Rev Econ Household DOI 10.1007/s11150-017-9379-8

The role of grandparenting in early childcare and child outcomes Daniela Del Boca1,2,3,4 Daniela Piazzalunga3,4,5 Chiara Pronzato1,3,4 ●

●

Received: 27 April 2016 / Accepted: 10 June 2017 © Springer Science+Business Media, LLC 2017

Abstract This paper examines the short and medium term impact of early childcare provision by grandparents and formal care settings on child cognitive outcomes, using data from the Millennium Cohort Study (UK). Compared with children placed in formal childcare, children cared for by their grandparents are better at naming objects, but perform worse on tests of basic concept development and non-verbal reasoning. These results mask strong heterogeneities. On the one hand, the positive association between grandparental care and child outcomes is stronger for children from more advantaged households; on the other, the negative association is only significant for those from more disadvantaged households. The results of OLS estimations used for our analysis are confirmed using panel methods and an instrumental variable approach. Keywords

Grandparents Childcare Child cognitive outcomes

JEL Classification

●

●

J13 D1 I21 ●

●

Electronic supplementary material The online version of this article (doi:10.1007/s11150-017-9379-8) contains supplementary material, which is available to authorized users. * Chiara Pronzato [email protected] Daniela Del Boca [email protected] Daniela Piazzalunga [email protected] 1

Department of Economics and Statistics “Cognetti de Martiis”, University of Turin, Lungo Dora Siena, 100A, Torino 10153, Italy

2

CEPR, Washington, DC, USA

3

IZA, Bonn, Germany

4

CHILD Collegio Carlo Alberto, Via Real Collegio, 30, Moncalieri, Torino 10024, Italy

5

FBK-IRVAPP, via S. Croce, 77, Trento 38122, Italy

D. Del Boca et al.

1 Introduction Interest in economic research on early childcare and child development has grown rapidly in recent years. According to the most recent PISA data (OECD 2016a) several advanced countries such as US, UK, and Italy are lagging behind and do not show any sign of improvement in terms of scholastic performance. This has been increasingly being attributed to delayed and inadequate investment—both public and private—in human capital (e.g. Carneiro and Heckman 2003). In most countries, attending pre-primary education is positively associated with better performance at age 15 (OECD 2016b). Cost-benefit analysis of human capital investments targeted towards different age groups indicates that investing in the earliest years yields the best outcomes. Later investments carry higher costs for remediating gaps in scholastic achievement and reducing detrimental behaviour. Not only are investments made during the preschool period less costly, they are also more effective. This is because the individual abilities of young children are more malleable, and early investments have a cumulative effect over time, unlike those made later in life (Carneiro and Heckman 2003; Cunha and Heckman 2007). Early childcare plays an important role in the production of cognitive skills. What determines cognitive ability and behavioural development early in life is of crucial policy importance. While early psychological theories stressed the need for maternal care, more recent studies from the fields of psychology, sociology, and economics show that other childcare arrangements do not necessarily produce negative outcomes. Whereas a substantial body of research has analysed the impact of parental and formal childcare, the influence of other family members such as grandparents has received little attention. However, data available from several countries indicate that grandparents do play an important role in childcare and show that quite a large proportion of grandparents provide some kind of care for grandchildren, even on a regular basis (Hank and Buber 2009; Aassve et al. 2012). For the UK, Gray (2005) estimates that about 30% of families with children under 5 receive some childcare support from grandparents. The objective of this paper is to explore the impact of grandparental and formal childcare on child cognitive outcomes, controlling for other forms of childcare. These are the two most important alternatives to parental care and represent two very different childcare choices. We utilize the Millennium Cohort Study (MCS) for the UK, which provides very detailed information about children, parents, grandparents, childcare choices, as well as several child outcomes. We find that the effect of grandparental care is always in the same direction of parental care, and that children looked after by their grandparents are better at naming objects but worse at nonverbal reasoning and at developing basic concepts compared to children looked after in formal childcare centres. The magnitude of the effects is similar to what has been found in other papers. However, the results are highly heterogeneous according to the socio-economic status of the family: the positive association of grandparents’ care with naming abilities is stronger for households with higher incomes, while the negative association with other abilities is stronger for households with lower incomes. In addition, there is a negative association of grandparents’ care with problem-solving ability only among children from lower income families.

The role of grandparenting in early childcare and child outcomes

2 Literature Much of the growing literature on childcare arrangements and child outcomes over the last few years has been influenced by the seminal work of Todd and Wolpin (2003) and James Heckman and co-authors (e.g. Carneiro and Heckman 2003), who modelled children’s outcomes (cognitive, health, and behavioural) as the result of a production function in which inputs are provided by families as well as by other individuals and institutions (schools, teachers, peers, society). In their framework, child development is considered the outcome of a cumulative process of knowledge acquisition, analogous to a firm production process. These inputs play a very significant role, since cognitive and non-cognitive outcomes are largely determined early in life. Estimates of the impact of maternal employment on several child outcomes, such as attainment and years of schooling, are very different across studies. Blau and Grossberg (1992), Ruhm (2004), and Bernal (2008) report negative effects, while Del Boca et al. (2016) and Vandell and Ramanan (1992) report non-significant impacts. The variability of these results depends on the different datasets used, the institutional background, and the childcare alternative chosen by mothers as a substitute for their own time. For instance, Gregg et al. (2005), using data from the UK, show that it is only the children whose non-parental care consists of informal care (provided by a friend, relative or neighbour) who experience detrimental effects of maternal employment. The use of formal childcare, instead, seems to protect children from these negative effects. These results demonstrate that the quality of parental care and the quality of non-parental care are tightly intertwined in determining the overall impact of maternal employment. Several studies have analyzed the impact of formal childcare attendance on child outcomes. Felfe and Lalive (2012) estimate the impact of childcare attendance between 0 and 2 years of age in West Germany and find that attending prekindergarten has positive effects on both language and social skills, especially for the children of mothers with less schooling. Datta Gupta and Simonsen (2010) analyse the impact of formal childcare in Denmark and find that preschool attendance at age 3 has the same positive effects as home care regardless of the mother’s level of education. Del Boca et al. (2016) have explored the link between formal childcare availability and child outcomes for the Italian case and report a positive impact on child outcomes in primary school. Other studies have focused on childcare provided by other family members as substitutes for the mother’s time. To start with, a large proportion of fathers are increasingly involved in childcare. Del Boca et al. (2014) have analysed the impact of mothers’ and fathers’ inputs in the child development production function, showing that fathers are as productive as mothers in childcare (especially as the children grow older), and make a significant contribution to their children’s cognitive outcomes. Another important substitute for mothers’ care is grandparenting, although the significance of this form of caregiving has received less attention (Fergusson et al. 2008). Bernal and Keane (2011), using US data and focusing on single mothers, have distinguished between formal center-based care, which has no adverse effect, and informal care (grandparents, siblings, relatives, non-relatives), which has a significant

D. Del Boca et al.

adverse effect. As discussed above, similar results emerge for the UK (Gregg et al. 2005). There is a general consensus in the literature that the availability of granparental care increases labor force participation of young mothers (Arpino et al. 2014; Compton and Pollak 2014; Garca-Morán and Kuehn 2017). However, while childcare provided by grandparents may support young parents labor supply, this could be at their own detriments. In a detailed analysis on US data, Rupert and Zanella (2014) show that childcare time provided by grandparents implies a reduction of labor market participation for grandmothers. Recently, scholars have investigated the potential negative impacts of time dedicated to grandchildren on grandparents wellbeing. Brunello and Rocco (2016) show that the addition of 10 hours of childcare per month increases the probability of developing depressive symptoms among grandparents. Arpino and Bordone (2016) find that regular provision of childcare to grandchildren has a significant negative effect on the number of social activities in which grandmothers participate (such as volunteering, engagement in educational or training courses and participation in political or community-related organisations). Ho (2015) discusses the alternatives of subsidizing childcare provided by grandparents or formal care, depending on which generation (parents or grandparents) should work longer. However, arguing for subsidizing one form of care should also take into account the effects of all different forms of care on child outcomes. Very little is known about the role of grandparents in child development. To the best of our knowledge, the only published paper that specifically considers the impact of grandparental childcare is Hansen and Hawkes (2009). Their study uses 2 waves of the Millennium Cohort Study to compare the impacts of different childcare alternatives in families with working mothers on vocabulary development and the school readiness test, as well as on behavioural development. They find that children who receive informal care from the mother’s partner or from grandparents do better on vocabulary tests, but are weaker in terms of school readiness than children who receive formal childcare. They also show that grandparental care results in more behavioural problems. A recent report by Bryson et al. (2012) extends the analysis to the third wave of the Millennium Cohort Study. They find that being cared for by grandparents has no significant effect on naming abilities at age 5, while the impact of grandparental care on behaviour varies according to the mother’s level of schooling. These results imply that the connection between different forms of early childcare and child cognitive outcomes in the short term disappears in the medium term. Our paper introduces several contributions to the literature. First, we consider a longer time frame, which allows us to follow child outcomes until the child is in primary school. Looking at a longer period is important, as the effects may change over time. Second, we include more cognitive outcomes. Third, we extend the analysis to include not only families with working mothers, but also those with non-working mothers. Finally, we attempt to address the potential endogeneity issue arising from the fact that the choice of childcare attended by the child is made by the parents. More specifically, we augment our data by including more details from a greater number of waves (child cognitive outcomes at 3, 5, and 7 years of age) and a larger number of cognitive outcomes. In addition to the School Readiness and the Naming

The role of grandparenting in early childcare and child outcomes

Vocabulary tests used to assess the child’s basic concept development and spoken vocabulary, we include the Picture Similarity and the Pattern Construction tests, which provide measures of the child’s non-verbal reasoning abilities and spatial problem-solving ability, and the Word Reading and Number Skills tests, which are indicators of the child’s language and math in later life. The consideration of more waves allows us to estimate the persistence of the impact in the medium run. This is of particular interest, as other studies show that the cognitive impacts of recent childcare programs often decline or, in some cases, disappear in the medium run (Duncan and Magnuson 2013). Despite this decline, formal childcare tends to have other beneficial effects in the medium-long term, on outcomes such as physical and mental health, criminal activity, earnings, and labour market status. The positive impacts are probably mediated by short-term improvements in non-cognitive skills, which are found to be “important determinants of later-life outcomes” (Elango et al. 2016, p. 259). The reasoning behind extending our analysis to include non-working as well as working mothers is that non-working mothers also rely on formal and informal forms of childcare in order to juggle other responsibilities, such as caring for other children, attending to other family matters, and job search activities. Finally, in order to attempt to address the potential endogeneity of childcare choices, we use panel data methods and instrumental variable techniques.

