Parental investment and the joint evolution of children’s cognitive and non cognitive skills evidences in vietnam

i Given the importance of skill set as a predictor of life outcomes, this study investigates the effect of parental investment on the development of cognitive and non-cognitive skills of

UNIVERSITY OF ECONOMICS ERASMUS UNIVERSITY ROTTERDAM

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

PARENTAL INVESTMENT AND THE JOINT EVOLUTION OF CHILDREN’S COGNITIVE

AND NON-COGNITIVE SKILLS:

EVIDENCES IN VIETNAM

BY

TRAN KHANH HOA

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, December 2017

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY THE HAGUE

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

PARENTAL INVESTMENT AND THE JOINT EVOLUTION OF CHILDREN’S COGNITIVE

AND NON-COGNITIVE SKILLS:

EVIDENCES IN VIETNAM

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

TRAN KHANH HOA

Academic Supervisor:

Dr PHAM KHANH NAM

HO CHI MINH CITY, December 2017

TRAN KHANH HOA

A CKNOWLEDGEMENTS

Academically, I would like to express my sincere gratitude to Dr Pham Khanh Nam,

Dr Truong Dang Thuy and Mr Do Huu Luat for the valuable comments, instructions and support, without which my thesis would have not been completed

Personally, I would like to thank my big family, my loved one and friends for the great care, understanding, encouragement, tolerance and patience during my tough time of writing the thesis

Tran Khanh Hoa

i

Given the importance of skill set as a predictor of life outcomes, this study investigates the effect of parental investment on the development of cognitive and non-cognitive skills of 2,000 children, who were tracked from age 1 to age 12, using longitudinal data from Young Lives project in Vietnam A model of skill formation

is estimated using ordinary least squares and full information maximum likelihood as well as instrumental-variables to cope with missing values and potential endogeneity Cognitive and non-cognitive skills are latent variables obtained from confirmatory factor analysis There are three important findings in this study Firstly, the results support that parental investment plays a crucial role in the development of both cognitive and non-cognitive skills in the first 12 years of child’s life Secondly, there

is compelling evidence for the presence of self-productivity of cognitive-skills as well

as cross-productivity from cognitive skills to non-cognitive skills and vice versa Thirdly, the time span until the age of 12 is found to be critical period for the development of cognitive skills while the age of 5 and the age of 8 are sensitive periods for cognitive skills relative to the age of 12 These findings support policy towards an education reform which help to promote not only cognitive skills but also non-cognitive skills; and policy towards an early childhood intervention programs, which aim at children living under disadvantage circumstances

JEL Classification:

Keywords: parental investment, cognitive skill, non-cognitive skills, critical and sensitive periods, children

ii

ABSTRACT i

ABBREVIATION iv

LIST OF FIGURES v

LIST OF TABLES vi

Chapter 1 INTRODUCTION 1

1.1 Problem statement 1

1.2 Research objectives 3

1.3 Scope of the study 3

1.4 Structure of the thesis 4

Chapter 2 LITERATURE REVIEW 5

2.1 Human capital theory: skills matter 5

2.2 Cognitive skills and non-cognitive skills: measures and impacts 6

2.3 Determinants of skill development 8

2.3.1 Genetic endowment 8

2.3.2 Environment 9

2.4 Dynamics of skill development 11

2.4.1 Critical and sensitive periods: timing of parental investment 12

2.5 Analytical framework 13

Chapter 3 RESEARCH METHODOLOGY 15

3.1 Methodology 15

3.1.1 Technology of skill formation 15

3.1.2 Genetic endowment – Initial conditions 17

3.1.3 Method in cognitive, non-cognitive skills measurement: Latent variables 18

3.1.4 Instrument Variables 18

3.2 Data 19

3.2.1 Data source 19

3.2.2 Variables description 21

Chapter 4 RESEARCH RESULTS 25

4.1 Overview 25

iii

4.2 Descriptive statistics 28

4.2.1 Summary statistics 28

4.2.2 Bivariate analysis 36

4.2.3 Empirical results 39

Chapter 5 CONCLUSIONS 57

5.1 Conclusion 57

5.2 Policy implication 59

5.3 Limitations and further research 59

REFERENCES 61

APPENDIX 67

iv

CDA Cognitive Developmental Assessment

OLS Ordinary Least Square

SDG Sustainable Development Goals

UIS UNESCO Institution of Statistics

v

Figure 1: A model of parent-child genetic and environment effects 8

Figure 2: A framework of skill development 12

Figure 3: Analytical framework 14

Figure 4: Expenditure on education from government and household, 2009-2013 27 Figure 5: Household average expenditure per student, all level of education, by income quantile, 2012 27

