Dropout Behaviour of Children: The Case of West Bengal

Abstract

This article examines the dropout behaviour of students below class 10 level (maximum age of 16 years) from two different perspectives. First, from the parents’ perspective, we identify the major characteristics of a household that forces the child to dropout from school, using a Probit analysis. Second, from the child’s perspective, we try to relate the reasons for dropping out (as specified by the child) with the background of the household the child belongs to and the school infrastructure provided to him/her, through a multinomial logit model. The data set used is the 71st round (January 2014 to June 2014) data on education expenditure and is provided by the National Sample Survey Office (NSSO). The analysis is done for the state of West Bengal, India, separately for boys and girls. It shows the significance of parental education and economic factors in children’s dropout behaviour, which is in line with the observations in existing literature. Additionally, this article offers children’s perspectives on such behaviour from across genders and economic classes.

Keywords

Dropout behaviour school education female dropouts Probit analysis multinomial logit model West Bengal

Introduction

Education is one of the basic requirements for human development. It is a foundation for individual and societal development. With education, employment opportunities are broadened and income levels are increased. The level of education of a country is measured by its literacy rate. Illiteracy is a serious obstacle for the well-being of individuals as well as for development. Despite governmental efforts in various forms, school dropout remains a major problem in many countries. Dropping out means that all that is invested by the state, society, community and family goes to waste, which involves waste of national labour and wealth, with adverse long-term consequences.

In India, the problem of dropout emerges as one of the major reasons of low literacy rate, as these dropped out children do not receive adequate education. Dropout problem is prevalent not only in schools but also at every level of the education system in India and at different age groups of children, although the reasons may vary at different stages. According to ‘Educational statistics at a glance’ report (2016) by the Ministry of Human Resources, India, the dropout rate among boys and girls are 4.53 per cent and 4.14 per cent, respectively. Although, Sarva Shiksha Abhiyan (2010) and Right to Education Act (RTE) (2009) have played major roles in the decline of dropout rate in India, still a huge number of children do not complete their education in schools.

There are a number of studies of empirical and theoretical nature (Chugh, 2011; Gouda & Sekher, 2014; Mishra & Chatterjee, 2017; Nithiya Amirtham & Kundupuzhakkal, 2017; Prakash et al., 2017) that address the dropout problem of India from different aspects. Chugh (2011) pointed out that family background (lack of financial resources and lack of interest in studies) are major reasons for drop out of children living in the slum areas in Delhi. She further noted that the dropout problem becomes acute at the below higher-secondary level. Nithiya Amirtham and Kundupuzhakkal (2013) studied the current situation of education system in the light of gender disparity. They observed that the dropout problem is more acute among girl children because of marriage at early age, confined social structure and lack of suitable facilities for female students. Gouda and Sekher (2014) observed that the dropout was high among the children belonging to Muslim, Scheduled Caste and Scheduled Tribe families. They also noted that dropouts among the children belonging to illiterate parents were four times higher than that of the literate parents and that, if parents were not working the possibility of dropout among their children was relatively high. In the context of lack of suitable facilities for female students, Prakash et al. (2017) provided a vivid picture based on villages in Karnataka where the vulnerability of adolescent girl children is another reason for discontinuing school. Adolescence is a period of acquiring new capacities. It is not only a time of opportunity but also of vulnerability to risky behaviour, which can have life-long consequences, especially on education, career and health. According to a joint study by UNESCO Institute for Statistics and the Global Education Monitoring, India has over a fifth of children between the ages of 6 and 11 out of school, followed by a third of adolescents belonging to the age group 12 and 14.¹

In case of West Bengal, Hati and Majumder (2012) found that 19.6 per cent dropout rate in primary schools for boys and 15.4 per cent for girls are higher than the overall dropout rate in India (6.4%) in 2009–2010. Further, due to financial constraints of the family, the dropped out children are often engaged in jobs in some unorganised sectors. A survey conducted by the Pratichi Trust (2017), West Bengal, pointed out that dirty toilets are a common feature for most schools. Unavailability of either stored water or running water inside the toilet was one reason for it. Children, particularly girls, were hardly found to be using these toilets. Thus, lack of basic cleanliness rendered most toilets in West Bengal into non-functional toilets. In all, unavailability of functional toilets (real or virtual) is a serious barrier for transition and retention of children, particularly girls, in secondary education. In this context, it is important to mention that hardly any school in West Bengal had any provision for the disposal of sanitary pad, which is also a serious impediment for the promotion of girls’ education at the secondary level.