3 The UK childcare market Historically, the UK childcare system has been much more heterogeneous than in other European countries (Sweden, Denmark, and France), where an affordable, fulltime public childcare system has been in place for several decades. The earliest childcare policies for 3- and 4-year-olds in the UK were introduced around the early 2000s. All 4-year-olds since 2000 and all 3-year-olds since 2005 have been entitled to free, part-time nursery education.1 In 2013, free part-time education was also extended to disadvantaged 2 year olds. Children in our sample are born in 1999/2000, thus they have been affected only by the free early childcare at age four. The fact that younger children have not been influenced by the policies imply that formal childcare at age two and before is mainly determined by parents’ choices. Formal childcare for the under-twos is mainly provided by the private market (West 2006) and mainly serves children from families with a higher socioeconomic status (Table 16 in the Appendix). In our sample, on average 13% of children is in formal care at 18 months, and about 30% at 36 months; by age 5, 97% have experienced formal care. However, there are large differences across income groups, especially for younger children. As can be seen from Table 16, at 18 months the largest care option for children in families in the 1st–9th income quantiles is the mother, who is not working. Next comes “working parents” (10th–14th income quantiles) and grandparents (15th–18th). The percentage of children in formal care increases along the wage distribution, especially in the second half, but formal care is the preferred choice only among top-income families. Additional details about childcare use are provided below, in section 5.6. 1

12.5 h per week for 33 weeks per year until 2010, then 15 h per week for 38 weeks.

D. Del Boca et al.

Strong differences persist across childcare types, their quality and use. Different kinds of childcare providers are available in the UK (public, private, childminders), and they all have very different characteristics in terms of training, wages, qualifications, hours of work, teachers-students ratio, etc. For an extensive description of the characteristics of UK childcare and its evolution over time, see West (2006). West has shown that, in terms of availability, more children were placed in preschool childcare in the mid 2000 than before, largely as a result of an expansion of private and voluntary providers, but that there is strong variability in terms of the quality of the provision. The evidence indicates that public nursery schools provide better “quality” than the service provided by private centres or childminders, mainly as a reflection of higher staff qualifications. Since childminders usually care for babies and children at their own home or at the child’s own home, no formal qualification was required. Only with the passage of the Childcare Act of 2006, childminders were first requested to have a minimum level of training under certain conditions.2 While formal childcare opportunities are limited, maternity leave policies are quite generous in terms of duration. At the beginning of the 2000, which is the relevant period for our analysis, a British mother is allowed up to 18 weeks of paid maternity leave. In other Western European countries maternity leave is often less generous in terms of duration, but often more in terms of benefits (in the UK the replacement rate is only 44% of average wages) (OECD 2001).3 Because of the paucity of childcare and the prevalence of part-time childcare opportunities, grandparental care has remained quite important, especially for lower income households. Gray (2005) compared different data sources and reports that about 30% of grandparents in the UK help with childcare. Analysis of childcare trends from the British Household Panel Survey (BHPS) suggests that support from grandparents has become an important complement to part-time formal care as pre-school places expanded since the early 2000s. Grandparents’ childcare allows mothers to enter the workforce and to work longer hours. While both grandmothers and grandfathers usually contribute to informal childcare, the main contribution comes from grandmothers. According to Wheelock and Jones (2002), the provision of childcare comes from women in their 50s and 60s and nearly a third of them are below retirement age. Recent policies regarding women over the age of 50 appear to go in conflicting directions. On the one hand, policies to encourage employment amongst the over 50s (UK Government 2014) may conflict with the role of younger grandparents providing childcare. On the other hand, the government has recently proposed policies which would encourage grandparents to take leave of absence from work (e.g. socalled “granny leaves”—see UK Government 2015). Some recent studies have 2

According to Ofsted (2016), if the person wants to care for children under 8 for more than 2 h a day (and s/he is paid), s/he needs to be registered with Ofsted (Office for Standards in Education, Children’s Services and Skills). To be registered, one needs to do a training course and a basic first aid course and to have an interview and home inspection. Under a number of exception, childminders are exempt from registration, in particular if they do not receive any payment or if they provide care for children in the children’s home. 3

In the same period, among the most generous countries in terms of duration are Sweden, Denmark, Italy and some Eastern European countries. The replacement rate is 100% in most European countries, and between 70% and 90% in the others. Only a few have lower replacement rate.

The role of grandparenting in early childcare and child outcomes

examined the implications of these policies. Cardia and Ng (2003) have argued in favour of the subsidization of grandparental childcare when the grandparent is retired, so that parents can devote more time to the labour market. Ho (2012) instead has argued for subsidization of formal childcare so that younger grandparents can devote more time to the labour market, with the positive effects of an active life and participation. However, more empirical evidence on the impact of grandparenting on child coutcomes is needed in order to discuss the potential outcomes of these policies.

4 Methodological framework The relationship between different types of childcare and child development can be described by a household production function that explains how parental inputs (time and money) and other inputs translate into child outcomes. In Todd and Wolpin (2003), both family and external inputs are important determinants of child development, and the impact depends on the timing of the investments. The human capital production function can be written as:
ð1Þ kija ¼ kijt1 ; Hija ; Sija where kija is a measure of the cognitive achievement of child i in household j at age a; kijt−1 is child i's past early cognitive ability, or child endowment; Hija is the vector of past and current family inputs up to age a; and Sija is the vector of present and past school inputs. The function that translates family and school inputs into child outcomes is allowed to depend on the child’s age a. In Todd and Wolpin’s model, the timing of the investment matters, since the same investment made at different ages could lead to different results. However, acquired skills are stable over time, and investments made in different periods do not interact with each other. Our specification corresponds to the cumulative specification in Todd and Wolpin (2003). The current output of the production process is also likely to depend on the history of previous inputs as well as on the child’s initial endowment. By including the past output of the child production function in the estimated equation kt−1, we control implicitly for the set of past inputs as well as for the child’s initial endowment. In our framework, child outcomes depend both on parental childcare as well as on other forms of non-parental care available. The availability of non-parental care can change the amount of time parents spend with their child and the effect it has on the child’s development. We assume, as in most empirical studies based on the production function approach, that k is an additive separable function of parental time, non-parental care, family characteristics, child characteristics and an i.i.d. disturbance. In our analysis, we distinguish between the two components of household input H: the inputs from the parents (Pij1) and inputs from the grandparents (Gij1). The education production function can thus be written as:
ð2Þ kija ¼ kij0 ; Pij1 ; Sij1 ; Gij1

D. Del Boca et al.

where a is the age at which outcomes are measured, 0 is the initial endowment (when the child is 9 months old), 1 is the period in which we measure inputs from family and external childcare (18 months). This production function is similar to those considered in previous works on child cognitive development, with the main difference being that it includes investments made by the grandparents themselves in addition to the usual inputs from parents and school. The estimation of the child production function implies several problems, with perhaps the most important being that family inputs are likely to be chosen by the family itself also in response to child achievements and child quality, and are therefore endogenous. Mothers’ and fathers’ time, in fact, may depend on the parents’ perception of the child’s needs, as proxied by the child achievements. For instance, if the child performs poorly, parents might react by spending more time with her/him. Even after controlling for past inputs, for child initial endowment and for current school inputs, the estimated effect of simultaneous parental time on child outcome might be inaccurate, namely because it could be both biased and inconsistent due to endogeneity issues. By assuming that the production function is additively separable, linear in its arguments and invariant during the period, we can estimate the following equation: 0 0 0 0 Ki;a ¼ β0 þ Ki;9 β1 þ Ci;18 β2 þ Xi;9 β3 þ Zi;a β4 þ εi;a

ð3Þ

where Ki,a is a vector of cognitive outcome of child i when s/he is a years old (a = 3, 5, 7); Ki,9 is a vector of early child endowment/development variables (motion, motor and communication development) of child i when s/he is 9 months old; Ci,18 is a vector of childcare modalities experienced by child i when s/he is 18 months old; Xi,9 is a vector of characteristics of child i, such as gender, and of her/his family, measured when s/he is 9 months old; Zi,a is a vector of time-varying characteristics of the child i and her family measured when s/he is a years old (a = 3, 5, 7); εi,a is a random error normally distributed; β0, β1, β2, β3, and β4 are the parameters to be estimated by OLS. The main interest of this paper is estimation of β2, which is the effect of different childcare modalities on child cognitive outcomes. In particular, we are interested in comparing the effect of grandparents’ care with that of formal care, while controlling for any other childcare input (parents, childminders, others). In this regard, the main assumption we are making is that the random error εi,a is uncorrelated with Ci,18. However, parents make choices about childcare, and there may be unobservable factors correlated with this decision that directly influence outcomes. For example, more “schooling oriented” parents may choose early formal childcare, but they may also buy books for the child, read to him/her, and engage him/her in more educational activities. These activities may be not observable and may influence child cognitive outcomes, biasing our estimates. We propose three further empirical specifications, characterized by weaker assumptions, which should reinforce (or not) the evidence of a causal link between care modalities and child cognitive outcomes. In the second specification, we add to the first model (Eq. 3) past decisions about childcare taken by parents and changes of childcare modalities over time. Most transitions in childcare modalities occur by the time the child is 3 years old, and are therefore recorded in the survey. We estimate the following equation: 0 0 0 0 0 Ki;3 ¼ β0 þ Ki;9 β1 þ Ci;18 β2 þ ICi;1836 βI2 þ Xi;9 β3 þ Zi;a β4 þ ωi;3

ð4Þ

The role of grandparenting in early childcare and child outcomes

where Ki,3 is a vector of cognitive outcomes of child i when s/he is 3 years old; ICi,18−36 indicates a new form of childcare introduced for child i between the age of 18 and 36 months (one not used before, at the age of 18 months); ωi,3 is a normally distributed random error; and β0, β1, βI2, β3, β4 are the parameters to be estimated by OLS. This second model is an enriched version of the first, and explains part of the unobserved heterogeneity across families by including early childcare decisions and childcare changes over time. Of interest here is estimation of βI2, which represents the effect of introducing a new form of childcare. For example, the effect of introducing formal care is identified by looking at children who start out in similar types of childcare and comparing the outcomes of those who are subsequently placed in formal care with the outcomes of those who are not. The assumption we rely on is still strong: ωi,3 needs to be uncorrelated with ICi18−36. The third specification employs a linear regression with fixed effects by exploiting the availability of one cognitive outcome repeated over time. We estimate: 0 0 0 0 Ki;a ¼ β0 þ Ki;9 β1 þ Ci;a β2 þ Xi;9 β3 þ Zi;a β4 þ ui þ μi;a

