## Abstract

Several countries in sub-Saharan Africa, including Kenya, have introduced free/subsidized secondary education. This paper examines the role of these free/subsidized education policies on teenage motherhood. To identify the causal effect, I exploit the timing of a national reform in Kenya that eliminated/subsidized secondary school fees using a difference-in-difference estimation design. Using the 2014 Kenya Demographic and Health Survey (DHS), I estimate that the likelihood of teenage motherhood decreased by approximately 5 percentage points after the policy's implementation. This study reiterates that the teenage period is crucial in terms of developing human capital through formal schooling. In most developing countries, parents often determine and fund human capital, which makes household wealth/income a critical factor in human capital accumulation and its intergenerational process. I also highlight positive externalities from educational-centered policies, such as long-term economic growth, poverty reduction, and reduction of social welfare dependency.

## Introduction

In 2013, the United Nations Population Fund (UNFPA 2013b; formerly the United Nations Fund for Population Activities) noted that teenage pregnancy1 remains a global challenge that requires urgent resolve. Demonstrating the gravity of teenage pregnancy, the UNFPA (2015) reported that approximately 16 million girls aged 15–19 become pregnant worldwide. Estimates for 2010 show that about 95% of adolescent births worldwide occur in developing countries, with about 36 million women aged 20–24 reporting having given birth before their 18th birthday and about 5.6 million reporting a birth before their 15th birthday (UNFPA 2013a, 2013b). In 2013, the highest incidence of teenage pregnancies occurred in sub-Saharan Africa. In this region, teenage births accounted for more than one-half of all births, an estimated 101 births per 1,000 women, which was almost double the global average (Odimegwu and Mkwananzi 2016). Additionally, among the 15 countries worldwide where more than 30% of 20- to 24-year-olds gave birth before age 18, 14 are in the sub-Saharan region, including Niger, Uganda, and Malawi. Although Kenya is not among this group, it is in the top 10 in terms of numbers of women aged 20–24 who gave birth by their 18th birthday (see Loaiza and Liang 2013).

Ramifications of teenage childbearing are enormous, one of the foremost being its high risk of adverse health outcomes for the mother and the child (Kennedy et al. 2011; Magadi 2006). For instance, whereas these births account for approximately 11% of births worldwide, they contribute approximately 25% of all the pregnancy- or childbirth-related mortality, with the mortality rate of 10- to 14-year-olds estimated to be fivefold that of adult women (Kennedy et al. 2011). Among 15- to 19-year-olds, childbirth complications are the leading cause of death (Godding 2008). Infants face similar adverse outcomes, such as preterm delivery, low birth weight, greater than a 50% increased risk of dying within the first month of life, and higher rates of perinatal morbidity compared with those born to adult women (Kennedy et al. 2011; Patton et al. 2009; Shah and Åhman 2012).

Most teenage mothers drop out of school and may not return; they are vulnerable to a variety of unfavorable outcomes, such as high rates of unemployment, low labor market earnings, lower prospects of marriage, and high welfare dependency and poverty rates (Godding 2008; Hao and Cherlin 2004; Micklewright and Stewart 1999). Consequently, these teenage mothers are unprepared for the psychological, emotional, and financial responsibilities and challenges of being a parent (Hoffman and Maynard 2008), and they are more likely to perpetuate these unfavorable outcomes to multiple generations. The phenomenon is not limited to developing countries: developed countries face similar challenges, with studies showing that children born to teenage mothers are likely to have lower levels of education and are later likely to become teenage parents themselves (Bonell et al. 2006; Jaffee et al. 2001; Moffitt and Team 2002; Pevalin 2003). However, the developed countries often have programs in place to ease teenage mothers' financial burdens or to provide them with avenues for self-reliance. Because many governments in developing countries are already budget-constrained, they may not be able to afford such programs.

In developing countries, the likelihood of teenage motherhood is influenced by myriad issues, such as education level, culture (ethnicity and religion), socioeconomic status, and demographic or sexual behavior factors (e.g., early marriage, age at sexual debut) (Pradhan et al. 2015). The United Nations Population Fund (UNFPA 2015) reported that girls in the poorest regions of the world are four times more likely to give birth in adolescence than girls in richer regions.

These observations are particularly relevant in sub-Saharan Africa, a region with the highest teenage pregnancy rates and a high proportion of girls with poor socioeconomic status (Odejimi and Young 2014). Therefore, identifying protective factors against teenage motherhood, especially in sub-Saharan Africa, could inform policy-making and help abate the adverse health and economic outcomes of teenage motherhood. Human capital improvement is critical, and reducing barriers to schooling (e.g., cost), especially for girls whose school enrollment rate drops during the teenage years (Lloyd et al. 2000), is especially important (Duflo 2004; Duflo et al. 2015; Keats 2014; Odejimi and Young 2014; Osili and Long 2008).

The concentration of adolescent girls aged 10–17 is projected to increase worldwide, with the most substantial change expected to occur in sub-Saharan Africa,2 where adolescent pregnancy is most common, the use of contraceptives is the lowest, and income levels are among the lowest in the world. In general, ages 6–18 are officially primary and secondary school ages. Unfortunately, many girls are out of school, especially in sub-Saharan Africa and other Asian regions, such as South and West Asia, where almost one-third of adolescents of secondary school age are out of school (Loaiza and Liang 2013).

Addressing adolescents' pregnancies and vulnerabilities parallel one of the United Nations Millennium Development Goals: to increase the level of formal education for all school-age children by 2015 and to provide both boys and girls with equal access to education at all levels.3 Addressing the initiative of education for all, several countries in sub-Saharan Africa have taken steps to eliminate or have already eliminated primary school fees in government-aided (public) schools: Malawi (in 1994); Uganda (in 1997); Tanzania (in 2000); and Cameroon, Burundi, Ghana, Rwanda, and Kenya (in 2003) (Grogan 2008). Other countries—such as Uganda (in 2007), Rwanda (in 2007), and Tanzania (in 2016)—have expanded these efforts to the secondary school level.

