Linear regression models of treatment differences between info and belief control treatments.

Model A shows differences in claimed mean payments and model B differences in claimed number of fives with respect to under- and overestimators and their treatment interactions. One case refers to the reported mean (model A) or reported number of fives (model B) over the sequence of twelve dice casts per period per subject (yielding a total of N = 480 cases for each model). Only periods 2, 3 and 4 are used because these are the periods after information feedback in the info treatment. Robust standard errors are used, which were clustered for subjects. T statistics are reported in parentheses, stars denote statistical significance with *p