ð5Þ

where Ki,a is the cognitive outcome of child i when s/he is 3 or 5 years old; Ci,a is a vector of childcare modalities experienced by child i by the age of 3/5; ui represents the personal/family effect of child i, fixed over time; and μi,a is a normally distributed random error. By subtracting the data available in the two waves, we can eliminate the personal/ family effect ui. That means that whatever is unobserved by the researcher (parental school orientation, for example) but is fixed over time is taken into account and does not bias our results. The assumption behind the third model is therefore much weaker than the assumptions behind the first and second ones. It must also be taken into account, however, that parents may learn more about their children over time and respond accordingly. For example, the parents’ realization that a child is not developing properly for her/his age may result in them sending the child to formal childcare and to take further action that may also influence the child’s development. To check whether this may be the case, we use an instrumental variables approach. We limit the sample to children cared for by their grandparents or looked after in a formal care centre, and estimate the following equations: 0 ^ i;18 β2 þ X 0 β3 þ Z 0 β4 þ vi;a Ki;a ¼ β0 þ Ki;9 β1 þ G i;9 i;a

ð6Þ

0 0 0 γ 1 þ D0i γ 2 þ Xi;9 γ 3 þ Zi;a γ 4 þ ei;a Gi;18 ¼ γ 0 þ Ki;9

ð7Þ

Gi,18 indicates that child i is mainly cared for by his/her grandparents (rather than attending a formal care centre) at the age of 18 months; Di is our instrumental variable, the distance (in minutes) between parents’ and grandparents’ homes; vi,a and ei,a are normally distributed random errors. Instruments need to be relevant and valid to have credible results. While a relationship between geographical distance and the probability of being cared for by grandparents is expected and easy to verify empirically, we need to argue the validity of the instrument. How close the parents live to the grandparents’ home represents a decision that has been made, and its unobservable determinants may also be related

D. Del Boca et al.

to child cognitive outcomes. Geographical distance has been often used (and criticized) when studying the impact of grandparental care on mother’s employment, because more work-attached mothers may choose to live near the grandparents so they can receive needed support and they can go out and work; on the other hand, they may also be more willing to move for job opportunities. Instead, it is more difficult to think of an example in our case. We argue that there is no reason for the parents of children with better/worse cognitive outcomes to choose to live closer to the grandparents. It is possible that more “capable” parents move farther away in order to seek better job opportunities, and that they transmit their knowledge and skills to their children. If this is the case, our instrument would be not valid. However, we control for maternal education in the preferred specification, and also for employment, hours of work, and wages in an additional specification. Parents of children with strong difficulties (e.g. disable children) may decide to stay closer to grandparents because their children need additional attention. Despite these children being only a minority, we will replicate our IV estimates excluding them, to be sure that this is not driving our results. There may be other differences—even if not systematic—between parents living closer to the grandparents and those who live farther away: differences in values and norms that might affect the way they rear their children. We try to explore potential differences by exploiting data on values and beliefs from the British Household Panel Study (BHPS). This dataset represents the longest panel available for the UK (1991–2008) and allows us to describe women’s moves, values and beliefs just before and after the birth of a child.

5 Data, sample, and variables 5.1 The Millennium Cohort Study The Millennium Cohort Study is a longitudinal survey conducted by the Centre for Longitudinal Studies that tracks the lives of a sample of about 19,000 babies born in the UK in the year 2000/2001. The survey is conducted in several waves, with the first one concentrating on the circumstances of the pregnancy and birth as well as the first few months of life. This first part of the survey also contains important information about the socio-economic background of the family to which the child was born. The second wave took place when the children were about 3 years old, and focussed mainly on continuity and change in the family as well as the parenting environment to extract information about the child’s development. In the third wave in 2006, the children were old enough to start primary school. The fourth wave took place in 2008. 5.2 The sample In wave 1, the survey consists of 18,552 children. We began by excluding the 256 twins, since childcare arrangements and their effects may be different when more children have to be looked after. Second, we decided not to consider the 40 children whose main caregiver (and who therefore answered the questions concerning the child) was not the mother. Approximately 10% of the remaining sample was later lost

The role of grandparenting in early childcare and child outcomes

between the second and the third waves, and another 10% between the third and the fourth surveys, due to attrition. Unfortunately, because of a filter error in the questionnaire, we lost another non-negligible number of mothers (around 1500) who did not answer questions on childcare in wave 2 that are pertinent to constructing our main variable of interest. The final sample is described in Table 1, where we have also eliminated children for whom dependent or independent variables were missing. The effects of sample selection on the size and characteristics of our final sample are shown in the Appendix (Table 17). Our analysis is based on approximately 9000 observations. Is this sample representative of the initial one? If we compare the average characteristics (Table 17), we observe some small but significant differences: in particular, we observe that the mothers in our final sample are older (30 years old rather than 29), better educated (41% completed academic upper secondary education vs. 33%), more likely to have had a job while pregnant (70 vs. 62%), more likely to be living with their partner (88 vs. 82%), and slightly wealthier (weekly equivalent income 332 rather than 298). 5.3 Control variables fixed over time We control for a large number of variables, measured when the child is 9 months old (see top of Table 1). We consider the child’s characteristics (ethnicity, gender, age, birth-weight, whether s/he was breastfed for at least 1 month, accidents at home, having been in a hospital, and three indicators of child development4); household characteristics (other siblings, real weekly equivalent income,5 if parents meet friends at least once a week, region of residence, presence of the father at home); and the mother’s characteristics (age, level of education,6 age she left education, whether she held a job while pregnant, whether she experienced post-partum depression, whether she had lived with a single mother during childhood, whether she has a chronic illness, cigarettes smoked per day, whether she drinks at least once a week). 5.4 Time-varying control variables In our empirical analysis we also control for a number of variables which may vary over time and may influence childcare decisions as well as child outcomes: age of the child (in months), the presence of the father at home, the presence of a new partner of the mother at home, real household income, and the arrival of new siblings (see bottom of Table 1). 4

We include three factors derived from analysing a battery of questions through a factor analysis. Examples of questions and answers are: “s/he waves bye-bye on her/his own when someone leaves”, “s/he can pick up a small object using forefinger and thumb only”, “s/he can sit up without being supported”; answers are “often”, “once or twice”, and “not yet”.

5

The equivalent income is the income of the household taking into account the number of people in the family and assigning weights. The one provided in the MCS follows the OECD equivalence scale, which assigns a value of 1 to the first household member, of 0.7 to each additional adult, and of 0.5 to each child.

6

The level of education is controlled for by including a dummy variable equal to 1 if the mother completed at least “academic upper secondary education”. This corresponds to having completed at least an “a/as/s level” (GCE Advanced Level) qualification, i.e. a secondary school leaving qualification that leads to higher education.

D. Del Boca et al. Table 1 Descriptive statistics Variable

Mean

Std. dev.

Min

Max

Wave 1 (9 months) Child British

0.87

0.34

0

1

Indian/Pakistani/Bangladeshi

0.06

0.24

0

1

Other ethnicity

0.07

0.26

0

1

Girl

0.50

0.50

0

1 12

Age child (months)

9.7

0.6

8

Birth-weight

3.4

0.6

0.4

6.8

Breastfed 1 month

0.50

0.50

0

1

Injuries

0.08

0.30

0

3

Hospital

0.17

0.51

0

10

−0.06

0.97

−3.63

4.32

Motor development

0.03

0.95

−5.15

2.21

Motion development

0.07

0.82

−12.85

1.01

Siblings

0.91

1.01

0

9

Equivalent weekly income

332

231

14

1254

Communicative devel

Household

Meet friends every week

0.72

0.45

0

1

England

0.63

0.48

0

1

Wales

0.16

0.36

0

1

Scotland

0.12

0.33

0

1

Northern Ireland

0.10

0.30

0

1

Father at home

0.88

0.33

0

1

15

48

Mother Age Academic upper sec. education Age left education

30 0.41 18

6 0.49

0

1

3

0

32

Job during pregnancy

0.70

0.46

0

1

Post-partum depression

0.24

0.43

0

1

Difficult childhood

0.28

0.45

0

1

Long illness

0.22

0.41

0

1

Cigarettes per day

2.9

5.9

0

40

Alcoholic drinks

0.40

0.49

0

1

Time-varying variables Wave 2 (age 3) Age child (months)

37

2

32

55

HH annual income

25,656

18,034

510

94,920

Father at home

0.84

0.37

0

1

Step-father at home

0.01

0.12

0

1

Newborn siblings

0.25

0.44

0

1

The role of grandparenting in early childcare and child outcomes Table 1 continued Variable

Mean

Std. dev.

Min

Max

Wave 3 (age 5) Age child (months)

63

3

53

74

HH annual income

26,435

17,835

469

85,835

Father at home

0.80

0.40

0

1

Step-father at home

0.03

0.17

0

1

Newborn siblings

0.18

0.39

0

1

Wave 4 (age 7) Age child (months)

87

3

77

98

HH annual income

27,451

18,519

438

94,151

Father at home

0.76

0.42

0

1

Step-father at home

0.04

0.21

0

1

Newborn siblings

0.11

0.32

0

1

Maximum number of observation: 9162

As expected, the percentage of separated parents increases over time, income increases, and the probability of having an additional child decreases. We control for the child’s age, since children are not all interviewed at exactly the same age, but this has to be taken into account when considering their outcomes.7 5.5 Child outcomes In the Millennium Cohort Study, child outcomes are measured in three ways. A first set of outcomes is measured by the interviewer through a standardized test focusing on the child’s ability to perform certain tasks; a second set is reported by the teacher and concerns the child’s abilities and behaviour at school; a third set is reported by the mother and regards the child’s behaviour at home. The analysis of each group of outcomes has different challenges that need to be addressed using different methodologies. When studying behaviour as reported by the mother, for example, we have to take into consideration that mothers may feel guilty about not staying at home and therefore be more lenient in evaluating their child’s behaviour. Instead, when considering teacher-performed evaluations, we need to keep in mind that teachers may be influenced by other children in the class, whose outcomes we cannot observe. In this paper, we focus on tests administered by the interviewer, which should be more objective. We look at outcomes from the following eight tests: the Bracken School Readiness Assessment (for 3-year-olds), Naming Vocabulary (for ages 3 and 5), Picture Similarity (age 5), Pattern Construction (ages 5 and 7), Word Reading, and Number Skills (age 7). The Bracken School Readiness Assessment evaluates the basic concept development in young children. BBCS-R measures the comprehension of 308 functionally relevant educational concepts in 11 subtests or concept categories. Only 6 subsets 7 Scores are standardized according to the child’s age, but within a 3-month interval. Thus, it is preferable to also control for age expressed in months (see Connelly 2013).

