Bryan Caplan  

The Hours and Academic Achievement

Surviving on the Margin... Low hanging fruit and the ineq...
Adults love controlling the way kids spend the hours of the day.  What's the payoff for all their meddling?  Hofferth and Sandberg's "How American Children Spend Their Time" (Journal of Marriage and the Family) provides some fascinating answers for kids ages 0-12. 

After compiling the basic facts about kids' time use from the 1997 Child Development Supplement to the PSID, H&S regress measures of academic achievement on time use, controlling for child's age, gender, race, ethnicity, head of household's education and age, plus family structure, family employment, family income, and family size.  All test scores have a mean of 100 and a standard deviation of 15, and all time use is expressed in hours per week. 



The big result is the lack of results.  Controlling for family and child background, time in school and studying barely help - and television viewing barely hurts.  Contrary to wishful assertions that exercising the body improves the mind, sports don't matter either.  Out of nineteen activities, only two predict greater academic success across the board: reading and visiting. 

The estimated effect of visiting is modest.  Reading, however, is a huge deal.  Ceteris paribus, 10 extra hours of reading per week raise letter-word comprehension by .5 SDs, and passage comprehension, applied problems, and calculations scores by .4 SDs.  Despite obvious worries about reverse causation - smart kids enjoy books more - much of this is plausibly causal.  After all, many smart kids don't read much, and H&S include a lot of solid control variables.  And you really can learn a lot from books.

I've long argued that the effects of parenting are overrated.  H&S's results lead to a separate but related result: How kids spend their time is overrated, too.  If adults really wanted to raise kids' test scores, they'd adopt the maxim, "If the kid has a book in his hands, leave him in peace."  Which, by sheer coincidence, was the maxim young Bryan Caplan vainly begged all the adults in his life to embrace.  Reading rules, school drools.

Comments and Sharing

COMMENTS (20 to date)
Fabio Rojas writes:

What is "visiting?"

Kevin Dick writes:

+ 1 to Fabio Rojas' question.

Also, does playing a MMORPG with your IRL friends while on Skype carrying on a conversation the whole time count as "visiting"?

blink writes:

Cool data set! I think you are right about a causal story for reading. While I understand the authors used the best available controls, my guess is that "eating", "church", and "visiting" are proxies for a good home environment.

I'm most surprised (pleasantly!) by "market work". Admittedly the effect is not statistically significant, but still suggestive and certainly evidence that paid work does not detract from learning.

Rick G writes:

The R^2 at the bottom: is that the explained proportion of total variance, or the explained proportion of variance left over after controlling for all the things you mentioned in the second paragraph? In either case, it's not a lot, but it's *really* not a lot if its the explained proportion of left over variance -- that would suggest there are a bunch of things neither controlled for nor measured (or that achievement test results are very noisy, which I don't believe).

Will Ambrosini writes:

The intercepts are telling us that the explanatory variables aren't centered or normalized. So keep in mind that the effect of school for the average kid (30 hours a week in school?) is likely close to the effect of reading for the average kid (a couple hours a week?). This, and the particular functional form being used to estimate these effects, also suggests something else to keep in mind: these first order effects are estimated at the sample averages. Because of the diminishing returns to doing anything, the first hour or second hour of reading may be more impactful than the 31st hour of schooling. A kid who reads a lot, like our young Bryan Caplan, might not be better off reading an additional hour.

Will Ambrosini writes:

Maybe more to the point: if the level of an activity is optimal with the respect to some outcome, you'd expect marginal increases to have zero effect. We've learned that the average kid should read more not that she should play sports less and that parents have somehow calibrated their kid's time use just right for most other activities.

Massimo writes:

Caplan is focused on the scenario of the intellectual gifted & talented book worm kid, with parents debating sending the kid to fancy activities. Raising a kid like that sounds like heaven. I am used to parents with the opposite type of child: kids who get caught shoplifting, fighting, using drugs, having unprotected sex, would never read books or practice a musical instrument, and are at the bottom of standardized testing, and constantly pressure parents for money and expensive food, clothing, and entertainment.

Paul writes:

This at least seems to justify somewhat my mother's overriding concern that I learn how to read well before entering kindergarten.

Brad writes:

I am not sure I would be so quick to waive off the reverse causality of reading.

They controlled for lots of family factors some of which would be correlated with IQ (parents income and education), but there probably is lot of variation left for reading for pleasure to be selecting for high IQ and intellectual curiosity.

