Discussion

5.1. Common factors in school attendance among cohort 1 and cohort 2

According to the results of binary logistic regression, it was discovered that the common determinants of school attendance between cohort 1 and cohort 2 were as follows: 1) pre-school experience, 2) mother’s education level, 3) home language, 4) wealth index, 5) a child with disability within the family, and 6) parental support for education. These factors had effects on school attendance regardless of the children’s age.

　　Firstly, both in cohort 1 and cohort 2, pre-school experience proved to be the most inluential factor regarding the likelihood of attending school. These results explain that if the child attends pre-school, the probability of attending public school also increases. These results were consistent with research by Reynolds et al. (2007) and Rumberger and Lim (2008), who found that pre-school experience had a positive impact on children’s school attendance because pre-school experience prepared children for school. The interesting point in this research was that the impact of pre-school experience on cohort 2 was approximately 3.6 times greater than the impact on cohort 1. This means that children’s attendance of

pre-Table 7: Results of Cohort 2

school will strongly inluence their primary school enrolment, particularly among adolescents.

This research has revealed that pre-school attendance has a great potential to increase the possibility of primary school enrollment among late entry adolescents.

　　Secondly, mother’s education had an impact on school attendance both among lower age and higher age children. Based on the results, children who had a more educated mother were more inclined to attend school than children who had a less educated mother, which was consistant with Suliman and El-Kogali (2005), Kabubo and Mwabu (2007), Huisman and Smits (2009), Onphanhdala (2010), and Alcott and Rose (2015). As with the results of Chevalier (2004), these results show that mother’s education level was more important than father’s education level. Furthermore, it found that the inluence of mother’s education was greater among cohort 2 than cohort 1.

　　Thirdly, home language was another common influential determinant of school attendance among cohort 1 and cohort 2. According to the results, children who speak non-ethnic languages, in other words, children who speak Swahili or English were more likely to attend public school. These results were consistent with claims previously made by Guimbert, Miwa, and Nguyen (2008). Descriptive statistics (Tables 4 and 5) show that the percentage of never-enrolled children who speak Swahili or English in both cohort 1 and 2 was only slightly over 30%, while the percentage of children who speak ethnic languages was approximately 70%. In Tanzania, the national language is Swahili, and the oficial language is English, but there are also 128 ethnic languages (Petzell, 2012). The medium of instruction in primary education is Swahili except in some private schools where the medium of instruction is English. As a result, children who speak a minority language at home tend to struggle in school with a language they are not used to. There was no major difference in the magnitude of impact between cohort 1 and cohort 2.

　　Fourthly, one of the common determinants of school attendance was the wealth index (satisfaction of basic goods). This variable can be described as whether or not the child’s household meets the standard wealth or not. Children who had basic goods such as a radio, newspapers, phone, books, and bicycle were more likely to enroll in school based on the result. This result was consistent with researchers such as Kabubo and Mwabu (2007), Huisman and Smits (2009), Ngware et al. (2009), Olaniyan (2011), and Gonsch (2016). This outcome demonstrated that children need to have a certain level of wealth in order to attend school. Although free primary education has been introduced since 2001, some amount of money is still required (to pay for school uniforms and the cost of school materials) to attend school (Dennis & Stahley, 2012). Therefore, children who cannot meet the standard, in short, children from extremely poor households, still did not have access to education.

　　Finally, parental support for education was found to be a statiscally significant determinant of school enrollment both among cohort 1 and cohort 2 in this study. As researchers such as Chimombo et al. (2000) mentioned, most of the decision to enroll children in school is determined by parents in the case of primary school. Therefore, children whose

parents were supportive of education were found to be more likely to attend school. However, the impact of this variable was very small among both cohorts.

5.2. Different factors of school attendance between cohort 1 and cohort 2 (1) Signiicant variables only among cohort 1

According to the results, there were three factors that were statistically significant only among cohort 1. These were child’s gender, occupation of the head of household, and number of children in the household. Contrary to many studies that have concluded that females are less likely to attend school (Guimbert, Miwa, & Nguyen, 2008; Rolleston, 2009; Gonsch, 2016), this research discovered that females tended to attend school more than males in the younger age cohort. This result was consistent with research in Tanzania by Hoogeveen and Rossi (2011), who discovered that 7-year-old females tend to enroll more in school because they have less opportunity cost than males. Additionally, girls are already monetarily and physically more ready to enroll in school by the age of 7 (Hoogeveen & Rossi, 2011). As they mentioned, it seems that girls have a lower opportunity cost than boys, particularly in rural areas, and thus they enrolled in school more than males. Additionally, this research found that a child’s gender does not matter among the older age children (cohort 2).

　　The occupation of the head of household was signiicant only in cohort 1, and children from households engaged in primary industry occupations were less likely to attend school.

The reason for this can be attributed to the characteristics of primary industry occupations, which require more human power than other occupations. Many households that engage in primary industry occupations depend on their children for help when they are busy, such as at harvest time (Mulkeen, 2005).

　　Schooling decisions were inluenced by the number of children in the household. As in research by Al-Samarrai and Peasgood (1998), Kabubo and Mwabu (2007), Olaniyan (2011), Alcott and Rose (2015), and Gonsch (2016), the number of children in the family impacted the school enrolment negatively although the magnitude of the inluence was very small. This result can be considered to be caused by the decrease in resource availability per child due to an increase in the number of children in the family. The reason for non-signiicance of the number of children in cohort 2 might be the reduction in resource share and childcare. There is a possibility that as a child’s age increases, their younger brothers and sisters do not need so much care, while their older brothers and sisters will generate income by themselves by working.

(2) Signiicant variables only among cohort 2

The factors that were only signiicant among cohort 2 (13- and 14-year-olds) were father’s level of education and gender of the head of household. According to the results, children who had a more educated father were more likely to attend school among the older age children, whereas it was not signiicant among younger age children. These results can be explained by the increase in children’s interaction with the father as they grow up. Children have more

interaction with their mothers, sisters, and female relatives until approximately ive years of age for boys, and until adolescence for girls (Kimambo & Temu, 1969). When they reach the age of 13 or 14, they tend to spend time with their fathers as well, and thus the inluence of the father becomes signiicant in cohort 2. However, the magnitude of the impact of mother’s education level was still greater than that of father’s education level.

　　Gender of the head of household had an impact on children’s school enrollment only among cohort 2. The results show that children who had a male head of household tended to attend school more than children with a female head of household. The results were consistent with Katapa (2005) that female-headed households are less advantaged than male-headed households. The point is that the gender of the head of household was only signiicant among cohort 2. This implies that children who have never enrolled even at 13 and 14 years of age tend to belong to female-headed households, which are often very poor in money and food.