Figure 6: Household average expenditure per student, by level of education and location, 2012 28

Figure 7: Scatter plots - Cross productivity 81

Figure 8: Scatter plots - Self-productivity 81

Figure 9: Scatter plots - Cognitive skills vs Expenditure 82

Figure 10: Scatter plots - Non-cognitive skills versus expenditure 84

vi

Table 1: Variable description 22

Table 2: Summary statistics - Round 1 (Age 1) 29

Table 3: Summary statistics - Round 2 (Age 5) 30

Table 4: Summary statistics - Round 3 (Age 8) 31

Table 5: Summary statistics - Round 4 (Age 12) 32

Table 6: Pairwise rank correlation matrix 35

Table 7: Reults - Round 1 - Age 1 40

Table 8: Results - Round 2 - Age 5 42

Table 9: Results on cognitive skills - Round 3 - Age 8 45

Table 10: Results on non-cognitive skills - Round 3 - Age 8 48

Table 11: Results on cognitive skills - Round 4 - Age 12 50

Table 12: Results on non-cognitive skills - Round 4 - Age 12 52

Table 13: Identification of sensitive periods 56

Table 14: Main characteristics of provinces selected 67

Table 15: Questions in measurement of non-cognitive skills 68

Table 16: Factor loading outcomes 70

Table 17: Results - Round 2 - Age 5 (Detailed expenditure) 71

Table 18: Results on cognitive skills - Round 3 - Age 8 (Detailed expenditure) 73

Table 19: Results on non-cognitive skills - Round 3 - Age 8 (Detail expenditure) 75

vii Table 20: Results on cognitive skills -Round 4 - Age 12 (Detail expenditure) 77Table 21: Results on non-cognitive skills - Round 4 - Age 12 (Detail expenditure) 79

“Human development is economic development.” James Heckman, Nobel Prize Laureate.

1

Chapter 1 INTRODUCTION

Development of a child from early childhood to adolescence has always been a great concern of parents and the society Evidence from recent interdisciplinary studies have showed that child development, particularly development of cognitive and non-cognitive skills, accounts largely for future life outcomes of a child In particular, development during childhood affect education, health, labor market outcomes, and social engagement (Francesconi & Heckman, 2016) For example, Heckman et al (2006) found out in the US that schooling decisions, wages and work experience are effected by cognitive and non-cognitive skills Another example from Sweden, Lindqvist and Vestman (2011) pointed out the role of cognitive and non-cognitive abilities in predicting labor force participation and wages among skilled and unskilled workers

Cognitive skills concern about child’s general intelligence while non-cognitive skills reflect child’s personality traits (Borghans et al., 2008) Skills are shaped and influenced by both genetic endowment and factors from the environment Although genetic endowment is critical and unchangeable, the expression of genes can vary depending on different environment factors Among the factors, factors from family environment such as parental characteristics, parenting styles, family socioeconomic condition, parental investment, etc., are the most important to child development Knowing when parental investment is most effective for child’s skill development is of policy importance Since skill development is a dynamic rather than a static process, there are two important features to be noticed Firstly, cognitive

2

and non-cognitive skills can reinforce each other in a mechanism called productivity and cross-productivity (Cunha & Heckman, 2008) Self-productivity (or cross-productivity) means a higher stock of skills in one period can create higher stocks of skills (or higher stocks of another skill) in the next period Secondly, there are critical and sensitive periods for skills acquisition because investment at different stages of the life cycle may have different impacts on cognitive and non-cognitive skills “Sensitive period” is a period where parental inputs are more productive than

self-in other stages And “critical period” is a period self-in which self-investment is only productive in that period Understanding of sensitive and critical periods are useful, especially to policy makers For example, policies which aim to subsidize investment

to disadvantaged children could be more effective when it is provided to the children

at the right time

So far, there is more supporting evidence from developed countries than from less developed countries on the discussion above regarding the timing of parental investment and the evolution of cognitive and non-cognitive skills throughout multiple stages of life Meanwhile, it is essential to learn about the topic in emerging economies because human development is the key to economic development

In Vietnam, a skilled workforce is crucially needed for economic modernization The STEP (Skills Towards Employment and Productivity) survey conducted in Vietnam by World Bank showed that the proportion of jobs that using analytical and interpersonal skills has increased significantly since 1998 As Vietnam

is shifting to more nonagricultural sectors, it is implied that more attention should be given to equip the workforce with essential skills such as cognitive and non-cognitive skills

