We are applying to use ALSPAC data as part of a large project for which we will be requesting funding from the ESRC. This project will not focus solely on ALSPAC data, but will also use data from the Millennium Cohort Study (MCS) which has already been made available to us, and the Infant Feeding Survey (IFS). We envisage our project as comprising four sub-projects. Not all of these will make use of ALSPAC data; however. In the proposal below, we include a brief outline of those parts of the project which involve other data sets, as this provides useful context for the investigations which do use ALSPAC.

There is already an extensive literature on the effects of breastfeeding on a number of outcomes: allergy and asthma; developmental delay in children; childhood blood pressure, and many others. There is also a recognition that standard multivariate analytical techniques may overestimate the positive effects of breastfeeding, due to a mother's propensity to breastfeed being correlated with unobservable (or simply unobserved) factors which are themselves associated with favourable child's outcomes. This problem of endogeneity is likely to lead to the effects of breastfeeding being overestimated. This project aims to carry out a causal analysis of the effects of breastfeeding: that is, an analysis which, as far as possible, overcomes the effects of endogeneity.

The ideal data with which to undertake such a causal analysis of the effects of breastfeeding would arise from a large-scale randomised experiment on infant feeding. However, the ethical considerations involved make such an experiment almost inconceivable. In order to study this issue, therefore, researchers must exploit variation in infant feeding patterns arising from other factors. Here, we propose to use two distinct analytical techniques to examine the extent to which estimates of the effects of breastfeeding are biased if one does not account for the endogeneity of the decision to breastfeed; and to attempt to uncover the causal effect of breastfeeding. We propose four inter-related projects, the first two of which examine children's cognitive outcomes, the third of which examines mothers' mental health, and the last of which examines parents' time off work looking after sick children.

1. The effect of breastfeeding on children's cognitive skills, using an Instrumental Variables approach

This project will identify the causal effect of breastfeeding on children's cognitive development, using an instrumental variables approach. The project will use data from the MCS, and will employ as instrument an exogenous variation in breastfeeding rates determined by whether or not the hospital at which a mother gave birth subscribed to the Unicef Baby Friendly Initiative (BFI). Several studies show that breastfeeding rates (initiation and/or duration) are higher in Baby Friendly hospitals. The IV method relies on the assumption that BFI status is exogeous. We believe this is a reasonable assumption: although mothers in the UK are free to choose the hospital at which they want to give birth, in most cases they are restricted in this choice by travel distance.

2. The effect of breastfeeding on children's cognitive skills, using a Propensity Score Matching approach

This project will also examine the relationship between breastfeeding and children's cognitive skills - but it will use a different data set (ALSPAC) and a different analytical technique (Propensity Score Matching, or PSM). This technique is described by Dehejia and Wahba (2002) and others, and has been used extensively in economics to assess causal effects. Using this technique, the analyst generates two samples of individuals who have and who have not undergone a "treatment" (here, breastfeeding), matched on a wide range of characteristics, as well as by their estimated propensity to breastfeed. The difference in outcomes between the two matched groups provides an estimate of the causal effect of breastfeeding. ALSPAC lends itself well to PSM, containing a wealth of data on parents' attitudes to breastfeeding, their prior intentions to breastfeed, and events during and after birth which may affect whether children are eventually breastfed.

The outcome variables on which we propose to focus are as follows:

* Key Stage 1 SATs results (we would like to use KS2 SATs results as well, but understand these are not yet available

* Results of school entrance assessments from the SATs files

* Teacher assessments: the child's ability group (Year 6); the Year 4 maths test; and teacher ratings of children's ability in Year 3.

* Parents' assessments of their children's development - particularly their assessments of vocabulary and other language-related skills - from the parent-completed questionnaires at 15, 24 and 38 months.

3. Breastfeeding and mothers' mental health: A PSM approach

This section would use data from both ALSPAC and the MCS to examine the relationship between breastfeeding and mothers' mental health. Kendall-Tackett (2007) presents evidence that breastfeeding protects against post-natal depression via two mechanisms: reducing stress, and reducing inflammation. We propose to build on this by investigating the relationship between breastfeeding and maternal wellbeing, using the analytical techniques described above to control for the possible endogeneity of the decision to breastfeed, and to identify the causal relationships involved.

In ALSPAC, the outcome variables of interest would be:

* Crown-Crisp Experiential Scores from birth until 33 months, and subscales where available

* Edinburgh Post-Natal Depression Scores from birth until 33 months

* Other indicators of wellbeing and relationship to child (for example, J151, mum has felt unattached to study child, in the mother's questionnaire at 47 months).

4. Breastfeeding and mothers' work: a cost-benefit analysis

There is a considerable body of research looking at the return to work as a determinant of breastfeeding behaviour. Here, we propose to look at the relationship between breastfeeding and mothers' paid work, by allowing for causal effects in both directions. A mother's decision to breastfeed might influence the time at which she chooses to return to work, but recent research has shown that the length of the period of maternity leave may in turn affect decisions about initiation and duration of breastfeeding.

In addition, we propose to examine the effect of breastfeeding on the number of days parents take off work to look after sick children. Existing research shows that breastfed babies are at lower risk of a number of childhood illnesses; we propose to examine whether this translates to the parents of breastfed children taking less time off work to look after sick children. This is not an easy question to address, since data on this aspect are not available in all waves of ALSPAC. However, by combining data on hospital admissions, the number of illnesses (including ear infections and attacks of gastro-enteritis), and the amount of time parents have taken off work to look after children with ear and other general health problems (asked at 33 and 47 months), we may begin to make some informed assumptions about whether breastfeeding has economic benefits by reducing working parents' absenteeism rates and therefore is beneficial for employers as well as mothers and children.