Education in Developing Countries - Supplementary Information

Published: April 2018

Evidence of effectiveness

Effects on labor market and pecuniary outcomes

There are a number of studies that use quasi-experimental methods to estimate the impact of primary and secondary education interventions on labor market outcomes. The following table gives a summary of these papers, and we discuss them in more detail below.

Program	Context	Study/studies	Evaluation design	Main results
Admission to public secondary school	Children completing primary school in rural Kenya, 1998	Ozier 2016	'Fuzzy' regression discontinuity design, with survey data from 2007-2009	Individuals with more schooling more likely to be employed aged 20-25.
Building primary schools	Major school construction program in Indonesia, 1973-1979	Duflo 2001 Duflo 2004	Difference in differences, with survey data on wages from 1995	Increased wages for individuals who directly benefited from the program. Decreased wages for cohorts who were too old to benefit from the program.
Elimination of primary school fees	Primary school students in Uganda, 1997	Keats 2016	Regression discontinuity design, with data from 2006, 2009, and 2011	Girls gained one year of schooling; were more likely to be employed and had more assets
Elimination of school fees in grades 1-10	Students in grades 1-10 in Ethiopia, 1993-1996	Chicoine 2017	Difference in differences, with data from 2005 and 2011	Girls gained over two years of schooling and were more likely to work
Elimination of secondary school fees	Secondary school students in Kenya, 2008	Brudevold-Newman 2016	Difference in differences, with data from 2014	Women more likely to work in skilled jobs rather than agriculture; no impact on wage work or self-employment.
3-year scholarship program	Secondary school students aged 12-16 in Cambodia, 2005-2010	Filmer and Schady 2014	Regression discontinuity design	Scholarships increased years in school; no effect on test scores, employment, or earnings.

In general, we think these studies provide weak supporting evidence to Duflo, Dupas, and Kremer 2017 and Bettinger et al. 2017 and indicate that education interventions may have positive effects on labor market outcomes, particularly for women (Keats 2016, Chicoine 2017, Brudevold-Newman 2016) and possibly with negative consequences for individuals that did not benefit from the education intervention (Duflo 2004).

• Ozier 2016 uses a 'fuzzy' regression discontinuity design to estimate the effects of secondary school on the occupational choice and fertility of young adults in Kenya.1 It uses the fact that students are more likely to gain admission to a government secondary school if they pass a national exam taken upon completion of primary school.2 Students just above this cutoff go to school for more years on average than those just below it, and Ozier 2016 compares the later life outcomes of these two groups using survey data.3 It estimates that secondary schooling in this context increased human capital and employment and decreased low-skill self-employment, though the paper does not estimate effects on earnings or income.4
• Esther Duflo estimates the effects of a school building program in Indonesia in the 1970s on labor market outcomes in two papers using a difference in differences approach.5 Duflo 2001 estimates that building additional primary schools in the 1970s increased educational attainment and resulted in increased wages in 1995 for individuals who directly benefited from the program.6 However, Duflo 2004 estimates the impact of the school building policy on cohorts that were too old to benefit from it and finds that it had a negative effect on their wages, suggesting that there were significant negative externalities of the program.7
• We are aware of three recent working papers that use the elimination of school fees in various contexts to estimate the effects of education, generally by comparing outcomes of cohorts that started school just before fees were eliminated with those who started just after fees were eliminated. Keats 2016 estimates the effects of the elimination of primary school fees in Uganda in 1997 using a regression discontinuity design.8 Keats estimates that girls 14 and younger when the program began gained about one year of education, were more likely to be employed by someone else, and had more valuable assets.9 Chicoine 2017 uses the elimination of school fees in grades 1-10 in Ethiopia between 1993 and 1996 to estimate that women who benefitted from the policy gained over two additional years of schooling, were more likely to work, and were less likely to work in agriculture.10 Brudevold-Newman 2016 evaluates the effects of the abolition of fees for public secondary schools in Kenya in 2008 and estimates that women who benefited from this change were more likely to work in skilled jobs rather than in agriculture, though it finds no impact on wage work or self-employment.11
• Filmer and Schady 2014 uses a regression discontinuity design to evaluate the effects of a three-year scholarship program in Cambodia. It estimates that the scholarships had large effects on educational attainment but found no evidence that they improved test scores, employment, or earnings.12

Effects on health outcomes

In addition to the two RCTs discussed in our main report on education, we have found three quasi-experimental studies with plausible causal identification strategies. These studies estimate larger effects on more significant health outcomes than the RCTs, though we place less emphasis on these papers.13

We summarize their key findings in the following table and discuss them in more detail below.

Program	Context	Study/studies	Evaluation design	Main results
Building primary schools	Major school construction program in Indonesia, 1973-1978	Breierova and Duflo 2004	Difference in differences, with survey data from 1995	Reduction in under-5 child mortality.
Elimination of primary school fees	Primary school students in Uganda, 1997	Keats 2016	Regression discontinuity design	Women with more schooling more likely to have a doctor or midwife assist in delivery of first child; more likely to vaccinate child; child less likely to be malnourished. No effect on length of breastfeeding, wasting, fever, diarrhea, or infant mortality.
Building secondary schools	Major school construction program in Taiwan, 1968	Chou et al. 2010	Difference in differences, with survey data from 1978-1999	Women's education reduced cases of low birthweight, neonatal mortality, postneonatal mortality, and infant mortality. Smaller/insignificant effects of men's education.

