ScHARR Public Health Evidence Report 9.4 Promoting the social and emotional wellbeing of vulnerable pre-school children Economic outcomes of early years programmes and interventions designed to promote cognitive, social and emotional development among vulnerable children and families. Part 1 - Econometric analysis of UK longitudinal data sets Authors: Mónica Hernández Alava, Gurleen Popli, Silvia Hummel, Jim Chilcott Public Health Collaborating Centre
30
Embed
ScHARR Public Health Evidence Report 9/file/preschool... · ScHARR Public Health Evidence Report 9.4 ... and the Millennium Cohort Study ... Jon Johnson at the Centre for Longitudinal
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ScHARR Public Health Evidence Report 9.4
Promoting the social and emotional wellbeing of vulnerable pre-school children
Economic outcomes of early years programmes and interventions designed to promote cognitive, social and emotional development among vulnerable
children and families.
Part 1 - Econometric analysis of UK longitudinal data sets
Authors: Mónica Hernández Alava, Gurleen Popli, Silvia Hummel, Jim Chilcott
Public Health Collaborating Centre
2
Commissioned by:
NICE Centre for Public Health Excellence
Produced by: ScHARR Public Health Collaborating Centre
Correspondence to: Vivienne Walker School of Health and Related Research (ScHARR) University of Sheffield Regent Court 30 Regent Street Sheffield S1 4DA [email protected]
This report should be referenced as follows: Mónica Hernández Alava, Gurleen Popli, Silvia Hummel, Jim Chilcott. (2011) Economic outcomes of early years programmes and interventions designed to promote cognitive, social and emotional development among vulnerable children and families. Part 1 - Econometric analysis of UK longitudinal data sets. ScHARR Public Health Evidence Report 9.4
About the ScHARR Public Health Collaborating Centre
The School of Health and Related Research (ScHARR), in the Faculty of Medicine, Dentistry and
Health, University of Sheffield, is a multidisciplinary research-led academic department with
established strengths in health technology assessment, health services research, public health,
medical statistics, information science, health economics, operational research and mathematical
modelling, and qualitative research methods. It has close links with the NHS locally and nationally
and an extensive programme of undergraduate and postgraduate teaching, with Masters courses
in public health, health services research, health economics and decision modelling.
ScHARR is one of the two Public Health Collaborating Centres for the Centre for Public Health
Excellence (CPHE) in the National Institute for Health and Clinical Excellence (NICE) established
in May 2008. The Public Health Collaborating Centres work closely with colleagues in the Centre
for Public Health Excellence to produce evidence reviews, economic appraisals, systematic
reviews and other evidence based products to support the development of guidance by the public
health advisory committees of NICE (the Public Health Interventions Advisory Committee (PHIAC)
and Programme Development Groups).
Contribution of Authors
Mónica Hernández Alava, Gurleen Popli, Silvia Hummel were the authors. Jim Chilcott was the
senior lead.
Acknowledgements
This report was commissioned by the Centre for Public Health Excellence of behalf of the National
Institute for Health and Clinical Excellence. The views expressed in the report are those of the
authors and not necessarily those of the Centre for Public Health Excellence or the National
Institute for Health and Clinical Excellence. The final report and any errors remain the
responsibility of the University of Sheffield. Elizabeth Goyder and Jim Chilcott are guarantors.
The analyses in this work are based wholly or in part on analysis of data from the 1970 British
Cohort Study (BCS) and the Millennium Cohort Study (MCS). The data was deposited at the UK
Data Archive by the Centre for Longitudinal Studies at the Institute of Education, University of
London. BCS and MCS are funded by the Economic and Social Research Council (ESRC). Data
on the highest educational qualification obtained by cohort members in the BCS was provided by
Jon Johnson at the Centre for Longitudinal Studies.
