The evidence from what is now a fairly lengthy series of Deliberative Polls suggests that deliberation frequently produces statistically and substantively significant change in policy attitudes, both at the individual level (gross change) and in the aggregate (net change). It also suggests that many of the participants learn a lot and that the net attitude change and the learning are frequently related, with the people who emerge knowing the most being disproportionately responsible for the net attitude change. But these analyses have considered learning only as explanatory variable. What explains how much the learning?
This paper enlists the data from 22 Deliberative Polls to estimate deliberation’s effect on learning, in the sense of increased factual knowledge about the issues at hand. Who learns how much? We consider the participant’s initial interest and knowledge, initial attitude extremity, reading of the briefing materials (provided some weeks in advance); his or her age, race, and gender; and such contextual variables as the knowledge level and heterogeneity of opinion within the participant’s small group, and the mode of the DP (face-to-face versus online). We estimate the effects of these and other variables based on a hierarchical model with the small group and the DP as hierarchical groupings. We also estimate within-DP hierarchical models to show that our results are robust.
