|
Logo

Early childhood education in India – A possible investment in better outcomes? A quantitative analysis using Young Lives India

1 Leave a comment on paragraph 1 0 This paper explores the relationship between early childhood education and academic outcomes for children in India by estimating the ability of preschool participation at age 5 to predict results on major cognitive assessments at age 12.  Initially looking at differences in means, it moves on to utilise regression analysis first in an uncontrolled model, and then in a model which controls for both gender and maternal education, as these have been deemed important inputs for academic attainment in the wider literature on human capital development. The sample used for this research is constructed from Young Lives (India), which from 2002 to 2017 surveyed two cohorts of children across Andhra Pradesh and Telangana, with a pro-poor sampling strategy. Surprisingly, the results of the analysis find that participation in early childhood education had a negligible effect on test scores, even when controlling for gender and maternal education. Meanwhile, maternal education emerged as a strong predictor of test results. These findings contradict much of the existing evidence that demonstrates associations between early childhood education and cognitive development, and, in turn, improved economic outcomes. Accordingly, it raises questions about the generalisability of the existing evidence and the quality of India’s ECE offering. The premise, method and findings of this paper are divided into nine sections, including an introduction, an explanation of Human Capital as the paper’s conceptual framework, a literature review, an overview of the context of ECE in India, a section on the paper’s data and variables, a methods section, an overview of the results, a discussion, and conclusions.

2 Leave a comment on paragraph 2 0 Keywords:

3 Leave a comment on paragraph 3 0 Early Childhood Education, India, Longitudinal Research, Academic Outcomes, Poverty Alleviation

4 Leave a comment on paragraph 4 0 Introduction

5 Leave a comment on paragraph 5 0 Research Context and Foundation

6 Leave a comment on paragraph 6 0 Research over the past half-century has demonstrated that the period of early childhood is the most critical phase in human development, and that the foundational capacities established during this time can beget improved outcomes across the life course (Black et al., 2016; Shonkoff and Phillips, 2000). Access to early childhood care, health and education (ECCE)[1] has been shown to nurture these capacities by enabling children to achieve improved cognitive outcomes (Attanasio, Meghir and Nix, 2015; Engle et al., 2007). Numerous studies have linked preschool[2] participation to increased earnings and progress toward poverty alleviation (see Black et al., 2016; Engle et al., 2007; Heckman, 2006 and Shonkoff and Richter, 2013). The subsequent literature, including Goal 1 of the Dakar Framework for Action (Education for All), and Sustainable Development Goal 4.2[3], has crystallised early childhood education (ECE)[4] as a critical lever for economic outcomes. But much of this evidence comes from developed countries, which have very different economic, cultural, and political contexts than those of low-and-middle income (LAMI) countries (Levin and Schwartz, 2006). This raises the question of how applicable the existing evidence is to those contexts (Woodhead, 2009; Yoshikawa and Nieto, 2013). This paper explores the relationship between ECE and cognitive outcomes for a sample of children in India, to determine whether India’s ECE programmes are capitalising on this critical period and whether the nature of the relationship between these variables matches wider patterns.

7 Leave a comment on paragraph 7 0 To date, research on ECE in India has been limited, and the Indian government has acknowledged that its ECE services are not up to standard (Alcott et al., 2018; Government of India, 2013). But with India’s rapidly growing economy and extensive early years infrastructure (Alcott et al., 2018), it offers interesting parameters for the assessment of ECE’s role. This paper’s findings are intended to contribute to the limited evidence and offer analysis that could be useful in improving India’s ECE provision.

8 Leave a comment on paragraph 8 0 The primary aim of this paper is to estimate the ability of preschool participation to predict cognitive outcomes, as they have been widely linked to improved economic opportunities. However, researchers on Indian early education have pointed out the need to go beyond establishing the simple effects of ECE (Kaul et al., 2017). In order to respond to this suggestion, I also consider two factors which have been identified in the literature as important to India’s educational experience: gender[5] and maternal education.

9 Leave a comment on paragraph 9 0 Drawing from the body of evidence on ECE, I test the hypothesis that preschool participation should emerge as a good predictor of cognitive outcomes. This hypothesis is tested using Indian students who participated in early childhood education, from the Young Lives (YL) study[6]. Young Lives is a renowned study on childhood poverty[7] consisting of quantitative and qualitative layers, undertaken to inform effective policy and interventions for children (Early Childhood Development: informing policy and making it a priority, 2018; Vennam et al., 2009).

10 Leave a comment on paragraph 10 0 The indicators selected to represent ‘early childhood education’ and ‘cognitive development’ are preschool participation at age 5 and test score results on the Peabody Picture Vocabulary Test (PPVT) and mathematics assessment (taken from the Trends in International Mathematics and Science Study or TIMSS) at age 12. These types of tests are in line with widely used indicators of cognitive development from related studies (see Cunha et al., 2006; Singh and Mukherjee, 2018; Woldehanna and Gebremedhin, 2012).

11 Leave a comment on paragraph 11 0 The research questions addressed are as follows:

  • 12 Leave a comment on paragraph 12 1
  • Is early childhood education a good predictor of scores on the PPVT and maths tests?
  • Do boys who attended preschool fare better than girls on these tests?
  • Do children with educated mothers fare better on these tests?
  • Once gender and maternal education are controlled, is preschool a useful predictor of test scores?

13 Leave a comment on paragraph 13 0 These research questions called for the use of empirical data, recorded as categorical and continuous variables. Accordingly, my methodological approach necessitated quantitative analysis[8], for which I employed statistical modelling to estimate the relationships between these variables.

14 Leave a comment on paragraph 14 0 Human Capital as a Conceptual Framework

15 Leave a comment on paragraph 15 0 Early Investments and Future Returns

16 Leave a comment on paragraph 16 2 The human capital perspective on early childhood offers useful theoretical parameters for exploring the economic justification for investment in early childhood education. Human Capital Theory (HCT) posits that early investments in education boost overall educational attainment (Heckman, 2011), and that ultimately education generates future returns (Becker, 1964).HCT thus informs and underpins this paper with these two notions, and helps to connect it to the vast literature on early investments in human capital. It also provides the foundation for this paper’s hypothesis by theorising that ECE should contribute to improved academic outcomes.

17 Leave a comment on paragraph 17 0 The notion that education generates future returns represents the initial conception of education by human capital theorists. Originally, the theory posited that educational attainment was a measure of cognitive development, which is ultimately of interest because of its link to economic returns: education economists showed that cognitive ability[9] was an important determinant of labour market outcomes[10] (Heckman, 1995).

18 Leave a comment on paragraph 18 0 Meanwhile, the notion that early investments maximise academic attainment, explored particularly in the research of James Heckman and Flavio Cunha, was predicated upon the idea of a ‘time profile’ in which early investments offered a longer period to realise future returns (Becker, 1962 and 1964; Mincer, 1958). Cunha et al. demonstrated the value of looking at childhood as two distinct time periods, showing that the rate of return (RoR) to a dollar of investment made in early childhood was higher than for the same dollar invested later (2006).

19 Leave a comment on paragraph 19 0 Human Capital and Women’s Education

20 Leave a comment on paragraph 20 0 Whilst human capital theory does not offer much on gender and ECE specifically, its framework for gender and wider education is robust. The writings of Paul Schultz in particular advocated for the importance of educating women, especially as a strategy for poverty alleviation (1993), and served as the foundation for other work endorsing this instrumental view (see Herz and Sperling, 2016; King and Hill, 1993). Other major thinkers on human capital have also endorsed this perspective; Erik Hanushek (2008) highlighted that gender equality in education was a human capital investment with important economic outcomes, and Harry Patrinos (2008) undertook specific econometric methodologies to calculate the rates of return to women’s education. The implications from the related literature are that women’s education is instrumental to a range of intergenerational social benefits (Unterhalter, 2007). Accordingly, HCT offers a lens through which to interpret the potential impact of maternal education (encompassing both ‘gender’ and ‘maternal education’) as an input to desirable economic outcomes, which could complement or cloud the role of preschool in those same outcomes.

21 Leave a comment on paragraph 21 0 Literature Review

22 Leave a comment on paragraph 22 0 ECE and its Related Policy Context

23 Leave a comment on paragraph 23 0 The origination of early childhood development (ECD)[11] as an area of study was neurological research highlighting the rapid pace of brain development during the first years of life (Karoly et al., 1998; Young, 2007). The amalgamation of other evidence on child development from neuroscience, psychology, sociology and health collectively highlighted the importance of the early years for cognitive development, and shortlisted a number of critical inputs including education (Campbell et al., 2001; Heckman, 2011; Kohlberg, 1968 and Shonkoff and Phillips, 2000). A landmark study which helped to catalyse global policy engagement was an ECD-focused series in The Lancet, which quantified the loss of development potential and impact on long-term outcomes for children who lacked strong starts in education, nutrition and material stability. Consequently, ECD became tied to international agendas proposing human capital approaches to development. Many studies on ECD captured ‘development’ indicators through measures related to schooling, such as test scores or grade completion. Schooling-related development studies found consistently that early education helped to shape opportunities across the life course, including improving educational attainment, earnings and market competitiveness (Becker, 1993; Carneiro and Heckman, 2003; Cunha et al., 2006; Cunha and Heckman, 2007 and Heckman, 2011)[12]. Therefore, the Education for All initiative inscribed ECCE as its ‘bedrock’ (Strong Foundations: Early Childhood Care and Education, 2006). Additionally, under the Sustainable Development Goals (SDGs), investment in ECE was seen not only as tackling inequality but also poverty alleviation (Morabito, Vandenbroeck and Roose, 2013; Richter et al., 2016). In this way, ECE came to have its own significance for development agendas linked to poverty reduction and socioeconomic mobility (Nadeau et al., 2011).