D. Del Boca et al.

were used in the MCS: colours (primary and basic), letters (knowledge of upper- or lower-case letters), numbers/counting (single and double-digits numbers and assigning a number to a set of objects), sizes (concepts that describe one, two and three dimensions), comparisons (the ability to match or differentiate objects), and shapes (including lines, circles, squares, cubes, and pyramids). The Naming Vocabulary verbal test assesses the spoken vocabulary of the child. Children are shown individual test items from booklets of colourful pictures of objects which they are asked to name. The scale measures language ability, but picture recognition is also crucial. Low scores may also reflect reluctance to speak. For the Picture Similarity test, children are shown a row of 4 pictures on a page and asked to place a card with a fifth picture under the picture most similar to it. This assessment measures children’s non-verbal reasoning abilities (Connelly 2013). The Pattern Construction test assesses the child’s accuracy and speed in constructing a design by assembling flat squares or solid cubes with black and yellow patterns on each side. The test assesses the spatial problem-solving ability of the child. In the Word Reading test the child is shown a series of words on a card and must read them out loud, providing a measure of his or her reading skills. The Number Skills test is adapted from the NFER Progress in Math test. In this assessment, children complete various task involving numbers, shapes, space and measures, and data handling. Table 2 reports the descriptive statistics of the outcomes of interest. The outcomes are age-based and standardized. In the norming sample, Naming Vocabulary, Pattern Construction, and Picture Similarity have a mean of 50 and a standard deviation of 10, while School Readiness, Word Reading, and Number Skills have a mean of 100 and a standard deviation of 15 (Connelly 2013). This means that our final sample has slightly better results than the norming sample, with the exception of Number Skills. 5.6 Childcare variables In the surveys, the main respondent was asked about the childcare choices made. In survey one, when the child was 9 months old, working mothers were asked to state the types of care being used at the time of the survey when they were at work. In the second survey, when the child was 3 years old, all mothers were asked details about the types of childcare that had been used since the first survey, including starting dates, stop dates, and the number of hours per week. We use information from the second survey to create a more representative sample (working and non-working mothers) and to gain information about the hours of used types of care. Obviously, working mothers are more likely to use external forms of childcare than non-working mothers. Table 3 describes childcare choices of working/ non-working mothers. 36% of working mothers entrust their children to the care of grandparents, 20% to formal childcare, and 11% to childminders. Instead, 6% of nonworking mothers entrust their children to the care of grandparents, 5% to formal childcare, and 2% to childminders. There are two additional categories: “other relatives or non relatives” (7% of working mothers and 2% of non-working mothers) and “working parents” (40% of working mothers and 3% of non-working mothers). “Working parents” includes the mother taking care for the child “while working at home or at the workplace” (stated like that in the questionnaire) or/and the husband/

The role of grandparenting in early childcare and child outcomes Table 2 Children outcomes

Variable

Observations Mean Std. dev. Min Max

Wave 2 (age 3) School Readiness

7463

106

16

56

149

Naming Vocabulary

7766

51

11

20

80

Wave 3 (age 5) Naming Vocabulary

8368

55

11

20

80

Pattern Construction 8343

51

10

20

80

Picture Similarity

56

10

20

80

8354

Wave 4 (age 7) Pattern Construction 8194

54

11

20

80

Word Reading

8109

112

18

55

145

Number Skills

8220

98

15

69

136

Notes: Naming Vocabulary, Pattern Construction, and Picture Similarity have been normalised using a norming sample of children to have a mean of 50 and a std.dev. of 10; School Readiness, Word Reading, and Number Skills have been normalised with a mean of 100 and a std.dev. of 15. Mean and std.dev in our sample are a bit different (mostly higher) than the original sample

Table 3 Childcare by mother working status

Childcare modality

Working mothers

Non-working mothers

Non-working mother

0.00

0.83

Working parents

0.40

0.03

Grandparents

0.36

0.06

Formal care

0.20

0.05

Childminder

0.11

0.02

Others

0.07

0.02

Observations

4959

4203

partner taking care of the child. This means that the working parent(s) is (are) able to take care of their child even if working (for example, if both parents work part-time). To these modalities, we add “non-working mother”, which corresponds to all nonworking mothers who do not use any childcare; this represents 83% of non-working mothers and, by definition, 0% of working mothers. In order to measure childcare inputs at the same time for all children in the survey, we choose a point in time between the first two surveys: we look at childcare choices when the child is 18 months old, which is before the second survey, when we measure the first outcomes (the youngest child who participates in the second survey is 25 months old at the time of the interview) and which is after the first survey (the oldest child who participates in the first survey is 12 months old at the time of the interview). Because we know the average weekly number of hours for each, we can also define the “main” type of childcare used. Table 4 describes parental childcare choices when the child is 18 months old: the first panel indicates the percentage of children receiving that kind of care, the second panel shows the average number of

D. Del Boca et al. Table 4 Childcare choices when the child is 18 months old

Variable

Mean

Std. dev.

Min

Max

Any care Non-working mother

0.38

0.49

0

1

Working parents

0.23

0.42

0

1

Grandparents

0.22

0.41

0

1

Formal care

0.13

0.34

0

1

Childminder

0.07

0.26

0

1

Others

0.05

0.21

0

1

Users’ hours of care Non-working mother

20.00a

0.00

20

20

Working parents

20.15

12.01

1

120

Grandparents

21.11

12.56

1

80

Formal care

22.52

12.73

1

80

Childminder

25.12

13.71

2

160

Others

19.70

13.49

1

72

Main care Non-working mother

0.38

0.49

0

1

Working parents

0.22

0.41

0

1

Grandparents

0.19

0.39

0

1

Formal care

0.11

0.32

0

1

Childminder

0.06

0.24

0

1

0.04

0.19

0

1

Others Observations

9162

Notes: “Users’ hours of care” (second panel) reports the average number of hours by type of care considering only children who receive that type of care a

Imputed

hours per child using that kind of care, and the third panel shows the percentage having a certain kind of care as main care. In particular, 22% of children are looked after by their grandparents, for an average of 21.1 h a week, and this is the main childcare arrangement for 19% of children. It is interesting to observe that among children not looked after by their parents, grandparental care is the most common choice. We also considered childcare choices at other points in time. However, childcare choices seem highly consistent over time, in particular when the child is very young (Table 5): for most children, the main type of childcare used at 12 months is the same as that used at 24 months (for 90% of children for all childcare modalities, with the exception of “others”). This is one of the reasons why we include childcare variables only at one point of time. We also decide not to include subsequent childcare/educational choices in the main results, since we only have one instrument. In the Appendix, we will show the robustness of our results to the inclusion of variables describing childcare decisions at other points in time.

The role of grandparenting in early childcare and child outcomes Table 5 Childcare choices between age 1 and age 2 of the child 2 years old

Non-working

Working

Grand-

Formal

Child-

1 year old

mother

parents

parents

care

minder

Others

Non-working mother

0.89

0.02

0.03

0.05

0.01

0.01

Working parents

0.01

0.92

0.01

0.03

0.02

0.01

Grandparents

0.03

0.01

0.91

0.04

0.01

0.01

Formal care

0.02

0.00

0.01

0.96

0.01

0.00

Childminder

0.05

0.00

0.01

0.06

0.88

0.01

Others

0.06

0.03

0.01

0.06

0.03

0.81

Notes: Observations: 9162

6 Empirical results 6.1 Cross section analysis Tables 6–8 show the main results. We estimate the effect of early care (when the child is 18 months old) on cognitive outcomes at age 3 (Table 6), at age 5 (Table 7), and at age 7 (Table 8).8 For each outcome, we show the estimated coefficients of three different models: in the first (“Any”), we estimate the effect of the child having experience of each type of childcare; in the second (“Hours”), we estimate the effect of weekly hours the child is cared for in each type of childcare; in the third model (“Main”), we estimate the effect of the main (in terms of hours) type of childcare (formal childcare is the excluded category). When the child is 3 years old, we observe a positive effect on School Readiness (1.56) of being looked after in a formal care centre, while an increasing number of hours spent with grandparents worsens it (Table 6). Even if the estimated coefficients are not significant in the “Main” column, the signs are all negative when compared to formal care (excluded category). On the contrary, being looked after by grandparents has a positive effect on Naming Vocabulary scores at age 3 (1.24 when looking at “Any” childcare), as found by Hansen and Hawkes (2009): being cared for within the family (also as the main type of childcare) has a much stronger positive effect on the child’s ability to name objects than formal childcare, and an increasing number of hours spent in formal childcare actually decreases it. One potential explanation for the positive effect of grandparental care on Naming Vocabulary is that grandparents might provide one-on-one care, meaning that the child might be addressed verbally by adults much more frequently than children in formal care. The fact that parental care has the same positive impact reinforces this interpretation. In addition, another explanation might be related to the fact that grandparents (as well as parents) provide a more stable relationship with children, whereas formal care centres are likely to have a rotating staff (Walker 1991). On the 8

In the online Appendix (Table B.1), we present the effects of the type of childcare on cognitive outcomes, progressively adding control variables. This exercise suggests that the results of the baseline specification (without any control variable) are biased, but that we are able to remove most of the bias by controlling for a (large) number of observable characteristics.