As noted earlier, the dropout problem is not confined to India. Studies on dropouts outside India include works by Bradley and Renzulli (2011), Rumberger (2001) and Duc and Tam (2013). Bradley and Renzulli (2011) found that apart from socio-economic and socio-demographic family background, race (Black, Latino and White) and academic performance (grade point average [GPA]) of a student are major factors for higher dropout rate in the USA. Rumberger (2001) provides a theoretical reasoning of infrastructural factors related to school, family background affecting dropouts of students in the USA. Duc and Tam (2013) point out that lack of interest in studies and household poverty are the major reasons for a child to discontinue studies. Further, through interviews of the dropped out children about their activities after discontinuing school, they find that dropped out children generally get engaged in the household chores (looking after younger siblings, working in fields, herding, etc.). For example, in case of boys, they get jobs in small factories, garages, etc., and in case of girls, they generally help their mothers in household chores, work as maids/help, etc., in other households.

In most of the empirical works, the incidence of dropout is analysed taking the dependent variable to be a dichotomous variable. It is, however, more important from policy perspective to track the paths for the chosen reasons for the child to drop out (Bradley & Renzulli, 2011).

In this article, we address the dropout problem from two different perspectives. First, keeping in mind that the socio-economic and socio-demographic backgrounds of a household are influential factors for the dropout decision, we use a Probit regression to identify those household factors which prevent the child from going to school, using household level information available in our dataset. Second, in line with the study by Bradley and Renzulli (2011), we classify the reasons, as stated by the students, into certain groups of reasons for dropout. A multinomial logit regression is used to identify the specific factors that influence the dropped out child’s particular choice. For this analysis, we use individual specific information available in our dataset. The 71st round (January 2014 to June 2014) education expenditure data, provided by the National Sample Survey Office (NSSO), for the state of West Bengal are used in this analysis.² The analysis is done separately for boys and girls.

The plan of the article is as follows. In the second section a brief description of empirical models used in this article is provided. The methodology is discussed in the third section. The fourth section describes the data. In the fifth section, the results of the Probit regression and multinomial logit regression are provided. Finally, the sixth section concludes the article with some remarks.

Model Specification

The dropout problem is complex and multidimensional. The related research works have found that not only individual specific reasons (no interest in studies, low grade points, etc.), the socio-economic background of the family, financial constraints are also reasons that influence a child to stop going to school. Keeping this in mind, the household specific reasons and individual reasons are analysed as follows:

Household Level Analysis:

A dichotomous variable Y is defined as

Y = 1, if the household has atleast one dropped out child

= 0, if otherwise.

The Probit regression equation is as follows:

\Pr o b (Y = 1) = Φ (X^{'} β),

(1)

where Φ is the cumulative density function of standard normal distribution. X is a vector of explanatory variables, which contain household specific information that influence a child to drop out and β is the vector of coefficients for the explanatory variables. β is estimated using maximum likelihood method.

Individual Level Analysis

We now look at the dropout issue from an alternative angle. Here the students come up with reasons for not going to school. The possible reasons, as stated by the students, are summarised below.

According to Bradley and Renzulli (2011), the ‘push’ factor group refers to the supply side factors that discourage a student from going to school. These factors are generally related to facilities provided by school authority: the environment of classes, medium of instruction, unable to cope up with studies,³ unfriendly atmosphere, quality of teachers, etc. In case of India, as per the recent District Information System for Education (DISE) report (2014), 30 per cent of primary and 15 per cent of upper primary schools have pupil–teacher ratio (PTRs) higher than 30:1 and 35:1, respectively. Some private schools in urban areas even have just one teacher for 65–70 students. Roy (2013) notes that high PTR is a reflection of poor class environment which is nonconductive for good learning. These percentages thus indicate that such ‘push’ factors may discourage many students from going to school and consequently there will be dropouts.⁴

The ‘pull’ factors refer to the reasons for dropout as a cost-benefit analysis (Stearns & Glennie, 2006). In the context of India, almost every household has more than one child. Normally, in middle-class and lower middle-class families, the elder child in the household takes more responsibilities for the household. For example, a student may have to work with family members in farming, take care of younger siblings or may have to go out to work for earning money for the family. As per the latest Census report (2011), there are 33 million child labourers in the age group of 5–18 years in India. Mukherjee (2012) notes that the parents send their child to work to be able to use their earning as a supplementary source of income. Moreover, the producers are also willing to employ children for low cost of labour. This increases the opportunity cost of staying in school as the student is forgoing the possible wage earning. Clearly, this implies that financial constraints, engagements in domestic and economic activities lead students to discontinue study.

‘Opted out’ factors refer to individual specific factors. Apart from push and pull factors a student may lose interest in studies or may think that a certain level of achievement in education is sufficient for him/her.

‘Other reasons’ refer to the reasons that do not fall into the above three categories. For example, it might be a tradition in the family of a student not to study further.

For a boy child, the reasons for dropping out, as stated by the child, are categorised into the aforementioned four groups-push factors, pull factor, opted out factor and other factors. For a girl child, another category is added to these four categories, which is ‘female-specific reasons’. Prakash et al. (2017) noted that in addition to the four reasons, there are some hurdles that only a girl child has to face in India. It might so happen that due to reasons like early marriage, lack of female teachers and improper toilet facilities, the girl child has to leave school. These reasons become particularly relevant when the girl child attends adolescence.⁵

A multinomial logit model, described below, is used for linking the choices with the actual variables of demand and supply side factors. The analysis is performed for boys and girls separately.

