The mass migration of Jews from the Russian Empire to the US is commonly believed to have been caused by two waves of pogroms (1881--1882 and 1903--1906). This view has recently been questioned by historians, but little quantitative evidence exists to support or refute it. I construct a data set that links hundreds of thousands of Jewish immigrants (1900--1914) and Jewish hometown-based associations (1861--1920) to their places of origin, and geo-locate hundreds of pogroms. I find no evidence that the Jewish migration was started by the first wave of pogroms; instead, subsequent migration continued along a pre-existing spatial trend and originated from districts that did not experience violence. The second wave of pogroms, however, did meaningfully increase the rate of migration from affected districts. I interpret these findings as an indication that prior existence of migration networks is a necessary condition for push factors to cause migration, and that these networks tend to spread through a process of spatial diffusion. This leads to a new understanding of the causes of the Jewish migration and poses a challenge to the push-pull consensus paradigm in the economic literature on the Age of Mass Migration.