For many countries, an immediate outcome of this policy has been a significant increase in school enrollment, perhaps an indication of its high impact or a grave state of many households that were previously unable to educate their children. As a point of reference, the average net enrollment ratio for primary school education in Africa increased from 56% in 1999 to 73% in 2007. Although the average net enrollment ratio for secondary education is lower, it showed a 9 percentage point improvement (from 18% to 27%) in the same period (Ohba 2011). The relatively small response at the secondary school level could be attributed to two factors. First, in Kenya, parents are customarily responsible for guiding and selecting a child's primary and sometimes secondary school level. Parents' incentives, however, may not be fully aligned with the children's long-term potential earnings (Baland and Robinson 2000), and parents may opt to enter the child into the labor market. Second, Jensen (2010) noted that misinformation on the financial returns to further schooling may lead individuals (in this case, parents) to underinvest in schooling. This point is particularly important in the context of Kenya, where some communities may hold a negative perception of formal education, especially in areas with high unemployment rates among high school graduates.

Despite evidence of increasing school enrollment resulting from free primary or secondary education, the full impact of free education is subject to research, and results from some research are still contested. For example, some studies have found that free primary education (FPE) increased public primary school enrollment and did not compromise the quality of education (Lucas and Mbiti 2012a). Some studies have found that although FPE improved both boys' and girls' primary school completion rates, it had a more substantial impact on boys, thereby increasing the gender gap in graduation (Lucas and Mbiti 2012b). Other studies have not found evidence of a net change in public school enrollment (Bold et al. 2011), perhaps because of a substantial increase in private school enrollment, particularly in urban areas (Dixon and Tooley 2012). Aiming to add to this literature, I examine the impact of a nationwide elimination or subsidizing of public secondary school fees on the likelihood of teenage motherhood.

I take advantage of the national free secondary education (FSE) program in Kenya and use a difference-in-difference framework to analyze the question. Although the mechanism through which the FSE program acted is not straightforward, I find evidence of a reduced likelihood of teenage motherhood. Although these results are irreconcilable with Filmer and Schady's (2014) findings for Cambodia, for example, they complement those of Baird et al. (2010) and Ferré (2009) showing that improving secondary school attendance in East Africa had a significant impact on teenage motherhood.

## Education System and Free Secondary Schooling

Before 1985, the Kenyan education system required seven years of primary education, four years of secondary education, two years of high school education, and a minimum of three years in the university. To create a more practical curriculum (vocationally oriented curriculum), the Kenyan government overhauled the education system in 1985 to require eight years of primary education, four years of secondary education, and a minimum of four years in the university (Amutabi 2003).4 Students are required to take the Kenya Certificate of Primary Education (KCPE) exam to complete primary education and the Kenya Certificate of Secondary Education (KCSE) exam to complete secondary education. Both exams are nationally standardized and are used for high school and university admissions, respectively. Because students are required to register for these exams in advance, school enrollment is essential for test taking. Historically, however, many schools denied students with outstanding school fees the opportunity to register for these exams, putting additional financial pressure on poor households to pay the outstanding fees or forfeit the students' opportunity to take the national exam.5

### Attendance Patterns and the Reform

Kenya has historically had fewer high schools than primary schools. When it gained its independence in 1963, Kenya had approximately 40 times more primary schools than secondary schools (Ohba 2011). This gap, however, shrunk to almost 4 times by 2011 (Glennerster et al. 2011) mostly due to community-built secondary schools (Ohba 2011). Because the establishment of these community-built schools depended on a community's economic development and strength, some areas were unable to establish their own.

In addition, although the number of secondary schools in Kenya has increased, significant bottlenecks in the education system remain. For instance, distance to the nearest secondary school is still an impediment, more so in some rural areas than in others. In some areas, less than 50% of individuals are within walking distance of a secondary school (Glennerster et al. 2011). According to Alderman and King (1998), walking distance to school has a more significant influence on females' schooling because of safety concerns. Other impediments to Kenya's education system include low transition rates to secondary school (Ohba 2009) and high dropout rates in primary schools (Somerset 2007). The cost of education remains an especially significant barrier for many households (Glennerster et al. 2011; Holla and Kremer 2009; Keats 2014).

Kenya introduced a free secondary education (FSE) program in February 2008. In his FSE introduction speech, President Kibaki clearly articulated that “. . . the main objective of providing free secondary education is to ensure that children from poor households acquire quality education that enables them access [to] opportunities for self-advancement and become productive members of society” (Ohba 2009:5). The FSE program covered 10,265 KES (approximately US\$100) per public secondary student per year in basic tuition expenses.6Glennerster et al. (2011) estimated that household spending on secondary education amounts to one-half of their expenditures, and they spend approximately 8 times more on education for secondary school students than for primary school students: an average of 25,000 vs. 3,000 KES per year. The government grant covered only approximately one-half of the cost per student, making households responsible for the costs of school uniforms, boarding fees, and infrastructure and therefore leaving significant financial burden on many families for whom the cost of school fees remains a significant barrier to schooling. Funds from the grant were distributed from the central government to individual schools based on the number of students.

Although the FSE program was national, regional variation in program intensity is likely, perhaps stemming from heterogeneity in pre-program transition rates from primary to secondary school across the country. Regions with low pre-FSE program transition rates to secondary school could have experienced larger increases in secondary school enrollment. For instance, North Eastern province may have experienced higher program intensity than Nairobi and thus may have experienced a larger decline in teenage motherhood. However, the data I use do not allow me to disentangle potential variance in intensity levels by region, and the results I present are a country average.

