This paper presents a comparative test of full profile (FP), original hierarchical information integration (HII-O), and integrated hierarchical information integration (HII-I) conjoint methods for modeling group preferences. It is hypothesized that in settings where groups need to decide about complex multi-attribute alternatives, HII-O will predict holdout profiles better than FP, and that HII-I will perform better than HII-O. The predictive ability of the three methods is tested for the case of housing preferences of housing co-ops, which are groups consisting of three to five people who jointly own a house. The results confirm that HII-I outperforms the other two methods and further suggest that FP and HII-O perform equally well. In addition, two variations of HII-I are developed. One of these provides independent estimates of the relative influence of group members on the decision outcomes, however at the expense of requiring larger designs. The two HII-I variations are also tested and found to be equivalent in predictive ability.