• Breierova and Duflo 2004 uses a quasi-experimental approach to estimate the effects of a large school construction program in Indonesia between 1973 and 1978 on fertility and child mortality (this is the same school construction program evaluated in Duflo 2001 and Duflo 2004, mentioned above).14 The authors estimate a large and significant (at the 1% level) negative effect of increased educational attainment on child mortality (0.10 fewer children died before age 5 for each additional year of average education of the parents, from a baseline of 0.21) and find no significant difference between the effects of male and female education.15 We interpret these results with caution because this remains a working paper, we typically place less emphasis on estimates using quasi-experimental methods relative to those using an RCT, and we have not read this paper in detail.16
• Keats 2016 uses the elimination of primary school fees in Uganda in 1997 to estimate the effects of education on labor market outcomes (as discussed in more detail above) and also on child health outcomes. It estimates that women with more schooling were much more likely to have a doctor or midwife assist in the delivery of their first child (69% versus 36%) and more likely to vaccinate their first-born children. The study found no significant increase in the number of months of breastfeeding.17 The first-born children of more educated women were also less likely to suffer from chronic malnourishment (stunting and anemia), though there is no statistically significant effect on wasting, fever, diarrhea, or infant mortality.18
• Chou et al. 2010 exploits regional variation in a school building program in Taiwan in 1968 to estimate the effects of parental education on infant birth outcomes between 1978 and 1999.19 It estimates large and significant negative effects of mother's education on low birthweight, neonatal mortality, postneonatal mortality, and infant mortality, with smaller and statistically insignificant effects estimated for father's education.20

Effects on social outcomes

There are a number of quasi-experimental studies that estimate the effects of education interventions on rates of fertility and marriage among females. We interpret them as evidence that is broadly supportive of the conclusions of the experimental studies, that education interventions can reduce rates of fertility and marriage. We summarize the quasi-experimental studies in the following table and discuss them in more detail below.

Program	Context	Study/studies	Evaluation design	Main results
Elimination of primary school fees	Primary school students in Uganda, 1997	Keats 2016	Regression discontinuity design	Women with more schooling delayed marriage and fertility and reduced overall fertility.
Elimination of school fees in grades 1-10	Students in grades 1-10 in Ethiopia, 1993-1996	Chicoine 2017	Difference in differences	Women with more schooling delayed marriage and fertility and reduced overall fertility (one extra year of schooling reduced fertility by 0.16-0.19 children)
Elimination of secondary school fees	Secondary school students in Kenya, 2008	Brudevold-Newman 2016	Difference in differences	Women with more schooling delayed first intercourse and marriage; were less likely to give birth before age 20.
Introduction of universal primary education	Girls in primary school in Nigeria, 1976	Osili and Long 2008	Difference in differences	Additional schooling reduced women's fertility at age 25 (one extra year of schooling reduced fertility by 0.26 children)
3-year scholarship program	Secondary school students aged 12-16 in Cambodia, 2005-2010	Filmer and Schady 2014	Regression discontinuity design	Scholarship increased girls' years in school; no effect on test scores, probability of getting married, or teenage pregnancy.
Building primary schools	Major school construction program in Indonesia, 1973-1978	Breierova and Duflo 2004	Difference in differences, with survey data from 1995	Small increases in age at first marriage; small decreases in number of children before age 25; no effect on overall fertility.

• Three recent working papers that use the elimination of school fees in various contexts to estimate the effects of education on labor market outcomes (discussed in more detail above) also estimate the effects on various social outcomes. Keats 2016 uses the elimination of primary school fees in Uganda in 1997 to estimate that women with more schooling delayed and reduced overall fertility and delayed marriage.21 Chicoine 2017 uses the elimination of school fees in grades 1-10 in Ethiopia between 1993 and 1996 to estimate that an additional year of education for women resulted in a lasting reduction in overall fertility (between 0.16 and 0.19 fewer births) and postponement of first birth and marriage.22 Brudevold-Newman 2016 evaluates the effects of the abolition of fees for public secondary schools in Kenya in 2008 and estimates that women who benefited from this change delayed first intercourse and marriage, with corresponding large reductions in the proportion of women who gave birth before age 20 (38% to 24%).23
• Osili and Long 2008 uses the introduction of universal primary education in Nigeria in 1976 to estimate the effects of female schooling on fertility. It estimates that increasing female education by one year reduced the average number of children a woman had before age 25 by 0.26 births.24
• Filmer and Schady 2014 uses a regression discontinuity design to evaluate the effects of a three-year scholarship program in Cambodia (as discussed above). It estimates that the scholarships had large effects on educational attainment but finds no evidence that they changed the probability of getting married or having a child in adolescence.25
• Breierova and Duflo 2004 uses a quasi-experimental approach to estimate the effects of a large school construction program in Indonesia between 1973 and 1978 on child mortality (as discussed above). The authors also estimate the effects of the program on fertility and marriage. They estimate that an additional year of education increased women's average age at marriage (from 18.4 to 18.78) and that women's education mattered more than men's education for age of marriage.26 They estimate a small reduction in the average number of children born before age 25 (from 1.37 to 1.32) but do not find a statistically significant effect of education on overall fertility in this context.27 As before, we interpret these quasi-experimental estimates with caution.