24 Leave a comment on paragraph 24 0 Gaps in Evidence

25 Leave a comment on paragraph 25 0 Longitudinal studies from the US[13] have found ECE to be a successful predictor of both academic attainment and economic outcomes (Campbell and Ramey, 1994; Currie and Thomas, 1995). In particular, the Abecedarian Study (Campbell and Ramey, 1994) and the HighScope Perry Pre-School Project (Currie, 2001; Heckman, 2011) from the US and the Effective Provision of Preschool Education Study from the UK became widely cited because they involved experimental research. Grouped with these were the ongoing results of two landmark ECCE programmes, HeadStart, launched in the USA in the 1960s, and Sure Start, its later British counterpart in the 1990s (Woodhead, 2006). This group of studies confirmed that ECE improved cognitive abilities and served as a strong foundation for academic success (Black et al., 2016). They also concluded that the most effective time to invest in education to equalise initial differences[14] in endowments was the early years (Currie, 2001).

26 Leave a comment on paragraph 26 0 However, though comprehensive, the evidence being primarily from developed countries is potentially problematic: there has been a noted lack of evaluation of preschool programs in developing countries (Currie, 2001; Woldehanna and Gebremedhin, 2012)[15]. This gap in the literature points to the hegemony of existing evidence.[16] The implications of this are that international policies are being informed by findings from a narrow group of contexts.

27 Leave a comment on paragraph 27 0 ECE Literature from India

28 Leave a comment on paragraph 28 0 This evidence gap[17] extends to ECE research on India. Only a small body of evidence[18] on the association between preschool participation and developmental outcomes exists, but it is cross-sectional rather than experimental, and limited in examination of longitudinal effects. Most of it also covers only particular regions or states, and is therefore limited in generalisability. For example, Arora, Bharti and Mahajan (2006) were able to link preschool participation to higher cognitive development, but their sample was limited to urban slums in Jammu city. A few cover wider geographies but are stock-taking studies, such as CECED’s 2013 review (Indian Ministry of Women and Child Development, 2013). A small number show the effect of preschool on primary school outcomes such as retention or school readiness (Kaul et al., 1993; NCERT, 1996). Others examine the comparative effects of government versus private pre-primary[19]. But overall, the research is still limited, with longitudinal and more complex evidence particularly scarce (Kaul et al., 2017). Aside from the comprehensive (but geographically limited) data from YL India, the only other sizeable evidence on ECE is from the ASER Centre, which publishes an ‘Annual Status of Education’ report (Early Childhood Development: informing policy and making it a priority, 2018) and, most recently, the India Early Childhood Education Impact (IECEI) study[20].

29 Leave a comment on paragraph 29 0 Perhaps closest to my research is the recent study by Singh and Mukherjee (2018), which takes YL data to examine the effect of private preschool on cognitive achievement and subjective wellbeing at age 12, but uses propensity score matching where this paper utilises OLS regression. While some of the findings overlap, their paper looks only at private ECE provision, and is therefore not able to comment on ECE in India more widely.

30 Leave a comment on paragraph 30 0 Evidence on Gender and Maternal Education

31 Leave a comment on paragraph 31 0 Historically, gender has been a regular predictor of disparity in educational and economic outcomes in India (Asadullah and Yalonetzky, 2012; Boserup, 1970 and Kaul et al., 2017). But other evidence suggests that this disparity is lessening; official statistics show almost identical enrolment for boys and girls in ECE and primary (“Education Statistics – All Indicators”, 2018), and in some studies, gender is no longer found a significant contributor to academic outcomes (Streuli, Vennam and Woodhead, 2011; Vennam et al., 2009). But other evidence argues that gender may affect the extent to which Indian children benefit from ECE (Garcia, Heckman and Ziff, 2018; Magnuson et al., 2016). Therefore, evidence is mixed[21]. I have consequently included gender as a control variable within this study, as further research may help to clarify.

32 Leave a comment on paragraph 32 0 Tied to the notion of gender is that of maternal education. There is a significant body of literature that demonstrates the intergenerational persistence of economic status[22]. Mothers’ education in particular has been associated with higher earnings and better educational outcomes for children (Aakvik et al., 2003; Rosenzweig and Wolpin, 1994). Studies using data from YL India have also shown association between maternal education and the completion of secondary school, which itself is linked to improved economic opportunities in other research (Singh and Mukherjee, 2015). This body of evidence supports the idea that maternal education has a role to play in the intergenerational transmission of human capital (Galab, Reddy and Himaz, 2008; Richter et al., 2016). It has therefore been included as a variable of interest within this paper.

33 Leave a comment on paragraph 33 0 Context of ECCE in India

34 Leave a comment on paragraph 34 0 Overview of Policies and Infrastructure

35 Leave a comment on paragraph 35 0 The provision of early childhood services in India has long been conceptualised as an investment in human capital (Mohite and Bhatt, 2008; Streuli, Vennam and Woodhead, 2011). This notion of ECCE informed India’s National Policy for Children and its launch in 1974 of the Integrated Child Development Services (ICDS), today the world’s largest publicly funded early childhood system. Through the ICDS, India provides universal access to health, nutrition and education services (Alcott et al., forthcoming; Indian Ministry of Women and Child Development, 2013). The operating infrastructure for these services includes 1.3 million anganwadi centres (preschools), which served over 104.5 million beneficiaries in 2014, from expectant mothers to children (Kaul et al., 2017; Richter et al., 2016).

36 Leave a comment on paragraph 36 0 But pre-primary enrolment in India is still only at 12.9% despite the sizeable infrastructure (“Education Statistics – All Indicators”, 2018). Moreover, evidence shows that the anganwadis are poor providers of early childhood education, due to insufficient teacher training, substandard facilities and lack of regulation[23] (Rao and Kaul, 2017; Singh and Mukherjee, 2018). Poor quality pre-primary provision has precipitated widespread lack of school readiness: evidence indicates that pre-literacy and pre-numeracy skills at age 5 are vastly below expected levels (Beyond Basics, 2018; Save the Children, 2009). As a result, disillusioned parents seek other options such as low-cost private preschools[24] (Alcott et al., forthcoming; Kaul et al., 2017 and Streuli, Vennam and Woodhead, 2011).

37 Leave a comment on paragraph 37 0 The Indian government has made improving ECE a priority by strengthening the policy framework; India’s twelfth Five-Year Plan (2012-2017) shifted attention toward early education (Singh and Mukherjee, 2017; Streuli, Vennam and Woodhead, 2011). Its 2013 National Early Childhood Care and Education Policy was specifically to expand the educational component of the ICDS, coupled with the National Curriculum Framework and Quality Standards for ECE (Kaul et al., 2017). However, while these demonstrate the political will to enhance the ICDS, the system still suffers from lack of resources, direction, and governance (Richter et al., 2016). Consequently, there remains a large gap between policy and practice, with lack of research on the specificities impeding the ability to optimise. (Alcott et al., forthcoming).

38 Leave a comment on paragraph 38 0 The Context of Andhra Pradesh

39 Leave a comment on paragraph 39 0 The state of Andhra Pradesh has a complex set of challenges[25]. It has a long-established government-run ECCE system but also a growing trend of private schooling (Asadullah and Yalonetzky, 2012). Yet students fall vastly below the expected levels of attainment (Singh and Mukherjee, 2016).

40 Leave a comment on paragraph 40 0 In many cases, children in Andhra Pradesh are first generation learners; this was the case for more children than not in the YL India sample (Streuli, Vennam and Woodhead, 2011). This has an impact on maternal education’s ability to improve outcomes for this generation. Secondly, the rapid growth of private schools is attracting parents with English-medium teaching, resonating with wider national patterns. The government of Andhra Pradesh is therefore under pressure to compete, and is now moving toward English-medium instruction in some secondary schools (Streuli, Vennam and Woodhead, 2011).

41 Leave a comment on paragraph 41 0 Overall, educational disparity in Andhra Pradesh is decreasing (Asadullah and Yalonetzky, 2012). However, the trends discussed above show that education in Andhra Pradesh is characterised by differentiation of preschool experiences by wealth status, location, gender, and parental education (Streuli, Vennam and Woodhead, 2011). These provide an important backdrop for this paper and justification for the examination of gender and maternal education’s roles.

42 Leave a comment on paragraph 42 0 Data

43 Leave a comment on paragraph 43 0 Construction of the Sample

44 Leave a comment on paragraph 44 0 This paper draws from the YL study in India, which surveyed families from undivided Andhra Pradesh (later Andhra Pradesh and Telangana). The sample used is the younger cohort[26], comprised of 2,000 children aged 1 at the start of the survey in 2002[27]. Within this sample, only children who participated in both Rounds 2 and 4 were included, as critical data was needed from both rounds. Finally, my sample included only children for whom responses to the key predictor variable of preschool participation were recorded[28]. Table 1 offers descriptive statistics on all variables.

45 Leave a comment on paragraph 45 0 Table 1: Frequencies and Percent of Sample for Each Input Variable

46 Leave a comment on paragraph 46 0 Source: Young Lives data, Round 2

47 Leave a comment on paragraph 47 0 Variable Selection and Description

48 Leave a comment on paragraph 48 0 Early Childhood Education (ECE)

49 Leave a comment on paragraph 49 0 The core concept of this paper is early childhood education. It is therefore the primary explanatory variable of this study[29]. In the YL surveys, parents were asked whether their child was attending preschool. ‘Preschool’ was thus encoded as a binary variable for this study[30]. In my sample, 56.8% of children were attending preschool during Round 2, when they were approximately aged 5.

50 Leave a comment on paragraph 50 0 It is important to note that local educational policies have different ages of entry for primary school, sometimes leading to early entry or nonlinear trajectories between preschool and primary.[31] This duality of eligibility posed a challenge to the estimation of this variable, but as it proved difficult to calculate or minimise, responses were taken at face value.

51 Leave a comment on paragraph 51 0 Cognitive Outcomes

52 Leave a comment on paragraph 52 0 Cognitive outcomes, as measured by academic performance, were selected as the main outcome variable for this study because of their strong link within the literature to desirable socioeconomic returns. Academic performance is therefore intended to be interpreted as a medium-term indicator of whether the return on investment in ECE is likely to be favourable.