Mark writes:

One problem that I have with the paper is that they do a bunch of significance tests and don't indicate whether they adjusted for multiple comparisons (or dealt with multiple comparisons specifically in a statistical model). My guess is the null hypothesis still stands and without a more sophisticated analysis (or higher criticism, or something of the sort) we can't tell if the reading effect is just noise.

For an illustration of why we need to do some adjustment before declaring any effects see this comic or this paper.

Grant Gould writes:

Clearly I should be doing a lot more "Intercept" with my daughter. Sounds like some sort of sports thing :-).

lemmy caution writes:

Smart people read books. Smart people are good with calculations. Not clear that reading books helps improve your calculations though.

Hana writes:

Since I didn't practice, did my weekly music lessons count as visiting?

gwern writes:

Mark: if this paper is having problems with multiple-comparison, that reinforces Caplan's point. (His major point is that almost all of the measured activities have ~0 correlation with test scores; if the remaining few non-zero estimates are themselves the products of multiple-testing, then his point becomes even stronger.)

Sieben writes:


No, he's not. The data set adjusts for all those things.

You might imagine that the kids who steal and screw around would do poorly regardless of the parental style.

The real problem with young people is that they have low self esteem. Society tells you you're worthless if you don't go to college, and going to college means playing a self-abasing role to a bunch of glorified babysitters. So you're worthless either way.

Society doesn't want anyone to truly achieve greatness in any sense of the word. They just want everyone to be middle class. We don't have any positive cultural values that adults can embody. Hard work and persistence are demonized by the meme of the greedy white businessman. Everyone is equal. Liberal arts majors aren't better or worse than anyone else.

I mean, I worked hard enough to send my kids to Montessori style schools, but you are an equally good parent even though you didn't work as hard and your kid has to go get his ego brutalized in some public school.

MG writes:

I have a conceptual question. If several activities are likely to be performed by the same "type" of child, would not this decrease the "impact" each activity would have on the measured outcome? (The input is activity hours.) And what if one aggregated activities into categories (subjective, but say, tiger mom type activities, detached single parent type activities, etc) and saw a larger impact at the category level. What would this mean?

Levi Russell writes:

Visiting is, according to the paper:

Visiting is included as a
structured activity because it includes participation
in (non-church–sponsored) youth clubs and
organizations. Such activities are expected to promote
children’s achievement and behavior (Task
Force on Youth Development and Community
Programs, 1992).

This seems to be things like 4-H and other non-sports municipal and county activities.

The fact that "visiting" and reading are the most important factors, it seems that homeschooling (since we're talking about young kids) is a pretty good idea. Homeschooling often takes the form of participation in things like 4-H and other structured social activities and a boat load of reading (including a focus on fostering a love of reading).

Thanks for sharing this study!

Floccina writes:

If we cannot get them to much learn more, we should concentrate on optimizing what they learn? Prioritize the most useful stuff to know. So how important are those test scores and reading?

branis panjoomby writes:

nice research: good readers like to read!
(good readers brains are somewhat pre-set to find reading skills a bit easier to acquire)

NB: "letter-word comprehension" variable is actually "letter-word identification"-

at the low end it's just identifying letters, then it's untimed isolated word reading - words ordered from easy to difficult, with fairly even item difficulty level gradients between items.

also, given the date, the test is probably the Woodcock-Johnson-Revised (1989)

Timothy Bates writes:

You might be interested in two papers where we hypothesised that aquiring the ability to read (as opposed to reading books) would raise IQ.

In a large cohort, controlling childhood SES, reading (and math) were associated with big gains in adult social status, through pretty complex links, including later IQ.

Collaborating with the TEDS cohort of several thousand identical twins studied in multiple waves from age 3 to age 16, we tested if one of the twins got better on a test of reading, would they score higher on IQ in later years? The Design includes the initial (moderately strong) association of these two traits, and nests all analyses within families to avoid those confounds).

The answer was yes: If one of two genetically identical twins acquires better reading skills, these are realised not only as enduring gains in reading, but in changed (increased) IQ in later waves.

These were not explained by reading of books: Rather learning to read seems to enable the mind to work better at a categorical level.

So, people differ greatly in IQ for genetic reasons, and literacy makes people smarter than they would have been.

Links to the papers are here:

The MZ differences method in general is greatly underused for disentangling environment and genetics. Usually the result is that associations are largely genetic, but not always.

Comments for this entry have been closed
Return to top