discussion on individual preferences. The first experiment was a standard “Deliberative Poll,” a type of poll designed to reveal the conclusions that would be reached if citizens had the opportunity to become more informed and more engaged with the issues. Prior to the deliberation, participants received a background briefing document containing factual information about the issues that would be discussed.

When participants arrived at survey sites, they answered a short survey covering demographic information and their stance on the issues to be discussed. At each site, participants and moderators were randomly assigned to discussion groups of 6 to 13 people. The day’s schedule included introductory videos on each discussion topic, a morning discussion on one of two topics (either economics or national security), a midpoint survey, an afternoon discussion on the remaining topic, a panel in which groups posed questions to a balanced group of experts, and a final survey.

The second and third evaluations replicated the first, but additionally randomly varied the group size, whether individuals shared their individual opinions with the group, and how actively the moderator intervened.

Results and policy lessons

Overall, researchers found that the composition of the discussion groups did not affect participants’ opinions. On average, group opinion influenced individual opinions in only 4 out of the 34 questions across all experiments. Neither the size of a group nor publicizing each person’s survey responses had any effect on individual opinions. Additionally, directing moderators to play active or passive roles had no influence on individual opinions. These results demonstrate that group deliberations do not necessarily induce conformity, even for individuals who have characteristics that traditionally affect one’s level of influence (including gender, race, education, and age). The research design could not isolate the relevant structural factors, but the results demonstrate that it is possible to construct a deliberative environment that does not induce group composition effects.

Farrar, Cynthia, Donald P. Green, Jennifer E. Green, David W. Nickerson, and Steven Shewfelt. 2009. "Does Discussion Group Composition Affect Policy Preferences? Results From Three Randomized Experiments." Political Psychology 30(4): 615-647.
Mendleberg, Tali. 2006. “Small group deliberation.” Paper presented at the Annual Meeting of the American Political Science Association, Philadelphia. PA.