Article : The effects of overcrowded housing on children's performance at school

Eric Maurin and Dominique Goux have studied the possibility that sharing the same room with a sibling might affect a child’s performance at school (in [2001] and [2004]). They used a large body of statistical data available in the French Labor Survey. Their goal was to determine if educational attainment was correlated with overcrowded housing, independantly of the other factors which can influence both simultaneously.

Their model rests on the three following assumptions:

- A child can be held back one year in primary or secondary school depending on her birth of month (the age relative to her peers has been shown to have an influence) and academic ability.

- The academic ability can depend on the child’s gender, the total number of children in the family, the amount of space available for each family member, and other unknown factors.

- Finally, the amount of space itself depends on the parent’s socioeconomic status, and the way they chose to share their budget between housing and other consumption.

Their very clever statistical modeling shows a clear correlation between housing conditions during childhood and performance at school, which cannot be totally explained in terms of differences of parental socioeconomic status and family size. They conclude that it is worth studying how housing policy could help improve the poorest children’s school performance.

While reading their papers, I was struck by the fact that intelligence was never mentioned. Since it is obviously an important factor in a child’s performance at school, it called for additional questions which are listed below:

- What role could intelligence play in the story? (Let us call this the intelligence hypothesis)

- Can this idea be rejected on the basis of Goux and Maurin’s data?

- Could it on the contrary provide a better explanation of their data?

- In case intelligence looks like the right way to go, what additional study would we need to eliminate any remaining skepticism?

Here is a brief discussion of these four questions.

1) What role could intelligence play in the story?

Since the model leaves room for some “unknown factors”, intelligence could be one of those. Another way to put it is that it is not enough to know a child’s gender, month of birth and environment to predict her educational attainment: intelligence counts. For the model to be invalidated, intelligence would need to be correlated with one of the other factors such as family size, housing conditions, or gender composition. How could this happen? Although it is a rather controversial issue (note the use of “could” below), intelligence is thought to have the following correlations:

- Intelligence could affect family size: more intelligent people having less children, independantly of their socioeconomic background, and the people with higher socioeconomic status having more children when compared to people of similar intelligence but lower socioeconomic status.

- Intelligence could affect professional occupation, income, and therefore housing conditions.

- Intelligence could affect cultural attitudes towards feminism and gender preference, thereby influencing the sibship’s gender composition.

- Intelligence could affect a woman’s probability of getting a divorce, and being alone to raise her children.

Whose intelligence are we talking about, the child’s or the parent’s? This does not necessarily matter, if it turns out that intelligence has a certain heritability. Without claiming that “intelligence runs in families”, it would be enough for intelligence to be somewhat heritable to account for the above. Therefore, an indicator of the parents’ intelligence is all that would be needed in addition to the data provided in the French Labor Survey.

2) Can the intelligence hypothesis be rejected on the basis of Goux and Maurin’s data?

It is impossible to give a firm negative answer without going through all the computations again, which has not been done. As a first step, however, it is possible to search the articles for any bit of information which would clearly contradict the intelligence hypothesis. Here are a few arguments which come close:

- In Chapter 3 of [2001], the authors discuss the assumption of independence for the unknown factors. They check that their results would be unaffected by any factor in the form of spending, different from housing expenses, which could affect the child’s outcome. This is not sufficient, however, to reject the intelligence hypothesis since it is an exogenous characteristic of the child. In order to factor intelligence into the model, it should be added as an independant variable to the list of a child's characteristic along with age and gender, or to his parent's characteristics along with socioeconomics status.

- In Chapter 4.4 of [2001], the authors show that a statistical test (Sargan) shows no correlation between the estimated residuals of their model and the instrument variables. They consider that the amount of time a mother spends at home with her children could have influenced schooling, but there is no trace of such an effect here. Again, part of the effects of intelligence are factored into the model through various correlations. All the test tells us is that the remaining variations in intelligence are independent from the other instrument variables, which is consistent with the intelligence hypothesis.

- In [2004], the authors introduce the notion of the two last children of a family being a boy and a girl. In Chapter 5, it is shown that only when the children come in this order does it have a positive influence on their outcome. The proposed reason is that parents are reluctant to place their youngest daughter in her brother’s bedroom. Admittedly, it could be tough to explain this phenomenon under the intelligence hypothesis. A tentative approach could be the following: some families in this configuration are those whose parents absolutely want a girl, and keep having children until they finally have a girl. This could be correlated with cultural preferences linked with intelligence as indicated above.

3) Does the intelligence hypothesis provide a better explanation of their data?

Generally speaking, the intelligence hypothesis should produce results very similar to what we see in both articles. Academic ability would be correlated with intelligence, and family size and overcrowding would be negatively correlated with intelligence because more intelligent people would tend to have at the same time more children, lower socioeconomic status, and therefore lower housing conditions. In order to decide whether the intelligence hypothesis is valid, it is therefore not sufficient to look at the data at hand as it is largely explainable by the overcrowding hypothesis. However, the intelligence hypothesis would have an edge when it comes to understanding why large families are correlated with lower socioeconomic status (Table 1 in [2004]).

4) What additional study would we need to eliminate any remaining skepticism?

Intelligence tests are not very well considered in social sciences, and it is unlikely that any such data has been collected in the French Labor Study. It would be possible to make such study in the United States using the data from the National Longitudinal Study of Youth. In France, the available data would be sufficient to test if socioeconomic status and educational attainment are correlated with the sibship’s gender composition. A very simple test would be to check whether there is a difference between families with one or many boys followed by a girl and families with one or many girls followed by a boy.

Conclusion

What is the point in bringing up the intelligence issue? Well, if there were some truth to the intelligence hypothesis, it would mean that Maurin and Goux’s results partly reflect the fact that less intelligent children do not do as well at school, while simultaneously having more siblings and poorer housing conditions. If this interpretation were right, then public policies aimed at improving poor children’s schooling by giving every child an individual bedroom could turn out to be simply very expensive and inefficient.