Let the reason for dropping out for a student be denoted by T. Here, a boy child has four categories of reasons (‘push’ factor, ‘pull’ factor, ‘opted out’ and ‘other reasons’) and a girl child has five categories of reasons (‘push’ factor, ‘pull’ factor, ‘opted out’, ‘female-specific’ and ‘other reasons’). The ‘other reasons’ category is taken as base outcome for both boy and girl child. The probability of choosing the jth reason category for lth child (P_lj) is denoted by

P_{l j} = \Pr o b (T_{l} = j t h r e a s o n) = \frac{e^{{Z^{'}}_{l j} γ_{j}}}{\sum_{j} e^{{Z^{'}}_{l j} γ_{j}}}, j = 1, \dots, J

(2)

Here, Z_lj represents a vector of individual specific information and some household characteristics that influence a child to choose a certain reason over the ‘other reasons’ category and γ_j is the vector of coefficients corresponding to jth reason. The qualitative dependent variable T_l can take any of the J possible values, each corresponding to a different reason category. Since each individual must select one reason category, only J – 1 sets of coefficients are uniquely defined. We will normalise by setting the coefficients for the Jth reason category to zero. The parameters of the model are estimated by the maximum likelihood method.

Methodology

There are three parts of the estimation procedure in this article.

First, the Probit model (Equation (1)) is estimated to analyse the important household specific features that cause child drop out. Here, the explanatory variables⁶ are (a) the log of monthly per capita expenditure (log(MPCE)) as the household’s financial profile, (b) education level⁷ of the father (Father_edu), (c) education level of the mother (Mother_edu) as the household’s educational background, (d) household size (hhsize), (e) proportion of male children of below 16 years (child_m), (f) proportion of female children of below 16 years (child_f), (g) age of the household head (hdage) as the household’s demographic information, (h) distances from primary, upper primary and secondary schools to the residence (distance1, distance2 and distance3, respectively) and (i) a dummy (d₁) which takes value 1, if any member of the household has computer,⁸ and takes value 0, if otherwise.

Next, the multinomial logit model (Equation (2)) for a male child is estimated. As already mentioned, here ‘other reasons’ category is chosen as the base outcome. In this case, the dependent variables are the log of odds ratio of choosing ‘push factor’ (denoted by push) vs. ‘other reasons’ (denoted by other), ‘pull factor’ (denoted by pull) vs. ‘other reasons’ and ‘opted out’ (denoted by opted) vs. ‘other reasons’. Thus, for ‘push factor’, ‘pull factor’ and ‘opted out’, the dependent variables are $\log [\frac{\Pr o b (T_{l} = p u s h)}{\Pr o b (T_{l} = o t h e r)}], \log [\frac{\Pr o b (T_{l} = p u l l)}{\Pr o b (T_{l} = o t h e r)}]$ and $\log [\frac{\Pr o b (T_{l} = o p t e d)}{\Pr o b (T_{l} = o t h e r)}]$ , respectively, for lth male child.

Here the explanatory variables are (a) the log(MPCE), (b) education level of the father (Father_edu), (c) education level of the mother (Mother_edu), (d) proportion of male children of below 16 years (child_m), (e) proportion of female children of below 16 years (child_f), (f) a dummy (d₂), which takes value 1, if the child has access to computer, and 0, if otherwise, (g) a dummy (d_institute), which takes value 1, if the school is a government school, and 0, if otherwise, (h) a dummy (d_language) which takes value 1, if the language used in the school and that at home are same, and 0, if otherwise, (i) a dummy (d_midday) which takes value 1, if the school provides midday meal, and 0, if otherwise, (j) age at the time of drop out (dropout_age) and (k) grade when dropped out (dropout_grade).

Finally, the multinomial logit model (Equation (2)) is estimated for a female child. Here, there is an additional option for the dependent variable, viz. ‘female-specific factors’ (denoted by fem) vs. ‘other reasons’. Thus, for ‘push factor’, ‘pull factor’, ‘opted out’ and ‘female-specific’ the dependent variables are $\log [\frac{\Pr o b (T_{l} = p u s h)}{\Pr o b (T_{l} = o t h e r)}], \log [\frac{\Pr o b (T_{l} = p u l l)}{\Pr o b (T_{l} = o t h e r)}]$ , $\log [\frac{\Pr o b (T_{l} = o p t e d)}{\Pr o b (T_{l} = o t h e r)}]$ and $\log [\frac{\Pr o b (T_{l} = f e m)}{\Pr o b (T_{l} = o t h e r)}]$ , respectively, for lth female child. The explanatory variables remain the same as in the case of a male child.