Sources

Document	Source
Bettinger et al. 2017	Unpublished
Breierova and Duflo 2004	Source
Brudevold-Newman 2016	Source
Chicoine 2017	Source
Chou et al. 2010	Source (archive)
Duflo 2001	Source (archive)
Duflo 2004	Source (archive)
Duflo, Dupas, and Kremer 2017	Source
Filmer and Schady 2014	Source (archive)
Keats 2016	Source
Osili and Long 2008	Source (archive)
Oye, Pritchett, and Sandefur 2016	Source
Ozier 2016	Source (archive)
Wikipedia, "Difference in differences"	Source (archive)
Wikipedia, "Regression discontinuity design"	Source (archive)
Wikipedia, "Regression discontinuity design: Extensions: Fuzzy RDD"	Source (archive)

• 1
  • "I estimate the impacts of secondary school on human capital, occupational choice, and fertility for young adults in Kenya. Probability of admission to government secondary school rises sharply at a score close to the national mean on a standardized 8th grade examination, permitting me to estimate causal effects of schooling in a regression discontinuity framework. I combine administrative test score data with a survey of young adults to estimate these impacts. My results show that secondary schooling increases human capital. For men, I find a drop in low-skill self-employment; for women, I find a reduction in teen pregnancy." Ozier 2016, Pg. 1.
  • "I use a regression discontinuity approach to identify the effect of secondary school on outcomes. As described in Section 2, Kenyan students who take the primary school leaving examination (KCPE) face an admission rule: below a cutoff score, c, it is more difficult to gain admission to secondary school. The identifying assumptions in my analysis are that all other outcome-determining characteristics except for the probability of secondary school attendance vary smoothly near the cutoff, and that outcomes change at the cutoff only because of the induced change in schooling. Because the probability of attendance does not jump from zero to one, this is a "fuzzy" regression discontinuity (Imbens and Lemieux 2008)" Ozier 2016, Pg. 13.

• 2

  "Since 1985, the Kenyan education system has included eight years of primary schooling and four of secondary (Eshiwani 1990, Ferré 2009). At the end of primary school, students take a national leaving examination, the KCPE. A score of 50% or higher—currently 250 points out of 500—is considered to be a passing grade. This examination is the chief determinant of admission to secondary schools (Glewwe, Kremer, and Moulin 2009).... Though an official letter of admission to a government secondary school is rare below this cutoff, it is still not guaranteed for those above it because the number of candidates passing the KCPE may exceed the number of spaces available in public schools." Ozier 2016, Pgs. 6-7.

• 3

  "The primary dataset used in this study is the Kenyan Life Panel Survey (KLPS), an ongoing survey of respondents originally from Funyula and Budalangi Divisions of Busia District, Kenya (Baird, Hamory, and Miguel 2008).... Because the outcomes of interest occur only for adult respondents, I mainly use the more recent round of survey data (KLPS2), treating it as cross-sectional data for 5,084 individuals.

  The KLPS2 survey is comprehensive, including questions on education, employment, and fertility, as well as cognitive tests. The education section includes yearly school participation, from which secondary school completion, grade repetition, and other measures can be constructed; it also includes self-reported KCPE scores for students who complete primary school." Ozier 2016, Pg. 8.

  Note that this survey data with outcomes information is combined with administrative test score data for the KCPE.

• 4
  • "The KLPS2 survey includes a commonly used test of cognitive ability—a subset of Raven’s Progressive Matrices—and an English-language vocabulary test based on the Mill Hill synonyms test… I standardize both outcomes so that they are measured in terms of standard deviations in the KLPS2 population, and show both OLS and 2SLS results for a combined Z-score and separately by test in Panel A of Table 3: completing secondary school improves performance on these tests by 0.6 standard deviations, with very similar estimates given by 2SLS and (potentially biased) OLS…. To the extent that subsequent outcomes depend on a mixture of human capital and signaling, this is evidence that secondary schooling in Kenya does not play a purely signaling role: students measurably gain skills from schooling." Ozier 2016, Pg. 21.
  • "OLS shows a fairly precise zero effect of secondary schooling on employment at this age. However, the regression discontinuity approach gives very different results: the coefficient on schooling is positive and significant depending on controls, shown in IV probit and bivariate probit specifications in columns 3-6… Depending on the specification, I find a rise in employment of between 24 and 43 percent in response to secondary schooling." Ozier 2016, Pg. 23.
  • "As in other labor market studies of relatively young men (Griliches 1977, Zimmerman 1992), I use sector of employment rather than wage to estimate the impact of secondary schooling… While secondary education and self-employment are negatively associated in the cross-section (columns 1 and 2), the causal impact of secondary schooling on low-skill self-employment is much larger; marginal effects from IV probit and bivariate probit estimation are in broad agreement with the 2SLS coefficients: a 40-50 percent lower probability of being self-employed among those who go to secondary school because they pass the KCPE cutoff." Ozier 2016, Pg. 24.

• 5
  • "This paper exploits a dramatic change in policy to evaluate the effect building schools has on education and earnings in Indonesia, a country where the GDP per capita in 1995 was only $720, 3.5 percent that of the United States. In 1973, the Indonesian government launched a major school construction program, the Sekolah Dasar INPRES program. Between 1973-1974 and 1978-1979, more than 61,000 primary schools were constructed - an average of two schools per 1,000 children aged 5 to 14 in 1971. Enrollment rates among children aged 7 to 12 increased from 69 percent in 1973 to 83 percent by 1978. This was in contrast to the absence of capital expenditure and a decline in enrollment in the early 1970’s.

