Using DHS data for six countries in Latin America and the Caribbean region, we estimate the relation between a mother’s teenage childbearing and that of her daughter’s. Our results show that restricting the estimating sample to mother-daughter matches in the data leads to large negative selection bias in the estimated effect because missing matches are non-random and potentially affected by the teen childbearing status of mothers and daughters. We deal with this selection bias by developing a methodology that uses all available data, including incomplete mother-daughter pairs, and allows missing observations to be endogenous. Our preferred specification shows that being the daughter of a teen mother increases the chances of being a teen mother between 7.4 and 22.2 percentage points (between 42 and 138%). In general, it is also associated with other negative outcomes such as lower educational achievement, acceptance of risky sexual behavior and submissive gender roles in sexual relationships.