Description of the Data

The data set used here is the 71st round data on participation and expenditure on education conducted by the NSSO, Government of India, for the state of West Bengal. The span of the data set is January 2014 to June 2014. The survey has been conducted over 29 states and 6 union territories (UTs). This is the latest data set available that contains detailed information on education expenditure. Moreover, from this data set a complete profile of household education expenditure pattern at the individual level can be obtained.

The dropout rate in the state of West Bengal is much higher than the figures at the national level. Also, in the context of marriage at early age, West Bengal (53%) comes fifth after Bihar, Jharkhand, Rajasthan and Andhra Pradesh (Ghosh & Kar, 2010). More importantly, there has been no systematic study of this nature on school dropouts in West Bengal. This has prompted us to select the state of West Bengal for this analysis. The sample size is 2,536 households in the rural sector and 2,304 households in the urban sector. Of these, 1,268 households in the rural sector and 936 households in the urban sector have at least one dropped out child. These constitute our data set.

Table 1 presents the distribution of male and female students (below class 10 level) by educational status. While the dropout rate for male children is 17.45 per cent in the rural sector and 13.51 per cent in the urban sector, the corresponding figures for the female children are 17.84 per cent in the rural sector and 15.29 per cent in the urban sector.

Table 2 presents the distribution of dropped out children by reason for dropping out, average dropout age and average grade of dropout separately for male and female students. Clearly, the ‘pull factor’ plays the major role from childrens’ perspective, followed by ‘opted out’ for males and ‘female specific’ reasons for females. Hence, as per the child’s perception, financial constraints, engagement in economic and domestic activities are the major reasons for dropping out for both male and female children in both sectors. It is also evident that on an average the children drop out at the 7th grade across all categories.

Table 1.

Distribution of Students (below class 10 level) by Educational Status

Sex of the Child	Educational Status	Sector
Sex of the Child	Educational Status	Rural	Urban
Male	Dropped out	996 (17.45%)	689 (13.51%)
	Studying	4,787 (82.55%)	4,506 (86.49%)
	Total	5,783 (100%)	5,195 (100%)
Female	Dropped out	972 (17.84%)	743 (15.29%)
	Studying	4,564 (82.16%)	4,223 (84.71%)
	Total	5,536 (100%)	4,966 (100%)

Source: Authors’ calculation.

Table 2.

Descriptive Statistics of Dropped Out Students

Reasons/Factors for Dropping Out	Male (1,685)		Female (1,715)
Reasons/Factors for Dropping Out	Rural (996)	Urban (689)	Rural (972)	Urban (743)
Push	45 (4.52%)	22 (3.19%)	40 (4.12%)	19 (2.56%)
Pull	605 (60.74%)	374 (54.28%)	520 (53.50%)	337 (45.36%)
Opted out	281 (28.21%)	246 (35.70%)	176 (18.11%)	186 (25.03%)
Female specific			180 (18.52%)	152 (20.46%)
Others	65 (6.53%)	47 (6.82%)	56 (5.76%)	49 (6.59%)
Average dropout age (years)	12	14	14	16
Average dropout grade	7	7	7	7

Source: Authors’ calculation.

Results and Discussion

Table 3 gives the results of the Probit regression for households with at least one dropped out child. The values of R² are 0.1238 and 0.1564 for the rural and the urban sector, respectively. The coefficients of distance1, distance2 and distance3 are non-significant in both sectors, which implies that the distance from the residence to the schools are not important factors for drop out of a child.⁹ The coefficients of log(MPCE) are negative, statistically significant implying that children belonging to the households with better financial status are less likely to drop out. The coefficients of hhsize are positive, significant in both sectors leading to the implication that children from larger households (in terms of household size) are more likely to drop out compared to smaller households. This is further supported by the positive, significant coefficients of child_m and child_f in both sectors. In other words, belonging to a large family and having many younger siblings negatively affects school participation, which corroborates the findings of earlier studies. This is possibly because children belonging to the households with more children (male and female) get less care in terms of education. The coefficients of Father_edu, Mother_edu and hdage are negative, significant in both sectors. This implies that children of the households with educated parents and households with older head (with experience and wisdom) are less likely to drop out. The coefficients of the dummy d₁ are negative, significant in both sectors, indicating that children belonging to the households having computer are less likely to discontinue studies. Children belonging to households owning a computer generally belong to affluent society and they become used to such facility.¹⁰ Hence, the probability for such a child to drop out is likely to be low.

Table 3.

