Many children from broken homes are born “losers” and so deprived of love that they grow up to be dysfunctional adults, according to a Catholic bishop. (Irish Independent, 16/8/2011).

Lone parents often receive poor media coverage and examples such as the above quote are common. In 2011, the British Press blamed poor parenting for a series of riots which took part in London and around the UK. Much media coverage was given to the moral collapse of British society and the breakdown of the family;

Mr Cameron said that family breakdown and poor parenting had played significant role. “In too many cases, the parents of these children – if they are still around – don’t care where their children are or who they are with, let alone what they are doing,” he said. (Telegraph, 11/8/2011).

In the Irish case, a prominent academic called for DNA testing to be used to track runaway dads who dodge financial responsibilities for their offspring, in order to tackle Ireland’s worsening economic situation. Accordingly, Dr. Ed Walsh noted that;

whether a community is wealthy or poor, where you have a predominance of lone parents, you have a predominance of crime (The Irish Examiner, 20/1/2011).

Yet, what is the empirical evidence to support such claims made about lone-parenting? International empirical evidence suggests that children who grow up living with both biological parents fare better than children who don’t (Acock and Demo 1994; Amato 2001; Amato and Booth 1997; Cookston 1999; Flewelling and Bauman 1990; Frisco, Muller, and Frank 2007; Mayer 1997; McKeown et al. 1997; Patten et al. 1997). A meta-analysis of 92 studies conducted during the 1980s (Amato and Keith 1991) and of 67 studies conducted during the 1990s (Amato 2001), for example, compared the wellbeing of children from divorced and two-parent families and found that children from divorced families had significantly lower scores on a range of outcomes including educational achievement, behaviour, psychological adjustment, self-concept, social competence and long-term health.

The empirical evidence on the differences in adolescent and child outcomes across family types may however result in a large part, or entirely, from selection bias. It is therefore possible that marital dissolution is rather inconsequential. In the Irish case, while divorce rates remain low, there are significant numbers of children growing up in never-married one-parent families. The selection view of lone-parenthood maintains that unmarried childbearing does not necessarily result in negative outcomes for children. In the Irish case, the majority of unmarried parents are mothers who tend to be younger and come from impoverished backgrounds when compared to the married populations. The adverse consequences of unmarried motherhood may therefore be an artefact of the pre-existing socio-economic disadvantages of these mothers. This chapter presents the empirical evidence on the association between family structure and children’s wellbeing for Ireland.

The Irish Case

Like most European countries, the number of children in Ireland growing up in one-parent families continues to increase (see Fig. 14.1). Between the 1986 and 2011 Census of Population in Ireland, the number of one-parent families almost tripled so that by 2011, 24.8% of all families were one-parent families. According to Eurostat, Ireland and Latvia had the highest percentage of children living with a lone parent in the EU (OECD 2007). The growth in lone-parenthood was driven by a dramatic rise in the proportion of births occurring outside of marriage; from 5% in 1980 to 32% in 2000 (Lunn, Fahey, and Hannan 2009). Marital breakdown has also risen in Ireland although it appears that the Irish are still adverse to divorce (Fig. 14.1). Between the 1986 and 2006 Census, the total number of people whose marriages had broken down increased five-fold, from 40,000 to just fewer than 200,000 although the rate is still low by international comparisons (Lunn and Fahey 2011).

Fig. 14.1
figure 1

The Changing Nature of Family Formations in Ireland 1971–2011 (Population aged 16–64) (Source: Minnesota Population Centre IPUMS; CSO Ireland 10% random census samples)

This aversion to divorce and the continuation of a late age of marriage makes Ireland an interesting case study. It points to a selection process which means that marriage is not entered into lightly. Higher rates of entry into marriage among the professional classes were a prevalent feature of Irish demography throughout the twentieth Century (Connolly 2015). On average, those who marry and then have children are older, more educated and more likely to have professional careers when compared to those who do not have children within marriage. Despite globalisation, secularisation and a rise in individualisation, the decision to marry is still constrained by economic considerations related to one’s social class. It appears that love is less “liquid” in Ireland(Bauman 2003).