53 Leave a comment on paragraph 53 0 Whilst there are a variety of ways of quantifying academic performance, test scores are most common. Most often, these are children’s scores on comprehension, reading or maths assessments[32]. In the case of YL, results on only a small selection of tests are recorded, of which I used two: one verbal reasoning test (PPVT) and one maths test (Attanasio, Meghir and Nix, 2015). Scores from two tests were included in this paper’s model in order to offer a broader scope of analysis and comparison. Amongst this sample, children’s mean score on the PPVT was 75.8% (SD = 13.61%), and mean score on the maths test was 44.6% (SD = 22.7%).

54 Leave a comment on paragraph 54 0 Gender

55 Leave a comment on paragraph 55 0 Literature suggests that ECCE interventions have the potential to compensate for the gender biases that have historically affected Indian education systems (Magnuson et al., 2016; Morabito, Vandenbroeck and Roose, 2013). Consequently, I examined whether gender was associated with preschool attendance for this sample[33], as well as whether it was associated with variation in score results. I also included gender as a control variable in the multivariate OLS regressions to determine whether it made up any part of preschool’s ability to predict results on the PPVT and maths tests.

56 Leave a comment on paragraph 56 0 Maternal Education

57 Leave a comment on paragraph 57 0 Human capital research has positioned women’s education as an intergenerational asset, by which parents’ education[34] is a determinant of their children’s cognitive development[35]. In research from India specifically, mother’s education has been shown to be positively associated with children’s educational outcomes (Jeong, Kim and Subramanian, 2018). Taking the lead of other research such as the IECEI, which grouped mother’s education into larger brackets (Kaul et al., 2017), I recoded this variable as binary[36] for the purposes of this study, with mothers falling either into having ‘some’ education, meaning anything including primary and upwards, or ‘none’[37].

58 Leave a comment on paragraph 58 0 Statistical Methods

59 Leave a comment on paragraph 59 0 I selected three methodologies based upon precedent[38], constraints from the data and possibilities for further research. Firstly, I compared mean test scores of the major groups (preschool attendance, gender and maternal education level). This offered an initial picture of any significant differences. Secondly, I conducted linear regression as an uncontrolled estimation of preschool’s ability to predict score outcomes. Thirdly, I utilised multivariate ordinary least squares (OLS) regression to estimate preschool’s ability to predict score outcomes, but with gender and mother’s education as conditions.

60 Leave a comment on paragraph 60 0 My bivariate analysis included Chi-square and t-tests[39], as well as looking at resulting values for r, r² and other measures of effect size. I used Chi-square tests to determine whether gender and maternal education covaried with preschool participation[40]. T-tests were used to determine whether the independent variables (preschool participation, and subsequently gender and maternal education) were associated with test score outcomes.

61 Leave a comment on paragraph 61 0 My multiple regression utilised OLS  to estimate coefficients[41]. The method of least squares was used to estimate the parameters associated with each of the explanatory variables. Each model was then assessed for its goodness of fit.

62 Leave a comment on paragraph 62 0 Whilst the majority of assumptions for linear regression were met[42], endogeneity in the form of omitted variable bias presented a sustained issue, affecting the ability to interpret results as causal (Schonemann and Steiger, 1976). This paper included only two variables as controls, maternal education and gender. But literature on ECE highlights socioeconomic status (SES) as an important predictor variable[43] of both preschool participation and academic attainment (Woodhead, 2009). Therefore, the absence of SES within the model could have confounded results[44].

63 Leave a comment on paragraph 63 1 One way to reduce bias would be to build a multivariate model that includes a greater selection of predictor variables, which could improve the validity of the results by showing, through partial correlations, which coefficients contribute the most to the model, as well as improving the model overall[45]. However, such a model would still not indicate causality, and might be overly cumbersome. I therefore selected only two variables which past research has deemed important specifically for the intergenerational transmission of human capital.

64 Leave a comment on paragraph 64 0 Results

65 Leave a comment on paragraph 65 0 The goal of the regression analysis was to estimate the role of preschool in predicting test score outcomes of children in Andhra Pradesh at age 12. The findings show that, for this sample, preschool participation and gender were ineffective predictors of test score outcomes, but maternal education emerged as a significant predictor of test scores.

66 Leave a comment on paragraph 66 0 Was Early Childhood Education a Good Predictor of Test Scores?

67 Leave a comment on paragraph 67 0 Initial results indicated that preschool participation did not have a statistically significant impact on PPVT scores, but did on maths scores, yet with only a small effect size. Figures 1a and 1b below depict mean scores on each test by children who did and did not attend preschool.

68 Leave a comment on paragraph 68 0 Figure 1a: Comparison of mean percent score on Maths test by preschool attendance (L)

69 Leave a comment on paragraph 69 0 Figure 1b: Comparison of mean percent score on PPVT by preschool attendance (R)

70 Leave a comment on paragraph 70 0 Source: Young Lives India, Rounds 2 and 4

71 Leave a comment on paragraph 71 0 These results contradict the existing body of evidence on ECE. Possible explanations could have to do with bias in the model due to endogeneity, or with the homogenisation of ‘preschool’ as a single ‘treatment’ rather than a further disaggregated variable.  Table 2 below summarises the findings.

72 Leave a comment on paragraph 72 0 Table 2:Summary of Test Score Means Compared by Preschool Attendance

73 Leave a comment on paragraph 73 0 Source: Young Lives India, Rounds 2 and 4

74 Leave a comment on paragraph 74 0 Did Boys Who Attended Preschool Fare Better Than Girls on These Tests?

75 Leave a comment on paragraph 75 0 Subgroup analysis by gender revealed that boys were not more likely than girls to score higher on either the PPVT or maths test. Additionally, differences in preschool participation rates for boys and girls were not significant. This confirms analysis from YL India conducted by Streuli, Vennam and Woodhead (2011) and also mirrors World Bank data which indicates that pre-primary enrolment across India is now essentially on par (“Education Statistics – All Indicators”, 2018). Table 3 below summarises the findings from the gender subgroup analysis, and Figures 2a and 2b below illustrate that there was no significant difference between mean test scores.

76 Leave a comment on paragraph 76 0 Table 3:Summary of Test Score Means Compared by Gender

77 Leave a comment on paragraph 77 0 Source: Young Lives India, Rounds 2 and 4

78 Leave a comment on paragraph 78 0 Figure 2a: Comparison of mean percent score on Maths test by gender (L)

79 Leave a comment on paragraph 79 0 Figure 2b: Comparison of mean percent score on PPVT by gender (R)

80 Leave a comment on paragraph 80 0 Source: Young Lives India, Rounds 2 and 4

81 Leave a comment on paragraph 81 0 Did Children With Educated Mothers Fare Better on These Tests?

82 Leave a comment on paragraph 82 0 As illustrated by Figures 3a and 3b below, analysis found statistically significant differences between test scores for children whose mothers were educated and those whose were not[46]. The results are summarised in Table 4. Effect sizes as measured by r also showed a medium effect.

83 Leave a comment on paragraph 83 0 Figure 3a: Comparison of mean percent score on Maths test by mother’s education (L)

84 Leave a comment on paragraph 84 0 Figure 3b: Comparison of mean percent score on PPVT test by mother’s education (R)

85 Leave a comment on paragraph 85 0 Source: Young Lives India, Rounds 2 and 4

86 Leave a comment on paragraph 86 0 Table 4: Summary of Test Score Means Compared by Maternal Education

87 Leave a comment on paragraph 87 0 Source: Young Lives India, Rounds 2 and 4

88 Leave a comment on paragraph 88 0 Results also revealed that the proportion of children that were not attending preschool was significantly higher amongst families with uneducated mothers. Mothers with some education were also more likely to have children attending preschool than not. 

89 Leave a comment on paragraph 89 0 Once Gender and Maternal Education Were Controlled, Was Preschool a Useful Predictor of Test Scores?

90 Leave a comment on paragraph 90 0 The results of the multivariate regression indicated that gender and preschool did not offer much value as predictors, but maternal education did. The models for predicting PPVT and maths scores are summarised in Table 5 below:

91 Leave a comment on paragraph 91 0 Table 5: Multivariate Model Summaries for Predicting Maths and PPVT Scores

92 Leave a comment on paragraph 92 0 Source: Young Lives India, Rounds 2 and 4

93 Leave a comment on paragraph 93 0 Summary of PPVT Model Findings

94 Leave a comment on paragraph 94 0 The PPVT model was found to be a ‘good fit’ for estimating test score outcomes. However, surprisingly, the model showed that preschool attendance had a negative, statistically significant coefficient in the conditional relationship. This is best interpreted as a ‘negligible’ finding. Meanwhile, gender was shown to have a slightly negative coefficient in the PPVT model, but the significance value of the t-test indicated that it was not a significant contributor,  confirming the results of the bivariate analysis. Lastly, the model’s results showed that mother’s education had a highly significant correlation (p = .000) with test outcomes for the PPVT. In fact, it was the only variable in this model with a positive part correlation (.279). Therefore, though the model as a whole was found to be a good method of estimation, the goodness of fit came primarily from the coefficient for maternal education.

95 Leave a comment on paragraph 95 0 Summary of Maths Test Model Findings

96 Leave a comment on paragraph 96 0 In the bivariate analysis, linear regression showed a statistically significant association between preschool participation and maths test results. The findings of that analysis were consistent with those of Singh and Mukherjee, who also found that preschool was associated with improved maths scores (2017). However, results from the multivariate model showed that the conditional association (with gender and maternal education controlled) was no longer statistically significant. As with the PPVT model, maternal education provided the bulk of the model’s ability to predict maths score outcomes. Though the F-test results found this model to be a ‘good predictor’ of maths scores, it is important to note that the r-squared value demonstrates that the model could only account for 11% of the variance in score results, and also that the original research question was about whether ECE (not the other conditional variables) was a useful way to estimate score outcomes, which this model indicates it was not.

