Contrary to rather a lot of current belief fathers actually matter. It’s this which explains – largely enough at least – why the children of teenage mothers do worse in life. Those fathers are of lower quality:
Why children of teen mothers do worse in life
Anna Aizer, Paul Devereux, Kjell G. Salvanes 22 December 2018
It is well documented that children who are born to teenage mothers have worse outcomes in terms of health, education, and earnings. This column compares the outcomes for children of sisters in Norway to investigate the drivers of this. The results suggest that the ‘quality’ of the fathers explains as much of the difference in child outcomes as economic resources. Policies that consider the role that fathers play in teenage childbearing and its consequences may be more effective than those that consider mothers only.
It is well documented that children who are born to teenage mothers have worse outcomes, including worse health, less schooling, and lower earnings in adulthood. However, negative maternal selection is likely to be an important factor explaining this result. Teen mothers are more likely to come from disadvantaged backgrounds and their offspring might be expected to accumulate less schooling and earn less in adulthood, regardless of teen childbearing. For example, using data from the 2003 Panel Study of Income Dynamics (PSID), Kearney and Levine (2015) report that 20% of US women give birth before the age of 20, but among those born into poverty, 49% give birth before the age of 20. The same pattern is observed in Norway, where teen childbearing is twice as likely in low socioeconomic status (SES) families than in high SES families. Because of maternal selection, less is known about whether there is a negative impact of teen parenting per se on the short- and long-run outcomes of children. Moreover, the literature so far has been silent about the role of the father of the child. We address these issues in our research.
Previous work has shown that negative selection into teen motherhood can explain most, if not all, of the observed negative effects on maternal outcomes. However, teen motherhood may still exert a negative and causal effect on child outcomes for several reasons. First, teenagers have lower levels of psychosocial maturity than their older and more developed counterparts – they are more likely to be depressed and more likely to report greater levels of stress (Hodgkinson et al. 2014), both of which have been linked to deficits in parenting behaviors (Reid and Meadows-Oliver 2007). Second, previous work on the impact of teen parenting on maternal education and income finds small long-term effects once selection has been considered, but with the negative effects on income declining over time (Hotz et al. 2005). As such, teen parents may be more resource constrained during their children’s earliest years, a period critical for child development (Almond and Currie 2011). Finally, teen birth may also be associated with lower paternal quality. If so, this can also result in worse outcomes for affected children through either genetic inheritance or fewer and lower quality paternal inputs.
Comparing children of sisters
In general, it is challenging to distinguish the causal effect on children of being born to teen mother from selection effects. We examine the impact of teen motherhood on child short-, medium- and long-term outcomes using sister fixed effects to control for negative selection into teen motherhood. Specifically, using Norwegian administrative data that links individuals across three generations, we compare the outcomes of children born to a teen mother with the outcomes of children born to her sisters (that is, we compare outcomes of cousins). In this way, we allow for negative selection into teen motherhood by controlling for all family background characteristics (observed and unobserved) of teen mothers that are common across sisters. Moreover, we assess the likelihood that any remaining estimated effects are driven by unobserved characteristics biasing our estimates, by calculating how much selection on unobservables must remain for the true effect of teen motherhood on child outcomes to be zero. More specifically, we conduct an exercise in which we assume that the remaining omitted variable bias is proportional to coefficient movements scaled by the change in R-squared when controls are included (Altonji et al. 2005, Oster 2016).
Effect on children in the short and long run
In line with the previous literature on teen motherhood and maternal outcomes, we find that much of the negative relationship between teen motherhood and child outcomes can be explained by teen mothers’ underlying levels of disadvantage. For instance, in our data, teen mothers are 15 percentage points more likely to come from a home in which the father has not completed high school (an important determinant of household resources), and they are also 25 percentage points less likely to have started academic high school by age 16, a measure of past academic achievement and future aspirations.
While the negative relationship between teen childbearing and child outcomes is very strong, when we include sister fixed effects that control for observed and unobserved differences in the teen mother’s family background, the estimates decline considerably but still suggest a negative relationship between teen childbearing and child outcomes. With all controls included, we find that children born to teen mothers have cognitive test scores that are 13 percent of a standard deviation lower, complete half a year less of schooling, have four percent lower earnings at age 30, and are three percentage points more likely to have a teen birth themselves. These estimated effects are generated from examining all teen births (ages 15-19). We find that the estimated negative effects are much larger for the youngest teenagers, those aged 15-17 at the time of birth.
To assess the likelihood that these remaining estimated effects are driven by unobserved characteristics biasing our estimates, we calculate how much selection on unobservables must remain for the true effect to be zero. For reasonable assumptions about the relative importance of included and excluded variables, the estimates still indicate a negative effect of teen childbearing on child outcomes.
The role of fathers and family resources
What explains the remaining negative relationship between teen motherhood and offspring outcomes? We consider three factors: teen mother behaviour (i.e. smoking while pregnant), household resources in early childhood, and paternal quality. All three appear to play a role in explaining the worse child outcomes still observed after controls for maternal selection are included. Teen mothers are more likely to smoke during pregnancy even after including sister fixed effects. Teen mothers also have lower levels of family earnings especially when the child is young, but over time, the size of the effect declines (consistent with existing work). Finally, paternal quality is much lower for children born to teen mothers even once we control for observable and unobservable characteristics of the teen mother. The ‘partners’ of teen mothers score significantly lower on a cognitive test at age 18, are shorter (consistent with diminished nutrition in childhood), and are less likely to have started academic high school at age 16. Once we control for the underlying characteristics of both teen mothers and their partners, the offspring of teen mothers generally fare little worse than offspring born to older parents with similar background characteristics.
While previous work has emphasised the importance of considering the underlying background characteristics of teen mothers, we conclude that when considering the outcomes of children born to teen mothers, the underlying characteristics of the fathers also play an important role in explaining their worse outcomes in both the short and long run. More precisely, our decomposition analysis indicates that the quality of the fathers explains as much of the difference in child outcomes as economic resources. This is the first work to emphasise this linkage. Our results suggest that policies that consider the role that fathers play in teenage childbearing and its consequences may be more effective than those that consider mothers only.
We have built on the previous literature by using three generations of population data to study the effects of teenage childbearing on the outcomes of children. A major advantage of our analysis is that our administrative data likely suffer from little measurement error as they contain information from birth registers rather than self-reports of pregnancy. We are also able to link to a wider range of medium- and long-run child outcomes than have been studied in the past. Finally, our data also include information on paternal characteristics that is mostly absent from previous analyses based on survey data.
Our estimates suggest that cross-sectional analysis significantly over-estimates the adverse consequences of teen childbearing on the next generation. Our preferred estimates using maternal sister fixed effects suggest much smaller negative long-run consequences for children. Our rich administrative data allow us to study mechanisms behind the remaining effects that have not been fully explored previously. We find that paternal selection plays an important role – a major reason that children of teen mothers do worse is that their fathers are more likely to have lower education levels and cognitive scores. We also find evidence that lower family resources during childhood may play an important role. The adverse consequences of teenage childbearing are larger for mothers from higher socioeconomic groups. Consistent with this, we find that adverse paternal selection is greater for higher SES moms. This underscores the importance of the role of fathers in mediating the effects of teen childbearing on child outcomes. Policies that target young first-time mothers, such as the Nurse Family Partnership programme, should also consider providing services to the fathers if their objective is to improve long-term offspring outcomes.