This research identifies two concerns about skill development in Vietnam First, the general education in Vietnam has long placed great emphasis on the acquisition

of knowledge or basic cognitive skills but not on soft skills or non-cognitive skills such as self-esteem or self-control Whereas, cognitive and non-cognitive skills should receive equal attention, in line with “cross productivity” mechanism Second,

3

parents, especially those with disadvantaged conditions, has neglected the importance

of early development of their children The children receive inadequate investment from their parents because the parents might not be aware of the importance of investment on children or because the lack of support from the government or social groups Hence, it is suggested that policies should help raise parents’ awareness and provide parents with necessary support from very early stages of raising their children The effectiveness of the intervention may lie on the timing of the investment Evidence for sensitive and critical periods of skill development may provide an answer to whether more investment should be made during early childhood

Given the problem stated above, this research has three main objectives:

- Firstly, to investigate the effect of parental investment on the development of cognitive and non-cognitive skills over multiple stages of childhood, along with other determinants such as household and community characteristics

- Secondly, to find evidence for the presence of the reinforcement mechanism called self-productivity and cross-productivity

- Thirdly, to identify critical and sensitive periods for investing in the development of cognitive and non-cognitive skills The finding might be useful in policy making process regarding designing interventions, remediation programs for disadvantaged children

To achieve the research objectives, the study investigates 2,000 children from 5 provinces in Vietnam (Lao Cai, Hung Yen, Da Nang, Phu Yen and Ben Tre) These children were born in 2001-2002 and were tracked from the age of 1 to the age of 12 The data is collected from four rounds survey of the Young Lives project in 2002-3, 2006-7, 2009-10 and 2013-14

4

The thesis is organized into five chapters Chapter 1 presents an introduction to the study with problem statement, research objectives and scope of study Chapter 2 reviews the theories and existing empirical evidence on skills formation Following

is the discussion on research methodology and data in Chapter 3 Chapter 4 provides the summary statistics, bivariate analysis and empirical results In sum, the conclusion along with policy implications and limitation of the research are presented

in Chapter 5

5

Chapter 2

The review captures literature on skill development and previous empirical

evidence of self-productivity, cross productivity, critical and sensitive periods The

terms “skill” and “ability” will be used interchangeably

Human capital is broadly defined as any stock of knowledge or characteristics the worker has (either innate or acquired) that increase his or her “productivity” (Acemoglu, 1999) There are several ways to view and classify human capital such

as the Becker view or the Gardener view Both views agree that human capital is valued in the job market and helps explain life outcomes The difference between the two views is that the Becker view regards human capital as a unidimensional object while the Gardner view states that human capital should rather be represented by multi-dimensional ones This view comes after the development of multi-intelligence theory developed by the social psychologist Howard Gardner in 1983 (Acemoglu, 1999) In addition to traditional linguistics intelligence and mathematical intelligence (cognitive skills), Howard Gardner added seven more intelligences, including interpersonal intelligence and intrapersonal intelligence (non-cognitive skills) (Gardner, 2011) Accordingly, this study examines both cognitive and non-cognitive skills as human capital, which is in line with the Gardner view

In terms of the formation of human capital, genetic factors and investments can determine the differences in the level of human capital acquired by individuals On one hand, investments foster and enhance the acquisition of human capital On the other hand, genetic factor helps explain the heterogeneity in human capital given the same level of investments So far, education and training have been deeply discussed

6

in many literature and considered as the most important investments in human capital (Becker, 1994) In addition to education and training investments, other non-schooling investments are also important In fact, when schooling is limited in less developed communities, non-schooling investments have found to be greater in magnitude than schooling investments (Behrman, 1987)

Cognitive ability has multiple facets Cognition or cognitive ability is defined in the American Psychological Association Dictionary as “all forms of knowing and awareness such as perceiving, conceiving, remembering, reasoning, judging, imagining and problem solving” In specific, cognitive skills involve language skills, memory skills, motor skills, thinking skills Moreover, cognitive ability is often considered as intelligence, which is divided into fluid intelligence (the rate at which people learn) and crystallized intelligence (acquired knowledge) Due to this fact, cognitive ability is commonly measured using scores on intelligence tests such as Peabody Picture Vocabulary Test (a measure of children’s verbal intelligence), Raven’s Progressive Matrices (a nonverbal test), Wechsler tests or Stanford-Binet test (measures of general intelligence) (Neisser et al., 1996) The term “IQ” (Intelligence Quotient) indicates the scores of these intelligence tests