0.73 (0.62)

0.07 (0.81)

0.19 (0.85)

Grandparents

Childminder

Others

−0.80 (0.60) −1.17 (0.78) −1.50 (0.96)

−0.05** (0.02)

−0.08*** (0.02)

−0.09*** (0.04) 0.23

0.23

−0.64 (0.60)

−0.03 (0.02)

7463

0.71 (0.51)

0.21

−0.35 (0.57)

0.02 (0.54)

1.24*** (0.42)

0.68 (0.45)

−0.69 (0.46) −0.51 (0.61)

0.00 (0.02)

−0.03 (0.03)

0.21

7766

−0.07*** (0.02)

−0.06*** (0.02)

−0.00 (0.01)

−0.02 (0.01)

−0.02 (0.02)

0.21

0.65 (0.62)

0.79 (0.52)

1.60*** (0.41)

1.41*** (0.41)

1.30*** (0.41)

Main

**p < 0.05; ***p < 0.01

Robust standard error in parentheses. Notes: Controlling for all the variables listed in section 5. Full table with controls available as online supplementary material (Table B.2)

R squared

0.23

0.68 (0.66)

Working parents

Observations

1.02 (0.75)

Non-working mother

−0.06*** (0.01)

1.56** (0.67)

Formal care

Hours

Any

Main

Any

Hours

Naming Vocabulary

School Readiness

Table 6 Early childcare and child outcomes (age 3)

D. Del Boca et al.

0.22

0.22

0.09

0.09

8343

0.00 (0.01)

0.95** (0.46)

Any

0.02 (0.42) 0.03 (0.54)

0.09

0.06

−0.27 (0.66) −0.23 (0.58)

0.21 (0.52)

−0.22 (0.39) −0.15 (0.42)

−0.14 (0.38)

−0.00 (0.39) −0.68 (0.49)

Main

Picture Similarity

0.06

8354

0.02 (0.02)

0.02 (0.02)

0.01 (0.01)

0.02 (0.01)

−0.02 (0.02)

0.05*** (0.01)

Hours

0.06

−0.86 (0.67)

−0.94 (0.54)

−1.27*** (0.41)

−1.03** (0.40)

−1.71*** (0.41)

Main

**p < 0.05; ***p < 0.01

Robust standard error in parentheses. Notes: Controlling for all the variables listed in section 5. Full table with controls available as online supplementary material (Table B.2)

R squared

Observations

8368

0.38 (0.50) 0.56 (0.51) −0.00 (0.02) −0.72 (0.62) 0.04 (0.56) −0.02 (0.02)

0.67 (0.51) −0.01 (0.02)

−1.02 (0.54) −0.05** (0.02)

Childminder

0.22

0.32 (0.38) 0.20 (0.40)

0.47 (0.39) −0.01 (0.01)

Grandparents

Others

0.07 (0.38) 0.40 (0.42) −0.01 (0.01)

0.20 (0.41) −0.02 (0.01)

0.00 (0.02)

0.00 (0.01)

Working parents

0.34 (0.42)

0.12 (0.47) −0.02 (0.02)

Non-working mother

−0.00 (0.39) 0.47 (0.48)

0.03 (0.43) −0.03** (0.01)

Formal care

Hours

Any

Main

Any

Hours

Pattern Construction

Naming Vocabulary

Table 7 Early childcare and child outcomes (age 5)

The role of grandparenting in early childcare and child outcomes

0.09

0.09

8194

−0.62 (0.57)

0.09

−0.84 (0.69)

−0.96 (0.88)

0.15

−1.91 (0.97) 0.15

8109

−0.05 (0.04)

−0.02 (0.03)

0.00 (0.02)

0.00 (0.02)

0.15

−0.50 (1.09)

−0.14 (0.86)

0.24 (0.67)

0.09 (0.66)

1.21 (0.67)

Main

0.10

−0.45 (0.87)

−0.35 (0.77)

0.48 (0.63)

−0.14 (0.64)

−0.22 (0.74)

0.98 (0.67)

Any

Number Skills

0.10

8220

−0.02 (0.04)

−0.01 (0.02)

0.00 (0.02)

−0.01 (0.02)

−0.02 (0.03)

0.02 (0.02)

Hours

0.10

−0.97 (0.98)

−1.40 (0.76)

−0.81 (0.60)

−1.36** (0.59)

−1.39** (0.60)

Main

**p < 0.05; ***p < 0.01

Robust standard error in parentheses. Notes: Controlling for all the variables listed in section 5. Full table with controls available as online supplementary material (Table B.2)

R squared

Observations

−0.01 (0.02)

−0.03 (0.03)

−0.08 (0.59)

−0.63 (0.63)

Childminder

Others

−0.93 (0.70)

−0.42 (0.43)

0.01 (0.01)

0.24 (0.46)

Grandparents

−0.85 (0.71)

−0.49 (0.42)

−0.00 (0.01)

0.12 (0.48)

0.05 (0.82)

0.05 (0.03)

−0.02 (0.02)

−0.97 (0.75) −0.72 (0.43)

0.02 (0.02)

−0.01 (0.02)

0.40 (0.49)

−0.15 (0.54)

Working parents

Non-working mother

Formal care

Hours

Any

Main

Any

Hours

Word Reading

Pattern Construction

Table 8 Early childcare and child outcomes (age 7)

D. Del Boca et al.

The role of grandparenting in early childcare and child outcomes

other hand, as suggested by Bryson et al. (2012), grandparents may not have the skills to foster their grandchildren’s school preparation and other cognitive abilities, whereas formal childcare workers are expected to have the training to do so. Table 7 reports the results of the same specifications at age 5. The positive effect of grandparental care on Naming Vocabulary is no longer significant, confirming results by Bryson et al. (2012); however, the hours spent in a formal care centre have a negative effect. Likewise, we do not find any effect on Pattern Construction, although we do observe a strong positive effect of formal care on Picture Similarity scores (0.95), which increases with the number of hours spent in formal childcare. Table 8 reports the results at age 7: the only positive effect observed is for formal childcare as the main type of care on Number Skills, compared to parental care. The decline of the effects on cognitive skills in the medium run is in line findings from other researches (see Del Boca et al. 2017 for a discussion). The significant effects are in the range of about one tenth of a standard deviation (slightly larger when we consider the main type of care). The magnitude of the effects, in addition to their direction, is similar to what has been found in other papers (e.g., Gregg et al. 2005; Bernal and Keane 2011) and corresponds to about 2–2.5 months ahead/behind the age group. Bernal and Keane (2011) also relates scores at early ages to later outcomes. We also present heterogeneous effects of the main childcare type for families with equivalent household income either below or above the median (Table 9).9 We still find a positive effect of grandparental care (compared to formal care) on Naming Vocabulary, but only for children from more advantaged backgrounds. The negative effects of grandparental care on School Readiness are confirmed only for children from more disadvantaged backgrounds. Negative effects of grandparents’ care on Pattern Construction emerge for children in more disadvantaged families. Thus, being cared in a formal care centre seems to be more important for children of relatively poorer families, as previous research highlighted (Cornelissen et al. 2016; Elango et al. 2016). However, interestingly, there are negative effects of grandparental care on Picture Similarity statistically significant for children from more advantaged backgrounds. This result may be due to the fact that it is more difficult to boost Picture Similarity compared to other outcomes, and thus only specifically trained staff can boost it, and in particular only well trained teachers and high quality formal care centre, options more often available only for high-income teachers. It is also possible that this heterogeneity is in part due to different childcare choices within high income families (as can be see from Table 16). The absence of positive effects on verbal skills for children from less advantaged background could be due to different reasons: the less rich grandparents’ vocabulary, the fact that grandparents spend more “passive” time with grandchildren, or the fact that they take care of more than one child at the same time. Unfortunately, with the available data we cannot properly test which one is the prevailing mechanism, especially with respect to the first two. Concerning the number of children looked after by grandparents, we know if the child has brothers/sisters, but not if the 9

Results are similar when dividing the sample by parental education, i.e. a group with no parents with academic upper secondary education and a group in which at least one parent has completed it.

1.32** (0.53) 0.25 (0.46) 0.29 (0.46) 0.67 (0.57)

−1.51** (0.66)

−1.29 (0.71)

−0.78 (0.73)

−0.79 (1.09)

Grandparents

Childminder

hh income

Non−working mother

hh income

median

−0.70 (1.05)

−1.52 (0.78)

−0.90 (0.75)

−1.08 (0.71)

hh income

median

Below

Above median

Below

Pattern Construction

Naming Vocabulary

Above

0.36 (0.59)

0.13 (0.45)

−0.22 (0.46)

0.64 (0.53)

hh income

median

−0.90 (1.08)

−0.98 (0.79)

−0.64 (0.76)

−1.81** (0.72)

hh income

median

Below

Picture Similarity

(0.71) (0.76) (0.77) (1.04) (1.00)

1.30 −0.44 −1.03 −1.13 −1.11 3753 0.12

−2.74*** (1.05) −2.86** (1.12) −2.30** (1.16) −2.48 (1.55) −3.98** (1.61) 3710 0.19

−0.37 −0.41 −0.32 −0.37 −0.63 3861 0.22

Below median hh income

Above median hh income

Below median hh income (0.81) (0.72) (0.70) (0.89) (1.23)

Naming Vocabulary

School Readiness

Working parents

PANEL B (age 5)

Non-working mother Working parents Grandparents Childminder Others Observations R squared

PANEL A (age 3)

Table 9 Heterogeneous effects of early childcare, by household income level

−1.02 (0.61)

−1.54*** (0.49)

−1.48*** (0.48)

−1.09 (0.56)

hh income

median

Above

2.35*** (0.57) 1.60*** (0.50) 1.75*** (0.50) 0.88 (0.60) 0.44 (0.83) 3905 0.10

Above median hh income

D. Del Boca et al.

hh income

hh income

0.07

0.06

0.12

4154

−2.73 (1.77)

−0.53 (1.81)

−0.24 (1.32)

−0.31 (1.27)

0.73 (1.21)

hh income

Above

0.10

3955

0.78 (1.44)

−0.35 (0.98)

0.01 (0.78)

−0.25 (0.79)

1.51 (0.85)

hh income

median

Above

0.06

4088

−0.76 (0.92)

hh income

median

0.08

4215

−1.57 (1.66)

−3.66** (1.60)

−2.00 (1.17)

−2.84** (1.11)

−3.16*** (1.03)

hh income

median

Below

Number Skills

0.05

4265

0.22 (1.07)

hh income

median

Below

Picture Similarity Above

0.06

4005

−1.63 (1.21)

−0.88 (0.86)

−0.85 (0.71)

−1.23 (0.72)

0.39 (0.81)

hh income

median

Above

0.05

4089

−1.84** (0.89)

hh income

median

**p < 0.05; ***p < 0.01

Robust standard error in parentheses. Notes: Effect of main care on child outcomes (formal care is the excluded category). Controlling for all the variables listed in section 5. Full table with controls available upon request

R squared

3994

−1.49 (0.93)

−1.07 (1.07)

Others

4200

−0.05 (0.51) −0.63 (0.65)

−1.71** (0.81)

−1.04 (1.15)

Grandparents

Childminder

0.52 (0.60) −0.21 (0.52)

−2.02*** (0.72)

hh income

hh income

−1.71** (0.77)

