I think this question was addressed many times in the literature but I was not able to make a clear picture for an answer. It is known that peculiar velocities of galaxies are too small to explain the formation of voids. Voids must have existed already before galaxies were in place with their usual speeds. However, I do not understand how this is consistent with two facts: (i) we do not expect to find clusters beyond z > 2. Clustering is also evolutionary, so why should the formation of voids follow a different history? (ii) Quasar distribution at z > 2 shows no clustering nor voids.