I recently finished the book “Freakonomics” by Steven D. Levitt and Stephen J. Dubner, and one of the final chapters in the book argues that what parents do has much less influence than they might think on how their kids will turn out. Rather, they support the idea that most outcomes that can be attributed to parental influence are essentially due to our genes [i.e., who we are (nature), not what we do (nurture)]. While they openly admit that people can always point out exceptions to this rule, they seem fairly convinced by their own argument. Among other interesting tidbits, they indicate that analyses of survey data show that reading lots of books to your children is not correlated with better performance on tests, whereas the mere fact of having more books in your home is (even if you never read them with your kids). Similarly, they report that there is no negative effect of allowing kids to watch more TV.
In reading this chapter, I couldn’t help but feel that something was not quite right. I employ statstics regularly in my own work, and I understand that they are simply analyzing available data from one large study and reporting the results. However, what kept popping into my mind is that applying statistical methods to a dataset will always give you some kind of answer, and sometimes that answer will be unexpected. The problem is that the results of a given analysis are only as good as the data that were analyzed. Now at the outset I will say that I have not gone into the details of how the ECLS study they discuss was conducted, but they report in the book that the data consist of measurements of academic performance, survey information provided by parents, and answers to questions posed to parents during interviews. Whereas test scores are concrete measurements, the latter two sources of information seem highly susceptible to error. Bear with me and I’ll try to explain.
Lets assume that the parents of the vast majority of the 20,000 or so children in this study love their kids and are trying to be the best parents that they can be. Conventional wisdom would have you believe that reading to your kids is good, and that letting them watch lots of TV is bad. If you were surveyed in this study either on paper, or even worse in an interview, my suspicion is that most parents would tend to exaggerate parenting qualities that are thought to be good (i.e., reading), while downplaying qualities that are though to be bad (i.e., letting kids watch TV). My guess is that this effect is more extreme in parents who are doing less of a supposedly good thing (reading), and more of a supposedly bad thing (TV watching). It is human nature for a parent to do this – we want our kids to succeed and we may even trick ourselves into believing we spend more time on positive parenting behaviors than we really do. So what does this mean for the conclusions in Freakonomics?
If parents overestimate how much they read to their kids, and parents who read less bump their estimate up even more than those who read alot (perhaps due to a sense of guilt or fear of judgment), then a correlation between reading and academic performance may not be all that accurate. I don’t know if the ECLS study had controls for this potential source of error, but I would be interested to find out. At the very least, this possibility was not addressed in Freakonomics, which is the source from which the vast majority of people will read about this supposed lack of a relationship. In a similar manner, if parents tend to downplay how much they let their kids watch TV, a similar problem would arise. The validity of these supposed correlations seems to be almost entirely dependent on how honest these parents are about their behavior. Now think about what you would do if you were sitting across from some interviewer and asked how much you read to your child. Would you be more likely to overestimate how many books you have in your house, or how often you actually read them together? Which carries less likelihood of judgment by the interviewer or someone reading your survey responses? Which is more likely to be subject to error? You can apply this same logic to questions regarding how much TV you let your kids watch. If you’re like me, the tendency to enhance the good and downplay the bad is hard to avoid when you’re simply trying to be the best parent you can be for your kids.
I personally have mixed feelings about the supposed detrimental effects of TV. When my 5-year-old son tells me out of the blue that the melting of ice into water is an example of reversible change (which he learned from Sid the Science Kid), it is clear that he has absorbed a lot from some of the TV shows he watches. However, I’m not buying the data showing that reading to your kids provides no future academic benefit. Maybe I’m just a sucker for the conventional wisdom, but my five years of hands-on experience as a parent leads me to believe otherwise. Ill be needing proof that the data analyzed in Freakonomics are good and valid before I put Dr. Seuss back on the shelf.