Below

median

median

Above

median

0.09

Below

Non-working mother

Observations

−0.53 (1.03) 4255

Word Reading

0.11

4097

hh income

Pattern Construction

0.20

4271

−1.03 (0.80)

median

median

−1.49 (1.02)

Below

Above

Below median

Pattern Construction

Naming Vocabulary

Working parents

PANEL C (age 7)

R squared

Observations

Others

PANEL B (age 5)

Table 9 continued

The role of grandparenting in early childcare and child outcomes

D. Del Boca et al.

grandparents are taking care of them, and we do not know if s/he has cousins, cared for by grandparents. Keeping in mind these limitations, we replicate the analysis excluding children with siblings 0–2 years apart. This results in similar findings,10 inducing us to exclude the last channel. While we do not observe any difference between grandparents’ and formal care on Pattern Construction when considering the whole distribution (Table 8), a negative effect of grandparents emerge when looking at children from low family backgrounds. Again, the explanation could be related to the quality of time these grandparents provide to their grandchildren, which may be influenced also by the number of grandchildren they are looking after. 6.2 Changes over time We try to exploit the transitions between different early childcare types between age 18 months and 3 years, and estimate the effect of introducing a new type of childcare on outcomes in wave 2 (age 3). By exploiting the usual questionnaire on childcare in wave 2, we reconstruct all types of childcare experienced by the child by age 3. For example, some children may have been in formal childcare already at 18 months, and being so in both periods, while some start formal childcare later on, or never. We can observe these transitions for all types of childcare: as expected, the most common childcare type introduced is the formal one (see Table 18 in the Appendix). Grandparental care and other modalities show much less variability over time. We therefore use this method only to test the effect of formal childcare on School Readiness (which was found positive, Table 6). From an econometric perspective, e.g., we estimate the effect of introducing formal childcare by age 3 (without being in formal childcare at 18 months) and of being in formal childcare at 18 months (as in our main specification) rather than never being in formal childcare. The results concerning School Readiness confirm the OLS ones (Table 10): introducing formal childcare between 18 months old and 3 years old has a positive effect, in addition to the positive effect of formal childcare at 18 months, on average and in particular for children at the bottom of the income distribution. Apart confirming results of our main specification, this finding indicates the existence of beneficial effects of formal childcare even when introduced later in age. 6.3 Panel data and IV methods The main results (Tables 6–9) may be biased because of unobservables correlated with childcare choice. In this section, we try and test whether this is so in our analyses by employing estimations with fixed effects and with instrumental variables. We restrict the sample of analysis to children looked after by grandparents or in formal childcare, and consider the outcomes for which these analyses are possible. We first employ a fixed effects model for analyzing Naming Vocabulary, which is the only outcome observed twice, and before age 5. While our childcare variables have been constructed from wave 2, here we use also information from wave 3: parents are asked whether the child has ever been in an early formal childcare centre 10

Available as online Appendix (Tables B.3 to B.7).

The role of grandparenting in early childcare and child outcomes Table 10 Change of childcare arrangements over time (age 3) School Readiness Average effects

Heterogeneous effects

Any

Below

Above

median

median

hh income

hh income

Formal care (18 m)

1.85*** (0.68)

4.17*** (1.27)

Non-working mother (18 m)

0.92 (0.77)

1.03 (1.25)

0.95 (0.81) 2.09** (1.02)

Working parents (18 m)

0.59 (0.67)

0.74 (1.17)

0.20 (0.82)

Grandparents (18 m)

0.71 (0.62)

1.87 (1.11)

−0.32 (0.77)

Childminder (18 m)

0.02 (0.81)

2.11 (1.43)

−1.02 (0.98)

Others (18 m)

0.20 (0.85)

0.40 (1.40)

−0.23 (1.08)

Start—Formal care (18–36 m)

1.17*** (0.43)

1.73*** (0.64)

0.28 (0.57)

Start—Non-working mother (18–36 m)

−0.50 (1.10)

−2.11 (1.35)

Start—Working parents (18–36 m)

−0.22 (0.70)

−2.13 (1.12)

0.71 (0.88)

0.72 (0.98)

1.88 (1.30)

−1.00 (1.53)

−0.99 (1.21)

−0.36 (1.63)

−1.63 (1.77)

0.47 (1.02)

0.76 (1.34)

Start—Grandparents (18–36 m) Start—Childminder (18–36 m) Start—Others (18–36 m) Observations R squared

7463 0.23

3710 0.20

1.83 (1.93)

0.35 (1.56) 3753 0.12

Robust standard error in parentheses. Notes: Effect of introducing a new type of childcare between age 18 months and age 3 years (reference category: not being in e.g. formal childcare in both periods (at 18 months, or between 18 months and 3 years old)). Controlling for all the variables listed in section 5. Full table with controls available upon request **p < 0.05; ***p < 0.01

or ever been looked after by a childminder.11 We can therefore observe how many children start having these two childcare experiences: in wave 2 (around 3 years old) 30% of children have experienced formal childcare and 6% have experienced a childminder; by wave 3 (around 5 years old) 97% of children have experienced formal childcare and 15% have experienced a childminder. By employing a linear regression with individual fixed effects, we find that children who move from grandparents’ to formal care experience a decline in their naming abilities compared to children looked after in the same way over time (Table 11). The detrimental effect of formal care with respect to grandparents’ care is driven by children from more advantaged families, as we have already seen with the OLS estimations (Table 9). Indeed, the fact that children develop less vocabulary 11 Unfortunately, we lack information about other modalities of care after age 3, i.e., we cannot show the effect of being cared for by grandparents (where formal care is the excluded category) in the panel analysis, as we did in the previous ones. In this case, the reference category are children with childcare modalities different from formal or childminder, i.e., mainly cared for by parents or grandparents. To get around this issue, we focus only on children in formal care or cared for by grandparents at age 3. In fact, we know that the probability of switching from grandparental care to parental care is rather low (see Tables 5 and 18 for younger children). Thus, we can interpret the results as the effect of being cared for in a formal centre, or by a childminder, compared to being cared for by grandparents.

D. Del Boca et al. Table 11 Fixed-effects model (Naming Vocabulary, age 3 and 5) Naming Vocabulary Average effects

Heterogeneous effects

Any

Below

Above

median

median

hh income

hh income

Formal care

−1.48*** (0.40)

−1.06 (0.69)

−1.77*** (0.49)

Childminder

−0.56 (0.68)

−0.44 (1.35)

−0.54 (0.80)

Observations

3307

1185

2122

Robust standard error in parentheses. Notes: Sample of children looked after by grandparents or in formal care, who may have started being cared in formal care or by a childminder between 3 and 5. Linear fixed effect model. Controlling for all the variables listed in section 5. Full table with controls available upon request **p < 0.05; ***p < 0.01

when moved to formal childcare mirrors the positive effects of grandparental care, and the mechanisms could be the ones discussed above.12 The previous method relies on the assumption that the introduction of a new type of care is not related to any factor which may influence the outcomes directly. To relax this assumption, we adopt an instrumental variable approach. As an instrument we use the distance—in minutes—between the parents’ and grandparents’ homes. The distance is available as a categorical variable: grandparents living less than 15 min away, 15–30 min away, etc. The distribution is shown in Table 12. We include as instruments one dummy variable for grandparents living less than 15 min away and one dummy variable for 15–30 min, while the other categories are left as a reference (i.e. in the reference group, grandparents are located 30 or more minutes away). To guarantee the validity of the instrument, in the regressions, we control for a large number of variables which could be correlated with the decision of living close to grandparents and could affect directly the outcomes of interest, such as mother’s education. However, there could be other unobservable characteristics that affect both processes. To explore this possibility, we use data from the British Household Panel Survey, which provides information about moves, reasons for moving, fertility, and other variables not included in the MCS. In Table 13, we compare the opinions of mothers living close/far from grandparents. To mimic the definition of our instrument, we use the threshold of 30 min to define living “close” or “far”. We select women close to motherhood (2 years before/after the birth of a child) and summarize opinions regarding the family and the children, assuming that potential differences in the opinions could reflect differences in childrearing methods. Table 13 shows that 12

There are alternative or complementary potential explanations: the moment when the child is placed in formal childcare could coincide with the period when the mother starts working or starts working more, or when other siblings are born. In both cases, when placed in formal care, the child would also be exposed to fewer interactions with the adult at home, for different reasons. However, replicating the estimations with different specifications yields the same results (excluding all children with younger siblings; including the working status and working hours by the mother at different waves), inducing us to exclude those interpretations.

The role of grandparenting in early childcare and child outcomes Table 12 Distance between parents’ and grandparents’ houses

Distance to grandparents

Freq.

Percent

<15 min

1397

50.58

−30 min

421

15.24

−60 min

244

8.83

>60 min (within UK)

422

15.28

Outside UK

109

3.95

Dead or lost

169

6.12

Observations

2762

Table 13 Percentage of people who agree with the following statements, by distance to grandparents Mothers living closea to grandparents

Mothers living fara from grandparents

Sign. diff.

Opinions 2001 Pre-school child suffers if mother works

24.5

25.3

Family suffers if woman works full time

30.4

31.6

Children need father as much as mother

83.1

83.9

Employers should help with childcare

85.9

88.2

Single parents are as good as couples

59.6

66.3

Observations

494

**

352

2002, 2006 Parents ought stay together for children Marital status is irrelevant to children Observations

5.4

7.5

56.8

50.9

903

898

Notes: Sample of women 2 years before/after the birth of a child “Living close to grandparents” refers to those women less than 30 min. away from grandparents. “Living far from grandparents” refers to those living 30 min or more

a

Source: BHPS 2001, 2002, 2006 **p < 0.05; ***p < 0.01

mothers living close or further from grandparents have the most opinions in common. Always using data from the BHPS, Table 14 shows that there is no tendency in the UK to move close to grandparents when women have a child. Women close to motherhood are more likely to move than other women, but for reasons having to do with housing and location. Table 15 at the top shows the impact of housing proximity on the probability of grandparental care (first stage): children living closer to the grandparents are more likely to be cared for by them. The F-test also confirms that the instrument is relevant. The bottom section of Table 15 compares OLS for the sub-sample used and IV estimates. In the first row, we report the OLS estimates of the effect of having

D. Del Boca et al.

Table 14 Descriptive statistics of moving choices, new mother and all other women Non-movers Movers Observations

Mothers

All other women

82.2

87.2

17.8

12.8

12,240

44,523

Reason for moving Job related

8.7

15.4

14.0

16.3

Move in with family

1.2

1.8

Move from with family

1.1

1.2

To be closer to family, friends

3.2

2.4

Friends or college related

0.7

7.3

Forced moving

5.8

6.8

56.7

41.3

Partner related

Accommodation or area reasons Others Observations

8.6 1969

7.5 5130

Notes: Women 2 years before/after the birth of a child vs. all women Source: BHPS 1991–2008

grandparents as the main source of care compared with being in a formal care centre. We can see a positive association with Naming Vocabulary at age 3, and a negative one with Picture Similarity at age 5. In the second row, the IV estimates confirm the effect on Picture Similarity. Additionally controlling for parents’ employment, wages, and hours of work—which could affect both the decision of where to live and the outcomes—yields the same results.13 The results hold even when we use three rather than two categories as excluded instruments: less than 15 min, 15–30 min, 30–60 min. Finally, parents of children with disabilities may decide to stay closer to grandparents because they need additional help to address the cognitive skills of the child. If this is the case, the exclusion restriction of our IV would be not valid. To test against this issue, we exclude children with longstanding health conditions that have been diagnosed by a health professional: this includes both cases in which such health conditions limit or not the child from joining in normal activities, and concerns about 15% of children in our sample.14 Results are confirmed (reported in the online Appendix, Table B.8). 6.4 Robustness checks In our main specification, we only include the childcare decision at one point in time: when the child is 18 months. This decision may be correlated with other subsequent childcare/educational choices, which probably affect child outcomes and which may “take away” the observed effect of early childcare at 18 months. 13

Available upon request.