In contrast, childbearing is more fluid where the dominant route into lone-parenthood is outside of marriage (Fig. 14.1).

Debating the Role of Parents

Following from this argument, it is not family structure per se which influences child development but the characteristics of the parents. Marriage is therefore not necessarily good for child development or indeed for society. Nor is the rise in parenting guides, a product of Beck’s risk society and the need for expert advice, necessarily going to improve child development. According to the selection argument, the key is the social, economic and cultural resources of the parent or parents, rather than their marital status/living arrangements.

Such an approach contrasts sharply with Brad Wilcox’s National Marriage Project in the US. In a series of articles which have received much press coverage, Wilcox and colleagues argue that marriage matters for child development, parental income and health, societal wellbeing and a state’s economic prosperity. The overall conclusion is that marriage deserves special care and should be promoted, a point made by groups opposing the Marriage Equality Bill in Ireland which came into legislation in September 2015.

According to Putnam, the class divide in marriage is widening as the traditional family thrives among the educated elites (Putnam 2015). In his book, Putnam details the wide range of factors which differentiate the parental resources found in the homes of one-parent and two-parent families. Current income is often used as an indicator of these resources yet income differences across family types are a consequence of antecedent factors. Demographic characteristics such as nationality and age, and socio-economic status, like educational attainment and socio-economic background (social class) are all important factors influencing family income, child development and the propensity to marry. Ideally, a study of marriage would need to include all these factors and their interplay in the analysis of the effects of marriage at the individual or societal level.

Data Source

This research draws on the first wave of the Growing Up in Ireland (GUI) child cohort study, a large-scale survey of nine-year-old children (N = 8568) sampled within primary schools in Ireland in 2008/2009. The GUI study is an extremely rich data source, incorporating school principal, teacher, parent and child questionnaires as well as time diaries, and some qualitative data collection. At the time of writing, only one wave of the data was available for analysis.

As noted by Lunn and Fahey (2011), while there is diversity in family structures, there remain a small number of dominant family types that account for the large majority of families in Ireland as a whole. Many of the less traditional family types are not really that common in Ireland and this was reflected in the GUI data. While there are good policy reasons for treating one-parent families as a single group in relation to financial support, in this analysis never-married one-parent families were distinguished from the separated, divorced and widowed parents who are not living with a partner.

In terms of two-parent families, cohabiting and married parents living together with their children were distinguished as separate groups. Step-families are however included with this cohabiting group and lone fathers are not separately analysed given the small numbers involved (see Table 14.1). In general there is an under-representation of lone fathers in the Growing up in Ireland data.

Table 14.1 Family types in the growing up in Ireland, child cohort study

The Importance of Selection

Before discussing the effects of family type on child development, it is important to detail the important role of selection effects. In common with Putnam (2015) data from the child cohort study (GUI) reveal that marriage is more common among older, more educated, and more religious mothers (see Table 14.2).

Table 14.2 Differences in the characteristics of the primary caregiver (usually the mother) across family types

In wave one of the GUI study, the average age of a married mother in a two-parent family was 40 compared to 34 in a never-married one-parent family. Almost 19% of married mothers had a degree or higher-level educational qualification compared to 12% of unmarried mothers (Table 14.2). Mothers in previously-married one-parent families had much in common with mothers in married two-parent families, given this selection into marriage. Separated, divorced or widowed lone-parents (the post married group) had an average age of 40 and 16% had a degree or higher-level qualification. Never-married mothers tend to have children at a relatively young age, compared to all other women in this child cohort study. More than half of all unmarried mothers (58.5%) were less than 25 years old when they had their child compared to only 13% of married mothers (Table 14.2).

The nature of family structure re-affirms the importance of taking account of selection effects when it comes to marriage. Of the married majority (76%) of parents in the sample, only 4% were cohabiting outside of marriage with the father of their child at childbirth, as most (84%) were married when they had their child. Cohabitation rates were higher at childbirth among those who are now separated or divorced with about 10% of those who are separated or divorced cohabiting at childbirth. Of never-married lone-mothers (9.6% of the sample), almost 1 in 4 were cohabiting with the father of the child at the time of birth.