    ...The exposure of an individual to the program was determined both by the number of schools built in his region of birth and by his age when the program was launched. After controlling for region of birth and cohort of birth effects, interactions between dummy variables indicating the age of the individual in 1974 and the intensity of the program in his region of birth are plausibly exogenous variables, and are used as instruments in the wage equation." Duflo 2001, Pg. 795.

  • "The basic idea behind the identification strategy can be illustrated using simple two-by-two tables. Table 3 shows means of education and wages for different cohorts and program levels. Regions are separated in “high program” and “low program” regions. The difference between the number of schools constructed per 1,000 children constructed in high and low program regions is 0.90. 2 In panel A, I compare the educational attainment and the wages of individuals who had little or no exposure to the program (they were 12 to 17 in 1974) to those of individuals who were exposed the entire time they were in primary school (they were 2 to 6 in 1974), in both types of regions. In both cohorts, the average educational attainment and wages in regions that received fewer schools are higher than in regions that received more schools. This reflects the program provision that more schools were to be built in regions where enrollment rates were low. In both types of regions, average educational attainment increased over time. However, it increased more in regions that received more schools. The difference in these differences can be interpreted as the causal effect of the program, under the assumption that in the absence of the program, the increase in educational attainment would not have been systematically different in low and high program regions." Duflo 2001, Pg. 798.

• 6

  "The estimates suggest that each new school constructed per 1,000 children was associated with an increase of 0.12 to 0.19 in years of education and 1.5 to 2.7 percent in earnings for the first cohort fully exposed to the program. This implies estimates of economic returns to education ranging from 6.8 to 10.6 percent." Duflo 2001, Pg. 796.

• 7

  "The school construction program led to an increase in education among individuals who were young enough to attend primary school after 1974, but not among the older cohorts. 2SLS estimates suggest that an increase of 10 percentage points in the proportion of primary school graduates in the labor force reduced the wages of the older cohorts by 3.8 - 10% and increased their formal labor force participation by 4 - 7%." Duflo 2004, Abstract.

• 8

  "This paper contributes new findings from Uganda on the impact of women’s education on both fertility and early child health, and provides evidence on a broad set of channels through which these effects may take place. In order to identify impacts, I implement a regression discontinuity design that takes advantage of the timing of Uganda’s Universal Primary Education (UPE) reform, which eliminated primary school fees beginning in 1997." Keats 2016, Pg. 2.

• 9
  • "In Uganda the UPE reform led to a surge in primary school enrollment, particularly for girls, with differential effects on educational attainment across birth cohorts. Using data from the Uganda Demographic and Health Surveys (DHS), Malaria Indicator Surveys (MIS), and AIDS Indicator Surveys (AIS), I show that prior to its implementation girls aged 14 and younger were disproportionately more likely to be attending primary school and thus benefit from the reform, while girls from older cohorts were not. In line with this, I find that the UPE reform led to a jump of nearly 1 year of additional schooling on average for women who were 14 and younger at the time of the reform. 3 Moreover, the data show gains in educational attainment at the cutoff for all grade levels through the end of secondary school, indicating that treated women from all points in the skills distribution were affected by the program." Keats 2016, Pgs. 2-3.
  • "In this paper, I find evidence supporting the hypothesis that an income effect is responsible for some of the observed shifts in behavior and outcomes. Women with additional schooling are more likely to be literate, have better jobs, and they are wealthier. In 2006, when women to the right of the cutoff were 23, they were 4 percent less likely to be working. However, conditional on working, they were 61 percent more likely to be employed by someone else and 13 percent less likely to be self-employed. This result is consistent with the theory and evidence in Ozier (2015), which presents a model in which it is optimal for educated young people to forgo self-employment opportunities so that they may continue to search for more remunerative formal employment for someone else. Indeed, by 2009, women at the cut-off are 36 percent more likely to be working for cash, rather than for no pay or payment in-kind, and they are more likely to have migrated to the capital, Kampala (where perhaps employment opportunities are better). By 2011, when women at the cutoff were 28 years old, these relative advantages had disappeared as women in the older cohorts eventually found paid employment, yet the gains from the earlier start appear lasting. In both 2009 and 2011 women with more education demonstrate greater wealth in terms of an index of assets and the quality of household construction." Keats 2016, Pg. 5.