97 Leave a comment on paragraph 97 0 Conclusion of Results

98 Leave a comment on paragraph 98 0 Overall, though both models were ‘good fits’, the findings did not support the original hypothesis that ECE would be a good predictor of test score outcomes. Correlation between preschool and maths results was found significant at the bivariate level; however, once gender and maternal education were conditioned out, the coefficient was no longer significant. The correlation coefficients themselves were small, indicating that interpreting a statistically ‘strong’ association should be done with caution, even when p-values are very small (<.001). Effect sizes were negligible in some cases, and only small-to-medium in others. R² values showed that no variation in PPVT outcomes and only 4% in maths outcomes could be explained by preschool participation, meaning that variance in outcomes is likely to be better explained, or further explained, by other predictor variables not included in the model. Lastly, the lion’s share of the ‘predicting’ comes from the wrong variable – mother’s education. These results contradict copious evidence supporting preschool’s positive effect on academic attainment.

99 Leave a comment on paragraph 99 0 Discussion

100 Leave a comment on paragraph 100 0 Three major themes emerged from this paper’s findings. The first was that preschool was not a valuable predictor of test scores amongst this sample, supporting the idea that the Indian evidence base for ECE needs further substantiation[47] in order to improve services and policies. The second was that maternal education has significant value as a predictor of academic outcomes, confirming other evidence[48] of its role in intergenerational educational attainment. The third was that gender disparity appears to be reducing in scope. These themes help to substantiate the evidence base on early education in India and clarify areas of progress and residual challenges.

101 Leave a comment on paragraph 101 0 Preschool Was Not a Good Predictor of Academic Outcomes

102 Leave a comment on paragraph 102 0 Surprisingly, the findings indicated that children amongst this sample who participated in ECE did not achieve well academically. These findings resonate with only a limited body of existing evidence. For example, the significant association between preschool and maths in the bivariate analysis was consistent with other YL research in Peru, Ethiopia and Vietnam (Early Childhood Development: informing policy and making it a priority, 2018). The overall conclusion that preschool did not ‘make a difference’ mirrors other results on ECE in India reported by Chopra (2012), Pattnaik (1996) and Manhas and Qadiri (2010). Reports from ASER (Beyond Basics, 2018) and Save the Children (2009) also reported low attainment on pre-literacy and pre-numeracy skills for 5-year-olds, which indicates inadequate preschooling. However, the results contradict the majority of evidence on ECE.

103 Leave a comment on paragraph 103 0 Two possibilities emerge as to why these findings do not match the patterns. Firstly, problems could have arisen with the transition from data collection to variable construction which obscured important indicators. Secondly, the wider evidence body itself might not be heterogeneous enough to account for variation in patterns on ECE.

104 Leave a comment on paragraph 104 0 Recapitulating the first reason, it is possible that problems with construction of the primary explanatory variable obscured variation because responses to ‘is the child currently attending preschool?’ were not readily disaggregated by private versus public. Given evidence that preschool type has been associated with variations in attainment (Alcott and Rose, 2015; Beyond Basics, 2018; Chopra, 2012 and Singh and Mukherjee, 2015 and 2018), this would have offered a valuable condition[49].

105 Leave a comment on paragraph 105 0 Additionally, it is possible that families’ responses to the question used for variable construction (either a ‘yes’ or ‘no’ to ‘is child attending preschool?’) obscured early entry to primary after attending ECE by the time Round 2 took place (see Alcott et al., forthcoming)[50].

106 Leave a comment on paragraph 106 0 A useful methodological approach to combat the potential threat of inconclusive findings based on these possibilities could therefore entail conducting regression analysis that disaggregates preschool experience by type (as in the research conducted by Singh and Mukherjee, 2018), or also by other determinants of quality. Such research might reveal under what circumstances, or in what conditions, ECE in India matches wider patterns of positive association with academic outcomes.

107 Leave a comment on paragraph 107 0 Recapitulating the second reason, it is possible that this is the case because wider patterns are not diverse enough. Literature authenticating this possibility comes from Woodhead (2006), who noted that there is a distinct lack of literature from LAMI countries on ECE’s role in academic attainment, and also from Woldehanna and Gebremedhin (2012). Most ECE studies have had certain commonality in terms of the relative quality of ECE provision, ambient notions of education, patterns of parental cognition, or issues of gender. Thus, there is a possibility that patterns on ECE in LAMI countries could be different from those in developed contexts, but it is difficult to corroborate this without further evidence.

108 Leave a comment on paragraph 108 0 If patterns in LAMI contexts were to differ from the wider evidence body, as these findings indicate may be the case, it is likely that ‘quality of ECE’ divides them. This conclusion is based on evidence from Rao and Kaul, who found that anganwadis offered insufficient pre-primary education (2017) and Kaul et al. (2017) who noted that even private pre-primary is developmentally inadequate. Accordingly, overall poor quality ECE provision in India could be contributing to lack of statistically significant results.

109 Leave a comment on paragraph 109 0 The Role of Gender in Predicting Attainment Appears to be Diminishing

110 Leave a comment on paragraph 110 0 These findings demonstrate that gender is not a good estimator of academic attainment. They corroborate existing evidence on multiple fronts: evidence that gender does not play a role in estimating a range of education-related outcomes has also been found using YL India data by Vennam et al. (2009) and again by Streuli, Vennam and Woodhead (2011). Meanwhile, official World Bank data (“Education Statistics – All Indicators”, 2018) supports this paper’s findings on preschool participation rates and also offers a backdrop to the evidence on equal attainment.

111 Leave a comment on paragraph 111 0 These results speak to the long-established international focus on reaching gender parity in education, which has been a particular priority since the inception of the Millennium Development Goals (MDGs) (“United Nations Millennium Development Goals”, 2018). They also suggest that efforts made by the Indian government to redress gender disparity in education, such as launching the National Programme of Education for Girls at Elementary Level, are making a difference (Streuli, Vennam and Woodhead, 2011). And whilst these results do not support the views of Kaul et al. (2017) that gender in education is still a point of disparity in India, or evidence from Magnuson et al. (2016) or Garcia, Heckman and Ziff (2018) who indicate that girls in some instances benefit less from pre-primary education[51], they represent an important finding that demonstrates progress is being made.

112 Leave a comment on paragraph 112 0 Maternal Education Was a Useful Predictor of Attainment

113 Leave a comment on paragraph 113 0 Results on maternal education[52] found it to be a strong predictor of academic outcomes amongst this sample. These findings support wide-ranging evidence from the human capital tradition that investing in mothers’ education pays intergenerational dividends (see Andrabi, Das and Khwaja, 2012; Becker, Murphy and Tamura, 1990; Black et al,. 2005; King and Hill, 1993; Magnuson, 2007; Rosenzweig and Wolpin, 1994 and Schultz, 1993). They also corroborate evidence from India that maternal education has an important role to play in educational attainment (Jeong, Kim and Subramanian, 2018; Kaul et al., 2017 and Singh and Mukherjee, 2015). These findings provide clear evidence of the intergenerational nature of maternal education as an ‘endowment’.

114 Leave a comment on paragraph 114 0 Conclusions

115 Leave a comment on paragraph 115 0 Theory and literature both indicate that participating in early education should improve overall attainment by providing children with the skills to succeed academically. However, much of the evidence available is from developed contexts, which have very different parameters of experience. The aim of this study was to explore the impact of preschool participation on academic attainment for children in India[53], which is of interest because of educational attainment’s link to important socioeconomic outcomes. This research problem was addressed through layers of quantitative analysis, including the creation of multivariate models controlling the possible influences of gender and maternal education.

116 Leave a comment on paragraph 116 0 Despite being able to successfully create multivariate models that were good estimators of test results, upon interpretation of the results, little to no association was found between preschool participation and academic outcomes. These findings highlight firstly that ECE in India is facing issues of quality and secondly that evidence on ECE is being generalised because it is convincing, and because there is not sufficient evidence from elsewhere to change this. Accordingly, this paper argues that ECE policies for LAMI countries should rely not just on precedent but take into consideration localised evidence. It also argues against viewing ‘preschool’ as a monolithic or uniform experience, because a plausible explanation for the inconclusive findings is that ‘preschool’ was not disaggregated. However, considering all preschool experiences as ‘uniform’ for this sample also had value, because it demonstrated that as a group preschools in Andhra Pradesh are not of good enough quality to enhance cognitive outcomes. This evidence underscores the need to continue reforming ECE services in India (Streuli, Vennam and Woodhead, 2011).

117 Leave a comment on paragraph 117 0 India’s government therefore needs to redress inadequate pre-primary services, particularly in light of the increasing competitiveness of the private market. Other researchers have also cited the need for a regulatory body to help with governing quality standards (Chopra, 2012). Increasing the effectiveness of India’s ECE models can best be accomplished by garnering more reliable, relevant evidence to inform their improvement.

118 Leave a comment on paragraph 118 0 In this vein, this paper has been an endeavour to help substantiate the limited evidence body on ECE in India, provide nuance to existing debates on topics such as gender, and increase awareness of the limits of generalisability of existing ECE literature[54]. Though the findings are, in themselves, not generalisable, their importance lies in their departure from the results of other research showing ECE’s effectiveness.

119 Leave a comment on paragraph 119 0 References

120 Leave a comment on paragraph 120 0 Aakvik, A., Salvanes, K. and Vaage, K. (2003). Measuring heterogeneity in the returns to education in Norway using educational reforms. C.E.P.R. Discussion Papers: No.          4088.

121 Leave a comment on paragraph 121 0 Alcott, B. and Rose, P. (2015). Schools and learning in rural India and Pakistan: Who goes           where, and how much are they learning?. Prospects, 45, pp.345-363.

122 Leave a comment on paragraph 122 0 Alcott, B., Banerji, M., Bhattacharjea, S., Ramanujan, P., and Nanda, M. (2018). One step            forward, two steps back: transitions between home, pre-primary and primary       education in rural India. Compare: A Journal of Comparative and International     Education. DOI 10.1080/03057925.2018.1527214

123 Leave a comment on paragraph 123 0 Alcott, B., Bhattacharjea, S., Ramanujan, P., & Nanda, M. (Forthcoming). Participation    trends in early childhood education programmes: Who goes where and why? In Early       childhood education in India; findings from a cohort study (S. Bhattacharjea & V. Kaul, eds). Springer.

124 Leave a comment on paragraph 124 0 Andrabi, T., Das, J. and Khwaja, A. (2012). What did you do all day? Maternal education and      child outcomes. Journal of Human Resources, 47(4).

