You Make Pearson Cry #3: Daycare Correlates March 28, 2007

A recent report on the NICHD study on early childcare argues that time spent in daycare is associated with disruptive behaviour:

keeping a preschooler in a day care center for a year or more increased the likelihood that the child would become disruptive in class — and that the effect persisted through the sixth grade.

The effect was slight, and well within the normal range for healthy children, the researchers found. And as expected, parents’ guidance and their genes had by far the strongest influence on how children behaved.

But the finding held up regardless of the child’s sex or family income, and regardless of the quality of the day care center. With more than two million American preschoolers attending day care, the increased disruptiveness very likely contributes to the load on teachers who must manage large classrooms, the authors argue.

On the positive side, they also found that time spent in high-quality day care centers was correlated with higher vocabulary scores through elementary school.

Controlling for genetics, parenting, family income, and gender is probably a good start, but even with such controls it’s not exactly difficult to think of other interpretations of the data:

1. Disruptive children get put in daycare more because their parents need time off.
2. Children in daycare have parents who generally have less time for them, in and out of daycare.
3. Parents who put their children in daycare differ from stay-at-home parents on some other third variable

I’m also a bit curious about how they controlled for parenting, since I imagine a pretty crucial factor in parenting styles is going to be whether you believe in daycare or not. But before I spend too much time trying to make sense of the data, what does it look like, exactly?

In 2001, the authors reported that children who spent most of their day in care not provided by a parent were more likely to be disruptive in kindergarten. But this effect soon vanished for all but those children who spent a significant amount of time in day care centers.

Every year spent in such centers for at least 10 hours per week was associated with a 1 percent higher score on a standardized assessment of problem behaviors completed by teachers, said Dr. Margaret Burchinal, a co-author of the study and a psychologist at the University of North Carolina.

The statistical trick of creating extreme groups and then testing for significance tells me that they failed to find a significant correlation between time in daycare and scores on the test. By taking the extreme cases you can squeeze out a significant result, but unfortunately, you’re no longer just comparing daycare to no daycare. You’re comparing a lot of daycare to no daycare at all, and this may make your groups less comparable on other variables since you’re looking at extreme groups only, without considering the bulk of the sample that is somewhere in-between.

The effect size is a bit of a joke. a 1 percent different in scores on a test? And this is a cause for concern? Even while admitting that there were positive correlations with vocabulary scores? Making the unlikely assumption that daycare is the cause of both these effects, perhaps scoring 1 percent higher in this disruptive behaviour measure is actually a worthwhile trade-off for improved vocabulary scores?

While a large sample is always better, I think this study raises the question of how small a significant effect can be to be considered meaningful. This $200 million project (no exaggeration) could obviously afford a huge sample, to be able to get significance on a 1 percent difference. But what does this mean? Should parents take this into consideration? It’s quite possible that there is a 1 percent difference in scores on this measure for disruptive behaviour linked to a whole pile of other, less politically loaded factors that we know nothing about, because no one is looking for them.

Ultimately, a major part of why so much of basic sociology is open to interpretation is the reliance on correlational measures. Sociology, at its best, should be able to inform policy – to tell us what the best option is, for parents and for policy-makers. In this case, with a $200 million budget, I could easily see how experimental measures could be employed instead: get a sample of lower-income families and offer them free day-care in exchange for participation in the study. Out of the people who volunteer, assign half to day-care, and half as a control group. The study wouldn’t generalise beyond that part of the socioeconomic spectrum, but at least you would be able to show causation.