Convincing explanations with QCA : the contribution of “essential” configurational models