Dataset derived from existing cross-sectional socio-economic data collected across marginalised rural areas throughout 31 states in Mexico between 1996 and 1999, the Survey of Household Socio-Economic Characteristics (Encuesta de Caracteristicas Socioeconomicas de los Hogares, ENCASEH). ENCASEH had been conducted to aid in the targeting of the PROGRESA welfare programme, collected data from all households in these communities and contains a rich cross-section of information on individual and household characteristics, along with locality data. The current dataset includes households with at least one 12-17 year old. This research will provide evidence for rural Mexico on whether smaller family sizes increase children's school enrolment and decrease children's work participation, which affect children's human capital accumulation and thus long-term poverty. All too often, policies to promote lower families in order to increase human capital accumulation are based on observed negative relationships between family size and education (parents with large families tend to invest less in education). These relationships can be misleading, because differences in education outcomes between large and small families may be for reasons other than family size. Pinning down whether or not whether family size actually affects education is important for formulating appropriate and effective policies in relation to family planning. To test whether family size has a causal effect on outcomes, one needs access to random variation in the propensity to have more children. Such variation is often provided by natural experiments. In this research, two natural experiments will be used - having twins in the second birth, and having children of the same sex in the first two births - both of which provide sources of random variation in family size.

Dataset derived from Survey of Household Socio-Economic Characteristics (Encuesta de Caracteristicas Socioeconomicas de los Hogares, ENCASEH), 1996-199, a household survey targeting all households in marginalised rural villages in Mexico, 1996-1999. Sample of households with at least one 12-17 year old child. The dataset contains 225 variables, including socio-economic characteristics of households and characteristics of localities where they reside. Total of 550,163 unique households and 955,057 unique individuals.