• 10
  • "This paper investigates the causal relationship between women’s education and fertility by exploiting variation generated by the removal of school fees in Ethiopia. The increase in schooling caused by this reform is identified using both geographic variation in the intensity of the reform’s impact and the temporal variation generated by the implementation of the reform. The model finds that the removal of school fees in Ethiopia led to an increase of over two years of schooling for women impacted by the reform, and that each additional year of schooling led to a lasting reduction in fertility." Chicoine 2017, Abstract.
  • "Individual level outcome data for Ethiopian women are from 2005 and 2011 rounds of the Ethiopian Demographic and Health Survey (DHS) (Central Statistical Authority - Ethiopia, 2005; 2011)." Chicoine 2017, Pg. 10.
  • "The second, the government’s official Education and Training Policy (ETP) was published the following year, 1994, and officially forwarded to the regional governments prior to the 1995 school year. The release of the transitional government’s official education policy, the ETP, was largely known at the time of Proclamation No. 41. General knowledge of the forthcoming ETP, and the proclamation’s decentralization of power, led to the functional implementation of the ETP at different times in different parts of the country between 1993 and 1996 (Negash, 1996; Oumer, 2009; World Bank, 2009; UNESCO, 2007). This variation will be part of what is taken into account in the paper’s empirical strategy… The central consequence of the ETP was that it required public education to be fee-free for grades one through ten. The identification strategy used in this paper focuses on the removal of these fees, and the idea that the fraction of students completing these grades during the pre-reform period is likely inversely related to the magnitude of the reform’s eventual impact." Chicoine 2017, Pg. 3.
  • "The increase in schooling is identified by combining two dimensions of variation, the timing of the reform and geographic variation in schooling outcomes for cohorts who completed their education prior to the reform. The identification strategy employed in this paper uses age, schooling, and location data that are readily available for countries in most parts of the world. Motivated by the work from Bleakley (2010), Lucas (2010, 2013), and Lucas and Mbiti (2012a,b), the identification utilizes the concept that, although the policy itself is applied uniformly across the country, the intensity of the reform in a specific location depends on the pre-existing characteristics of that area. In this setting, removing school fees from an area of high educational attainment will have a small impact relative to removing the same fees in an area with low pre-reform educational attainment." Chicoine 2017, Pg. 1.
  • "The results in Table 7 examine the impact of the increased schooling on labor market outcomes, in columns (1) through (4), and changes in a woman’s ideal number of children, in column (5). Although income data are not included in the DHS, the first four results show that each additional year of schooling led to statistically significant increases in the likelihood of working, earning cash payment for their work, not working in subsistence agriculture, and not working in agriculture more broadly. All of these measures yield evidence of an increase in productivity for women following the reform. This is direct evidence that some level of quality persisted through the expansion of enrollment generated by the reform; the additional years of schooling led to improvements in later in life labor market outcomes for Ethiopian women." Chicoine 2017, Pg. 18.

• 11
  • "This paper examines the impacts of a national free secondary education (FSE) program in Kenya on educational attainment and achievement, and uses the program as an instrument to examine the impact of education on fertility behaviors and labor market outcomes… My identification of causal impacts exploits region and cohort-specific variation in the treatment intensity of individuals exposed to the program. Regional variation in treatment intensity stems from heterogeneous pre-program primary to secondary school transition rates across Kenya: regions with low pre-FSE primary to secondary transition rates experienced larger increases in secondary schooling rates as a result of the program. The cohort variation arises from the timing of the program: individuals above secondary schooling age at the time of the program’s implementation in 2008 would have had to return to school to take advantage of FSE rather than simply continue their schooling from primary to secondary school. I use these sources of variation to measure the impact of FSE on educational attainment using a difference-in-differences framework." Brudevold-Newman 2016, Pg. 3.
  • "This paper uses two main datasets: the 2014 Kenya Demographic and Health Survey (DHS) and an administrative dataset of secondary school completion examination results." Brudevold-Newman 2016, Pg. 12.
  • "Using the same instrumental variables approach, I also use exposure to the FSE program to examine impacts of education on labor market outcomes. My estimates show that post-primary education shifts young women into more productive sectors: decreasing the probability of agricultural work and increasing the probability of skilled labor while potentially delaying entry into the labor force… My results for women complement this earlier work: while I find no impact on wage work or self-employment, I find that education shifts women across sectors, decreasing the likelihood of working in agriculture and increasing the probability of skilled work." Brudevold-Newman 2016, Pgs. 4-5.
  • "I also find that post-primary schooling shifts young women across labor market sectors. Education increases the likelihood of work in skilled labor by 28% and decreases the likelihood of working in agriculture by almost 80%. These findings for women complement similar existing findings for men (Ozier, Forthcoming). This shift towards more productive sectors is suggestive of potential growth consequences of the program." Brudevold-Newman 2016, Pg. 29.

• 12
  • "We evaluate the medium-term effects of a program that provided scholarships for three years to poor children upon graduation from elementary school in Cambodia, a low-income country. To do this we use a sharp regression discontinuity design. We show that scholarships have substantial effects on school attainment. By the time children would have been in grade 11 had they remained in school, two years after they stopped being eligible for scholarships, those who were offered scholarships have attained 0.6 more grades of completed schooling. Nevertheless, we find no evidence that scholarships had significant effects on test scores, employment, earnings, or the probability of getting married or having a child in adolescence." Filmer and Schady 2014, Abstract.
  • "The CSP scholarship program, which we analyze in this paper, works as follows. The
    government fi rst selected 100 lower secondary schools throughout the country (from a
    total of approximately 800) to participate in the program. These CSP- eligible schools
    were chosen because they served poor areas, as indicated by a poverty map, and because
    there appeared to be high levels of school nonenrollment and dropout, as indicated by
    administrative data; schools covered by other scholarship programs were excluded.
    Next, each primary “feeder” school was mapped to a CSP- eligible secondary school. 4
    Finally, in every feeder school, all students in sixth grade, the last year of primary
    school, fi lled out an “application form” for the CSP scholarship program—regardless of
    whether children or their parents had expressed an interest in attending secondary school.
    Application forms consisted of 26 questions that were easy for sixth graders to
    answer and for other students and teachers to verify. In practice, the form elicited
    information on household size and composition; parental education; the characteristics
    of the home; availability of a toilet, running water, and electricity; and ownership of
    a number of household durables." Filmer and Schady 2014, Pg. 665.
  • "Separately for each CSP school, applicants were then ranked by their dropout- risk score. As requested by the Ministry, in “large” CSP schools, with total enrollment above 200, 50 students with the highest value of the score were then offered a scholarship for grades 7, 8, and 9; in “small” CSP schools, with total enrollment below 200 students, 30 students with the highest value of the score were offered scholarships. In total, just over 3,800 scholarships were offered." Filmer and Schady 2014, Pg. 666.
  • "The dropout- risk score perfectly predicts whether an applicant was offered a scholarship, with the exception of a very small number of changes that resulted from the public complaint mechanism. This is therefore a case of sharp (as opposed to fuzzy) RD [regression discontinuity]. Also, because we focus on the impact of being offered a scholarship, rather than that of actually taking up a scholarship, these are Intent- to- Treat (ITT) estimates of program impact. (For convenience, in the paper we interchangeably refer to children with values of the dropout- risk score above the cutoff, all of whom were offered scholarships, as children who were eligible for scholarships, recipients, or treated children.)" Filmer and Schady 2014, Pg. 669.

