Marriage migration in India is the largest permanent migration in the world, yet it is largely unstudied. Across India two thirds of all women have migrated for marriage, around 300 million women, on average moving approximately three and a half hours from their place of birth. This paper provides the first general description of marriage migration in India. It also shows that the existing explanations for marriage migration are either wrong or inconsequential. Marriage migration is unrelated to consumption smoothing (Rosenzweig and Stark, 1989). I show that there are almost no transfers between households linked by marriage, and so there is no marriage linked consumption smoothing. Marriage migration is not driven by geographically imbalanced sex ratios: migration is 30 times larger than necessary to completely equalize sex ratios across India, and increases rather than decreases the imbalanced geographic distribution of women. Marriage migration is also unrelated to caste fractionalization. Instead, I suggest a model of geographic spousal search. The model explains the differences between regions, and emphasizes the central role that the value placed on women plays in marriage migration.