## Possible Link Between Education and Pregnancy

I first offer a simple mechanism on how a household's wealth can impact schooling and teenage motherhood. Suppose a teenage population is split between those who reside in poor households, P, and those that reside in wealthy households, W. Assume that teenager i is endowed with ability αi and lives in a household with assets Ait at time t. I assume that αi varies such that students from wealthy households are more likely to perform well on the KCPE exam because they can purchase textbooks or pay for tutors or because they are less likely to miss meals. However, credit constraint does not affect the ability to reach the required KCPE cutoff mark to progress to secondary education: in the absence of credit constraint, students from poor and wealthy households would exhibit similar transition rates to secondary school. A teenager's ability and the household's accessible assets determine schooling investment. An individual's achievable schooling level is $SP*(αi,Ait)$ for poor households and $SW*(αi,Ait)$ for wealthy households, where $SW*$i, Ait) >$SP*$i, Ait). Wealthy households invest in higher levels of schooling for their teenagers; teenagers in poor households remain trapped at low levels of schooling and consequently maintain a poor standard of living and, in this case, I will show that teenagers in poor households face a greater likelihood of teenage motherhood.

A threshold level of assets, Â, exists such that SW*i, Ait) = SP*i, Ait.). Households with an initial asset endowment lower than Â are budget constrained and cannot invest in higher schooling levels. Policies such as the FSE in Kenya could address this need by elevating poor households beyond the threshold asset level. The FSE would also likely affect transition rates to secondary school for teenagers in wealthy households, but it is more likely to have a significant impact for teenagers in poor households who achieved the required exam cutoff marks and would, in the absence of the FSE, otherwise not attend secondary school because of credit constraint.

In this conceptual mechanism, the removal of financial barriers potentially provides access to secondary education, especially for the poor, at an age when most girls would benefit from formal education. Secondary school enrollment provides girls with opportunities for growth beyond simple domestic life, especially in sub-Saharan Africa, where early marriage and childbearing are customary (UNFPA 2013b). Secondary school could provide them with opportunities that may compete with customary norms and limitations, such as early marriage and early childbearing (Caldwell et al. 1983; Westoff 1992). A Kenya Demographic and Health Survey conducted in 1998 found that women with at least a secondary education marry later than those with lower than secondary school education (22 vs. 17 years of age) and have fewer children (National Council for Population and Development et al. 1999). In addition, studies have shown that education may change individuals' preferences toward fewer children with a better quality of life (e.g., Grossman 2006). Other studies have shown that a supervised environment, such as that provided by schools, contributes to lower pregnancy rates. For instance, in a study in South Africa, Rosenberg et al. (2015) found lower pregnancy rates during the school term than during summer holidays.

Educated women are more likely to use contraceptive methods (Ferré 2009), and secondary school education provides an additional level of education. Contraceptives are discussed in classes at the teacher's discretion, in biology classes, and in the curriculum on the biology of AIDS and HIV transmission and prevention (Duflo et al. 2015). Contraceptive knowledge and access could lower the likelihood of both school dropout and early marriages that could have been an outcome of pregnancy (Ferré 2009; LeVine et al. 1996).

Another possible link between education and pregnancy is the increased status that a secondary education would confer. As Caldwell and Caldwell (1987) noted, educated women in sub-Saharan Africa, regardless of marital status, are more respected and enjoy a higher status in society. This higher standing based on education diverges from earlier, traditional cultural trends in which marriage and childbearing were highly regarded. Consequently, education may be perceived as a channel through which girls can elevate their status in society (Caldwell et al. 1983).

Education also influences the timing of reproductive events, such as delayed sexual initiation. Pregnancy in sub-Saharan Africa is a frequent explanation for school dropout, and although pregnant girls are allowed to return to school after birth in some countries, the fraction actually returning is often low (Lloyd and Mensch 2008). If having a child precludes a girl from continuing with schooling, girls may delay sexual activities to ensure they complete their schooling.

Regardless of the mechanism through which schooling influences pregnancy, reduced rates of teenage childbearing could have long-term benefits for the child, mother, and society at large.

## Estimation

### Isolating the Impact of the Reform

Examining the impact of education on teenage motherhood requires acknowledgment of causality issues—that is, determining whether teenage motherhood reduces girls' formal education attainment or that their low education levels lead to teenage motherhood. I address this endogeneity issue by taking advantage of a quasi-experiment setting (the policy change). Although the provision of this free/subsidized secondary school education has a societal-level impact, it is more likely to have a direct and immediate effect on teenage motherhood. Isolating this effect presents methodological challenges. For example, period effects may be contemporaneous with the introduction of the FSE program. In this case, a comparison group to the target population is necessary, but then there may be persistent societal factors that could produce differences between the target group and a comparison group.

To address these issues, I evaluate first births for women aged 15–18 from two cohort groups: (1) the treated group, those impacted directly by the FSE (born in 1990–1993); and (2) the control group, those not impacted by the FSE (born in 1985–1988). Both cohorts were surveyed in 2014, when individuals in the treated group were aged 21–24 (i.e., they were aged 15–18 at policy implementation) and those in the control group were aged 26–29 (i.e., they were aged 20–23 at policy implementation). Figure 1 displays a Lexis diagram of the analysis and helps create cross-sections of the birth cohort in 2008 (when the policy was implemented) and 2014 (when the survey was conducted). Given my focus here on teenage motherhood and the 2014 survey's question on respondent's age at first birth, I consider only those women who were at least age 20 in 2014.