The low divorce rate in Ireland is possibly reinforcing this selection into marriage. Lunn and Fahey (2011) suggest that cohabiting unions (6.4% in the sample) and non-marital child-bearing have served to select less stable relationships out of marriage and concentrate union instability into these non-marital family situations. Marital separation is, however, much more common than divorce, so that that the overall levels of marital instability in Irish families as a whole are much closer to a European mean (Fahey 2012).

Table 14.2 includes the full range of factors which have been found to differ across families, including the height of mothers as it reflects social class background. The next section moves away from socio-economic differences in the background of mothers prior to the birth of the study child and focuses on behaviour during birth and upon the birth of the child which differs across family settings.

Confounding Factors

Selection is present when the factors that select individuals into marriage also influence child development. These factors are referred to as confounding factors; variables that affect both the probability of selection into a group and the outcome of interest. The GUI study allowed for the analysis of a range of confounding factors, including those factors detailed in Table 14.2.

In addition, the behaviour of mothers during pregnancy has been found to vary across socio-economic groups and to influence the development of the child, both in the long-term and short-term. Table 14.3 lists a number of other confounding factors taken into account here. Williams et al. (2010), for example, found that smoking during pregnancy is associated with lower educational levels among mothers and is related to a range of problems among children. Most mothers in the Growing up in Ireland study, regardless of family type, did not report to drinking any alcoholic beverages while pregnant, whereas smoking during pregnancy has a clear gradient with 13% of married mothers smoking while pregnant compared to twice that number of cohabiting or previously-married mothers and almost a third of all never-married lone-mothers admitted to smoking during pregnancy (see Table 14.3).

Table 14.3 Differences in the ex-ante characteristics of mothers across family types

A history of long-term illness, including depression among mothers, puts an increasing strain on family life and has been found to influence rates of marital breakdown as well as child development. Rates of chronic illness prior to the birth of the study child were not randomly distributed in our data with 1 in 10 never-married mothers suffering from a chronic illness prior to childbirth, compared to 1 in 20 married mothers. A more detailed breakdown reveals that mothers who are now separated from their husbands and rearing children alone reported the highest rate of illness prior to the birth of the child (at 21%).

A series of other confounding factors are highlighted in Table 14.4 which show the importance of controlling for confounding factors when trying to assess the impact of family structure on child health. Low birth weight is more common among never-married lone-mother (8% of children weighted less than 2.5 kg), compared to 5% born to a married mother (Table 14.4). This is related to the fact that never-married lone-mothers are more likely to have their child prematurely (2.9% compared to the sample average of 1.8%).

Table 14.4 Differences in the characteristics of the study child across family types

Elective caesareans were more common among married mothers with almost 1 in 10 opting for a caesarean, compared to 3.5% of unmarried lone-mothers. The mode of delivery at birth has the potential to influence the long-term development of the child, for example, Huh et al. (2012) found a link between the risk of obesity and caesarean section. Rates of breastfeeding also vary across family types. The lowest propensity to breastfeed was found among never-married cohabiting parents and never-married one-parent families (only one in three breastfed) compared to almost half (47%) of married mothers and 44% of previously-married lone-mothers who breastfed their child (table 14.4).

The differences reported in Tables 14.2, 14.3 and 14.4 points to the role of confounding factors and the importance of taking them into account when measuring differences in child outcomes across family types. There are many unobservable differences not reported here, as well as important missing confounding variables such as social class, but bearing these in mind, it is important to assess the size of the differential in a range of child outcomes controlling for the confounding factors discussed here.

School Clustering

In addition to the above selection effects into non-marital childbearing, lone-mothers tend to be spatially concentrated in areas of disadvantage. Family setting influences child development through socio-economic resources and these resources influence the schools and neighbourhoods in which children reside. Rather than seek to explain underachievement among children from never-married and post-married families by focusing on the living arrangement of the mother, or the individual characteristics of the mother alone, Williams (1989) argued that the impact of the social structure must also be taken into account. All the models presented in this chapter take account of the clustering of children within schools.