125 Leave a comment on paragraph 125 0 Arora, S., Bharti, S., and Mahajan, A. (2006). Evaluation of non-formal pre-school            educational services provided at Anganwadi Centres (Urban Slums of Jammu City).       Indian Journal of Social Science, 12(2), pp.135-137.

126 Leave a comment on paragraph 126 0 Asadullah, M. and Yalonetzky, G. (2012). Inequality of educational opportunity in India: Changes over time and across states. World Development, 40(6), pp.1151-1163.

127 Leave a comment on paragraph 127 0 Attanasio, O., Meghir, C. and Nix, E. (2015). Human capital development and parental     investment in India. Cowles Foundation Discussion Papers. New Haven, CT: Cowles    Foundation for Research in Economics, Yale University.

128 Leave a comment on paragraph 128 0 Barro, R. (1989). Economic growth in a cross section of countries. Working paper no. 3120.        Cambridge, Mass. NBER.

129 Leave a comment on paragraph 129 0 Becker, G. (1962). Investment in human capital: A theoretical analysis. Journal of Political          Economy, 70(5), pp.9-49.

130 Leave a comment on paragraph 130 0 Becker, G. (1964). Human Capital: A Theoretical and empirical analysis, with special       reference to education. National Bureau of Economic Research, New York.       Distributed by Columbia University Press.

131 Leave a comment on paragraph 131 0 Becker, G. (1993). Human Capital. 3rd ed. Chicago: University of Chicago Press.

132 Leave a comment on paragraph 132 0 Becker, G., Murphy, K. and Tamura, R. (1990). Human capital, fertility, and economic     growth. Journal of Political Economy, 98(5, Part 2), pp.S12-S37.

133 Leave a comment on paragraph 133 0 Belfield C., Nores, M., Barnett, W., and Schweinhart, L. (2006). The High/Scope Perry    Preschool program: Cost-benefit analysis using data from the age-40 followup.  Journal of Human Resources, 41(1), pp.162–90.

134 Leave a comment on paragraph 134 0 Ben-Porath, Y. (1967). The production of human capital and the life cycle of earnings.      Journal of Political Economy, 75(1), pp.352-365.

135 Leave a comment on paragraph 135 0 Beyond Basics. (2018). Annual Status of Education Report (Rural) 2017. New Dehli: ASER        Centre.

136 Leave a comment on paragraph 136 0 Black, M., Walker, S., Fernold, L. and Andersen, C. (2016). Early childhood development           coming of age: science through the life course. The Lancet, 389(10064), pp.77-90.

137 Leave a comment on paragraph 137 0 Black, S., Devereux, P. and Salvanes, K. (2005). Why the apple doesn’t fall far:     Understanding intergenerational transmission of human capital. The American       Economic Review, 95(1), pp.437-449

138 Leave a comment on paragraph 138 0 Boserup, E. (1970). Woman’s role in economic development. London: Earthscan.

139 Leave a comment on paragraph 139 0 Britto, P., Engle, P. and Super, C. (2013). Early childhood development: Translating research       to global policy. In: P. Britto, P. Engle and C. Super, ed., Handbook of Early Childhood Development Research and its Impact on Global Policy. Oxford: Oxford     University Press, pp.3-23.

140 Leave a comment on paragraph 140 0 Britto, P., Engle, P. and Super, C. ed., (2013). Handbook of Early Childhood Development           Research and its Impact on Global Policy. Oxford: Oxford University Press.

141 Leave a comment on paragraph 141 0 Bryman, A. (2014). Social Research Methods. 4th ed. Oxford: Oxford University Press,    pp.312-315.

142 Leave a comment on paragraph 142 0 Campbell, F. and Ramey, C. (1994). Effects of early intervention on intellectual and         academic achievement: A follow-up study of children from low-income families.       Child Development, 65, pp.684-698.

143 Leave a comment on paragraph 143 0 Carneiro, P. and J. Heckman. (2003). Human Capital Policy. Cambridge, MA: MIT Press.

144 Leave a comment on paragraph 144 0 Chopra, N. (2012). A study of early childhood education programmes in Dehli, India.       International Journal of Early Years Education, 20(2), pp.159-174.

145 Leave a comment on paragraph 145 0 Cohen, J. (1988). Statistical Power Analysis for the Behavioural Sciences. 2nd ed. New York:      Lawrence Erlbaum Associates.

146 Leave a comment on paragraph 146 0 Cunha, F. and Heckman, J. (2007). The technology of skill formation. American Economic          Review, 97(2), pp.31-47.

147 Leave a comment on paragraph 147 0 Cunha, F., Heckman, J., Lochner, L. and Masterov, M. (2006). Interpreting the evidence on          life cycle skill formation. In: E. Hanushek and F. Welch, ed., Handbook of the           Economics of Education. New York: Elsevier.

148 Leave a comment on paragraph 148 0 Currie, J. (2001). Early childhood education programs. Journal of Economic Perspectives,            15(2), pp.213–38.

149 Leave a comment on paragraph 149 0 Currie, J. and Thomas, D. (1995). Does a head start make a difference?. American Economic        Review, 85(3), pp.341-64.

150 Leave a comment on paragraph 150 0 Das, J. and Zajonc, T. (2010). India shining and Bharat drowning: Comparing two Indian states to the worldwide distribution in mathematics achievement. Journal of      Development Economics, 92(2), pp.175-187.

151 Leave a comment on paragraph 151 0 Datta, S., Boratne, A., Cherian, J., Joice, Y. and Vignesh, J. (2010). Performance of          anganwadi centres in urban and rural area: a facility survey in coastal South India.       Indian Journal of Maternal and Child Health, 12(4), pp.9.

152 Leave a comment on paragraph 152 0 Dearden, L., Stephen, M. and Howard, R. (1997). Intergenerational mobility in Britain.     Economic Journal, 107(440), pp.47-66.

153 Leave a comment on paragraph 153 0 DeCicca, P. and Smith, J. (2013). The long-run impacts of early childhood education:       Evidence from a failed policy experiment. Economics of Education Review, 36,        pp.41-59

154 Leave a comment on paragraph 154 0 Denison, P. (1985). Trends in American economic growth. Washington, DC: The Brookings        Institution, pp.1929-1982.

155 Leave a comment on paragraph 155 0 Durbin, J. and Watson, G. (1950). Testing for serial correlation in least squares regression.           Biometrika, 37(3-4), pp.409-428.

156 Leave a comment on paragraph 156 0 Early Childhood Development: informing policy and making it a priority. (2018). Young Lives Impact Case Study. Oxford: Young Lives.

157 Leave a comment on paragraph 157 0 Education Statistics – All Indicators. DataBank. (2018). Retrieved from            https://databank.worldbank.org/Data/indicator/UIS.E.0.T?id=c755d342&report_nam            =EdStats_Indicators_Report&populartype=series

158 Leave a comment on paragraph 158 0 Engle, P., Black, M., Behrman, J., Mello, M., Gertler, P., Kapiriri, L., Martorell, R., Young,         M. and the International Child Development Steering Group. (2007). Child      development in developing countries 3: Strategies to avoid the loss of developmental potential in more than 200 million children in the developing world. The Lancet, 369,          pp.229–42.     

159 Leave a comment on paragraph 159 0 Engle, P., Fernald, L., Alderman, H., Behrman, J., O’Gara, C., Yousafzai, A., Cabral de    Mello, M., Hidrobo, M., Ulkuer, N. and Ertem, I. (2011). Strategies for reducing inequalities and improving developmental outcomes for young children in low   income and middle income countries. The Lancet, 378 (9799), pp.1339-1353.

160 Leave a comment on paragraph 160 0 Field, A. (2009). Discovering Statistics Using SPSS. 3rd ed. London: SAGE.

161 Leave a comment on paragraph 161 0 Galab, S., Reddy, P. and Himaz, R. (2008). Young Lives Round 2 Survey Report Initial    Findings: Andhra Pradesh, India. Oxford: Young Lives.

162 Leave a comment on paragraph 162 0 Garcia, J., Heckman, J. and Ziff, A. (2018). Gender differences in the benefits of an          influential early childhood program. European Economic Review, 109, pp.9-22.

163 Leave a comment on paragraph 163 0 Garcia, J., Heckman, J., Ermini, D. and Prados, M. (2016). The life-cycle benefits of an    influential early childhood program. NBER Working Papers. National Bureau of     Economic Research

164 Leave a comment on paragraph 164 0 Goodman, A. and Sianesi, B. (2005). Early education and children’s outcomes: How long do       the impacts last?. Institute for Fiscal Studies. 26(4), pp.513–48

165 Leave a comment on paragraph 165 0 Gormley Jr., W., Phillips, D., Newmark, K., Welti, K. and Adelstein, S. (2011). Social      emotional effects of early childhood education programs in Tulsa. Child  Development, 82(6), pp.2095-2109.

166 Leave a comment on paragraph 166 0 Government of India, Ministry of Law and Justice (2009). The Right of Children to Free and       Compulsory Education Act. New Dehli: Government of India Press.

167 Leave a comment on paragraph 167 0 Government of India. (2013). National Early Childhood Care and Education Policy. Ministry       of Women and Child Development. New Delhi: GoI.

168 Leave a comment on paragraph 168 0 Grantham-McGregor, S., Cheung, Y., Cueto, S., Glewwe, P., Richter, L., Strupp, B., and the        international Child Development Steering Group. (2007). Developmental potential in     the first 5 years for children in developing countries. The Lancet, 36, pp. 60–70.

169 Leave a comment on paragraph 169 0 Handbook on Measuring Equity in Education. (2018). Montreal: UNESCO Institute for    Statistics.

170 Leave a comment on paragraph 170 0 Hanushek, E. (2008). Schooling, gender equity, and economic outcomes. In: M. Tembon and       L. Fort, ed., Girls’ Education in the 21st Century. Washington, D.C.: The World            Bank.

171 Leave a comment on paragraph 171 0 Harmon, C. and Walker, I. (1995). Estimates of the economic return to schooling for the  United Kingdom. The American Economic Review, 85(5), pp.1278-1286.

