Cost-Benefit Analysis of INPRES primary schools
This interactive dashboard allows users to explore how different parameters are affecting the projected net benefits of INPRES primary schools construction. Net benefits are available in terms of: (i) Living Standards and (ii) Taxes. Toggle "hide" or "show" to see one or both of them in action.
Estimates of benefits are based on "Long-term and Intergenerational Effects of Education: Evidence from School Construction in Indonesia" (Akresh, Halim, Kleemans 2021).
Interactive visualization dashboard was prepared by Sean Hambali.
Read more details how cost-benefit analysis is conducted
School construction cost approximately 782 million 2016 US dollars, around 1.5 percent of the 1973 Indonesian GDP (Duflo, 2001). Schools were expected to recruit three teachers and accommodate 120 students. Using estimates by Daroesman (1971), training three teachers across 61,800 schools would have cost the government 11.7 million in 2016 dollars.
Benefits include both taxes paid directly to the government (household-level taxes as well as estimates for value-added-taxes the government would have collected on household purchases) and improvements in overall living standards.
How long INPRES schools last?
Daroesman (1971) and Duflo (2001) report that schools were expected to last for 20 years. But many INPRES schools are still in-use as of 2016.
Individuals start paying taxes by age
If individuals start paying taxes after finishing Upper Secondary or Tertiary education, then that will be age 19 and 23, respectively.
Number of students and teachers per school
Duflo (2001) assumes 120 students/school and 3 teachers/school. These imply a class size of 20 students across six grades of primary education and one teacher per grade. The latter is reasonable given that schools often run two sessions per day: morning and afternoon classes.
World Development Indicators collects real interest rates in Indonesia between 1987 and 2017. It averages 5.77 percent per year. Since it does not extend as far as our sample period in 1973, we assume a constant annual discount rate of 5 percent.
Annual GDP growth
GDP per-capita growth is obtained from the World Bank’s World Development Indicators. The average between 1961 and 2017 is 3.25 percent per year.
Teachers’ salary growth
We first assume there is no real salary growth over the years and use Duflo (2001)’s reported teacher’s salary in 1973. Subsequently, we allow for linear growth using teacher’s salary observations in 1970 by Daroesman (1972), 1973 by Duflo (2001), Intercensal Surveys 1976 and 1995, and Labor Force Surveys 2000, 2005 and 2010. Teachers are paid for the lifetime of the schools.
Individuals’ tax payments and living standards generally follow an inverted-U shape, where they peak at around age 40-50. In our SUSENAS data, we observe individuals at their peak. To model the lifetime curvature of tax payments and living standards, we assume the same average effect on taxes and living standards across ages but different means at different ages. An average household with a 20-year old father, for instance, only spends $2,373 annually, compared to the mean in households with a 50-year old father, $3,597.
Ratio of Operating Costs over Salaries Costs
Duflo (2001) assumes 25% recurrent administrative or operating costs in addition to teachers’ salaries.
While we do not directly observe living standard and tax effects on the second generation, we do observe the education benefits on the second generation. We scale the second generation’s living standard benefits by the relative ratio of second generation education effect and first generation education effect. We further assume that the first child is born one year after the average age of first marriage (i.e. 25 for fathers and 21 for mothers). Since the average household size in the sample is 4.09, we assume two children per household and that the second child is born 3 years after the first child.
World Development Indicators suggest an average of 56.6 years of life expectancy at birth for individuals born between 1968 and 1980. Conditional on making it to primary school age, the life expectancy is likely higher. We assume a life expectancy of 60 years throughout.
Share of men and women in affected cohorts
We construct a weighted average of the treatment effects on men and women. The share of women in the affected cohort is 0.498. For simplicity, we assume an equal share of men and women.