The Propensity Score Matching Method

Propensity score matching techniques were employed to adjust for selection bias, controlling for the clustering of children within schools, in order to assess the effect of family structure on numerous outcomes in young children pertaining to their health, psychological wellbeing and educational development.

Propensity score matching methods use an estimate of the counterfactual group to adjust for selection bias (Rosenbaum and Rubin 1983). In order to reduce the bias of confounding factors, we need to know the answer to the counterfactual question, such as what level of an outcome would a child have gained had their parents married? By definition, the counterfactual cannot be empirically observed. However, it is possible to estimate the counterfactual by matching cases that are similar on confounding factors but differ on the focal independent variable (i.e. family type).

Cases are matched on multiple confounding factors, usually between two groups: a treatment group and a control group. One could imagine that two children are matched on the same pre-existing characteristics, one of whom is living in a one-parent family and the other in a two-parent family. Matching can, in principle, be done on a range of variables, but the more variables available, the more difficult it becomes to find a matched child. Instead, matching is carried out on the propensity score, which reflects the probability of receiving treatment assignment.

The predicted probabilities of receiving the treatment were calculated from a logit model which served to match the treatment and control groups based on pre-existing observed covariates i.e. the confounding variables. Then, using the propensity scores, a sample of treatment groups and their matched cases was generated. This normally consists of two groups, for instance, unmarried mothers and their matched cases in the married group. The current study extended this common technique to match multiple treatment groups: unmarried one-parent families, previously-married one-parent families and cohabitating families were matched to a control group of married two-parent families (groups are defined in Table 14.1).

Three issues arise in implementing propensity score matching which affect the matching process. First, matching with replacement was employed as it minimises the propensity score distance between the matched comparison units and the treatment unit. This is beneficial in terms of bias reduction. Secondly, the comparison units were chosen based on the nearest-neighbour method, which selects the m comparison units whose propensity scores are closest to the treated unit in question. Nearest neighbour matching faces the risk of bad matches if the closest neighbour is far away. This was avoided by imposing a tolerance level on the maximum propensity score distance (calliper).

The performance of different matching estimators was assessed prior to the settling on nearest neighbour (50), calliper (0.01). The fundamental assumption for the validity of matching is that, when observable characteristics are balanced between the family types, the groups are balanced with respect to all the characteristics relevant to the child outcomes. This assumption, however, relies on having a large number of available pre-intervention characteristics. A concern therefore was the success of the matching process given the limited list of available confounding variables.

The existence of a substantial overlap between the characteristics of the treated and control groups (the issue of common support) was assessed. To satisfy the requirement of common support, observations outside the maximum propensity score for our treatment groups were dropped (treated: off support ranged from N: 3 to 14 or a maximum of 3.3% of cases were dropped for the never-married one-parent group).

The second issue assessed was the percentage of bias reduction between the treated groups and the control group in the matched sample which was found to be, in general satisfactory (a range of 60–98% bias reduction) except for the case of the cohabiting group of parents where the matching process was not ideal. Caution is therefore advised in relation to this smaller group of families.


The results are split into three substantive areas; measuring the influence of growing up in a one-parent family on a child’s educational development, before and after matching. Secondly, measuring the effects of family structure on a child’s physical health and finally, on their emotional wellbeing. The results for cohabitation will not be discussed here given the limitation discussed in the previous section in relation to the matching process.

Educational Development

Figure 14.2 plots the size and significance of the negative effects of growing up in a never-married one-parent family in relation to the educational performance of children at age 9. The graph shows the average difference, that is, the point estimates in relation to three measures of educational performance in the never-married lone parent group compared to the married control group, with the lines reflecting 95 and 99% confidence intervals.

