An important issue to take into account when interpreting these differences in school progress is the question of ‘causality’. When two characteristics are related – in this case family structure and educational performance – this does not automatically mean that one causesthe other. To illustrate the point, it is clear that children who live in a rented home perform worse at school, not because renting a home is a particularly impactful experience, but rather because of other characteristics typical of people who rent their home. For instance, people who rent a home normally have less money than people who own a home and a higher proportion of them live in disadvantaged neighbourhoods. It might therefore be that differences in school completion between these two groups are because of money and not because of their housing status.
The same argument applies to families where children live without their biological father or mother. One of the striking elements of the changes in family life observed over the last decades is that families of less-educated people have changed much more than the families of more-educated people. In 2011, almost 40% of children whose mother did not have educational qualications did not live with their biological father, compared to 17% of those who had a university-educated mother. It is a well-known fact that children of educated parents are more likely to do well at school (a pattern similar to that already illustrated in figure 3). Better-educated parents transmit skills and attitudes to their children and can help them navigate successfully through the schooling system. If many children living with single parents have less educated parents, this could be the reason why they are more likely to not finish school on time. Therefore, ascertaining whether the number of parents that children live with really matters for their school outcomes, or whether they do worse at school due to other characteristics, is not so straightforward.

A first strategy for dealing with this issue is to statistically account for background characteristics that might possibly explain differences in school progress. In other words, the estimates of school completion can be adjusted according to the different levels of education of children’s mothers. This exercise is displayed in figure 5.
The first set of bars reproduces the unadjusted differences in compulsory education completion already shown in figure 3. The second set of bars in figure 5 shows these differences, but this time accounting for a set of background characteristics such as maternal education, region, and whether children are foreign born. Here it can be seen that the differences between the groups become slightly smaller, indicating that background characteristics are part of the reason why children living with a single parent perform worse at school in comparison with children living with two biological parents.

Nonetheless, after accounting for these characteristics, a 4% gap in school completion remains between both groups of children. Of course, there are many other possible background characteristics that could theoretically account for this gap which are not available in the census data. It is therefore still possible that this 4% difference between groups of children is due to third factors. However, it is also likely that part of this difference indeed reflects something that is particular to children living with one biological parent.
It was noted earlier that children living with one biological parent often receive different parenting, are more likely to live in an economically poor household, and might have needed to adjust psychologically to their parents’ separation in the past. The third set of bars in figure 5 enable evaluation of the importance of one of these three explanations: differences in economic situation. This third set of bars statistically accounts for differences in the mother’s employment and a set of durables children have in their home (in addition to the background characteristics covered by the second set of bars): heating, internet, and the number of rooms their home has. It can be seen that the differences between groups are further reduced to 3% once this small set of economic factors are accounted for. Economic factors are therefore likely to be an important reason why children living with one biological parent instead of two do slightly worse at school.
Remaining differences could possibly be explained by better measurements of economic resources, such as family income, or measurements of parenting and psychological well-being. Such information, however, is not available in the census.