I argue that several key epistemic values in the Evidence-Based Policy (EBP) paradigm stand in a trade-off relation with a wide range of moral values that policy-makers may be interested in pursuing. The reason is that while standard EBP methods are informative about average treatment effects they remain silent on heterogeneity in agents’ response to policy interventions. Because of this, these methods are uninformative on the distributive consequences of policy. This makes it difficult for policy-makers to pursue distributive values such as equality or priority for the worst-off. I sketch out how this challenges both value-freedom and neutrality in EBP.