Fig. 14.2
figure 2

Estimated educational differences between children in never-married one-parent families compared to their married counterparts (matched data) (Note: Confidence intervals calculated from robust standard errors for the matched sample)

The first point to note is that children from never-married lone-parent families scored on average 10.6 percentage points less on a math test, 7.8% less on a reading test and missed school on average 2 days more than their married counterparts (weighted raw differences plotted in Fig. 14.2).Footnote 1 Taking account of selection bias, reduced the size of these differentials so that children living in never-married one-parent families were scoring 5.2% less in the math test, 3.03% less in the reading test and missing 1.5 days more at school after propensity score matching. In other words, the maths differential was reduced by half and the reading differential was 60% smaller in the matched data (as plotted in the “PS-matched” line in Fig. 14.2). Selection effects therefore account for much of the differences in the study child’s educational outcomes .

The clustering of children within schools/neighbourhood also plays a role. Figure 14.2 plots the effect sizes from using multi-level modelling techniques which take account of the clustering of children within schools (the line labelled “Clustered raw” in Fig. 14.2).This modelling takes account of the fact that children within the same school will perform similarly on the tests administered in the study and have similar rates of absenteeism. Taking account of the school clustering alone accounts for 37% of the difference in math scores, 34% of the difference in reading scores and 36% of the difference in school attendance rates between the never married one-parent families and their married counterparts (Fig. 14.2).

We will not turn to the effects of growing up in a previously-married parental setting on the educational development of children. Children in previously married one-parent families (mostly as a product of marital breakdown rather than divorce or widowhood) also appeared to have substantial differences in educational outcomes at age 9 (Fig. 14.3). These 9-year olds scored on average 6 percentage points less on the maths and reading tests, and missed school 2 days more when compared to their married counterparts. Once selection bias is taken into account the size of the differential is reduced to 3.4 percentage points for maths, 1.6 percentage points for reading and 1.5 days absent. As before, it points to the important role of selection particularly in relation to reading scores (Fig. 14.3). Maths performance is a different story with high between-school differences in the amount of time dedicated to maths.

Fig. 14.3
figure 3

Estimated educational differences between children in previously-married one-parent families compared to their married counterparts (matched data) (Note: Confidence intervals calculated from robust standard errors for the matched sample)

Figure 14.3 shows again that school clustering is important when it comes to educational performance. Focusing on reading test scores, selection bias on observed pre-existing characteristics accounts for three-quarters of the difference in reading scores between children in previously-married one-parent families compared to their married counterparts. School clustering alone accounts for about 40% of the initial raw difference (Fig. 14.4). In all cases, the propensity score matching and school effects estimates are smaller than the initial weighted raw or raw difference (labelled “Raw difference” in figs 14.2 and 14.3) across families.

Fig. 14.4
figure 4

Estimated physical health differences between children living in never-married one-parent families compared to their married counterparts (Note: Confidence intervals calculated from robust standard errors for the matched sample)

Physical Health

A number of child outcomes were analysed in order to assess differences in the physical health of 9 year olds growing up in differing family types. In the GUI study, the primary caregiver was asked about the number of nights which the nine-year-old had spent in hospital as an in-patient over his/her lifetime, excluding neonatal care. They were also asked to indicate how often the Study Child visits the dentist. The Child’s height and weight were measured by the interviewer in the course of the interview and this was used to construct BMI. Children’s dietary intake was also assessed in the study via parental recall of the Study Child’s eating habits in the preceding 24-hour period using a 20-item semi-quantitative food frequency questionnaire. One aspect of this food diary is plotted in the graphs below, the amount of fruit consumed.

The analysis revealed that children from never-married lone-parent families were more likely to have spent time in hospital as an in-patient and to have higher BMI when compared to children from married two-parent families (Fig. 14.4). The increased likelihood of hospital stays and higher BMI was however related to the deprived socio-economic backgrounds of children from these families and not to lone-parenthood per se (the matched differences reported in Fig. 14.4 are close to zero and not statistically significant). There was no evidence of difference in rates of dental visits or the amount of fruit consumed in the diet between children living with lone versus married parents.

The physical health of children living in previously married one-parent families and their married counterparts differed only slightly (Fig. 14.5). Compared to educational outcomes, the differences in terms of size (effect size) were small in relation to a child’s health as measured in the outcomes plotted in Figs. 14.4 and 14.5.