Probit Regression Results for Households with at Least One Dropped Out Child

Explanatory Variables	Rural	Urban
Log(MPCE)	–0.239* (0.031)	–0.580** (0.000)
Hhsize	0.371** (0.000)	0.302** (0.000)
child_m	0.990** (0.001)	0.625** (0.000)
child_f	0.034** (0.002)	0.906** (0.001)
Father_edu	–0.027** (0.003)	–0.048* (0.028)
Mother_edu	–0.037** (0.001)	–0.042** (0.001)
hdage	–0.028* (0.021)	–0.032** (0.002)
distance1 (Distance of residence from primary school)	–0.048 (0.119)	–0.147 (0.201)
distance2 (Distance of residence from upper primary school)	0.061 (0.102)	0.053 (0.134)
distance3 (Distance of residence from secondary school)	–0.033 0.502)	–0.059 (0.147)
d₁ (dummy = 1, if household has computer access, and 0, if otherwise)	–0.250** (0.002)	–0.244** (0.002)
Constant	0.668** (0.000)	0.847** (0.000)
No. of observations	2536	2304
Pseudo R²	0.1238	0.1564

Source: Authors’ calculation.

Notes: Figures in the parentheses are the p-values.

** and *Significant at 1% and 5% level, respectively.

In Table 4, the results of the multinomial logit regression for reasons to drop out for the male children in both sectors are provided. The salient points that emerge from Table 4 are as follows:

For the ‘push factor’, all coefficients except ‘child_f’ in the rural sector and ‘child_f’ and ‘child_m’ in the urban sector are significant with meaningful signs. For a male child in the rural sector, belonging to an affluent household with larger number of male siblings and educated parents have negative impacts on the probability of choosing this option. In the urban sector, while the other implications remain the same, the number of siblings does not play any role. On the supply side, access to computers, going to non-public schools that provide midday meals, same language as the medium of instruction and that spoken at home, have negative impacts on the probability of choosing ‘push factor’ as a reason for dropout compared to ‘other factors’. The coefficients of ‘dropout_age’ and ‘dropout_grade’ are positive. That is, male children who drop out at a higher age or higher grade, have higher probability of choosing ‘push factor’ as a reason for dropout.

For the ‘pull factor’, all coefficients relating to household characteristics are significant in both sectors. On the supply side, except for d₂ (computer access) in the rural sector and ‘d_language’ in both sectors other coefficients are significant. The coefficients of ‘dropout_age’ and ‘dropout_grade’ are positive in both sectors. Thus, a male child, belonging to an affluent household with smaller number of siblings and educated parents is less likely to choose this option in the rural sector. In the urban sector, a male child, belonging to an affluent household with smaller number of male siblings, higher number of female siblings and educated parents is less likely to choose this option. This is possibly because the female siblings take care of the household activities. On the supply side, language does not play any role in either sector. Access to computers has no role in the rural sector, while in the urban sector male children having computer access are less likely to choose this option. Finally, probability of choosing this option is higher for children going to public schools with no midday meal in both sectors. Thus, for children choosing the ‘pull factor’ as a reason for dropout, the school infrastructure is not adequate to motivate them to study and they are left to be engaged in domestic activities by household level requirement/pressure.

For the ‘opted out’ option, in the rural sector, the significant variables are household income, parents’ education (with plausible negative coefficients) on the household side and most of the supply side variables (except ‘language’). In other words, for a male child, going to public schools that provide midday meals, have negative impacts on the probability of choosing ‘opted out factor’ as a reason for dropout compared to ‘other factors’. In the urban sector, in addition to the above household factors, having larger number of male siblings has a negative impact. On the supply side, computer access and provision of midday meal have negative impacts on the probability of choosing this option, while going to public schools has a positive impact in the rural sector. In the urban sector, in addition to the above factors, different language for medium of instruction at school and that spoken at home has a positive impact.

Table 4.

Reasons for Male Children Dropout (multinomial logit model)

Explanatory Variables	Reasons for Dropping Out (base outcome: other)
	Rural			Urban
	Push	Pull	Opted Out	Push	Pull	Opted Out
Log(MPCE)	–1.961*** (0.006)	–0.958** (0.012)	–1.531*** (0.009)	–0.575** (0.025)	–0.014** (0.012)	–0.041*** (0.005)
child_m	–0.092*** (0.005)	0.182** (0.022)	–0.140 (0.301)	0.211 (0.145)	0.380*** (0.001)	–0.592*** (0.001)
child_f	0.101 (0.127)	0.093*** (0.002)	0.183 (0.711)	–0.594 (0.102)	–0.495*** (0.008)	0.478 (0.109)
Father_edu	–0.084*** (0.001)	–0.057** (0.031)	–0.012** (0.031)	–0.001*** (0.007)	–0.018*** (0.002)	–0.133*** (0.001)
Mother_edu	–0.035** (0.011)	–0.079*** (0.001)	–0.078*** (0.001)	–0.058*** (0.002)	–0.102*** (0.001)	–0.028*** (0.001)
d₂ (=1, if child has computer access, and 0, if otherwise)	–0.217*** (0.005)	–0.109 (0.101)	–0.331*** (0.001)	–1.201** (0.021)	–0.336** (0.015)	–0.875** (0.021)
d_Inst (=1, if public school, and 0, if otherwise)	0.279*** (0.001)	1.753* (0.081)	2.805*** (0.001)	1.096*** (0.002)	0.864*** (0.002)	0.687*** (0.001)
d_language (=1, if same language in school and home, and 0, if otherwise)	–0.732*** (0.002)	0.203 (0.501)	0.084 (0.778)	–1.647*** (0.001)	–0.090 (0.154)	–1.092*** (0.009)
d_midday (=1, if midday meal provided, and 0, if otherwise)	–0.485** (0.011)	–0.885** (0.011)	–1.091* (0.098)	–1.819*** (0.002)	–0.136** (0.011)	–0.732*** (0.001)
dropout_age	0.147*** (0.001)	0.075*** (0.001)	0.022*** (0.005)	0.292** (0.021)	0.051*** (0.002)	–0.167*** (0.001)
dropout_grade	0.140*** (0.001)	0.119*** (0.001)	0.064*** (0.001)	0.432*** (0.003)	0.024*** (0.002)	–0.041*** (0.001)
Constant	1.664*** (0.001)	–0.615*** (0.001)	1.044*** (0.001)	–0.125*** (0.031)	0.029*** (0.001)	0.274*** (0.001)
No. of observations		996			689
Pseudo R²		0.1764			0.1545

