By reviewing and improving previous empirical works on this topic, the present paper investigates the dynamic externalities of agglomeration in China. Taking China’s top three municipalities (i.e., Beijing, Shanghai, and Tianjin) as sample regions, it assesses empirically and compares how three types of dynamic externalities—namely MAR (Marshall–Arrow–Romer), Jacobs, and Porter externalities—affect manufacturing productivity. The main findings of this paper are threefold. First, all three types of dynamic externalities measured in labor productivity can be found in the three sample regions, but large differences in the degrees and directions of the effects exist among them. Second, the degree and sign of the effects of each type of externality vary with changes in time lags. Third, the positive effects of these externalities seem to be substitutable for one another. Specifically, if MAR externalities contribute more to productivity growth in one city, Jacobs or Porter externalities do so to a lesser degree and vice versa.