Another potential issue is that some students will not have achieved a sufficient score on the KCPE exam to permit their transition to secondary school regardless of the ability to finance it. I expect this to affect the control group and the treated group similarly; therefore, although all secondary school students were eligible for treatment after the FSE policy, only those who successfully completed primary school potentially received the treatment. Therefore, the methodological approach provides estimates akin to the average treatment effect on the treated. The FSE exposure period differs among respondents. For example, women born in 1993 were exposed longer to the policy than those born in 1990 (i.e., they were in their first year of secondary school at policy implementation, whereas the latter were in their final years of secondary school). I collapse the birth cohorts into two groups and consider the average exposure, enabling me to compare trends in first births between the two cohorts.

Human capital investment (i.e., formal education) often takes place during the teenage years. However, as noted earlier, the high cost of education in developing countries regularly hinders this process. In poor households, many teenagers may be forced to quit school, and girls' educational opportunities may be sacrificed in favor of providing such opportunities for boys. Therefore, children in poor households are less likely to attend school, and girls are often disadvantaged compared with boys. I use Demographic and Health Survey (DHS) data on household wealth to categorize women in the sample into those who are most likely to be impacted by FSE (those from poor households) and those who are less likely to be impacted by FSE (those from wealthy households). Using this information and a difference-in-difference method, I compare trends (1) between treated and control groups and (2) between poor and wealthy households. Ideally, this method controls for unobserved factors between the cohorts, which could affect the differences between the treated and the control cohorts. I attribute these difference-in-difference estimates to the FSE program after controlling for observed individual and household characteristics.7

### Data

I use data from the 2014 Kenya Demographic and Health Survey (DHS) (Kenya National Bureau of Statistics [Nairobi] and ICF 2015). DHSs are national household sample surveys with questions on health, socioeconomic, and anthropomorphic indicators. The survey interviews all members of the selected household, usually men and women of reproductive age (15–49). The DHS uses a multistage sampling design, first selecting primary sampling units (clusters) and then randomly sampling listed households within the randomly selected clusters (excluding families living in institutional facilities, such as boarding schools, hospitals, army barracks, or police camps).

### Descriptive Statistics and Underlying Patterns in the Data

Summary statistics of key variables are presented in Table 1. The table reveals some differences in several variables, such as age at first birth (although I consider only whether a birth occurred) and education. Age at first birth is higher for the younger cohort, suggesting that the young cohort is more likely to delay first birth than the older cohort. For the younger cohort, education level is approximately one year lower, perhaps because some in this cohort are still enrolled in school. Overall, most sampled women identified as Protestant (as opposed to Catholic, Muslim, or having no religion), reflecting the high concentration of Protestants in the country. Reflecting the country's major ethnic groups, most of the sampled women also identified as Kalenjin, Kamba, Kikuyu, Luhya, Luo, and Kisii (as opposed to Maasai or Somali). On average, the two cohorts have relatively similar characteristics.

Evidence that secondary school admissions increased with the introduction of the FSE program can be found in Brudevold-Newman (2016: figure 2). Building on this evidence, I present a visual representation of trends in teenage birth rates in Figure 2. Approximately 35% to 40% of the sampled women were teenage mothers, but the trend decreased over time such that younger cohorts have lower teenage birth rates. Importantly, teenage birth rates are, on average, higher for women born before 1990 than for women born later.

Figure 3 plots teenage birth rates by rural/urban status, religion, and province. The figure shows that rural women were 10 percentage points more likely to be teenage mothers than urban women. This gap, however, was larger for the older cohort than for the younger cohort. The figure is encouraging: it shows a decline in the overall rate of teenage motherhood and helps to decipher which groups are driving the results. Women who identify as Muslim were more likely to be teenage mothers than were those identifying as Catholic or Protestant. This difference is larger in the pre-FSE period (1985–1988), with Muslim women approximately 10 percentage points more likely to be teenage mothers. Still, this gap declined remarkably for the younger cohort. At the tail end of the period covered, Muslim women were less likely to be teenage mothers than were Protestant women.

Trends in teenage motherhood between provinces could be related to cultural differences between ethnic groups, which often share similar migration patterns. Although ethnic groups are not as concentrated within particular provinces as they used to be, a high degree of this homogeneity remains, especially in the North Eastern, Nyanza, and Central provinces. Figure 3 shows that, on average, Nyanza had the highest rate of teenage motherhood (between 42% and 52%), and these high levels persisted over time; by contrast, Central and Nairobi provinces had the lowest rates of teenage motherhood. The observed significant drop for the North Eastern province and a corresponding increase in Nairobi around 1989 may be the result of an influx of Somali people, who are concentrated in the North Eastern province (bordering Somalia), into some Nairobi suburbs.

Figure 4 presents trends by household wealth status (low, middle, and high). DHS data divide wealth index into five categories, with higher values indicating wealthier households. I define households with a wealth index of 0–2 as low income, households with a wealth index of 3 as middle income, and households with a wealth index of 4 as high income. Figure 4 shows that women in high-income households were less likely to be teenage mothers than women from low-income households. For the control group (the 1985–1988 birth cohort), the trend for low-income households is relatively stable, but the trend for high-income households declines around the 1987 birth cohort. I do not expect this difference to impact the results; if it does, then it will bias the results downward, making them understated. For the treated group (the 1990–1993 birth cohort), low-income and high-income households seem to have a similar trend. Trends for those in high-income households are relatively similar before (1985–1988) and after (1990–1993) the reform.

Overall, rates of teenage motherhood vary across characteristics, such as place of residence or religion. What is common, however, is that teenage girls from low-income households or who live in rural areas are more likely to be teenage mothers than are their urban or high-income counterparts, respectively.

