In this paper we disentangle the role of cooperative preferences and beliefs for explaining MPCR and group size effects in public goods games. To achieve this, we use the ABC approach, which explains cooperation as a function of cooperative attitudes and beliefs. We measure cooperative attitudes using the incentive-compatible strategy method by Fischbacher et al. (2001, Economics Letters, 71-3, 397–404)(FGF). However, to keep the incentives in the strategy method equal across all group sizes (which FGF does not), we also compare FGF with a version of the strategy method that is scalable to any group size. We find that preference types are similar across strategy methods, group sizes of 3 and 9, and MPCRs of 0.4 and 0.8. Further experiments with group sizes of 3 and 30 again find similar distributions of conditional preferences. The ABC approach predicts actual cooperation in all conditions and for both strategy methods and reveals that, controlling for elicited preferences, differences in cooperation levels observed across the various games are mostly due to differences in beliefs.