4
Contents
1 Introduction 5
2 Econometric analysis: datasets and methodology 5
2.1 Choice of data set 5
2.2 Early childhood: analysis from the MCS 7
2.3 Long run outcomes: analysis from the BCS 12
3 Results 14
3.1 MCS Results 14
3.2 BCS Results 16
3.3 Limitations of the analysis: 17
4 References 19
Appendix 1 Schematic diagram representing the econometric analysis methods 21
Appendix 2 Econometric models: specification and derived equations for predictions 22
Appendix 3 Millennium Cohort Study Question Response Categories 23
Appendix 4 Millennium Study Model Results 25
Appendix 5 BCS Model Results 29
5
1. Introduction
It is now widely accepted that the ‘early years’1 matter, with children from disadvantaged
backgrounds having a lower probability of completing their education, higher probability of being
involved in crime and lower life time earnings potential (Patterson et al 1990; Heckman and
Masterov 2007). Poverty is often associated with lower cognitive development and higher
behavioural problems in children as young as 5 years (Bor et al 1997; Feinstein 2003).
Interventions have been trialled in the UK with the aim of improving outcomes for infants,
particularly from vulnerable populations. These studies are the subject of an accompanying
systematic review (Systematic review of UK evaluation studies of the effectiveness of early years
programmes and interventions designed to promote cognitive, social and emotional development
among vulnerable children and families). All of these studies had limited follow-up of between 12
and 18 months, the majority reporting outcomes when children were still in infancy. Analysis of
longitudinal data of children through to adulthood was therefore undertaken with the following
objectives:
- to understand the factors determining the formation of ability in early childhood;
- establish a link between early childhood development and adult outcomes;
- to allow the effects of childhood interventions on long term outcomes to be predicted.
To accomplish the first two objectives, two econometric models were developed and estimated
using two nationally representative UK longitudinal data sets, the Millennium Cohort Study (MCS)
and the 1970 British Cohort Study (BCS). The resulting models were used to predict outcomes at
age 5 and at age 38 (details in Appendix 2), with and without interventions to be able to fulfil the
third objective. The final step was the development of a mathematical model to estimate the
economic consequences of the interventions. The use of the models developed in the econometric
analysis in an economic model to determine the long term outcomes of the intervention are
described in a separate document - Part 2.
1 While it is widely accepted that ‘early years’ matter, what we mean by ‘early years’ is not well defined.
Figures often quoted in the literature are: ‘from conception to age six’ (McCain and Mustard 1999), ‘first eight years’ (Cunha et al 2006); and ‘up to the age of ten’ (Hopkins and Bracht 1975).
6
2. Econometric analysis: datasets and methodology
2.1 Choice of data set:
There are four datasets for the UK which follow children from birth. In reverse chronological order,
these are:
MCS:
- This dataset is nationally representative;
- has information on relevant measures from birth;
- is the most recent and thus most relevant for the children growing up now.
BCS:
- This dataset is nationally representative;
- does not have the relevant information from birth (so cannot be used for modelling
childhood outcomes): it started collecting information on cognitive and non-cognitive
measures from the age of 5;
- is the most recent available for the adult outcomes.
Avon Longitudinal Study of Parents and Children (ALSPAC) 2:
- This dataset is not nationally representative (has under representation of poor and non-
white families);
- does not have as detailed measures of cognitive and non-cognitive assessment of children
as MCS.
The 1958 National Child Development Study (NCDS) 3:
- This dataset is not nationally representative (has under representation of ethnic minorities);
- started collection of information on ‘educational and social development’ only from age 7;
- has longer follow up for modelling adult outcomes, but is from an earlier cohort than BCS,
reflecting less well the UK society of today.
For the analysis undertaken here we use the MCS and BCS. The two main reasons for the using
these datasets are their epidemiological coverage (i.e. these surveys are nationally representative)
and their temporal applicability (i.e. these are the most recent surveys, so most relevant for the
current population). Furthermore, the MCS is the only dataset which follows children from birth
and oversamples children from disadvantaged backgrounds, which means it is the only dataset
which gives us enough observations to perform an empirical analysis on disadvantaged/vulnerable
children.
2 For details on ALSPAC see www.bristol.ac.uk/alspac.
problems in the mother, number of other children in the household, longer TV watching times,
and being in poverty when young are associated with a higher probability of a teenage
pregnancy. The reading frequency to the child, mother’s education and the levels of cognitive
development are associated with lower probabilities of receiving benefit at age 38.
- Economically active (males only). This relationship has a number of highly insignificant
variables. However, this is expected given the high proportion of males in the sample who are
economically active. Higher behavioural problems and being in poverty when young are
associated with a lower probability of being economically active at age 38. The reading
frequency to the child and the levels of cognitive development as measured by the Copying
Designs Test are associated with a higher probability of being economically active at age 38.