Source: Authors’ calculation.

Notes: Figures in the parentheses are the p-values.

***, ** and *Significant at 1%, 5% and 10% level, respectively.

In Table 5 the results of the multinomial logit regression for reasons to drop out for female children in both sectors are provided. The values of the pseudo R² are 0.1925 and 0.1894 in the rural and the urban sector, respectively. From Table 5 the following observations are in order:

For the ‘push factor’, all coefficients are significant with meaningful signs in both sectors. To be specific, for a female child, belonging to an affluent household with large number of siblings and educated parents have negative impacts on the probability of choosing this option. On the supply side, access to computers, going to public schools that provide midday meals, same language as the medium of instruction and that spoken at home, have negative impacts on the probability of choosing ‘push factor’ as a reason for dropout compared to ‘other factors’. The coefficient of ‘dropout_age’ is positive, while that for ‘dropout_grade’ is negative. That is, female children who drop out at a higher age, have higher probability of choosing ‘push factor’ as a reason for dropout, but female children who drop out at a higher grade have lower probability of choosing ‘push factor’ as a reason for dropout. Higher grade is attained at higher age but not vice-versa. It has been observed that students face the difficulty in continuing study after repeating in a class. It creates a psychological barrier for a child to study further in class with younger fellows (Chugh, 2011).

For the ‘pull factor’, all coefficients are significant in both sectors, except for d₂ (computer access) and ‘dropout_age’ in the urban sector. Except for ‘child_m’, ‘child_f’ and ‘dropout_grade’, all other variables have similar impact on the probability of choosing this option as in case of the ‘push factor’. Here larger number of siblings have a positive impact. It is also interesting to note that female children who drop out at a higher grade have higher probability of choosing ‘pull factor’ (as opposed to the case in ‘push factor’) as a reason for dropout. These point to the fact that female children are forced to drop out not because they are incapable or are unable to cope with studies, but because they have to be possibly engaged in domestic/economic activities and take care of the siblings.

For the ‘opted out’ option, in the rural sector, the significant variables are parents’ education (with plausible negative coefficients) on the household side and most of the supply side variables. In other words, going to public schools that provide midday meals, same language as the medium of instruction and that spoken at home, have negative impacts on the probability of choosing ‘opted out factor’ as a reason for dropout compared to ‘other factors’. In the urban sector, for a female child, belonging to an affluent household with large number of male siblings and educated father have negative impacts on the probability of choosing this option. However, lager number of female siblings and mother’s education have a positive impact. One explanation could be that with a larger number of female children the mother decides that the girl child has had enough education. The positive impacts of ‘dropout_age’ and ‘dropout_grade’ in both sectors seem plausible for the choice for this option.

For ‘female specific factors’, on the household side in the rural sector, for a female child, belonging to an affluent household with smaller number of siblings and educated parents have negative impacts on the probability of choosing this option. On the supply side, only the fact that the child goes to a public school, has a positive impact on the probability of choosing this option. This points to the fact that the schools possibly suffer from improper toilet facilities or lack of female teachers. In the urban sector, belonging to an affluent household with small number of female siblings and educated parents have negative impacts on the probability of choosing this option. On the supply side, computer access has a negative impact and public school (providing midday meals, but not having adequate female specific infrastructure) have positive impact on probability of choosing this option. The positive impacts of ‘dropout_age’ and ‘dropout_grade’ in both sectors seem plausible for the choice for this option. Thus, ‘inadequate female specific infrastructure’ seems to turn out as a major factor that contributes to female children dropout and this applies generally to children from poorer families, as children from richer families do not go to such schools.

Table 5.