14

Longstanding health conditions which limit normal activities concern instead about 2.5% of the sample.

0.29*** (0.02)

Grandpar. childcare

−0.17 (0.41)

2568

[0.000]

81.99

−2.18 (1.55)

2561

[0.000]

82.37

0.76 (1.54)

−0.54 (0.41)

Constr.

Pattern

Picture

2563

[0.000]

81.96

−4.34*** (1.68)

−1.47*** (0.44)

Simil.

2517

[0.000]

83.80

−0.14 (1.70)

−0.62 (0.46)

Constr.

Pattern

Age 7 Word

2481

[0.000]

83.69

−1.16 (2.66)

−0.07 (0.71)

Reading

Number

2523

[0.000]

85.50

−3.87 (2.42)

−1.21 (0.64)

Skills

**p < 0.05; ***p < 0.01

a The first stage is estimated separately for each outcome (since the number of observations varies), but we report the results only for the first one, because they are very similar. Controlling for all the variables listed in section 5. Full table with controls available upon request

Robust standard error in parentheses. Notes: Effect of grandparents care as main care when the child is 18 months old (excluded category: formal care). OLS and IV estimations (excluded instruments: living 15 min far away from grandparents, 15–30 min)

2457

[0.000]

[0.000]

2371

83.98

86.75

F-test

Observations

1.31*** (0.44)

0.94 (1.62)

−1.17 (0.64)

−1.67 (2.30)

Grandpar. (OLS)

Vocab.

Read.

Grandpar. (IV)

Naming

Naming

School Vocab.

Age 5

Age 3

Distance: 15–30 min 0.26*** (0.03)

Distance: <15 min

First stagea

Table 15 IV estimates (Grandparents vs. formal care)

The role of grandparenting in early childcare and child outcomes

D. Del Boca et al.

We then repeat the regressions for outcomes measured at age 5, where we include a variable indicating early formal care between the ages of 3 and 5 years; and for outcomes at age 7, where we also include a variable indicating attendance of school reception class at age 5. We report the estimated effects in Table 19 in the Appendix: the positive effect of experiencing any formal care at 18 months on Picture Similarity is still present. Finally, we chose childcare modalities when the child is 18 months old. We have also measured childcare types at 12 and 24 months, but we cannot include them in the regressions because we would have multi-collinearity problems. However, we can check how sensitive results are to this choice. Table 20 in the Appendix shows that including childcare at 12 or 24 rather than at 18 months does not change the overall results.

7 Conclusions and discussions In this paper, we analyse the effect of early childcare provision by grandparents and by formal childcare centres on child cognitive outcomes in the UK. We find that, on average, children looked after by their grandparents at the age of 18 months appear to have better grasp of vocabulary. This result is potentially related to the more stable relationships and one-on-one interactions with an adult provided by the grandparents. However, children looked after by their grandparents perform worse on other cognitive tests than children in formal childcare (school readiness and non-verbal reasoning, as tested with Picture Similarity). In addition, children who attended early formal childcare perform better in mathematical concepts. We also explore if the effects depend on the socio-economic status of the family. Our results indicate that that the positive association of grandparental care with naming abilities is stronger for households with higher incomes. In contrast, the negative association with other abilities is stronger for households with lower incomes. The average effects appear to be significant when measured at ages 3 and 5, but decline at age 7, when the child is in primary school, although the effects are still significant when estimated separately by socio-economic background. The fact that the negative effect of grandparental care on cognitive outcomes is diminishing over time is somewhat reassuring and may be linked with the positive effect of introducing formal care even at later ages (when the child is between 18 months and 3 years). Nevertheless, the emergence of heterogeneous effects in the medium term underlines the importance of early childhood education for disadvantaged children. Moreover, recent studies have demonstrated that early childhood education fosters non-cognitive skills, which are important predictors of longterm outcomes. Our study contributes to the literature by analyzing early childcare and child outcomes in relation to grandparental care vs. formal childcare. An understanding of the effects of different forms of care on child outcomes may help inform policymakers decisions about which (or whether) to subsidize. Our findings support policies promoting formal childcare rather than informal care in order to improve school readiness, especially among children from more disadvantaged families. This conclusion is consistent with previous evidence that disadvantaged children are the ones who benefit most from early childhood education.

The role of grandparenting in early childcare and child outcomes

The recent introduction of policies providing free childcare is in fact moving in the direction of increasing the length and the hours of early childhood education. Acknowledgements We are grateful to the participants to seminars at Collegio Carlo Alberto, NYU, Dondena (Bocconi), University of Lyon, ISER (University of Essex), University of Turin, International Workshop on Intergenerational Relations (University of Bolzano) and to participants at the 2014 Alp-Pop Conference in La Thuile and the 28th ESPE Conference in Braga for helpful discussions and suggestions. We thank Collegio Carlo Alberto for financial and technical support. Compliance with ethical standards Conflict of interest The research leading to these results has received funding from the European Union’s Seventh Framework Programme (FP7/2007–2013) under grant agreement no. 320116 for the research project FamiliesAndSocieties.

8 Appendix Tables 16–20

Table 16 Childcare choices by quantile of household income Quantile Obs. Equivalent HH income (w1)

Nonworking mother

Working parents

Grandparents Formal care Childminder

Others

1

459

0.69

0.09

0.14

0.02

2

458

94.03

0.73

0.08

0.10

0.07

0.01

0.04

3

463

109.90

0.81

0.06

0.06

0.04

0.02

0.03

4

465

131.67

0.68

0.14

0.12

0.05

0.02

0.03

5

448

154.65

0.64

0.18

0.10

0.04

0.02

0.04

6

458

178.68

0.53

0.23

0.15

0.07

0.04

0.04

7

456

196.31

0.49

0.23

0.16

0.06

0.03

0.05

8

474

223.79

0.37

0.34

0.20

0.08

0.03

0.04

9

444

241.69

0.38

0.29

0.19

0.09

0.07

0.06

10

462

262.88

0.26

0.34

0.32

0.08

0.06

0.06

11

478

291.76

0.25

0.35

0.27

0.10

0.07

0.09

12

455

311.71

0.27

0.32

0.28

0.10

0.07

0.06

13

445

350.28

0.25

0.29

0.32

0.11

0.11

0.05

14

461

371.63

0.19

0.33

0.32

0.17

0.08

0.04

15

447

411.15

0.19

0.30

0.31

0.17

0.11

0.05

16

458

449.22

0.16

0.29

0.32

0.19

0.12

0.06

17

462

514.90

0.20

0.21

0.26

0.23

0.15

0.05

18

465

575.10

0.17

0.23

0.28

0.26

0.13

0.05

19

446

704.93

0.16

0.21

0.26

0.30

0.12

0.04

20

458 1019.74

0.20

0.13

0.22

0.38

0.12

0.04

0.38

0.23

0.22

0.13

0.07

0.05

Total

9162

60.83

332.18

0.07

0.03

D. Del Boca et al. Table 17 Sample selection Variable

Wave 1 Singleton mother resp.

Wave 2

Wave 2 Care variables

Wave 3

Wave 4 No missing variables

Child British

0.81

0.83

0.84

0.85

0.87

Indian/Pakistani/Bangladeshi

0.09

0.09

0.08

0.07

0.06

Other ethnicity

0.10

0.08

0.08

0.08

0.07

Girl

0.49

0.49

0.49

0.49

0.50

Age child (wave 1) (months)

9.7

9.7

9.7

9.7

9.7

Birth-weight

3.4

3.4

3.4

3.4

3.4

Breastfed 1 month

0.45

0.47

0.48

0.49

0.50

Injuries

0.08

0.08

0.08

0.08

0.08

Hospital

0.20

0.19

0.19

0.18

0.17

Communicative devel.