Fig. 14.5
figure 5

Estimated physical health differences between children living in previously married one-parent compared with married counterparts (Note: Confidence intervals calculated from robust standard errors for the matched sample)

Psychological Wellbeing

The GUI study collected information from the children themselves on how they viewed themselves. The children were asked to complete a detailed set of 35 questions known as the Piers- Harris Self-Concept Scale (Piers 1984). It gathered information about how children perceive themselves across the six domains of:

  1. 1.

    Behavioural Adjustment (e.g. ‘I am well behaved in school.’ and ‘I do many bad things.’)

  2. 2.

    Intellectual and School Status (e.g. ‘I am smart.’ and ‘In school I am a dreamer.’)

  3. 3.

    Physical Appearance and Attributes (e.g. ‘I have nice hair.’ and ‘My classmates in school think that I have good ideas.’)

  4. 4.

    Freedom from Anxiety (e.g. ‘I get worried when we have tests in school.’ and ‘I am often afraid.’)

  5. 5.

    Popularity (e.g. ‘My classmates make fun of me.’ and ‘I am popular with boys/girls.’)

  6. 6.

    Happiness and Satisfaction (e.g. ‘I am a happy person.’ and ‘I am cheerful.’).

Significant differences were found in relation to how children viewed themselves. In terms of weighted raw differences, children from lone-parents families scored substantially lower on the total overall score (Figs. 14.6 and 14.7) indicating more negative self-concept when compared to children living with married parents. These differences are greatly reduced once selection bias is taken into consideration as children from lone-parent families scored lower on the total score in the main because of the differential selection into marriage compared to all other family states. Selection effects account for between 70 and 75% of the initial weighted differences in total Piers-Harris scores.

Fig. 14.6
figure 6

Estimated differences in Pier Harris Self-Concept between children living in never-married one-parent families compared to their married counterparts (Note: Confidence intervals calculated from robust standard errors for the matched sample)

Fig. 14.7
figure 7

Estimated differences in Pier Harris Self-Concept between children living in previously-married one-parent families compared to their married counterparts (Note: Confidence intervals calculated from robust standard errors for the matched sample)

Analysis of the subscales reveals some interesting differences from the child’s perspective. Children from never-married lone-parent families were significantly more likely to report negative attitudes across all subscales, except for physical appearance. Children in never-married one-parent families had a lower average score (0.39 point lower post matching) than their married counterparts on the “Freedom from Anxiety” subscale, indicating that they more frequently reported feelings of anxiety than children with married parents, despite taking account of selection effects (“FrAnx” in Fig. 14.6). However, the size of the effects is greatly reduced in the matched data.

Children who had previously lived with their married parents but now live in a one-parent family had a lower average score (0.20 point lower post matching) than their married counterparts on the Happiness/Satisfaction subscale, indicating they reported pre- and post-matching more negative attitudes about their happiness (“Happy” in Fig. 14.7). Yet in terms of all the subscales, the differential is close to zero post –matching (see Fig. 14.7). It appears therefore that the role of selection bias in understanding differences in children’s self-concept is powerful.


There are clear systematic differences in the wellbeing of children across family types in Ireland. Much of the adverse consequences of childbearing outside of marriage pertaining to the child’s health, self-concept and educational development appear however to be an artefact of pre-existing socioeconomic disadvantages. Compared to married mothers, lone-mothers are on average younger and they are more likely come from impoverished backgrounds (as indicated in their recollection of difficulties to make ends meet reported in Table 14.2), with lower levels of educational attainment on average when compared to marriage mothers. The matching results reported here suggest that when faced with similarly adverse conditions when growing up, children from one-parent families fare similarly in most regards to children from married families. However, some significant and modest differences do remain especially in terms of math scores and school attendance among children from never-married one-parent families.

