Does Discussion Group Composition Affect Policy Preferences? Results from Three Randomized Experiments

We report the results of three randomized experiments designed to assess whether participants' political attitudes are affected by the prediscussion opinions of their discussion mates. The experiments took place during three multisite Deliberative Polls in 2004 and 2005. Participants discussed current political issues within 330 groups of approximately 10 people each. Because the composition of the groups was determined randomly within each location, the demographic and ideological complexion of the discussion groups varied randomly. This paper examines the effects of randomly varying group composition on political opinions expressed after daylong discussion. We find only sporadic evidence of group composition effects. These results run counter to—or at least qualify—a substantial body of theory and laboratory evidence on the influence of group discussion.