172 Leave a comment on paragraph 172 0 Hauser, R. and Logan, J. (1992). How not to measure intergenerational occupational         persistence. American Journal of Sociology, 97(6), pp.1689-711.

173 Leave a comment on paragraph 173 0 Heckman, J. (1995). Lessons from the bell curve. Journal of Political Economy. 103(5),   pp.1091

174 Leave a comment on paragraph 174 0 Heckman, J. (2006). Skill formation and the economics of investing in disadvantaged       children. Science, 312(5782), pp.1900-1902.

175 Leave a comment on paragraph 175 0 Heckman, J. (2011). The economics of inequality: The value of early childhood education.           American Educator, 35(1), pp.31-35.

176 Leave a comment on paragraph 176 0 Heckman, J. and Masterov, D. (2007). The productivity argument for investing in young  children. Applied Economic Perspectives and Policy, 29(3), pp.446–493.

177 Leave a comment on paragraph 177 0 Herz, B. and Sperling, G. (2016). What Works in Girls’ Education. Washington, D.C.: The           Brookings Institution.

178 Leave a comment on paragraph 178 0 Indian Ministry of Women and Child Development (2013). National early childhood care and      education (ECCE) Curriculum Framework.

179 Leave a comment on paragraph 179 0 Jeong, J., Kim, R. and Subramanian, S. (2018). How consistent are associations between  maternal and paternal education and child growth and development outcomes across       39 low-income and middle-income countries?. Journal of Epidemiology and       Community Health, 72(5).

180 Leave a comment on paragraph 180 0 Karoly, L., Greenwood, P., Everingham, S., Houbé, J., Kilburn, M., Rydell, C., Sanders, M.         and Chiesa, J. (1998). Investing in our children: What we know and don’t know about  the costs and benefits of early childhood interventions. Santa Monica, CA: RAND.

181 Leave a comment on paragraph 181 0 Kaul, V. and Sankar, D. (2009). Early Childhood Care and Education in India. Education for       All – Mid-Decade Assessment. New Dehli: National University of Educational          Planning and Administration (NUEPA).

182 Leave a comment on paragraph 182 0 Kaul, V., Bhattacharjea, S., Chaudhary, A., Ramanujan, P., Banerji, M. and Nanda, M.     (2017). The India Early Childhood Education Impact Study. Dehli: CECED,  Ambedkar University Dehli, and ASER Centre.

183 Leave a comment on paragraph 183 0 Kaul, V., Ramachandran, C. and Upadhyaya, G. C. (1993). Impact of Early Childhood     Education on Retention in Primary Grades: A Longitudinal Study.

184 Leave a comment on paragraph 184 0 King, E. and Hill, A. (1993). Women’s Education in Developing Countries. Baltimore:      Published for the World Bank [by] the Johns Hopkins University Press.

185 Leave a comment on paragraph 185 0 Kingdon, G. (1996). Private schooling in India: Size, nature, and equity-effects. Economic           and Political Weekly, 31(51), pp.3306-3314.

186 Leave a comment on paragraph 186 0 Kline, R. (2004). Principles and Practice of Structural Equation Modelling. 2nd ed. New York: The Guilford Press, p.29.

187 Leave a comment on paragraph 187 0 Kohlberg, L. 1968. Early education: A cognitive-developmental view. Child Development,           pp.1013-1062.

188 Leave a comment on paragraph 188 0 Kumra, N. (2008). An assessment of the Young Lives sampling approach in Andhra Pradesh,       India. Young Lives Technical Notes. Oxford, England: Young Lives.

189 Leave a comment on paragraph 189 0 Leibowitz, A. (1974). Home Investments in Children. Journal of Political Economy, 82(5),          pp.1111-31.

190 Leave a comment on paragraph 190 0 Levin, H. and Schwartz, H. (2006). Costs of early childhood care and education programmes.       Background paper prepared for the Education for All Global Monitoring Report 2007    Strong foundations: early childhood care and education. New York: National Cente         for the Study of Privatization in Education and UNESCO.

191 Leave a comment on paragraph 191 0 Literacy rate, adult female. (2018). Retrieved from            https://data.worldbank.org/indicator/SE.ADT.LITR.FE.ZS?view=chart.

192 Leave a comment on paragraph 192 0  Loeb, S., Bridges, M., Fuller, B., Rumberger, R. and Bassok, D. (2005). How much is too           much? The influence of pre-school centers on children’s social and cognitive    development. Economics of Education Review, 26(1), pp.52–66.

193 Leave a comment on paragraph 193 0 Magnuson, K. (2007). Maternal education and children’s academic achievement during    middle childhood. Developmental Psychology, 43, pp.1497–151.

194 Leave a comment on paragraph 194 0 Magnuson, K., Kelchen, R., Duncan, G., Schindler, H., Shager, H. and Yoshikawa, H.      (2016). Do the effects of early childhood education programs differ by gender? A  meta-analysis. Early Childhood Research Quarterly, 36, pp.521-536.

195 Leave a comment on paragraph 195 0 Magnuson, K.A., Meyers M., Ruhm, C.J. and Waldfogel, J. (2004) Inequality in pre-school          education and school readiness. American Educational Research Journal, 41(1),        pp.115–57.

196 Leave a comment on paragraph 196 0 Manhas, S. and Qadiri, F. (2010). A comparative study of preschool education in early      childhood education centres in India. Contemporary Issues in Early Childhood, 11(4).

197 Leave a comment on paragraph 197 0 Marope, P. and Kaga, Y. (2015). Investing against Evidence: The Global State of Early    Childhood Care and Education. Paris: UNESCO.

198 Leave a comment on paragraph 198 0 Meghir, C. and Rivkin, S. (2011). Econometric Methods for Research in Education. In: E.            Hanushek, S. Machin and L. Woessmann, ed., Handbook of the Economics of      Education. New York: Elsevier.

199 Leave a comment on paragraph 199 0 Mincer, J. (1958). Investment in Human Capital and Personal Income Distribution. Journal of      Political Economy, 66(4), pp.281-302.

200 Leave a comment on paragraph 200 0 Mohite, P. and Bhatt, N. (2008). From home to school: Mapping children’s transition in the         Indian context. Research in Comparative and International Education, 3(3), pp.308         316.

201 Leave a comment on paragraph 201 0 Morabito, C., Vandenbroeck, M. and Roose, R. (2013). ‘The Greatest of Equalisers’: A     critical review of international organisations’ views on early childhood care and         education. Journal of Social Policy, 42(3), pp.451-467.

202 Leave a comment on paragraph 202 0 Mulligan, C. (1999). Galton versus the human capital Approach to inheritance. Journal of Political Economy, 107(6), pp.184- 224.

203 Leave a comment on paragraph 203 0 Nadeau, S., Elder, L., Kataoka, N., Neuman, M. and Valerio, A. (2011). Investing in Young         Children : An Early Childhood Development Guide for Policy Dialogue and Project   Preparation. Washington, D.C.: The World Bank.

204 Leave a comment on paragraph 204 0 Nagaraja, G. and Anil, M. (2014). A sociological study of children irregularity and dropout          from Anganwadi Centre of Kolar district, Karnataka State. International Journal of        Health Sciences and Research, 4(3), pp.23-28.

205 Leave a comment on paragraph 205 0 NCERT. (1996). Identifying Numeracy and Reading Readiness Levels of Entrants to Grade I:      A Synthesis. Report under DPEP, Department of Elementary Education. New Delhi.

206 Leave a comment on paragraph 206 0 Patrinos, H. (2008). Returns to education: The gender perspective. In: M. Tembon and L. Fort, ed., Girls’ Education in the 21st Century. Washington, D.C.: The World Bank.

207 Leave a comment on paragraph 207 0 Pattnaik, J. (1996). Early childhood education in India: History, trends, issues, and            achievements. Early Childhood Education Journal, 24(1), pp.11-16.

208 Leave a comment on paragraph 208 0 Pratham (2010). Annual Status of Education Report, 2009. New Delhi, India: Pratham and           ASER Centre.

209 Leave a comment on paragraph 209 0 Rao, N. and Kaul, V. (2017). India’s integrated child development services scheme:           challenges for scaling up. Child: Care, Health and Development, 44(1), pp.31-40.

210 Leave a comment on paragraph 210 0 Rao, N., Sun, J., Wong, J., Weekes, B., Ip, P., Shaeffer, S., Young, M., Bray, M., Chen, E.           and Lee, D. (2013). Early Childhood Development and Cognitive Development in        Developing Countries: A Rigorous Literature Review. London: DFID.

211 Leave a comment on paragraph 211 0 Richter, L., Daelmans, B., Lombardi, J., Heymann, J., Lopez Boo, F., Behrman, J., Lu, C.,           Lucas, J., Perez-Escarnilla, R., Dua, T., Bhutta, Z., Stenberg, K., Gertler, P. And       Darmstadt, G. (2016). Investing in the foundation of sustainable development:   pathways to scale up for early childhood development. Retrieved from    https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(16)31698-1/fulltext.

212 Leave a comment on paragraph 212 0 Rosenbaum, P. (2006). Differential effects and generic biases in observational studies.      Biometrika, 93(3), pp.573–586.

213 Leave a comment on paragraph 213 0 Rosenzweig, M. and Wolpin, K. (1994). Are there increasing returns to the intergenerational        production of human capital? Maternal schooling and child intellectual achievement.   The Journal of Human Resources, 29(2), pp.670-693

214 Leave a comment on paragraph 214 0 Save the Children. (2009). Situational Analysis: Early Childhood Development in India.  India, New Delhi: Save the Children.

215 Leave a comment on paragraph 215 0 Schonemann, P. H. and Steiger, J. H. (1976). Regression component analysis. British Journal      of Mathematical and Statistical Psychology, 29, pp.175-189.

216 Leave a comment on paragraph 216 0 Schultz, T. (1993). Investments in the schooling and health of women and men: quantities            and returns. Journal of Human Resources, 23(4), pp.694-734.

