In the “blue lie game”, subjects can decide to lie in favor of themselves and also of their ingroup. We find that blue lies are pervasive and detrimental for the society. They decrease in the presence of an ingroup norm of truth-telling, but increase when the same norm is exhibited by the outgroup. Stronger ingroup identity determines a higher sensitivity to the ingroup norm. Silence substantially reduces the number of blue lies, weakens the effects of norms, and increases social welfare. We structurally estimate a model that disentangles the possible drivers of blue lies and evaluates social welfare.