### Empirical Strategy

Because the policy reform has a direct impact on a specific age cohort, I isolate its effect by comparing birth rate changes of women affected by the policy (i.e., younger cohort) with those of women not affected by it (i.e., older cohort) and by comparing low-income with high-income households. Subscripts L, H, O, and Y denote women from low-income households, high-income households, the older age cohort, and the younger age cohort, respectively. Equation (1) gives the change in first-birth rates, Ratet, for the treated (the younger cohort) between those from low-income households and those from high-income households, and Eq. (2) gives the change in first-birth rates, Ratec, for the control (the older cohort) between those from low-income households and those from high-income households.
$∆Ratet=Rate(Y,L)−Rate(Y,H).$
(1)
$∆Ratec=Rate(O,L)−Rate(O,H).$
(2)

The difference between the two then provides the estimated impact of free/subsidized secondary school education on teenage motherhood. This is the difference-in-difference (DD) approach.

$DD=∆Ratet−∆Ratec.$
(3)

Table 2 presents the estimated mean first-birth rates. Teens are defined as 15–18 years old in panel A and as 15–19 years old in panel B. In each panel, the first column contains mean first-birth rates for the older cohort (women not directly affected by the policy), the second column contains mean first-birth rates of the younger cohort (women directly affected by the policy), and the third contains the difference between the cohorts. Row 1 of each panel in the table considers women from high-income households only, and row 2 of each panel considers women from low-income households only. Women from high-income households had a 0.7 percentage point increase (15.5% vs. 16.2%) in first births compared with a 4.6 percentage points decrease (43.7% vs. 39.1%) for women from low-income households. The DD estimates of these rates, shown in bold, corresponds to a (statistically significant) 5.3 percentage point relative decline in first-birth rates after the reform for women from low-income households compared with women from high-income households.

In a sensitivity analysis, I conduct a similar exercise but define teens as 15 to 19-year-olds.8 The results, illustrated in panel B of Table 2, show a statistically significant 4.3 percentage point decrease in first-birth rates for women from low-income households compared with women from high-income households. Although this DD estimate indicates a declining trend and is statistically significant, it is smaller in magnitude relative to the estimate shown in panel A. Nonetheless, the evidence from the DD estimates thus far provides preliminary evidence supporting the hypothesis that teenage motherhood responded to the policy reform.

### Regression Model

I formalize the empirical strategy in a regression framework to account for differences in individual and household characteristics. Moreover, controlling for individual and household characteristics produces more efficient estimates. I estimate a DD estimator of the following form:
$Yi,t=β0+β1Cohortt+β2Lowi+β3(Cohortt×Lowi)+β4Xi,t+εi,t,$
(4)

where Yi,t is an indicator of whether a woman was a teenage mother (equal to 1) or not (equal to 0). $Lowi$ equals 1 for women from low-income households and 0 otherwise; $Cohortt$ equals to 1 for women in the younger cohort and 0 otherwise. Xi,t is a vector of individual and household characteristics: education (in years), province of residence, rural/urban residence, ethnicity, and religion. The primary coefficient of interest, β3, captures the policy's impact on teenage motherhood and is interpreted as the DD estimator. εi,t is an idiosyncratic error term.

## Empirical Results

Tables 3 and 4 present the results without individual and household controls (in columns 1 and 3) and with these individual and household characteristics (in columns 2 and 4). In Table 3, I define teens as aged 15–18; in Table 4, I relax this age restriction and consider teens to be aged 15–19, in line with the idea that graduation could be delayed, either because of delayed school entry or grade repetitions. In addition to comparing women from low-income households with those from high-income households, I conduct a sensitivity analysis comparing women from low-income households (columns 1 and 2) with those from middle-income households (columns 3 and 4) for sensitivity analysis.

In column 1 of Table 3, the coefficient on Cohort  ×  Low, which provides the policy's impact on teenage motherhood, is −0.052 and is statistically different from zero. This figure changes marginally to −0.053 when I include additional controls in column 2, but it remains negative and statistically significant at the 1% level of significance. These results suggest that the FSE policy led to a 5.3 percentage point reduction in teenage motherhood. If the average rate of teenage motherhood for poor women before the policy is approximately 48%, as shown in Figure 4, then a 5.3 percentage point decrease is equivalent to approximately a 12% drop in teenage motherhood. I consider this to be a consequential impact because, as discussed earlier, reduced teenage motherhood could translate into higher rates of secondary school completion, lower rates of child marriage, reduced incidence of complications from childbirth, and reduced intergenerational transmission of poverty. In addition, this decline in teenage motherhood provides justification for a targeted educational policy like FSE, especially given the projected increase of teenagers in developing countries.

The results shown in Table 3 may be underestimates if 19-year-olds are regarded as non-school-aged. As mentioned earlier, most students graduate high school at age 18, but graduating at age 19 is not uncommon. I test the sensitivity of the estimates to the definition of teens by considering those aged 19 or younger. The results are presented in Table 4. Teenage motherhood declined by approximately 4.3 percentage points, or by 4.4 percentage points with additional controls included. Importantly, the coefficients remain negatively signed and statistically significantly different from zero.

Overall, the findings for both definitions of teenagers hold up and suggest that the introduction of the free/subsidized secondary school policy has contributed to the decline in teenage birth rates. Notably, these findings mirror the data presented in the figures shown earlier and lend additional credibility to the hypothesis.

In a sensitivity analysis, I compare women from low-income households with those from middle-income households (columns 3 and 4 in Tables 3 and 4). These specifications, however, do not reveal significant differences, further showing that household income plays a large role in teenage pregnancy.

To show that the identifying assumption of parallelism is satisfied, I run a falsification test with 2003 as a “fake” year of FSE implementation. I use the 2008–2009 Kenya Demographic and Health Survey (DHS) (Kenya National Bureau of Statistics [Nairobi] and ICF 2010). Following the same approach as earlier, I evaluate first births for women aged 15–18 from two cohort groups: (1) the treated group, those impacted directly by the FSE (born in 1985–1988); and (2) the control group, those not impacted by the FSE (born in 1980–1983). Both cohorts were surveyed in 2008, when individuals in the treated group were aged 15–18 at policy implementation and therefore aged 21–23 at interview in 2008; those in the control group were aged 20–23 at policy implementation and therefore aged 25–28 in 2008. Given the focus here on teenage motherhood and the 2008 survey's question on respondent's age at first birth, I consider only those women who were at least age 20 in 2008.