The effect size varies across the range of child outcomes discussed here and the role of selection bias differs, explaining less of the differences in maths scores. The matching exercise had several limitations related to a lack of longitudinal data employing only one wave of data. The matching approach adopted here addresses selection effects driven by differences in observable characteristics between mothers but there remain important unobservable differences that are not accounted for especially in relation to pre-existing factors that were not recorded in the GUI study. The analysis therefore has only partially addressed the ‘selection’ issue by examining the size of the gap in outcomes between children growing up in lone-parent families and those in married families after controlling for observable factors. The role of unobservable factors is not clear. An individual’s decision to marry or not, and have children, will depend on many factors such as group norms, perhaps personality characteristics and availability of partners. None of these factors were taken into consideration here.

Yet, despite these limitations, the matching models have in the main substantially reduced the size of the differential in child outcomes scores. The evidence presented here strongly suggest that selection effects matter and thus point to concern with policy advise which would suggest that promoting marriage will reduce child disadvantage. The final section for this chapter will now discuss this policy context.

Families, Policy and Research

Our understanding of the social structure of inequality and the regeneration of inequality across families is growing giving the increased use of a life-course perspective as it relates to families. The availability of longitudinal data and methodological advancements has helped us to unravel the mechanisms by which families matter. Much work remains to be done especially in relation to translating these findings into a policy context.

The finding that selection matters is not unique to the Irish case (see Dohoon 2010 for a US analysis or Goodman and Graves 2010 for the UK). International research has tended to agree that children are at increased risk of adverse outcomes in one-parent families and that negative outcomes can persist into adulthood, but once account is taken for confounding factors the difference between children from one and two-parent families is a small one. While the antecedents, processes and sequelae of family breakdown vary across countries, its impact on children appears fairly consistent. In the Irish case, very few one-parent families are the result of divorce and few children are growing up in step-families, yet the findings point to an increased risk of educational under-performance compounded for children in one-parent families while the protective factors are reinforced for children in two-parent married families.

The evidence therefore indicates that much of the association between lone-parenthood and negative child outcomes is related to selection bias. The apparent benefits to marriage may not be all they seem. Rather, it is the interplay of a number of risk and protective factors at the family, school and neighbourhood level that serve to influence child development in children. As highlighted in this book, family structure is a process and not just a single event and given the lack of a longitudinal focus of this chapter, the dynamic nature of change in family relationships is missing. However, it appears that the benefits of marriage are overstated as they relate to child wellbeing. Parents who are married differ from those who never-married in very substantial ways, particularly relating to educational and socio-economic status of lone-mothers. The differential selection into marriage is therefore a key factor which policy-makers must consider. The results presented here strongly indicate that encouraging parents to formally marry will do little to improve child wellbeing rather policy should focus on mother’s education and socio-economic status.

In the Irish case, interventions to help children and parents of one-parent families have mostly involved income supports. According to the OECD (2006), Ireland had the third highest public spending on family benefits as a percent of 2011 GDP. Cash benefits to families were highest in Sweden and then Ireland. Some argue that the reliance on income supports has led to welfare dependence amongst the range of lone-mother families (lone fathers are more likely to be employed). Recent initiatives, in particularly a labour market activations policy, aim to have lone-mothers working when the child reaches the age of 7. In other words, the welfare reform is forcing lone-mothers (and not all mothers) into the labour force; despite the evidence showing that on average lone mothers have lower educational levels and poorer employment prospects as well as a lack of formal childcare for all mothers in the Irish context (Murphy 2014). The findings presented here point to the importance of back-to -educational supports for lone mothers, many of whom have their child at a relatively young age.

Another approach is to direct supports to the children through school services. The current roll-out of an Area Based Childcare programme is an initiative aimed at providing early childhood interventions at a multi-agency level. Such programmes acknowledge that a school intervention, such as a homework club, can provide some benefits but that interventions and approaches within disadvantaged areas should be integrated with mainstream services such as health, education and family supports. In other words, the public sector is seen to take on more of a role in child education, childcare and other labour that now falls into the remit of the lone-mother. In an era of increasing concern about income inequality and growing rates of child poverty, the implications of family research for policy formation needs increased visibility.