217 Leave a comment on paragraph 217 0 Shabana, K., Mohhamad, B., Sushil, D., Mahajan, H. and Muneer, M. (2013). A cross       sectional study on client satisfaction and utilization of ICDS services in Kashmir         Division of Jammu and Kashmir State. Indian Journal of Public Health Research and          Development, 4(4), pp.206.

218 Leave a comment on paragraph 218 0 Shonkoff, J. and Phillips, D. (Eds) (2000). From Neurons to Neighborhoods: The Science of         Early Childhood Development. Washington D.C., USA: National Academy Press.

219 Leave a comment on paragraph 219 0 Shonkoff, J. and Richter, L. (2013). The powerful reach of early childhood development: A         science-based foundation for sound investment. In: P. Britto, P. Engle and C. Super,           ed., Early Childhood Development Research and its Impact on Global Policy. Oxford:    Oxford University Press, pp.24-34.

220 Leave a comment on paragraph 220 0 Singh, R. and Bangay, C. (2014). Low fee private schooling in India – More questions than          answers? Observations from the Young Lives longitudinal research in Andhra          Pradesh. International Journal of Educational Development, 39, pp.142–50.

221 Leave a comment on paragraph 221 0 Singh, R. and Mukherjee, P. (2015). Determinants of Successful Completion of Secondary           Education: Evidence from Andhra Pradesh. Young Lives Working Paper 143, Oxford:            Young Lives.

222 Leave a comment on paragraph 222 0 Singh, R. and Mukherjee, P. (2016). Education Trajectories: From Early Childhood to Early         Adulthood in India. Oxford, England: Young Lives.

223 Leave a comment on paragraph 223 0 Singh, R. and Mukherjee, P. (2017). Comparison of the Effects of Government and Private           Preschool Education on the Developmental Outcomes of Children: Evidence from     Young Lives India. Oxford, England: Young Lives.

224 Leave a comment on paragraph 224 0 Singh, R. and Mukherjee, P. (2018). Effect of preschool education on cognitive achievement        and subjective wellbeing at age 12: evidence from India. Compare: A Journal of          Comparative and International Education, pp.1-19.

225 Leave a comment on paragraph 225 0 Solon, G. (1999). Intergenerational mobility in the labor market. In: O. Ashenfelter and D.           Card, ed., Handbook of Labor Economics. Amsterdam: Elsevier Science, pp.1761      1800.

226 Leave a comment on paragraph 226 0 Streuli, N., Vennam, U. and Woodhead, M. (2011). Increasing choice or inequality?          Pathways through early education in Andhra Pradesh, India. The Hague, The   Netherlands: Bernard van Leer Foundation.

227 Leave a comment on paragraph 227 0 Strong Foundations: Early Childhood Care and Education. (2006). EFA Global Monitoring          Report. Paris: UNESCO.

228 Leave a comment on paragraph 228 0 Sylva, K., Melhuish, E., Sammons, P., Siraj-Blatchford, I. and Taggart, B. (2011) Pre-school        quality and educational outcomes at Age 11: Low quality has little benefit. Journal of  Early Childhood Research, 9(2), pp.109-124.

229 Leave a comment on paragraph 229 0 Todd, P. and Wolpin, K. (2004). On the specification and estimation of the production      function for cognitive achievement. Economic Journal, 113(485), pp.F3–33.

230 Leave a comment on paragraph 230 0 Tooley, J. and Dixon, P. (2003). Private Schools for the Poor: A Case Study from India.   Reading: CfBT.

231 Leave a comment on paragraph 231 0 UNESCO (2000). Dakar Framework for Action: Education for All. Meeting Our Collective            Commitments. World Forum on Education, Dakar, Senegal, 26-28 April, 2000,     UNESCO, Paris.

232 Leave a comment on paragraph 232 0 United Nations Millennium Development Goals. (2018). Retrieved from            http://www.un.org/millenniumgoals.

233 Leave a comment on paragraph 233 0 Unterhalter, E. (2007). Gender, Schooling and Global Social Justice. New York: Routledge.

234 Leave a comment on paragraph 234 0 Vennam, U., Komanduri, A., Cooper, E., Crivello, G. and Woodhead, M. (2009). Early    Childhood Education Trajectories and Transitions: A Study of the Experiences and     Perspectives of Parents and Children in Andhra Pradesh, India.

235 Leave a comment on paragraph 235 0 Woldehanna, T. and Gebremedhin, L. (2012). The Effects of Pre-school Attendance on the          Cognitive Development of Urban Children aged 5 and 8 Years: Evidence from           Ethiopia. Oxford: Young Lives.

236 Leave a comment on paragraph 236 0 Woodhead, M. (2006). Changing perspectives on early childhood: theory, research and    policy. Paper commissioned for the EFA Global Monitoring Report 2007, Strong foundations: early childhood care and education.

237 Leave a comment on paragraph 237 0 Woodhead, M. (2009) Pathways Through Early Childhood Education in Ethiopia, India and         Peru: Rights, Diversity and Equity. Oxford: Young Lives.

238 Leave a comment on paragraph 238 0 Woodhead, M., Ames, P., Vennam, U., Abebe, W. and Streuli, N. (2009). Equity and quality?      Challenges for early childhood and primary education in Ethiopia, India and Peru.       Working Paper 55, The Hague: Bernard van Leer Foundation.

239 Leave a comment on paragraph 239 0 Woolridge, J. (2016). Introductory Econometrics: A Modern Approach. 6th ed. Boston:    CENGAGE Learning.

240 Leave a comment on paragraph 240 0 World Inequality Database on Education • India. (2018). Retrieved from    https://www.education            inequalities.org/countries/india#?dimension=all&group=all&year=latest.

241 Leave a comment on paragraph 241 0  Yoshikawa, H., and Nieto, A., (2013). Paradigm shifts and new directions in research on early childhood development programs in low and middle income countries. In Britto, P.R., Engle, P.L. and Super, C. (2013), Handbook of early childhood development research and its impact on global policy (pp.487-497). Oxford: Oxford University Press.

242 Leave a comment on paragraph 242 0 Young Lives (2017). Cohort Maintenance – Tracking and Attrition. A Guide to Young Lives       Research. [online] Oxford: Young Lives. Retrieved from            http://www.younglives.org.uk/sites/www.younglives.org.uk/files/GuidetoYLResearc         -S12-CohortMaintenance.pdf.

243 Leave a comment on paragraph 243 0 Young Lives Survey Design and Sampling in India. (2014). Young Lives. Retrieved from            https://www.younglives.org.uk/content/young-lives-survey-design-and-sampling    andhra-pradesh-india.

244 Leave a comment on paragraph 244 0 Young, M. ed., (2007). Early child development from measurement to action: a priority for          growth and equity. Portland: The World Bank.

245 Leave a comment on paragraph 245 0 Young Lives India. (2018). Retrieved from https://www.younglives-india.org.

246 Leave a comment on paragraph 246 0 [1] ECCE summarises the inputs and processes that create the building blocks of later social, cognitive, emotional and economic development (Kaul et al., 2017).

247 Leave a comment on paragraph 247 0 [2] Preschool is used within this paper interchangeably with ECE.

248 Leave a comment on paragraph 248 0 [3] Stipulates early childhood development, care and education for all children by 2030 (UNESCO, 2000).

249 Leave a comment on paragraph 249 0 [4] ECE refers to regularly attended education in a setting outside of a child’s home for children aged 3-5, during the period immediately preceding primary school (Woldehanna and Gebremedhin, 2012).

250 Leave a comment on paragraph 250 0 [5] In India, 25% of girls are out of school, and 17% of girls have never been to school at all. As these girls grow into motherhood, there are further intergenerational consequences (“Education Statistics – All Indicators”, 2018).

251 Leave a comment on paragraph 251 0 [6] As education research in India has sometimes been hampered by unreliable data (Kingdon, 1996), YL’s study offered rare information collected with rigorous sampling and response management procedures, resulting in good quality data with minimal attrition (Singh and Mukherjee, 2016; Young Lives, 2017).

252 Leave a comment on paragraph 252 0 [7] There are two limits to generalisability due to usage of YL’s data. The first is geographic: participants were only from one (later two) states: Andhra Pradesh and Telangana. Together these comprise only 7% of India’s population (“Young Lives India”, 2019). The second limitation is due to YL’s purposive sampling, employed to explore poorer communities (Attanasio, Meghir and Nix, 2015). However, the YL team limited bias by including households across the socioeconomic spectrum and ensuring a large data set (Kumra, 2008; Young Lives Survey Design and Sampling in India, 2014).

253 Leave a comment on paragraph 253 0 [8] Whilst the original YL research also included qualitative data, its use for a mixed methods approach was not possible due to access constraints.

254 Leave a comment on paragraph 254 0 [9] Measures of cognitive ability commonly included test scores or years of schooling (Rosenzweig and Wolpin, 1994). For example, researchers such as Denison (1985), Barro (1989), and Becker (1964) used years of schooling to ‘explain’ variance in per capita earnings distributions.

255 Leave a comment on paragraph 255 0 [10] These outcomes were calculated via their rate of return (RoR), and education economists demonstrated that the RoR to schooling was a critical factor in individual income as well as GDP (Harmon and Walker, 1995).

256 Leave a comment on paragraph 256 0 [11] ECD is “A multifaceted concept from an ecological framework that focuses on the child’s outcome (development), which depends on characteristics of the child and the context, such as health, nutrition, protection, care and/or education.” (Britto, Engle and Super, 2013).

257 Leave a comment on paragraph 257 0 [12] A number of studies on the effects of early intervention programs on economic outcomes is summarised neatly by Cunha (Cunha et al., 2006).

258 Leave a comment on paragraph 258 0 [13] See Belfield, Nores and Schweinhart, 2006; DeCicca and Smith, 2013; Engle et al., 2007; Goodman and Sianesi, 2005; Heckman and Masterov, 2007; Magnuson et al., 2004 and Sylva et al., 2011 for examples from the US and Europe.

259 Leave a comment on paragraph 259 0 [14] A large portion of the research on ECE underpinned by human capital tested the hypothesis that interventions in early education could remediate circumstantial disadvantages for children (Woodhead, 2006). 