The level of cognitive development as measured by the English Picture Vocabulary Test seems
to have a counterintuitive sign although it does not appear to be significant as conventional
significance levels.
3.3 Limitations of the analysis:
Apart from the usual noise/ error associated with estimated parameters in any econometric
analysis, the following are the limitations of this analysis that should be kept in mind:
18
- Other than the test scores for cognitive ability, a lot of the data used here is self-reported,
mainly by mother. Some of the variables are more prone to systematic misreporting than
others.
- The relationship between the observables (e.g. test scores) and the unobservables (e.g. latent
cognitive ability of the child), is fraught with measurement errors in both the MCS and BCS
analysis. While the approach used in the MCS analysis (dynamic factor analysis) limits this
problem it cannot completely eliminate it.
- The skill production function (i.e. the relationship between cognitive development at age 3 with
that at age 5; and similarly for the behavioural development) is assumed to be linear (results in
Table 1). This may not be the case. (However, evidence from the simulations carried out by
Cunha and Heckman (2008) seem to indicate that the linear production function is not a limiting
assumption).
- Endogeneity of inputs: in the MCS model we have assumed that parenting behaviour
influences child’s development; it could be that child’s development is influencing the parenting
environment (see Kiernan and Huerta, 2008). For example reading more to the child is
associated with a higher cognitive score of the child; but it could be that parents read more to
the child who is inherently more able and interested. Similarly, endogeneity problems are also
present in the BCS analysis and need to be considered in the interpretation of the results.
Taking endogeneity into account will tend to reduce the size of the long run effect.
- Long-run linkages: we are using the information in the BCS at age 5, to make long-run
predictions (age 38) because it is the only age the two surveys have in common. One can not
assume a linear trajectory from age 5 to age 38 because at age 5 children are still developing
at a very high pace. Studies by Cunha and Heckman (2008) and Heckman et al (2006)
indicate that the best possible way to make long run predictions is:
o Make them frequently before the mid-teens. Cunha and Heckman (2008) make
predictions in three stages: from ages (I) 6-7 to 8-9, (II) 8-9 to 10-11 and (III) 10-11
to 12-13.
o From mid-teens the long run predications make more sense. Heckman et al (2006)
make predictions from age 14 to age 30. Here too however the analysis has to take
into account the endogeneity of the choices related to: schooling, occupational
choices, fertility etc.
19
4 References Bor, W., J.M. Najman, M.J. Andersen, M. O’Callaghan, G.M. Williams, and B.C. Behrens, (1997), ‘The Relationship Between Low Family Income and Psychological Disturbance in Young Children: An Australian Longitudinal Study’, Australian and New Zealand Journal of Psychiatry, 31(5): 664–75.
Cunha, F. and J.J. Heckman (2006), ‘Investing in Young People,’ Working Paper, University of Chicago.
Cunha, F. and J.J. Heckman (2007), ‘The technology of skill formation,’ American Economic Review, vol. 97(2), pp.31-47.
Cunha, F. and J.J. Heckman (2008), ‘Formulating, Identifying and Estimating the Technology of Cognitive and Noncognitive Skill Formation,’ Journal of Human Resources, vol. 43(4), pp. 738-782. Cunha, F., J.J. Heckman, L. Lochner and D. Masterov (2006), ‘Interpreting the Evidence on Life Cycle Skill Formation,’ in Handbook of the Economics of Education, edited by E. Hanushek and F. Welch, pp. 697-812, North Holland: Amsterdam.
Feinstein, L. (2003) ‘Inequality in the Early Cognitive Development of British children in the 1970 Cohort’, Economica 70(277): 73–97. Heckman, J.J., J. Strixrud and S. Urzua (2006), ‘The Effects of Cognitive and Noncognitive Abilities on Labor Market Outcomes and Social Behavior, Journal of Labor Economics, vol. 24(3), pp. 411-482.
Heckman, J.J. and D.V. Masterov (2007), ‘The productivity argument for investing in young children,’ Review of Agricultural Economics, vol. 29(3), pp.446-493.