Reasons for Female Children Dropout (multinomial logit model)

Explanatory Variables	Reasons for Dropping Out (base outcome: other)
	Rural				Urban
	Push	Pull	Opted Out	Female Specific	Push	Pull	Opted Out	Female Specific
Log(MPCE)	–0.488*** (0.001)	–0.056*** (0.009)	0.007 (0.129)	–0.885*** (0.000)	–0.595** (0.011)	–0.156*** (0.001)	–0.307*** (0.005)	–0.665** (0.012)
child_m	–0.454*** (0.003)	0.349*** (0.001)	–0.532 (0.801)	0.404*** (0.008)	–1.048*** (0.001)	0.462** (0.034)	–0.414*** (0.001)	–0.417 (0.608)
child_f	–0.509*** (0.001)	0.503*** (0.006)	0.503 (0.651)	0.532*** (0.001)	–0.667*** (0.000)	0.474** (0.011)	0.451** (0.018)	0.370*** (0.001)
Father_edu	–0.005** (0.029)	–0.008** (0.011)	–0.027*** (0.001)	–0.011** (0.031)	–0.151*** (0.000)	–0.054** (0.012)	–0.036** (0.013)	–0.001*** (0.001)
Mother_edu	–0.097*** (0.009)	–0.112** (0.021)	–0.106*** (0.001)	–0.177*** (0.006)	–0.146*** (0.005)	–0.096*** (0.004)	0.042*** (0.001)	–0.069*** (0.005)
d₂ (=1, if child has computer access, and 0, if otherwise)	–0.076*** (0.002)	–0.098*** (0.007)	–0.652 (0.809)	1.009 (0.109)	–0.807*** (0.001)	0.089 (0.801)	0.093 (0.301)	–1.237*** (0.002)
d_Inst (=1, if public school, and 0, if otherwise)	0.008*** (0.009)	0.235*** (0.001)	0.045* (0.076)	1.205*** (0.001)	0.103*** (0.002)	0.054*** (0.001)	0.706*** (0.001)	0.034*** (0.001)
d_language (=1, if same language in school and home, and 0, if otherwise)	–0.514*** (0.008)	–0.175*** (0.001)	–0.394*** (0.001)	–0.487 (0.269)	–0.550*** (0.003)	–0.362*** (0.002)	–0.858*** (0.001)	–0.819 (0.507)
d_midday (=1, if midday meal provided, and 0, if otherwise)	–0.074*** (0.001)	–0.705** (0.011)	–0.183*** (0.001)	0.694 (0.451)	–0.114** (0.010)	–0.317** (0.015)	0.704 (0.191)	0.564* (0.091)
dropout_age	0.231*** (0.005)	0.054*** (0.002)	0.083*** (0.008)	0.272*** (0.002)	0.292*** (0.004)	–0.001 (0.998)	0.226** (0.026)	0.134*** (0.001)
dropout_grade	–0.137* (0.091)	0.023*** (0.001)	0.017** (0.031)	0.095*** (0.001)	0.113*** (0.001)	0.078** (0.021)	0.259*** (0.001)	0.341*** (0.001)
Constant	–0.905*** (0.001)	1.004*** (0.001)	1.067*** (0.001)	0.092*** (0.001)	1.055*** (0.001)	1.009** (0.023)	0.089** (0.043)	1.053*** (0.001)
No. of observations	972				743
Pseudo R²	0.1925				0.1894

Source: Authors’ calculation.

Notes: Figures in the parentheses are the p-values.

***, ** and *Significant at 1%, 5% and 10% level, respectively.

Concluding Remarks

In this article, the drop out behaviour of children (male and female) is studied both from the point of view of households and from that of children. The 71st round (2014) data on education expenditure, provided by the NSSO, for the state of West Bengal are used in this study.

In terms of incidence of dropout, children from affluent families and educated parents are less likely to drop out in both the rural and the urban sector. However, belonging to a large family and having many younger siblings negatively affect school participation in the rural sector, while for urban children these variables do not make any difference.

From the children’s perspective, for boys who drop out, reporting about ‘push factor’, ‘pull factor’ or ‘opted out’ are less probable compared to ‘others’ if they belong to affluent families and have educated parents. They are also less likely to report these factors if the school is a non-public school providing midday meals.

The same implications hold for girls with similar family background, in terms of the above choices. The additional choice, viz. ‘female specific reasons’ is also less likely, compared to ‘other’ factors, as these families are generally liberal (in terms of marrying off a girl at an early age) and these girls generally go to schools with adequate infrastructure. This is corroborated by the fact that ‘going to a public school’ has a positive impact on the probability of choosing this option, and children from poorer section of the society go to public schools, in general. It is interesting to note that mid-day meal plays no significant role in reporting ‘female specific reasons’. Thus, public school infrastructure emerges as a major factor contributing to female dropout because of ‘female specific reasons’. Given that infrastructure includes ‘adequate toilet facilities’, which are lacking in most public schools,¹¹ this study contributes to the justification for the demand of proper toilet facilities for female children at school in West Bengal.

