A main research focus in many Social Dilemma Games is the suitability of external institutional treatments in inducing socially optimal outcomes. It is likely that participating subjects exhibit unobserved heterogeneity in their reaction to these treatments. This type of “institutional heterogeneity” has to date not found much attention in the experimental literature. We propose a Hierarchical Bayesian estimation framework to highlight these heterogeneity effects. We illustrate that models that ignore treatment-specific heterogeneity can severely under-estimate the variability in treatment-induced decisions amongst the subject population. The resulting misleading picture of comparative treatment effects can lead to sub-optimal institutional choices and related policy decisions.