We tested for an effect in each of the levels of the significant moderators from Soderberg et al. (2015) to investigate whether any subgroup showed an effect different from zero. This indicates evidence for an effect only in 1 out of 11 subgroups (field studies). However, there were only 6 field studies included, potentially diminishing the effectiveness of publication bias correction methods in this subset. We further find evidence for heterogeneity in 8 out of 11 subgroups.