Hopkins, K.D. and G.H. Bracht (1975), ‘Ten-Year Stability of Verbal and Nonverbal IQ Scores,’ American Educational Research Journal, Vol. 12(4), pp. 469-477. Kiernan, K.E. and M.C. Huerta (2008), ‘Economic deprivation, maternal depression, parenting and children’s cognitive and emotional development in early childhood,’ The British Journal of Sociology, vol. 59(4), pp. 783-806. Kiernan, K.E. and F.K. Mensah (2010), ‘Poverty, family resources and children’s early educational attainment: the mediating role of parenting,’ British Educational Research Journal, available online 26th February 2010. McCain, M.N and J.F. Mustard (1999), Reversing the Real Brain Drain: Early Years Study, Final Report, Toronto: Ontario Children’s Secretariat.
Patterson, C.J., J.B. Kupersmidt and N.A. Vaden (1990), ‘Income level, gender, ethnicity, and household composition as predictors of children’s school-based competence,’ Child Development, vol. 61(2), pp.485-494.
Todd, P.E. and K.I. Wolpin (2006), ‘The Production of Cognitive Achievement in Children: Home, School and Racial Test Score Gaps,’ Working Paper, University of Pennsylvania.
Ford, R.M., S.J.P. McDougall, and D. Evans (2009). Parent-delivered compensatory education for children at risk of educational failure: Improving the academic and self-regulatory skills of a Sure Start preschool sample. British Journal of Psychology, 100(Pt 4), 773-797.
Toroyan, T., A. Oakley, G. Laing, I. Roberts, M. Mugford, and J. Turner (2004). The impact of day care on socially disadvantaged families: an example of the use of process evaluation within a randomized controlled trial. Child: Care, Health and Development, 30 691-698.
Wiggins, M., A. Oakley, I. Roberts, H. Turner, L. Rajan, H. Usterberry, R. Mujica, and M. Mugford, (2004). The Social Support and Family Health Study: a randomised controlled trial and economic evaluation of two alternative forms of postnatal support for mothers living in disadvantaged inner-city areas. Health Technology Assessment, 8(32), 1-+.
Plewis, I and C. Kallis, (2008). Changing economic circumstances in childhood and their effects on subsequent educational and other outcomes. DWP Working Paper No. 49. http://www.dwp.gov.uk/asd/asd5/report_abstracts/wp_abstracts/wpa_049.asp
Appendix 1 Schematic diagram representing the econometric analysis methods
22
Appendix 2 Econometric models: specification and derived equations for predictions. Please see attachment
23
Appendix 3 Millennium Cohort Study Question Response Categories Questions are answered by the main carer of the child (which in the majority of cases is the
mother, and will be referred to as such). Unless otherwise specified the questions are asked at
both age 3 and age 5:
- How often do you read to the child?
0 = Not at all
1 = Once/twice/less a month
2 = once/twice a week
3 = Several times a week
4 = Every day
- How often does child paint/draw at home?
0 = Never
1 = Occasionally
2 = once\twice a week
3 = Several times a week
4 = Every day
- How often help child learn alphabet?
The above question is asked at age 3. Its equivalent question for age 5 is: how often is the
child helped with reading?
0 = Never
1 = Occasionally
2 = once\twice a week
3 = Several times a week
4 = Every day
- How often is the child helped with writing? This is asked only at age 5.
0 = Never
1 = Occasionally
2 = once\twice a week
3 = Several times a week
4 = Every day
24
- Regular bedtime:
At age 3 the question asked is: child has regular bedtimes?
At age 5 the question asked is: regular bedtime on term-time weekdays?
0 = never or almost never
1 = sometimes
2 = usually
3 = always
- TV watching:
At age 3 the question asked is: hours a day child watches tv/videos?
At age 5 the question asked is: hours per term-time weekday watching tv/dvd?
0 = None
1 = Up to one hour
2 = [1,3) hours
3 = >3 hours
- Smack child if being naughty?
0 = Never
1 = Rarely
2 = >1 a month
- Shout at child if being naughty?
0 = Never
1 = Rarely
2 = >1 a month
Question asked of the father/father figure in the household:
- How often do you read to the child?
0 = Not at all
1 = Once/twice/less a month
2 = once/twice a week
3 = Several times a week
4 = Every day
25
Appendix 4 Millennium Study Model Results
Please refer to end of this Appendix for variable name descriptors.