While the conclusions about the importance of economic variables and parental education are similar to those established in the literature, it is the analysis of the children’s perspectives through factors like ‘push factor’ and ‘pull factor’ that contributes to the literature of drop out behaviour from a different angle. To be able to resolve a problem like drop out, it is important for a policymaker to have a complete picture of the facts from various aspects and opinions of all concerned with this issue.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Appendix

Table A2.

Explanation of the Explanatory Variables

Explanatory Variables	Explanation of the Variables
Log(MPCE)	Log of the monthly per capita expenditure
[Log(MPCE)]²	Squared value of log(MPCE)
hhsize	Household size
child_m	Fraction of the household size of below age of 16 years (male)
child_f	Fraction of the household size of below age of 16 years (female)
Father_edu	Education level of father
Mother_edu	Education level of mother
hdage	Age of the household head
distance1	Distance from primary school to the residence
distance2	Distance from upper primary school to the residence
distance3	Distance from secondary school to the residence
d ₁	d₁ = 1, if the household has access to computer, and =0, if otherwise
d ₂	d₂ = 1, if the child has access to computer, and =0, if otherwise
d_Inst	d_Inst = 1, if the school is government school, and =0, if otherwise
d_language	d_language = 1, if the languages used in the school and in the house are same, and =0, if otherwise
d_midday	d_midday = 1, if the school provides midday meal, and =0, if otherwise
dropout_age	Age when the child dropped out
dropout_grade	Grade when the child dropped out

Source: The 71st Round Schedule of NSSO.

Notes

References

Bradley

C. L.

, & Renzulli

L. A.

(2011). The complexity of non-completion: Being pushed or pulled to drop out of high school. Social Forces, 90(2), 521–545.

Brown

J. D.

(1988). Understanding research in second language learning: A teacher’s guide to statistics and research design. Cambridge University Press.

Chugh

(2011). Dropout in secondary education: A study of children living in slums of Delhi. NUEPA (Occasional Paper 37, pp. 1–53). National University of Educational Planning and Administration.

Dreze

, & Geeta

G. K.

(2001). School participation in rural India. Review of Development Economics, 5(1), 1–24.

Duc

L. T.

, & Tam

T. N. M.

(2013). Why children in Vietnam drop out of school and what they do after that (Vol. 102, Working Paper). Young Lives. http://www.younglives.org.uk/sites/www.younglives.org.uk/files/YL-WP102_Le%20Thuc%20Duc_why%20do%20children%20leave%20school%20early.pdf

Ghosh

, & Kar

A. M.

(2010). Child marriage in rural West Bengal: Status and challenges. Indian Journal of Development Research and Social Action, 6(1–2), 1–23.

Gouda

M. S.

, & Sekher

T. V.

(2014, November-December). Factors leading to school dropouts in India: An analysis of national family health survey-3 data. IOSR Journal of Research & Method in Education (IOSR-JRME), 4(6), 75–83.

Government of India (2014). School Education in India, National University of Educational Planning and Administration. Ministry of Human Resource Development Government of India, Department of School Education & Literacy.

Government of India (2016). Educational statistics at a glance. Ministry of Human Resource Development, Department of School Education & Literacy.

10.

Hati

K. K.

, & Majumder

(2012). Proximate determinants of school dropout: A study on rural West Bengal (MPRA Paper No. 45850). http://mpra.ub.uni-muenchen.de/45850/

11.

Mishra

, & Chatterjee

(2017). School dropout susceptibility mapping with fuzzy logic—A study in the district of Purulia, India. Human Geographies, 11(1), 111–128.

12.

Mukherjee

(2012). Schooling, child labor, and reserve army evidences from India. Journal of Developing Societies, 28(1), 1–29.

13.

Nithiya Amirtham

, & Kundupuzhakkal

(2017). Gender issues and dropout rates in India: Major barrier in providing education for all. Educationia Confab, 2(4), 226–233.

14.

Prakash

, Beattie

, Javalkar

, Bhattacharjee

, Ramanaik

, Thalinja

, & Collumbien

(2017). Correlates of school dropout and absenteeism among adolescent girls from marginalized community in North Karnataka, South India. Journal of Adolescence, 61, 64–76.

15.

Pratichi Trust Report (2017). Formative study to enhance the understanding about reasons for smooth transition among boys and girls to secondary schools: West Bengal Report. ERU Consultants Private Limited.

16.

Roy

(2013). A study on the dropout problem of primary education in Uttar Dinajpur district, West Bengal. https://ssrn.com/abstract=2601969,https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.2601969

17.

Rumberger

R. W.

(2001, January 13). Why students drop out of school and what can be done. Paper Prepared for the Conference, ‘Dropouts in America: How Severe is the Problem? What Do We Know About Intervention and Prevention?’ Harvard University.

18.

Stearns

, & Glennie

E. J.

(2006). When and why dropouts leave high school. Youth & Society, 38(1), 29–57.