Other personal attributes which cannot be measured using intelligence tests, are considered non-cognitive skills, i.e self-control, self-esteem, persistence, etc In the literature, non-cognitive skills are expressed under different names such as personality traits, soft skills, character skills, and socio-emotional skills One thing to

be noted is that “non-cognitive” is not just juxtaposed with “cognitive” since many aspects of non-cognitive skills are the outcome of cognitive skills and vice versa (Borghans, 2008) Regarding the measures of non-cognitive skills, there are two measurement approaches which are self-report questionnaires and conventional economic preference parameters (such as time preference, risk aversion and preference for leisure) Self-report questionnaire is commonly used among psychologist to construct latent variables using factor analysis

7

The role of cognitive and non-cognitive skills in predicting success in academic and career is apparent Previously, only cognitive skills were examined as a determinant of school attainments and wages For example, school performance was showed to be affected significantly by cognitive skills among Canadian High school students in a study of Riding and Agrell (1997) In addition, Murnane, Willett, and Levy (1995) found out an evidence of a growing importance of cognitive skills in determining wage level among American high school seniors in the 1980s Ree, Earles, and Teachout (1994) also showed that cognitive ability is the best predictor for job performance criteria in US Army Not until 2000s are there more studies investigating in the role of both cognitive and non-cognitive skills on the school performance and earning One of the most well-known work is an intensive study of

J J Heckman et al (2006) They found evidence in the US that both cognitive and non-cognitive skills affect schooling and work experience In fact, non-cognitive skills can raise wages thanks to their direct effect on productivity Similar evidence

is found in other developed countries such as Germany (Heineck & Anger, 2010), Sweden (Lindqvist and Vestman (2011), Lundborg, Nystedt, and Rooth (2014)), the

UK (Carneiro, Crawford, & Goodman, 2007), Finland (Viinikainen, Kokko, Pulkkinen, & Pehkonen, 2010) Indeed, there are some evidence that non-cognitive skills outperform cognitive skills in predicting academic success (Duckworth & Seligman, 2005)

Besides schooling and earnings, cognitive and non-cognitive skills also bring other benefits For instance, in the work of Hanushek and Woessmann (2008), cognitive skills were proved to have a powerful impact on economic growth They used data from International Association for the Evaluation of Educational Achievement (IEA) economic indicators from the OECD countries for investigation, and the results were remarkably robust In terms of non-cognitive skills, they are important predictor not only for academic and career success but also for life well-being such as health behaviors, marriage, and criminal offending A study by Moffitt

et al (2011) provides solid evidence on life-wellbeing The study is extraordinary since it followed cohort of 1000 children in Dunedin, New Zealand from birth to age

8

32 The results showed that children with low self-control are more likely to have health problems (27% vs 11%), commit crime (43% vs 13%) and be single-parent (58% vs.28%)

There are several factors affecting the skill development process The factors can be put into two groups: genetic factors and the inputs from the environment (such

as home, school and community) A model in Figure 1 illustrates the role of parents

in setting these factors in their children GP, GC denotes genetic factors of parents and children respectively

A child’s characteristics (PC) are determined through genes being transmitted from parents At the same time, parental characteristics (PP) affect the children through providing inputs into the living environment of the children (EC)

𝐺𝑃

𝐺𝐶

9

40% to 70% of IQ variance Likewise, results from a study on over 12,000 Swedish twin pairs suggest that genetic factors can account for half of the variance in personality (Floderus-Myrhed, Pedersen, & Rasmuson, 1980)

2.3.2 Environment

Secondly, environment is crucial in determining abilities In facts, heritability discussed in the previous section can be diluted by different environmental circumstances Since inputs from the environment can causes genes to express themselves differently, environment factors may outweigh genetics factors

2.3.2.1 Home environment:

Parental investment

Parents play a vital role in skills development of a child, especially during the early stages of life Inputs from parents or so-called parental investments involve both financial and non-financial elements On one hand, parental investment is commonly regarded as expenditure being spent on child’s well-being such as education, health and clothing Otherwise, in many of the cases, family income may be taken into consideration as a proxy for parental investment The empirical literature supports that families with higher level of permanent income on average invest more in their children and have children with greater skills (Carneiro & Ginja, 2016; Francesconi

& Heckman, 2016) Moreover, the divergence in cognitive and non-cognitive skills among children growing up in families with different level of income is observed at early ages of childhood (J Heckman and Carneiro (2003); Cunha, Heckman, Lochner, and Masterov (2006)) However, the empirical challenge in taking family permanent income as a proxy for parental investments is that the level of income is highly correlated with family background factors such as parental education and skills Therefore, caution needs to be taken when interpreting the effect of permanent income on skill development On the other hand, parents invest not only money in their children but also time and effort The importance of parental time investments

to child development has been mentioned in early literature such as Hill and Stafford

