This paper examines the significance of environmental regulation in an economy where an eco-product supplied by a single producer is differentiated from a conventional product generating negative externalities. We develop two types of the model: one is a static model without learning effect of eco-product planning, and the other is a dynamic model with learning effect. We show that the regulation should be adopted when the marginal cost of the eco-product production is high enough in a static setting. In a dynamic model, however, whether the regulation improves social welfare is dependent not only on current marginal costs of the eco-product but also on the degree of dynamic learning effect. Particularly, the regulation could improve social welfare when learning effect is either small or large enough, while it could deteriorate social welfare in an intermediate case. Although intuitions tell us that the value of the regulation appears to be monotonically increasing in learning effect, our results suggest that the value possesses a nonmonotone U-shaped feature with respect to learning effect. The optimal decision of the regulation in a dynamic setting could be converse to that of a static setting, providing important policy implications of learning potentials.