Liquid Democracy or Direct Democracy? One Theoretical Result and Two Experiments

Proponents of participatory democracy praise Liquid Democracy: decisions are taken by referendum, but voters delegate their votes freely. When better informed voters are present and the electorate is finite, we show theoretically that delegation can always strictly increase the probability of a correct decision. However, delegation must be used sparingly because it reduces the information aggregated through voting. In two different experiments -- a tightly controlled lab experiment and a perceptual task run online -- we find that subjects choose very high rates of delegation, and the theoretically possible improvements fail to materialize. The experimental evidence favors Direct Democracy, whether with or without abstention. We study the perceptual task, where signals' precisions are not known, both as a test of the robustness of the lab results and as an independent methodological contribution. We argue that tests under ambiguous information are valuable and under-used tools in studying collective decision-making.