10

(1974), Leibowitz (1974) and Leibowitz (1977) Despite that fact, empirical evidence

on parental time investments is still limited due to data constraint Some recent research demonstrating a positive relationship between parental time and child cognitive skills are studies of Carneiro and Rodrigues (2009) and Hsin and Felfe (2014) A few studies look at the impact of parental time on non-cognitive studies The work of Bono et al (2016) is one of the most recent study which shows that the more time mothers spend with their children the higher level of cognitive and non-cognitive skills children (aged 3 to age 7) attain In general, both money and time should be taken into consideration when measuring parental investment

Besides parental investment, there are several other home factors which affect child development such as parental education, household size and number of child’s siblings

2.3.2.2 School and neighborhood

As children grow up, the influences of parents on them are diluted because the older the children get, the freer they are to select their own environment Apparently, school and neighborhood play a vital role in determining the well-being of children, especially among older children or adolescence Based on sociology literature, there are several mechanisms in which neighborhood may affect child development Empirically, some models predict that higher socioeconomic status neighbors have a negative effect on children development, while other models predict the opposite (Mayer & Jencks, 1989) A research on Vancouver neighborhoods by Hertzman (2004) revealed strong relationship between children’s development and the socio-economic status and other features of the neighborhoods in which they lived Thus, socioeconomic status of the neighborhood is taken into consideration when estimating child development Although reported neighborhood effect on child outcome are often modest, controlling for the effect is essential In addition, the longitudinal property of Young Lives survey may allow us to see whether neighborhood effects hold for both young children and adolescents as is found in a study of Brook et al (1993)

11

2.4 Dynamics of skill development

According to J Heckman and Carneiro (2003), life cycle skill formation is a dynamic rather than static process The underlying mechanism of the process are self and cross productivity In order words, a skill acquired at one stage can affect the acquisition of that skill (or another skill) at a later stage of life

In the early literature, the role of non-cognitive ability is neglected, and only cognitive ability is considered as a part of human capital accumulation process With

a growing body of psychology literature, economists have recognized the importance

of non-cognitive ability in human capital and integrated it into economic model (see Borghans (2008); Almlund, Duckworth, Heckman, and Kautz (2011) for a profound review of the impact of non-cognitive ability) Heckman and his co-authors proposed

a model of technology which capture the interaction between cognitive and cognitive skills

non-Empirical evidence for self-productivity is straightforward while that for cross- productivity is a mix Applying the formulation of technology of skill formation, self-productivity of skills is found in various studies such as that of Cunha and Heckman (2008) using US Longitudinal dataset ; that of Helmers and Patnam (2011) using Young Lives India Survey; or that of Coneus, Laucht, and Reuß (2012) using Germany Manheim study of children Regarding cross-productivity, Cunha and Heckman (2008) found that cognitive skills in one period can be reinforced by the accumulation of non-cognitive skills in the previous period; meanwhile, evidence for the reverse case was not found In contrast, Helmers and Patnam (2011) found the evidence in favor of cross-productivity of cognitive on non-cognitive skills but not vice versa

12

Figure 2: A framework of skill development

Source: Kautz et al (2014)

As discussed in the previous sections, family together with extrafamilial inputs play a vital role in shaping skills, especially during childhood Indeed, development process relies on the accumulation of these inputs and the reinforcement of skills A general framework of skill formation is illustrated in figure 2

2.4.1 Critical and sensitive periods: timing of parental investment

In Figure 2, child development is treated as a single period However, Carneiro, Cunha, and Heckman (2003) argued that there should be a distinction between early and late childhood since the effects of family and environment inputs

on children during different periods might be different Indeed, based on the strength

of the effect of parental investments, Heckman introduced the concept of critical and sensitive periods

13

A sensitive period is a period in which the parental investments are more productive than in other periods A critical period is the only period in which parental investments are productive Phillips and Shonkoff (2000) emphasize that different stages of the life cycle are critical to the formation of different types of abilities

Most available evidence support that early period of life is important to cognitive development Hopkins and Bracht (1975) suggested that before age 10 is sensitive period for cognitive development Recent evidence from Coneus et al (2012), using Mannheim Study of Children in Germany, also demonstrated that early childhood is a sensitive period for cognitive skill relative to late childhood and that critical period to cognitive skills is the first 4 years of a child

Regarding non-cognitive skills, empirical evidence showed that the sensitive periods for noncognitive skills occur later than do sensitive periods for cognitive skills (Cunha and Heckman (2007); Borghans (2008)) In particular, Carneiro et al (2003) found that noncognitive skills are more malleable at later ages while cognitive skills are quiet stable by the age of eight In addition, evidence from Almlund et al (2011) showed that noncognitive skills are more responsive to interventions during adolescent than in early childhood

