Evaluating the effects of social programs (Brazil, Taiwan, China)

In 1971, Brazil first extended social security payments to people over 65 in rural areas. The payments represented a large cash transfer to rural household heads, (roughly doubling the income of the average recipient) and were not tied to previous contributions or to the labor force participation of the recipient. I take advantage of the discontinuity created by the rules of the program (which allows me to easily identify those affected by the program). Using data from before implementation (1970) and from after the program was introduced (1980), I look at the effect of these transfers on the labor force participation of the recipient, family size, labor force participation of other household members and educational attainment of children in the household. I find that for a 66 year old in 1980, the cash transfer lowered the probability of labor force participation of the recipient by 22 to 27 percent. However, I find no effect on other outcomes.; The extension of basic schooling from six to nine years in 1968 was the largest expansion of education in Taiwan's modern history. More than 140 new junior high schools were opened in 1968 under this program, increasing the number of junior high schools by 70 percent from 1967 to 1968. I evaluate the effect of this program on education and wages by analyzing cohort differences in educational attainment induced by the timing of the program and by combining these cohort differences with differences across counties in the number of schools built. These estimates suggest that children who were between the ages of 6 and 11 in 1968 received 0.5 to 0.6 additional years of education for every school constructed per 1000 children between the ages of 12 to 14. Using the exogenous variation in schooling due to this program, I construct instrumental variable (IV) estimates of the returns to education. I find that IV estimates based on cohort differences in education are lower than the corresponding OLS estimates, but IV estimates based on regional differences in inter-cohort patterns are not significantly different from the OLS estimates.