As shown in Table 5, the results are not significantly different from zero. As expected, the hypothetical program had no significant effect on the outcome of interest.

## Conclusion

The main objective of abolishing/subsidizing school fees is to increase access to education for low-income households. Reducing cases of unplanned teenage pregnancies, especially in low-income households, may have been an unexpected outcome but was a significant positive externality. Estimating the causal effect of such a countrywide policy change raises research challenges, which I attempted to address in this study. First, societal changes may have occurred contemporaneously with the reform in question. If these societal-level changes do not directly affect all women, then I can estimate the effect of the FSE program on the impacted group. Based on this argument, I designated a targeted group, a control group, low-income households, and high-income households. I then compared these groups to estimate the policy's impact on teenage motherhood.

I found a decrease of approximately 4–5 percentage points in teenage motherhood as a result of the FSE policy—approximately a 12% drop. The magnitude of these results demonstrates the consequences of the cost of education in settings where the cost is prohibitive, especially for low-income households. I presume that if the government increases these subsidies, then teenage motherhood might decrease further. Teenage motherhood has immediate and lasting consequences for girls' health, education, and income potential. These outcomes alter girls' lives, but the influence of such outcomes may also depend on how young they are at first birth. For example, the risk of maternal death for teens under 15 years in low-income countries is twice that of older women. Studies show that girls who remain in school longer are less likely to be teenage mothers (Ferré 2009; UNFPA 2013b). Education also reduces the likelihood of child marriage and, along with delayed childbearing, leads to healthier birth outcomes (Jejeebhoy 1995; UNFPA 2013b). Therefore, investing in girls' education has an immediate and long-term effect on the economy, given that having better-educated girls results in improved productivity and better income. By contrast, not investing in girls comes at a higher cost and often impacts multiple generations.

The results presented here have favorable implications for addressing income and wealth inequality and particularly speak to social mobility, which is at the heart of many social policy discussions. Studies show that high-income parents typically invest more in their children's human capital than lower-income parents do, which in turn creates a positive correlation between the social-economic outcomes for parents and children (Borjas 2010). As such, the human capital accumulation process is intergenerational. In most developing countries, human capital is largely determined and funded by parents, and poor households cannot afford to educate children or may not see the value of education because of their own low levels of education. Thus, government policies, such as subsidized education, provide opportunities for low-income households that could reduce poverty across generations.

I found variation in rates of teenage motherhood within the country, perhaps due to differences in poverty rates, rural/urban settings, and local customs, such as child marriage. Although I did find a positive effect of the FSE on abating teenage motherhood, understanding the underlying differences within communities could help policy-makers in developing and implementing customized intervention plans that cater to the various needs of communities across the country. For example, legislators and policy-makers could consider a review of customary laws, such as child marriage or schooling for teenage mothers; the effect of educating communities on girls' perceptions of early marriages, especially in light of international human rights standards; and the documented importance of an educated society. Fostering such an educationally supportive culture among community members could prevent girls from entering teenage motherhood or child marriage, thereby safeguarding their future well-being.

The study design used here does not explicitly examine mechanisms through which the reform in question may have changed teenage behaviors. I found that school enrollment increased as a result of the FSE policy and that this increased secondary school enrollment. The increased enrollment provided girls with the opportunity to enroll in school or stay in school longer, thus decreasing their chances of entering child marriage, perhaps expanding their horizons, and setting higher goals. School attendance may have provided the needed supervision for girls who otherwise would have been at higher risk if not monitored. Whether this FSE program will be maintained is unclear, dependent on its financial practicality. In the interim, however, the program's impact on teenage motherhood is promising.

## Acknowledgments

I thank three anonymous referees and the Editor for constructive comments. The findings, interpretations, and conclusions expressed are entirely mine.

## Notes

1

Teenage pregnancy is defined as pregnancies among girls aged 10–19. Developed countries often use age to define teenage pregnancy. Although developing countries also use age to define teenage pregnancy, they consider teenage pregnancy to be an issue only when it involves unmarried girls (Cherry et al. 2009). In this paper, I use the term teenage motherhood, defined as teenage childbearing regardless of marital status. Additionally, I prefer the term motherhood as opposed to pregnancy because the data do not provide information on abortions, those currently pregnant, or miscarriages.

2

It is projected that by 2030, approximately 1 in every 4 adolescent girls will reside in sub-Saharan Africa. Countries with the highest projected increases are Nigeria (9.2 million), United Republic of Tanzania (3.7 million), Democratic Republic of the Congo (3.3 million), Uganda (2.5 million), and Kenya (2.3 million).

4

In Kenya, secondary education is a single stage in the education system, following primary school education but preceding the tertiary level. In other countries, education may be divided into lower/junior and upper/senior levels. This study examines a single stage of secondary education designed for students to begin primary school at age 6 or 7, although some students delay their entry into formal schooling or withdraw and reenroll when they are older (Kramon and Posner 2016).

5

Registration for the exams, taken in October and November, occurs in March. Those failing to register in time are often denied the opportunity to take the exams and must wait a year for a second chance.

6

This program aid was not available to students attending private schools.

7

I am not aware of any other national program that was enacted in 2008 that would have affected these groups differently. However, if there were public health initiatives passed in 2008 or later, they would likely affect groups similarly, and they would be accounted for by the difference-in-difference approach used here.