As a result, researchers suggest that intervention program to disadvantaged children should be carried out in child early childhood Otherwise, intervention during adolescents and adulthood should focus more on noncognitive skills instead

of cognitive skills

2.5 Analytical framework

Based on the framework of skill development presented in the previous chapter, figure 3 illustrates the analytical framework being used in this study The framework allows us to investigate the effect of investments on skills; the cross-productivity and self-productivity effects; and the critical and sensitive periods of skill formation Empirical evidence for the relationships depicted in the diagram are discussed in chapter 2 The works from Cunha and Heckman (2008), Coneus et al (2012), Helmers and Patnam (2011) and Hernández-Alava and Popli (2017) provide evidence for self-productivity, cross productivity and the impacts of parental investments on

14

skill development The first two papers examine the context of developed countries while the latter examines a case of developing country In terms of child’s characteristics, genetic factors are found to account for a large part of the variance in

IQ and personality in the study of Scarr and Weinberg (1983) and Floderus-Myrhed

et al (1980) Besides, parental investment, other inputs from the household such as household size, household economic status, household social networks are worth to take into account when investigating the skill development, as suggested by Duncan and Brooks‐Gunn (2000); Pelto et al (1991) and Fafchamps and Gubert (2007) In terms of community characteristic, the study includes neighborhood social economic status and social problem into the analysis of skill formation given the literature from (Mayer & Jencks, 1989), (Brooks-Gunn et al., 1993)

Figure 3: Analytical framework

Child characteristics

Household characteristics

Community characteristics

Cognitive

Cognitive

Non-cognitive

Non-cognitive

15

Chapter 3

3.1.1 Technology of skill formation

Cunha and Heckman (2007, 2008) propose a model to investigate the dynamics

of skill formation where a child’s current level of skills is explained by his past level

of skills, parental investment and other contemporaneous variables such as caregiver, household and community characteristics Since the time span between each period

in the sample is 3-4 years, this thesis assume that parental investments have an immediate impact on skills ‘development, which is in line with Cunha et al (2006) The general technology of skill formation is:

𝜃𝑡𝑘 = 𝑓𝑡(𝜃𝑡−1𝑘 , 𝐼𝑡 , 𝑋𝑡) (1) where 𝜃𝑡𝑘 denotes a child’s level of skill k for age t, 𝐼𝑡 denotes parental investment

at age t (𝑡 ∈ {1, … 𝑇}), and 𝑋𝑡 denotes a vector of child, caregiver, household and community characteristics

In accordance to Cunha and Heckman (2007), the function in (1) is specified as

a linear specification as below:

This model allows for the examination of the interaction between cognitive and non-cognitive skills, between current and past level of skills, in order words examining the self-productivity and cross productivity

16

In terms of critical and sensitive periods, I follow the mathematical definition

of Cunha et al (2006) Period t is defined as a critical period for skill 𝜃𝑡𝑘 if

Equation (3) demonstrates that t is critical period for the development of skill k

if parental investment is only productive in period t but not in any other period t+j≠0 Meanwhile, equation (4) shows that t is a sensitive period relative to period t+j if, at the same level of skill 𝜃̅̅̅ and inputs i, parental investment is more productive in 𝑘period t than in another period t+j

Empirically, to identify critical periods, we look at the periods when coefficients for investments are significant different from zero To identify sensitive periods, we estimate the difference of 𝑏𝑡𝑘 − 𝑏𝑡+𝑗𝑘 Bootstrap is used for the difference to check the hypothesis if 𝑏𝑡𝑘− 𝑏𝑡+𝑗𝑘 > 0 t is a sensitive period if the hypothesis cannot be rejected

Due to the unavailability of data on cognitive and non-cognitive skills at the age

of 1, in round 1, child health is estimated against psychosocial risk factors (PRF) According to Walker et al (2007), PRF comprise of parenting factors (caregiver sensitivity) as well as contextual factors (maternal depression and exposure to violence) It is expected that high PRF adversely affects child health, leading to child stuntedness Then, child health enters the next period as an independent variable, explaining the level of skills at the age of 5 Since, the instruments used to measure

3.1.2 Genetic endowment – Initial conditions

As previously mentioned in section 2.2.1, genetic endowment is a vital determinant of skill development The genetic endowment is reflected through an individual initial skill set denoted as 𝜃0 = (𝜃𝑜𝐶, 𝜃0𝑁) However, this endowment is unobserved and heterogeneous across children Instead of using first differences suggested by Todd and Wolpin (2007) to capture the unobserved characteristics of genetic endowment, this study follows Helmers and Patnam (2011) to use several control variables such as child height and weight, maternal health during pregnancy and child birth’s location Height is one of the control variables since the researchers show that genes may explain up to 90% of variation in height (Weedon et al., 2007)

