Social decision schemes and choice shift: An analysis of group decisions among bets

Abstract

The present study analyzes group decisions in terms of current models of social decision scheme (SDS) research and in terms of two new models which simulate pairwise majority voting over all alternatives, the full paired comparison (FPC) and the reduced paired comparison (RPC) models. The latter two models take into account the subjects' rank orders over all alternatives (i.e., they are Condorcet functions). On the basis of their individual preferences in the preexperimental phase, 275 subjects were systematically assigned to 55 five-person groups according to theoretically useful configurations of occupancy numbers. Each five-person group had to choose one of six different bets with equal expected values. In the P-condition the probabilities for the bets were equally spaced; in the M-condition the distances between the money values were kept constant. The results show systematic deviations in at least one condition for all models except the best fitting RPC model. Furthermore a comparison with regard to the precision of the models indicates that RPC is significantly more precise than various other models, whereas no other model does significantly better than RPC. For the RPC model neither systematic bias nor precision varied with condition (P vs M). In addition the RPC model can account for risky shift and general choice shift both on the individual and on the group level. Results are discussed within the framework of SDS and choice-shift research.