• 13

  We also understand that Lant Pritchett and Justin Sandefur are working on a literature review of non-pecuniary effects of education which will in part focus on the health effects of schooling. They have an existing policy paper, Oye, Pritchett, and Sandefur 2016, that uses observational data to show that countries with more school participation of girls have lower child mortality rates and that this reduction is larger in countries with better learning outcomes. We have not read this paper in detail and do not place any emphasis on these results because of their use of observational methods, which do not have a good track record of estimating the causal impacts of policies on outcomes.

• 14
  • "This paper takes advantage of a massive school construction program that took place in Indonesia between 1973 and 1978 to estimate the effect of education on fertility and child mortality. Time and region varying exposure to the school construction program generates instrumental variables for the average education in the household, and the difference in education between husband and wife." Breierova and Duflo 2004, Abstract.
  • "The data used in this paper come from the 1995 intercensal survey of Indonesia (SUPAS), matched with administrative data on the number of schools sanctioned for each district (kabupaten)." Breierova and Duflo 2004, Pg. 4.

• 15
  • "The estimates suggest a strong and significant effect of education on child mortality, but no significant difference between the effects of male and female education. For fertility, the estimates suggest a very different picture, where the difference in education has a strong effect, suggesting that the wife’s education is a stronger determinant of fertility decisions than husband’s education." Breierova and Duflo 2004, Pg. 3.
  • "In table 7 we present the child mortality results. We obtain very similar results for total number of child who died, mortality before one month, mortality before one year and mortality before 5 years. Average education in the household has the effect of reducing child mortality, and there is no significant effect of the difference between husband’s and wife’s education. When we restrict the sample to death occuring before the woman was age 25, we find negative estimates as well, although they are less significant." Breierova and Duflo 2004, Pg. 16.
  • See Table 1 of Breierova and Duflo 2004 for full statistics on baseline child mortality and Table 7 for full estimates of effects on child mortality.

• 16
  • The key results reported in Table 7 of Breierova and Duflo 2004 do not include full regression results.
  • The authors discuss the assumptions underpinning their causal identification and how they test the robustness of these assumptions in section 2.3: "Under the assumption that, in the absence of the program, the pattern of fertility and child mortality across cohorts would not have been different in regions that got more schools than in regions that got fewer schools, we can compare the change in fertility or mortality across regions and over time (as we did for education). Under the assumption that the program itself did not affect anything else than the quantity of education, the interactions of time and the level of program can then be used as instruments for education for the outcomes of interest.

    There are several potential problems with these assumptions. First, there may be differential time trends across regions, not due to the program…. Second, the fertility and child mortality histories are not measured over a period of the same length for older and younger women…. A final problem is potential sample selection..." Breierova and Duflo 2004, Pgs. 7-8.

• 17
  • "Women with more schooling both delay and reduce overall fertility, increase early child health investments, and have less chronically malnourished children. In terms of mechanisms, education increases wealth as well as knowledge and practice of family planning, particularly when women are young. Schooling also delays marriage, but does not appear to alter who women marry or more general bargaining power within the household." Keats 2016, Abstract.
  • "Education increases the probability that a woman had a doctor or midwife assist in the delivery of her first child. According to the IV estimate, each year of schooling increases the likelihood of a trained health practitioner present at the birth of the first child by 33 percentage points. This is a large effect considering that only 41% of the sample had a trained professional assist in the birth. Women with more schooling also breastfeed longer. According to the World Health Organization (WHO), greater consumption of breast milk improves child health since mother’s milk is believed to be free from contaminants and to contain antibodies that may provide children immunity from some diseases. On average women in the sample breastfed their first-born children for approximately 13 months. For each additional year of schooling, women breastfed more than 1 month longer, though this effect is not statistically significant at the usual confidence levels." Keats 2016, Pg. 22.
  • "As shown in the bottom panel of Table 6, women’s education increases the likelihood that first-born children who have reached 1 year of age are fully immunized. Each additional year of schooling increases the probability the first-born child received the tuberculosis vaccine by 12 percentage points, the measles vaccine by 16 percentage points, the complete polio vaccine by 11 percentage points, and the DPT vaccine by 10 percentage points (although the latter two are not statistically significant). Moreover, higher mother’s education also increases the chances that a first-born child receives a vitamin A supplement. Periodic dosing with this supplement prevents severe vitamin A deficiency, which can lead to blindness and also exacerbate common illnesses such as diarrhea and the measles." Keats 2016, Pg. 23.
  • See Table 6, Pg. 22 of Keats 2016 for full results on having a trained assistant during delivery of first child.