18

3.1.3 Method in cognitive, non-cognitive skills measurement: Latent variables

There are several measures for cognitive and non-cognitive skills Thus, latent variables are used instead of a single measure Latent variables are variables that are not directly observed but are inferred from other indicators that are observed Given the observed indicators, latent variables are estimated using confirmatory factor analysis A one-factor model is employed to estimate the latent variables The model

is specified as below:

𝒁𝑖,𝑡𝐶 = 𝛼𝑖,𝑡𝐶 + 𝛽𝑖,𝑡𝐶𝜃𝑡𝐶 + 𝜀𝑖,𝑡𝐶 (11a)

𝒁𝑖,𝑡𝑁 = 𝛼𝑖,𝑡𝑁 + 𝛽𝑖,𝑡𝑁𝜃𝑡𝑁 + 𝜀𝑖,𝑡𝑁 (11b)

𝒁𝒊 denotes different observed indicators of the latent variable 𝛼𝑖,𝑡 and 𝛽𝑖,𝑡 represents

a measure-specific intercept and factor loading respectively 𝜃 and 𝜀 are unobserved The assumptions behind the model are: (1) the error terms 𝜀 are mutually independent and independently distributed over time and across children, (2) the factor 𝜃 and error term 𝜀 are uncorrelated and have an expected value of zero, (3) the factor 𝜃 and observed variables 𝑍 are assumed to be linearly related

3.1.4 Instrument Variables

Although endogeneity due to measurement error can be mitigated by the estimation of latent variables, endogeneity may still arise due to the potential simultaneity between parental investment and child’s skill level Parental investment and skill level may be codetermined On one hand, higher level of parent al investment may result in higher level of skill On the other hand, a child with a higher level of skill may lead parents to reduce or increase the investment on that child To avoid the simultaneity bias, IV strategy is employed

Learning from Helmers and Patnam (2011), I choose household specific shocks and a child birth’s order as instruments The argument for the validity of the

Likewise, parental investment in children is likely to be affected by child birth order Parental investment is believed to act as a channel for birth order to pose an impact on child development instead of a direct impact Parents tend to adjust investment towards first child and later-born children The empirical evidence from Lehmann, Nuevo-Chiquero, and Vidal-Fernandez (2016) suggest that changes in parental investment can be explained using birth order and that parental investment are plausible explanation for differences in cognitive achievement among different birth orders

3.2.1 Data source

The data is a panel data retrieved from Young Lives study in Vietnam through four survey rounds in 2002, 2006, 2009 and 2013 The survey follows two cohorts of children The older cohort comprises 1,000 children who were born in 1994-1995 and the younger cohort comprises 2,000 children who were born in 2001-2002 Young Lives surveys were conducted in five provinces (Lao Cai, Hung Yen, Da Nang, Phu Yen, Ben Tre) which represent five out of nine socio-economic regions of Vietnam

20

(North-East Region, Red River Delta, Cities, South Central Coast and Mekong River Delta) The main characteristics of the five selected provinces are summarized in the Table 14 in the appendix In each province, four sentinel sites are selected Among which, two are from the poor group, one from the average and one from the above-average group This selection of sentinel sites leads Young Lives survey to oversample the poorer household as compared to other national surveys such as Demographic and Health Survey (DHS) or Vietnam Household Living Standards Survey (VHLSS) Despite this fact, Young Lives dataset is child-focused and aims to cover the diversity of children in the country (Young Lives, 2014) Then, within each sentinel site, 50 children born in 1994-95 and 100 children born in 2001-02 were selected using random sampling technique The children are tracked throughout all survey rounds The attrition rates (drop-out rates) is 3.6% and 11.3% for the younger cohort and older cohort respectively This study employs only the younger cohort The survey questionnaire for the younger cohort will be discussed below

The survey comprises of a core child questionnaire, a household questionnaire and a community questionnaire First, given the differences in child characteristics over various stages of life, the child questionnaire is adjusted in every round Overall, the questionnaire for each child includes sections on school and work; time-use; health; feeling, attitudes and perceptions; social networks, social skills and social support; and cognitive tests The child questionnaire is only available to children from age 8 onwards For children age 1 and age 5, the questions were answered by the caregivers Second, the household questionnaire investigates household circumstances and includes sections on parental background; household and child education; livelihoods and asset framework; household food and nonfood consumption and expenditure; social capital; economic changes and recent life history; socio-economic status; Child health; anthropometry; caregiver perceptions and attitudes At age 1, household questionnaire also includes information on the pregnancy, delivery and breast-feeding on the mother Finally, community questionnaire is answered by community representatives to collect information about