8

Most students graduate high school at the age of 18, but it is not uncommon for some to do so at age 19.

## References

Alderman, H., & King, E. M. (
1998
).
Gender differences in parental investment in education
.
Structural Change and Economic Dynamics
,
9
,
453
468
.
Amutabi, M. N. (
2003
).
Political interference in the running of education in post-independence Kenya
.
International Journal of Educational Development
,
23
,
127
144
.
Baird, S., Chirwa, E., McIntosh, C., & Özler, B. (
2010
).
The short-term impacts of a schooling conditional cash transfer program on the sexual behavior of young women
.
Health Economics
,
19
(
S1
),
55
68
.
Baland, J. M., & Robinson, J. A. (
2000
).
Is child labor inefficient?
Journal of Political Economy
,
108
,
663
679
.
Bold, T., Kimenyi, M., Mwabu, G., & Sandefur, J. (
2011
).
Why did abolishing fees not increase public school enrollment in Kenya?
(CGD Working Paper No. 271).
Washington, DC
:
Center for Global Development
.
Bonell, C., Allen, E., Strange, V., Oakley, A., Copas, A., Johnson, A., & Stephenson, J. (
2006
).
Influence of family type and parenting behaviors on teenage sexual behavior and conceptions
.
Journal of Epidemiology & Community Health
,
60
,
502
506
.
Borjas, G. J. (
2010
).
Labor economics
.
Boston.
MA: McGraw-Hill/Irwin
.
Brudevold-Newman, A. (
2016
).
The impacts of free secondary education: Evidence from Kenya
(Working paper).
College Park
:
Department of Agriculture and Resource Economics, University of Maryland
.
Caldwell, J. C., & Caldwell, P. (
1987
).
The cultural context of high fertility in sub-Saharan Africa
.
Population and Development Review
,
13
,
409
437
.
Caldwell, J. C., Reddy, P. H., & Caldwell, P. (
1983
).
The causes of marriage change in South India
.
Population Studies
,
37
,
343
361
.
Cherry, A. L., Byers, L., & Dillon, M. (
2009
).
A global perspective on teen pregnancy
. In Ehiri, J. (Ed.),
Maternal and child health: Global challenges, programs, and policies
(pp.
375
397
).
New York, NY
:
Springer
.
Dixon, P., & Tooley, J. (
2012
).
A case study of private schools in Kibera: An update
.
,
40
,
690
706
.
Duflo, E. (
2004
).
The medium run effects of educational expansion: Evidence from a large school construction program in Indonesia
.
Journal of Development Economics
,
74
,
163
197
.
Duflo, E., Dupas, P., & Kremer, M. (
2015
).
Education, HIV, and early fertility: Experimental evidence from Kenya
.
American Economic Review
,
105
,
2757
2797
.
Ferré, C. (
2009
).
Age at first child: Does education delay fertility timing? The case of Kenya
(Policy Research Working Paper, No. WPS 4833).
Washington, DC
:
The World Bank
.
Filmer, D., & Schady, N. (
2014
).
The medium-term effects of scholarships in a low-income country
.
Journal of Human Resources
,
49
,
663
694
.
Glennerster, R., Kremer, M., Mbiti, I., & Takavarasha, K. (
2011
).
Access and quality in the Kenyan education system: A review of the progress, challenges and potential solutions prepared for the Office of the Prime Minister of Kenya
Godding, J. (
2008
).
Emergency contraception
.
Emergency Nurse
,
16
(
4
),
22
25
.
Grogan, L. (
2008
).
Universal primary education and school entry in Uganda
.
Journal of African Economies
,
18
,
183
211
.
Grossman, M. (
2006
).
Education and nonmarket outcomes
. In Hanushek, E. A. & Welch, F. (Eds.),
Handbook of the economics of education
(Vol.
1
, pp.
577
633
).
Amsterdam, the Netherlands
:
North Holland
.
Hao, L., & Cherlin, A. J. (
2004
).
Welfare reform and teenage pregnancy, childbirth, and school dropout
.
Journal of Marriage and Family
,
66
,
179
194
.
Hoffman, S. D., & Maynard, R. A. (Eds.). (
2008
).
Kids having kids: Economic costs & social consequences of teen pregnancy
(2nd ed.).
Washington, DC
:
The Urban Institute Press
.
Holla, A., & Kremer, M. (
2009
).
Pricing and access: Lessons from randomized evaluations in education and health
(CGD Working Paper No. 158).
Washington, DC
:
Center for Global Development
.
Jaffee, S., Caspi, A., Moffitt, T. E., Belsky, J., & Silva, P. (
2001
).
Why are children born to teen mothers at risk for adverse outcomes in young adulthood? Results from a 20-year longitudinal study
.
Development and Psychopathology
,
13
,
377
397
.
Jejeebhoy, S. J. (
1995
).
Women's education, autonomy, and reproductive behaviour: Experience from developing countries
.
Oxford, UK
:
Clarendon Press
.
Jensen, R. (
2010
).
The (perceived) returns to education and the demand for schooling
.
Quarterly Journal of Economics
,
125
,
515
548
.
Keats, A. (
2014
).
Women's schooling, fertility, and child health outcomes: Evidence from Uganda's free primary education program
. Unpublished manuscript,
Department of Economics, Wesleyan University
,
Middletown, CT
.
Kennedy, E., Gray, N., Azzopardi, P., & Creati, M. (
2011
).
Adolescent fertility and family planning in East Asia and the Pacific: A review of DHS reports
.
Reproductive Health
,
8
,
11
. https://doi.org/10.1186/1742-4755-8-11
Kenya National Bureau of Statistics (Nairobi), & ICF
. (
2010
).
Kenya Demographic and Health Survey 2008-09
.