• 18

  "I consider both chronic (upper panel) and acute (lower panel) health outcomes. In terms of long-term health, I find large and statistically significant impacts of education. First born children of women with additional schooling are nearly 15 percentage points less likely to display signs of chronic malnutrition through stunting (representing a 37% reduction at the cutoff). In line with this, these children are also 11 percentage points (or 14%) less likely to be anemic, a condition of low levels of hemoglobin in the blood brought about mainly by poor nutrition and, among children under 5, by prolonged exposure to malaria. In contrast, education does not appear to have an effect on the probability that a first-born child shows signs of wasting, which is an indicator of more recent malnutrition or illness that caused weight loss in the period immediately prior to measurement. Indeed, mothers on both sides of the cutoff were equally likely to report their first-born child had fever or diarrhea in the two weeks prior to the survey. Finally, there is also no difference in the probability that the first-born child died before reaching 1 year of age." Keats 2016, Pgs. 23-24.

• 19

  "In 1968, the Taiwanese government extended compulsory education from six to nine years and opened over 150 new junior high schools at a differential rate among regions. Within each region, we exploit variations across cohorts in new junior high school openings to construct an instrument for schooling and employ it to estimate the causal effects of mother’s or father’s schooling on infant birth outcomes in the years 1978–1999." Chou et al. 2010, Abstract.

• 20
  • "Low birthweight has extremely strong positive associations with infant morbidity and mortality. Neonatal deaths pertain to deaths within the first 27 days of life, while postneonatal deaths pertain to deaths between the ages of 28 days and 364 days. Infant deaths are the sum of those occurring in the neonatal and postneonatal periods. We distinguish between neonatal and postneonatal mortality because their causes are very different. Most neonatal deaths are caused by congenital anomalies, prematurity, and complications of delivery, while most postneonatal deaths are caused by infectious diseases and accidents. Infants who die within the first 27 days of life are excluded when the probability of postneonatal death is the outcome." Chou et al. 2010, Pg. 43.
  • Results reported in full in Table 5, Pg. 55 of Chou et al. 2010. "Perhaps the most interesting set of results in the table pertains to the structural infant health equations estimated by weighted two-stage least squares (WTSLS, see panel C). These estimates treat schooling as endogenous and employ the interaction between treatment status and program intensity as an instrument. Although the schooling coefficients have much larger standard errors in the WTSLS regressions, all are negative and all are significant at the 10 percent level. Only the neonatal mortality coefficient loses its significance at the 5 percent level." Chou et al. 2010, Pgs. 54-55.
  • Results reported in full in Table 6, Pg. 56 of Chou et al. 2010. "When father’s schooling is treated as exogenous, all four coefficients are negative, significant at the 1 percent level, and similar in magnitude to the corresponding mother’s coefficients (see panel B). These coefficients retain their negative signs in WTSLS, but only the low-birthweight coefficient is significant at the 10 percent level, and none of the coefficients are significant at the 5 percent level (see panel C)." Chou et al. 2010, Pg. 56.

• 21

  "Women with more schooling both delay and reduce overall fertility, increase early child health investments, and have less chronically malnourished children. In terms of mechanisms, education increases wealth as well as knowledge and practice of family planning, particularly when women are young. Schooling also delays marriage, but does not appear to alter who women marry or more general bargaining power within the household." Keats 2016, Abstract.

• 22
  • "This paper investigates the causal relationship between women’s education and fertility by exploiting variation generated by the removal of school fees in Ethiopia. The increase in schooling caused by this reform is identified using both geographic variation in the intensity of the reform’s impact and the temporal variation generated by the implementation of the reform. The model finds that the removal of school fees in Ethiopia led to an increase of over two years of schooling for women impacted by the reform, and that each additional year of schooling led to a lasting reduction in fertility. Although more educated women are found to marry more educated and economically active men, there is no evidence of an increase in empowerment or husbands encouraging any type of contraception use. The decline in fertility is found to be generated by a postponement in first birth and marriage, as well as a reduction in the ideal number of children, which is associated with an increase in the labor market activity of Ethiopian women" Chicoine 2017, Abstract.
  • "The results from the first stage demonstrate the strength in the instrument’s ability to identify the increase in schooling generated by the removal of school fees in Ethiopia. Estimating the second stage of the 2SLS model focuses on the relationship between the predicted level of education and birth rates, as described in equation (6). The results in Table 5 include the same seven variations of the model used to estimate the first stage, and all seven estimates find a statistically significant and negative relationship between schooling and women’s lifetime number of births. Estimates including linear trend controls and DHS data all yield similar estimates; each additional year of schooling leads to between 0.155 and 0.190 fewer births. Estimates without linear trends, column (4), and those using the census data, column (7), find evidence of even larger reductions in fertility. The findings are consistently larger in magnitude than the OLS estimate from Table 3." Chicoine 2017, Pg. 14.
  • "Figure 5 provides an interesting insight into behavioral changes associated with the timing of a woman’s first birth. The effect of an additional year of schooling on the timing of a woman’s first birth (dashed bars), first sexual intercourse (white), and marriage (black) are shown. The first statistically significant change is the 5.7 percentage point reduction in the likelihood of first birth by 19 for each additional year of schooling, and a 6.4 percentage point reduction by the age of 20. Furthermore, the additional schooling also led to reductions in the likelihood on marriage at the ages of 20 (6.4 percentage points) and 21 (5.3 percentage points). The effect on both first birth and marriage by the age of 20 coincide with an increase in the effect on total births seen in Table 6. The most noticeable effect seen in Figure 5 is the statistically significant reduction in the likelihood of first birth through the age of 24 (4.8 percentage points). This is a significant effect, especially taking into account that only 12.6 percent of Ethiopian women in the pre-reform cohorts had not had their first child by the age of 24; each additional year of schooling led to a nearly 40 percent increase in the number of women who had not had a child by the age of 24." Chicoine 2017, Pg. 15.