0.01

−0.01

−0.03

−0.04

−0.06

Motor development

0.01

0.01

0.02

0.02

0.03

Motion development

0.00

0.02

0.03

0.05

0.07

Household Siblings

0.94

0.94

0.94

0.93

0.91

Equivalent weekly income

298

311

318

324

332

Meet friends every week

0.70

0.71

0.71

0.71

0.72

England

0.62

0.63

0.63

0.63

0.63

Wales

0.15

0.15

0.15

0.15

0.16

Scotland

0.13

0.12

0.12

0.12

0.12

Northern Ireland

0.10

0.10

0.10

0.10

0.10

Father at home

0.82

0.85

0.86

0.87

0.88

29.10

29.47

29.68

29.85

30.01

0.33

0.36

0.37

0.38

Mother Age Academic upper sec. education Age left education

17

17

18

18

0.41 18

Job during pregnancy

0.62

0.65

0.67

0.68

Post-partum depression

0.25

0.24

0.24

0.24

0.70 0.24

Difficult childhood

0.30

0.29

0.29

0.28

0.28

Long illness

0.21

0.22

0.21

0.21

0.22

Cigarettes per day

3.3

3.2

3.1

3.0

2.9

Alcoholic drinks Observations

0.36

0.38

0.38

0.39

0.40

18,256

14,668

12,639

11,278

9162

Notes: Sample selection due to attrition and other selection criteria: only singletons, mother respondent when the child is 9 months old, control variables not missing, at least one child outcome not missing. Child’s development variables (communicative, motor, and motion) are factor points derived from factor analyses

The role of grandparenting in early childcare and child outcomes Table 18 Childcare choices between age 18 months and 36 months of the child 3 years old

Non-working

Working

Grand-

Formal

Child-

18 months old

mother

parents

parents

care

minder

Others

Non-working mother

0.74

0.03

0.03

0.18

0.02

0.02

Working parents

0.02

0.84

0.02

0.09

0.02

0.01

Grandparents

0.05

0.01

0.83

0.10

0.01

0.01

Formal care

0.06

0.01

0.01

0.90

0.01

0.01

Childminder

0.07

0.01

0.02

0.16

0.73

0.01

Others

0.09

0.02

0.01

0.17

0.04

0.67

0.22

0.09

0.06

0.09

0.15

8089

9.16*** (2.07)

1.82 (1.31)

−0.29 (0.77)

0.10

8199

6.48*** (1.61)

0.54 (1.04)

−0.49 (0.88)

**p < 0.05; ***p < 0.01

Robust standard error in parentheses. Notes: it corresponds to coefficients from Tables 7and 8 (first column), adding controls for subsequent childcare modalities. Controlling for all the variables listed in section 5. Full table with controls available upon request

R squared

8354

0.49 (0.71) 8173

8343

0.25 (0.73)

−1.96** (0.97)

−0.97 (0.89)

−0.64 (0.63)

−0.05 (0.59)

0.03 (0.54) −0.22 (0.58)

0.48 (0.63)

−0.09 (0.64)

−0.81 (0.71) −1.00 (0.70)

0.15 (0.48) 0.24 (0.46)

0.02 (0.42)

−0.16 (0.74)

0.98 (0.67)

Skills

Number

0.12 (0.82)

−1.01 (0.75)

Reading

Word

−0.15 (0.42)

Observations

8368

0.39 (0.49) −0.12 (0.54)

3.12** (1.26)

0.98 (0.72)

0.95** (0.46) −0.68 (0.49)

Reception class (age 5)

Formal care after age 3, before age 5

1.45 (0.75)

0.56 (0.51) 0.06 (0.56)

0.68 (0.51)

−0.99 (0.54)

Childminder (18 m)

Others (18 m)

0.39 (0.42) 0.21 (0.40)

0.20 (0.41)

0.49 (0.39)

0.47 (0.48)

0.33 (0.42)

Working parents (18 m)

0.13 (0.47)

Non-working mother (18 m)

Grandparents (18 m)

0.01 (0.43)

Formal care (18 m)

Construction

Pattern

Construction

Vocabulary

Similarity

Pattern

Naming

Picture

Age 7

Age 5

Table 19 Childcare/educational decisions over time

D. Del Boca et al.

The role of grandparenting in early childcare and child outcomes Table 20 Early childcare in different points in time Main results Childcare

Childcare

Childcare

18 months old

12 months old

24 months old

Any

Any

Any

1.56** (0.67)

1.60** (0.74)

1.56** (0.58)

1.24*** (0.42)

1.38*** (0.44)

1.46*** (0.38)

0.95** (0.46)

1.00** (0.50)

0.79 (0.41)

(Tables 6–8)

School Readiness (age 3) Formal care Naming Vocabulary (age 3) Grandparents Picture Similarity (age 5) Formal care

Robust standard error in parentheses. Notes: Childcare modalities with significant coefficients from Tables 6–8 (first column), substituting childcare choices at 18 months old with childcare choices at 12 and 24 months old. Controlling for all the variables listed in section 5. Full table with controls available upon request **p < 0.05; ***p < 0.01

References Aassve, A., Meroni, E., & Pronzato, C. (2012). Grandparenting and childbearing in the extended family. European Journal of Population, 28(4), 499–518. Arpino, B., & Bordone, V. (2016). Regular provision of grandchild care and participation in social activities. Review of Economics of the Household, doi:10.1007/s11150-016-9322-4. Arpino, B., Pronzato, C., & Tavares, L. P. (2014). The effect of grandparental support on mothers’ labour market participation: An instrumental variable approach. European Journal of Population, 30(4), 369–390. Bernal, R. (2008). The effect of maternal employment and child care on children’s cognitive development. International Economic Review, 49(4), 1173–1209. Bernal, R., & Keane, M. P. (2011). Child care choices and children’s cognitive achievement: the case of single mothers. Journal of Labor Economics, 29(3), 459–512. Blau, D., & Grossberg, A. (1992). Maternal labor supply and children’s cognitive development. Review of Economics and Statistics, 74(3), 474–481. Brunello, G., & Rocco, L. (2016). Is childcare bad for the mental health of grandparents? Evidence from SHARE. IZA Discussion Paper N.10022, Institute for the Study of Labor. Bryson, C., Brewer, M., Sibieta, L., & Butt, S. (2012). The role of informal childcare: A synthesis and critical review of the evidence. London: Nuffield Foundation. Cardia, E., & Ng, S. (2003). Intergenerational time transfers and childcare. Review of Economic Dynamics, 6(2), 431–454. Carneiro, P., & Heckman, J. J. (2003). Human capital policy. In J. J. Heckman, A. B. Krueger & B. M. Friedman (Eds.), Inequality in America: What role for human capital policies? (pp. 77–239). Cambridge, MA: MIT Press. Compton, J., & Pollak, R. A. (2014). Family proximity, childcare, and women’s labor force attachment. Journal of Urban Economics, 79, 72–90. Connelly, R. (2013). Interpreting test scores. Millennium Cohort Study Data Note 2013/1, Centre for Longitudinal Study. Cornelissen, T., Dustmann, C., Raute, A., & Schönberg, U. (2016). Who benefits from universal child care? Estimating marginal returns to early child care attendance. Mimeo, https://docs.wixstatic.com/ ugd/6247dc_19372ccede0f40a28a291291abad1882.pdf Cunha, F., & Heckman, J. J. (2007). The technology of skill formation. American Economic Review, 97(2), 31–47.

D. Del Boca et al. Datta Gupta, N., & Simonsen, M. (2010). Noncognitive child outcomes and universal high quality child care. Journal of Public Economics, 94(1–2), 30–43. Del Boca, D., Flinn, C., & Wiswall, M. (2014). Household choices and child development. Review of Economic Studies, 81(1), 137–153. Del Boca, D., Martino, E. M., & Piazzalunga, D. (2017). Investments in early education and child outcomes: Short and long term effects. ifo DICE Report, 15(1), 43–48. Del Boca, D., Pasqua, P., & Suardi, S. (2016). Child care, maternal employment, and children’s school outcomes. An analysis of Italian data. European Journal of Population, 32(2), 211–229. Duncan, G., & Magnuson, K. (2013). Investing in preschool programs. Journal of Economic Perspectives, 27(2), 109–132. Elango, S., Garca, J. L., Heckman, J. J., & Hojman, A. (2016). Early childhood education. In R. A. Moffitt (Ed.), Economics of means-tested transfer programs in the United States (Vol. 2, pp. 235–297). Chicago, IL: University of Chicago Press. Felfe, C., & Lalive, R. (2012). Early child care and child development: for whom it works and why. IZA Discussion Paper N. 7100, Institute for the Study of Labor. Fergusson, E., Maughan, B., & Golding, J. (2008). Which children receive grandparental care and what effect does it have. Journal of Child Psychology and Psychiatry, 49(2), 161–169. Garca-Morán, E., & Kuehn, Z. (2017). With strings attached: Grandparent-provided child care, fertility and female labor market outcome. Review of Economic Dynamics, 23, 80–98. Gray, A. (2005). The changing availability of grandparents as carers and its implications for childcare policy in the UK. Journal of Social Policy, 34(4), 557–577. Gregg, P., Washbrook, E., Propper, C., & Burgess, S. (2005). The effects of a mother’s return to work decision on child development in the UK. The Economic Journal, 115(501), F48–F80. Hank, K., & Buber, I. (2009). Grandparents caring for their grandchildren: Findings from the 2004 survey of health, ageing, and retirement in Europe. Journal of Family Issues, 30(1), 53–73. Hansen, C., & Hawkes, D. (2009). Early childcare and child development. Journal of Social Policy, 38(2), 211–239. Ho, C. (2012). Optimal disability insurance with informal child care. Centre for Silver Security WP CSSWP-04-201. Ho, C. (2015). Grandchild care, intergenerational transfers, and grandparents’ labour supply. Review of Economics of the Household, 13, 359–384. OECD (Ed.) (2001). Balancing work and family life: Helping parents into paid employment. In OECD Employment Outlook 2001. Paris: OECD Publishing. OECD (2016a). PISA 2015 results in focus. Paris: OECD Publishing. OECD (2016b). Education at a Glance 2016: OECD Indicators. Paris: OECD Publishing. Ofsted (2016). Early years and childcare registration handbook. Available at https://www.gov.uk/ government/uploads/system/uploads/attachment_data/file/551134/EY_and_childcare_reg_handbook. pdf. Ruhm, C. (2004). Parental employment and child cognitive development. Journal of Human Resources, XXXIX, 155–192. Rupert, P., & Zanella, G. (2014). Grandchildren and their grandparents’ labor supply. University of Bologna Department of Economics Working Paper No. 937. Todd, P., & Wolpin, K. (2003). On the specification and estimation of the production function for cognitive achievement. The Economic Journal, 113(February), F3–F33. UK Government (2014). Fuller working lives: A framework for action. Policy Paper by the Department for Work and Pensions. Available online at https://www.gov.uk/government/publications/fullerworking-lives-a-framework-for-action. UK Government (2015). Chancellor announces major new extension of shared parental leave and pay to working grandparents. Available online at https://www.gov.uk/government/news/chancellor-a nnounces-major-new-extension-of-shared-parental-leave-and-pay-to-working-grandparents. Vandell, D., & Ramanan, J. (1992). Effects of early and recent maternal employment on children from low income families. Child Development, 63, 938–949. Walker, J. R. (1991). Public policy and the supply of child care services. In D. Blau (Ed.), The Economics of child care (pp. 51–78). New York, NY: Russel Sage Foundation. West, A. (2006). The pre-school education market in England from 1997: Quality, availability, affordability and equity. Oxford Review of Education, 32(3), 283–301. Wheelock, J., & Jones, K. (2002). Grandparents are the next best thing: Informal childcare for working parents in Urban Britain. Journal of Social Policy, 31(3), 441–463.