Nairobi, Kenya, and Calverton, MD
:
KNBS and ICF
Kenya National Bureau of Statistics (Nairobi), & ICF
. (
2015
).
Kenya Demographic and Health Survey 2014
.
Nairobi, Kenya, and Rockville, MD
:
KNBS and ICF
Kramon, E., & Posner, D. N. (
2016
).
Ethnic favoritism in education in Kenya
.
Quarterly Journal of Political Science
,
11
,
1
58
.
LeVine, R. A., Levine, S., Dixon, S., Richman, A., Keefer, C. H., Leiderman, P. H., & Brazelton, T. B. (
1996
).
Child care and culture: Lessons from Africa
.
Cambridge, UK
:
Cambridge University Press
.
Lloyd, C. B., & Mensch, B. S. (
2008
).
Marriage and childbirth as factors in dropping out from school: An analysis of DHS data from sub-Saharan Africa
.
Population Studies
,
62
,
1
13
.
Lloyd, C. B., Mensch, B. S., & Clark, W. H. (
2000
).
The effects of primary school quality on school dropout among Kenyan girls and boys
.
Comparative Education Review
,
44
,
113
147
.
Loaiza, E., & Liang, M. (
2013
).
Adolescent pregnancy: A review of the evidence
(UNFPA report).
New York, NY
:
United Nations Population Fund
.
Lucas, A. M., & Mbiti, I. M. (
2012a
).
Access, sorting, and achievement: The short-run effects of free primary education in Kenya
.
American Economic Journal: Applied Economics
,
4
(
4
),
226
253
.
Lucas, A. M., & Mbiti, I. M. (
2012b
).
Does free primary education narrow gender differences in schooling? Evidence from Kenya
.
Journal of African Economies
,
21
,
691
722
.
2006
).
Poor pregnancy outcomes among adolescents in South Nyanza region of Kenya
.
African Journal of Reproductive Health
,
10
(
1
),
26
38
.
Micklewright, J., & Stewart, K. (
1999
).
Is the well-being of children converging in the European Union?
Economic Journal
,
109
,
692
714
.
Moffitt, T. E., & Team, E.-R. S. (
2002
).
Teen-aged mothers in contemporary Britain
.
Journal of Child Psychology and Psychiatry
,
43
,
727
742
.
National Council for Population and Development (NCPD), Central Bureau of Statistics (CBS) (Office of the Vice President and Ministry of Planning and National Development) [Kenya], and Macro International Inc. (MI)
. (
1999
).
Kenya Demographic and Health Survey 1998
(Report).
Calverton, MD
:
NCPD, CBS, and MI
Odejimi, O., & Young, D. B. (
2014
).
A policy pathway to reducing teenage pregnancy in Africa
.
Journal of Human Growth and Development
,
24
,
135
141
.
Odimegwu, C., & Mkwananzi, S. (
2016
).
Factors associated with teen pregnancy in sub-Saharan Africa: A multi-country cross-sectional study
.
African Journal of Reproductive Health
,
20
(
3
),
94
107
.
Ohba, A. (
2009
).
Does free secondary education enable the poor to gain access? A study from rural Kenya
(CREATE Research Monograph No. 21).
Brighton, UK
:
Consortium for Research on Educational Access, Transitions and Equity
Ohba, A. (
2011
).
The abolition of secondary school fees in Kenya: Responses by the poor
.
International Journal of Educational Development
,
31
,
402
408
.
Osili, U. O., & Long, B. T. (
2008
).
Does female schooling reduce fertility? Evidence from Nigeria
.
Journal of Development Economics
,
87
,
57
75
.
Patton, G. C., Coffey, C., Sawyer, S. M., Viner, R. M., Haller, D. M., Bose, K., . . . Mathers, C. D. (
2009
).
Global patterns of mortality in young people: A systematic analysis of population health data
.
Lancet
,
374
,
881
892
.
Pevalin, D. J. (
2003
).
Outcomes in childhood and adulthood by mother's age at birth: Evidence from the 1970 British Cohort Study
(ISER Working Papers, No. 2003-31).
Essex, UK
:
Institute for Social and Economic Research
.
Pradhan, R., Wynter, K., & Fisher, J. (
2015
).
Factors associated with pregnancy among adolescents in low-income and lower middle-income countries: A systematic review
.
Journal of Epidemiology & Community Health
,
69
,
918
924
.
Rosenberg, M., Pettifor, A., Miller, W. C., Thirumurthy, H., Emch, M., Afolabi, S. A., . . . Tollman, S. (
2015
).
Relationship between school dropout and teen pregnancy among rural South African young women
.
International Journal of Epidemiology
,
44
,
928
936
.
Shah, I. H., & Åhman, E. (
2012
).
Unsafe abortion differentials in 2008 by age and developing country region: High burden among young women
.
Reproductive Health Matters
,
20
(
39
),
169
173
.
Somerset, A. (
2007
,
September
).
A preliminary note on Kenya primary school enrollment trends over four decades
. Paper presented at the CREATE symposium at the UKFIET conference,
Oxford, UK
.
UNFPA
.(
2013a
).
Adolescent pregnancy. A review of evidence
.
New York, NY
:
United Nations Population Fund
.
UNFPA
. (
2013b
).
Motherhood in childhood: Facing the challenge of adolescent pregnancy
(State of the World Population 2013 report).
New York, NY
:
United Nations Population Fund
.
UNFPA
. (
2015
).
Girlhood, not motherhood: Preventing adolescent pregnancy
.
New York, NY
:
United Nations Population Fund
.
Westoff, C. F. (
1992
).
Age at marriage, age at first birth, and fertility in Africa
(World Bank Technical Paper No. 169).
Washington, DC
:
The World Bank
.
This is an open access article distributed under the terms of a Creative Commons license (CC BY-NC-ND 4.0).