260 Leave a comment on paragraph 260 0 [15] The noted paucity of global evidence on ECD outcomes has given rise to an important World Bank initiative entitled SABER-ECD, which now collects, analyses, and disseminates related information. There is growing evidence from LAMI countries; see Grantham-McGregor et al., 2007; Rao et al., 2013 and the 2016 Lancet series on economic outcomes linked to early years investments (Engle et al., 2011).

261 Leave a comment on paragraph 261 0 [16] One of the few exceptions includes the compendium published by Engle and colleagues highlighting the status of ECD globally, though this covers not only education but also health and social welfare (Britto, Engle and Super, 2013).

262 Leave a comment on paragraph 262 0 [17] Alcott et al., 2018; Chopra, 2012; Kaul et al., 2017; Kaul and Sankar, 2009 and Singh and Mukherjee, 2017 have all documented this scarcity.

263 Leave a comment on paragraph 263 0 [18] See Arora et al., 2006; Datta et al., 2010; Nagajara and Anil, 2014 and Shabana et al., 2013 for examples.

264 Leave a comment on paragraph 264 0 [19] See Kingdon, 1996; Pratham, 2010; Singh and Mukherjee, 2017 and Tooley and Dixon, 2003.

265 Leave a comment on paragraph 265 0 [20] However, even the IECEI is limited as it only covers a four-year period.

266 Leave a comment on paragraph 266 0 [21] The importance of measuring gender’s impact is also evidenced by the regular disaggregation of outcomes such as enrolment, grade completion, and test scores by gender across global education research (Handbook on Measuring Equity in Education, 2018).

267 Leave a comment on paragraph 267 0 [22] see Asadullah and Yalonetzky, 2012; Hauser and Logan, 1992 and Solon, 1999 for examples.

268 Leave a comment on paragraph 268 0 [23]Anganwadis are noted to have poorly trained workers, with basic education. Moreover, qualitative research from YL shows that teachers generally seem disinterested and disengaged.

269 Leave a comment on paragraph 269 0 [24] Private preschools range from low-cost to highly-priced, but almost all promise English medium instruction, which parents see as a path to upward socioeconomic mobility (Indian Ministry of Women and Child Development, 2013, though research shows more boys than girls are sent to private preschool (Streuli, Vennam and Woodhead, 2011).

270 Leave a comment on paragraph 270 0 [25] Comprising just under 10% of India’s population, Andhra Pradesh is largely rural, but home to the capital of India’s IT sector, Hyderabad. Telugu is its major language, spoken by 85% of the population (Vennam et al., 2009).

271 Leave a comment on paragraph 271 0 [26] Selected because information on both preschool participation at age 5 and test score results at age 12 was collected.

272 Leave a comment on paragraph 272 0 [27] The original sample sizes were n=2,011 in Round 1, n=1,950 in Round 2, n=1,930 in Round 3, and n=1,915 in Round 4 (Singh and Mukherjee, 2016).

273 Leave a comment on paragraph 273 0 [28] Though constraining the sample in this way could have introduced some bias, the non-response count on this question was low.

274 Leave a comment on paragraph 274 0 [29] It should not be assumed that I consider ECE to be the only explanatory variable accounting for cognitive development. Variance not explained by these models could be due to genetic endowment (see Todd and Wolpin, 2004), parental education, SES, or other inputs.

275 Leave a comment on paragraph 275 0 [30] See Meghir and Rivkin (2011) for a discussion of methodologies for binary education choice, as analysed by Heckman, LaLonde and Smith, 1999.

276 Leave a comment on paragraph 276 0 [31] Despite the RTE Act’s stipulation that children should enter Grade 1 at age 5-6, only at age 8 does correct enrolment by age stabilise (Kaul et al., 2017).

277 Leave a comment on paragraph 277 0 [32] Using test scores could constrict the meaning of ‘cognitive ability’ as outcomes from a range of subjects are not included; however, maths, comprehension, and reading are usually deemed the most critical skills to assess and are also the most comparable amongst educational studies.

278 Leave a comment on paragraph 278 0 [33] Descriptive statistics on this paper’s sample show that 57.8% of boys were attending preschool, while 55.7% of girls were attending. Of the children who were attending preschool, 54.3% were boys and 45.7% were girls.

279 Leave a comment on paragraph 279 0 [34] Not accounting for other education levels in the household presented a risk for bias. But as the majority of literature cites mother’s education as a paramount factor, I opted to follow this precedent.

280 Leave a comment on paragraph 280 0 [35] See Andrabi, Das and Khwaja, 2012; Asadullah and Yalonetzky, 2012;  Becker, Murphy and Tamura, 1990; Black et al., 2005 and Magnuson, 2007.

281 Leave a comment on paragraph 281 0 [36] One fundamental issue in estimating the effect of maternal education is unmeasured abilities, or ‘endowments’, of children (Rosenzweig and Wolpin, 1994). Because this is a highly complex matter to calculate, this paper did not account for bias resulting from endowments.

282 Leave a comment on paragraph 282 0 [37] Amongst this sample, 49.6% of mothers reported having no education, while 50.2% of mothers reported having completed some education. Of the children attending preschool, 43.4% had mothers without any education, and 56.6% had mothers with some education.

283 Leave a comment on paragraph 283 0 [38] Meghir and Rivkin (2011) provided a useful review of methodologies used in the economics of education. A number of these have been used to identify ‘causal’ relationships, but as this paper is not evaluating the results of experimental research, these models would not be appropriate to replicate in this context. However, they offer important conceptual background.

284 Leave a comment on paragraph 284 0 [39] The level of statistical significance used was a p value of < .05. All tests were two-tailed.

285 Leave a comment on paragraph 285 0 [40] Yates’ correction was applied to mitigate against Type 1 errors, as variables were dichotomous. (Field, 2009).

286 Leave a comment on paragraph 286 0 [41] Woldehanna and Gebremedhin (2012) used propensity score matching substantiated with OLS and IV regression to estimate the effects of pre-school attendance on cognitive development in Ethiopia with YL data. As it uses data from the same wider study on the same topic type, it provided a good reference point for my use of OLS methodology. See Loeb et al., 2005 from the US context for other examples of use of OLS on a similar topic.

287 Leave a comment on paragraph 287 0 [42] The assumption regarding multicollinearity could have presented an issue, given there was a correlation shown between mother’s education and preschool. However, Field (2012) advises that this assumption is only problematic if any of the input variables correlate highly, above .8 or .9. This was not the case.

288 Leave a comment on paragraph 288 0 [43] Relatedly, as SES is a known correlative of other variables within the models, this represents a violation of the assumption of no correlation with external variables. However, problems of correlation within the error term were checked for using the Durbin-Watson test (Durbin and Watson, 1950).

289 Leave a comment on paragraph 289 0 [44] Other variables such as location, nutritional status, birth order and caste could also have been ‘omitted variables’ (Singh and Mukherjee, 2018), but were not included to reduce overburdening the model and detracting from the research question.

290 Leave a comment on paragraph 290 0 [45] Another way to limit these biases would be to conduct further research utilising differential effects. This methodology could be used to compare each input variable’s explanatory potential, which could help to correct unobserved biases associated with both variables (Rosenbaum, 2006).

291 Leave a comment on paragraph 291 0 [46] However, due to the issue of external variables, it is possible that there was some bias in the coefficient for maternal education resulting from a correlation between SES and maternal education, and between SES and preschool participation.

292 Leave a comment on paragraph 292 0 [47] See Alcott et al., 2018; Kaul and Sankar, 2009; Kaul et al., 2017 and Singh and Mukherjee, 2017 for other documentation of this need.

293 Leave a comment on paragraph 293 0 [48] See Andrabi, Das and Khwaja, 2012; Becker, Murphy and Tamura, 1990; Black, Devereux and Salvanes, 2005; Black et al,. 2005; Magnuson, 2007; Rosenzweig and Wolpin, 1994 and also Jeong, Kim and Subramanian, 2018 for evidence from India, and Singh and Mukherjee, 2015 for evidence using YL India data.

294 Leave a comment on paragraph 294 0 [49] However, documented issues with ‘nonlinearity’ of early educational trajectories might have complicated this anyway, as many children have been shown to complete some government as well as some private preschool (Alcott et al., 2018).

295 Leave a comment on paragraph 295 0 [50] The YL team should have mitigated against this response bias, given ‘early entry’ is a documented phenomenon. Perhaps a question such as ‘did this child complete at least two full years of ECE between the ages of 3-5?’ would have allowed for a more accurate analysis of the effects of preschool on attainment.

296 Leave a comment on paragraph 296 0 [51] A complexity in these findings could arise from the possibility that variation between preschool type and gender was masked by the methodology involved in documenting preschool attendance, which did not account for private versus public. For example, other research, including qualitative research from YL India, has indicated that parents are more likely to invest in private education for sons (Woodhead et al., 2009).

297 Leave a comment on paragraph 297 0 [52] Because ‘maternal education’ was recoded for this paper as a binary variable, it is interesting to note that having any education at all made a difference to variation in test score outcomes on both tests. This is an important point for policymaking in India, as the country still has an adult female literacy rate of only 59% and a large quantum of first generation learners (“Literacy rate, adult female”, 2018).

298 Leave a comment on paragraph 298 0 [53] In order to determine how existing ECE services can be improved, further research into what elements or types of ECE covary with improved outcomes should be explored (Engle et al., 2007; Engle et al., 2011 and Yoshikawa and Nieto, 2013).

299 Leave a comment on paragraph 299 0 [54] Another opportunity for further research could be to widen the models in this paper by adding other highly pertinent predictor variables, thereby increasing the validity of the analysis. Using partial correlation could improve understanding of which variables are truly playing an explanatory role (Kline, 2004). This could also help to reduce some of the omitted variable bias or endogeneity (Woolridge, 2016). Making use of propensity score matching, as used by Woldehanna and Gebremedhin (2012), could also offer experimental-style research for comparing the effects of various input variables.

Page 9

Source: https://corerj.soc.srcf.net/?page_id=412