21

the characteristics of the locality such as the economy and issues affecting well-being

of the children living in the community

3.2.2 Variables description

Table 1 presents the description of variables The dependent variables are cognitive skills, non-cognitive skills and child health (in round 1) These variables are latent variables being estimated using confirmatory factor analysis (CFA) CFA

is favorable since it allows us to estimate one-factor model and test the validity of the model The primary independent variables are expenditures, quality of relationship between caregivers and the child, and PRF (for round 1 only) In terms of control variables, children, caregivers, household and neighborhood characteristics are included

The estimation of cognitive abilities is based on the results from cognitive tests which had gone through pilot testing to ensure their suitability and validity The cognitive tests being used are Peabody Picture Vocabulary Test (PPVT), Cognitive Developmental Assessment (CDA), Early Grade Reading Assessment (EGRA), Mathematics test and Reading comprehensive test The PPVT test assesses the vocabulary acquisition of a person from 2.5 years old to adulthood 1 CDA is a cognitive test developed specifically to assess cognitive achievement of pre-primary children CDA has several subtests but only quantity subtest is used in YL2 EGRA

is designed to measure the literacy acquisition of children in their early grades 3 The math test includes two sections which aim to measure basic quantitative and number

1 PPVT test is un-timed and orally administered The examinee selects the picture that best represents the meaning of a stimulus word presented orally by the examiner (Cueto et al., 2009)

2 The quantity subtest asks the children to pick an image from a selection of three or four that best reflected the concept verbally presented by the examiner (i.e few, most, nothing, etc.)

3 The EGRA adapted for YL has three subtests: familiar-word identification, passage reading and listening comprehension The final score comprises of number of correct words read in a minute in familiar-word identification; number of correct words read in a minute in passage reading; number of correct responses to reading comprehension items in passage reading; and number of correct

responses to reading comprehension and listening comprehension (Cueto & Leon, 2012)

22

notion; and ability to work on basic mathematics operations with number These cognitive tests are not available in round 1 since the child is too young In round 2, at the age of 5, PPVT and CDA are used The reliability and validity of the two tests in round 2 are confirmed in a paper by Cueto et al (2009) In round 3, at the age of 8, the tests for cognitive skills are PPVT, EGRA and a math test In round 4, at the age

of 12, the children take part in PPVT, a math test and a reading test

Regarding the measurements of non-cognitive skills, they are estimated based

on a Likert-scale self-administered questionnaire This questionnaire is only available

to 1-year old and 8-year old children, who are able to answer the questionnaire by themselves The indicators being measured are self-esteem (based on Rosenburg 1965), self-efficacy (based on Rotter 1966 and Bandura 1993), friendliness and social trust The questions are adjusted to be relevant to children A full set of questions for each indicator is presented in the appendix (Table 15) For each indicator, there are 4-8 questions The validity of these measures is confirmed by Dercon and Krishnan

(2009)

Table 1: Variable description

Measurement scale

𝜃𝐶𝐻 Factor score

23

Measurement scale

to child stunted and underweight

PRF Factor score

Independent variables

Clothing expenditure Spending on clothing and footwear

for index child, based on the assumption that all children in the household receive an equal share

of spending

‘000 VND

Education

expenditure

Spending over the last 12 months

on education for index child, based

on the assumption that all children

in the household receive an equal share of spending

‘000 VND

Presents or treats

expenditure

Spending over the last 12 months

on present or treats for index child, based on the assumption that all children in the household receive

an equal share of spending

Index

Other Control variables

Child siblings Number of siblings of the child nsiblings Count Preschool 1= Attend preschool 0 = Do not

Child years of

schooling

Number of schooling year

yrsch Count Parental altruism Normalized on a scale from 0-1 A

combination of the questions on Paltruism Index

24

Measurement scale

why parents feel it is important to have children (Only available for Round 2)

Caregiver education Education level of child’s primary

caregiver (category: none, primary, secondary, highschool or vocational, university or higher)

Careed1 Careed2 Careed3 Careed4

community, how active it is and level of trust within the

alcoholism )

CM SSP Index

Instruments

Household shocks The household experience any

economic shocks that resulted in serious loss of wealth over the last four years In round 1, a dummy variable is used In round 2, 3, 4, household shock represents the change in household wealth over two periods

hhshock Dummy/

Continuous

Child birth order The order of the child birth in the

household relative to his siblings birthorder Nominal