• 23
  • "In this section, I use the relationship between FSE and increased educational attainment established in Section 5 as the first stage of an instrumental variables approach to examine the impact of educational attainment on various demographic and occupational choice variables." Brudevold-Newman 2016, Pg. 21.
  • "Figure 8 and Table 7 present the coefficients from the instrumental variable estimates for the probability of first intercourse, birth, and marriage before each teenage age. The coefficients for first intercourse are all negative with the magnitude of the estimated coefficients increasing as the age cutoff increases. Each additional year of education is estimated to decrease the probability of having first intercourse at age 16 by around 2%, rising to 7% by age 18 and 15% by age 20 on base rates of 23%, 46% and 70% suggesting a decrease of between 10-20% at each age. The results appear to be driven by large effects for women as the coefficients for the pooled sample where men are included are lower than the women’s only sample. Larger impacts are estimated for age of first marriage. The coefficients for women indicate a decrease of about 50% in the likelihood of being married before age 18 and of 38% for being married before age 20. The results for age of first birth are most pronounced at the older age range where the estimated coefficients indicate a decreased likelihood of having a first child by age 19-20 of between 31-36%." Brudevold-Newman 2016, Pg. 23.
  • Reduction in proportion of women who gave birth before age 20 calculated using Table 7, Pg. 48 of Brudevold-Newman 2016.

• 24
  • "This paper tests whether the relationship between fertility and education is indeed causal by investigating the introduction of universal primary education in Nigeria. Exploiting differences in program exposure by region and age, the paper presents reduced form and instrumental variables estimates of the impact of female education on fertility. The analysis suggests that increasing female education by one year reduces early fertility by 0.26 births." Osili and Long 2008, Abstract.
  • "In Panel B, we present the IV results, which account for the endogeneity of the schooling decision by instrumenting for years of schooling using UPE primary classroom construction funds per capita in the state of education interacted with year of birth indicators. The IV regressions
    suggest that a one-year increase in female schooling reduces fertility by 0.26 to 0.48 births — close to an 11 to 19% reduction in fertility depending on the speci-
    fication of interest." Osili and Long 2008, Pg. 71.
  • The full estimates can be seen in Table 8, Pg. 72 of Osili and Long 2008.

• 25

  "We evaluate the medium-term effects of a program that provided scholarships for three years to poor children upon graduation from elementary school in Cambodia, a low-income country. To do this we use a sharp regression discontinuity design. We show that scholarships have substantial effects on school attainment. By the time children would have been in grade 11 had they remained in school, two years after they stopped being eligible for scholarships, those who were offered scholarships have attained 0.6 more grades of completed schooling. Nevertheless, we find no evidence that scholarships had significant effects on test scores, employment, earnings, or the probability of getting married or having a child in adolescence." Filmer and Schady 2014, Abstract.

• 26
  • "OLS and 2SLS deliver similar results for the age at marriage and the probability that the woman is currently married. Age at marriage is significantly associated with education (the 2SLS estimate is 0.38 for the average education in the household, suggesting that each year of education is associated with an increase of 0.38 in the age at marriage), and women’s education matters more than men’s education (conditioning on the average education, a greater difference in education between husband and wife reduces marriage age). Education does not seem to be correlated with current marriage status." Breierova and Duflo 2004, Pg. 16.
  • See Table 1 of Breierova and Duflo 2004 for full statistics on baseline marriage and fertility and Table 6 for full estimates of effects on marriage and fertility.

• 27
  • "The 2SLS results on the number of children ever born are somewhat noisy: the point estimate of the effect of average education on the number of children ever born is similar to the OLS (-0.09), but is not significantly different from 0. The 2SLS estimate effect of the difference in education is almost as large (in the opposite sense), but not significant either. The results of the number of children born before the woman turned 15 and 25 are more interesting. In the case of the woman born before age 15, both average education and the difference in education matter, suggesting once again that women’s education has a larger impact on early pregnancy than men’s education. In the case of the number of children born before the woman turned 25, the average education does not seem to matter, but the difference in education does matter. In other words, when the education of the man increases relative to that of his wife, the number of children in the household is predicted to increase." Breierova and Duflo 2004, Pg. 16.
  • See Table 1 of Breierova and Duflo 2004 for full statistics on baseline marriage and fertility and Table 6 for full estimates of effects on marriage and fertility.