When Parents Get Paid for Homework
Poor kids don't do nearly as well in school as the children of the affluent. There's a vicious cycle when you talk about this, where education reformers blame the teachers, teachers blame the parents and the economic conditions of the children, and everyone sort of gives each other the side eye while glumly agreeing that something really needs to be done.
Adoption studies seem to indicate that parenting does matter. Unfortunately, it's not clear what that actually tells policy makers. Reforming schools is harder than it sounds, but persuading principals and teachers to change what they do looks like a trivial exercise compared with getting millions of people to radically alter the hours they spend each day with their children in the privacy of their own homes. For one thing, we're paying the teachers and can threaten to cut off the checks if they don't change.
A team of superstar economists -- Roland Fryer, Steven Levitt and John List -- decided to see what effect it could have by paying the parents. They designed an expensive yearlong intervention, setting up a parent academy that distributed nearly $1 million to 257 families, including a control group that received minor sums for participating in periodic assessments. The academy was situated in a low-income neighborhood where the overwhelming majority of the schoolchildren qualify for free or reduced-price lunches. The researchers developed a curriculum focused on helping students develop both cognitive and noncognitive skills and then taught it to the parents, who were rewarded for "attendance at early childhood sessions, completing homework assignments with their children, and for their child's demonstration of mastery on interim assessments."
Here are the results from the abstract:
This intervention had large and statistically significant positive impacts on both cognitive and non-cognitive test scores of Hispanics and Whites, but no impact on Blacks. These differential outcomes across races are not attributable to differences in observable characteristics (e.g. family size, mother's age, mother's education) or to the intensity of engagement with the program. Children with above median (pre-treatment) non-cognitive scores accrue the most benefits from treatment.
The paper is fascinating throughout. Before I go on, I will issue, as always, my standard disclaimer: This is one study of one program. Anyone who bases a global opinion about policy based on a single study richly deserves the ridicule and heartbreak they will get.
That said, I will now discuss what this study tells us if the results hold true.
The effects are real, but not miraculous. The first thing to note is that across the whole group, the improvements in both cognitive skills and noncognitive skills are positive and nontrivial, but not huge, either: about a tenth of a standard deviation on cognitive skills, about two-tenths that for noncognitive. I know that some of my readers may not be familiar with the manifold splendors of the normal distribution and the standard deviation, so the easiest way to explain is to offer a few examples. The average IQ is 100, and with a standard deviation of 15, that means 68 percent of the population will have IQs between 85 and 115. The mean height of an American male is 70 inches, or 5-foot-10, with a standard deviation of 4 inches, so 68 percent of American men are somewhere between 5-foot-6 and 6-foot-2.
That gives you a sense of the level of improvement they recorded. As they note, the cognitive improvements are not statistically distinguishable from zero. The noncognitive difference is significant, but it was pretty expensive to obtain. They spent nearly $4,000 per family, but 100 of them were in the control group, which received $100 to participate in the assessments. So the cost per family in the treatment groups was more like $6,300. That's a lot of money to spend for fairly modest gains in noncognitive skills. Now, you can make a decent argument that this money would be well-spent, that the savings in reduced crime, higher worker productivity and improvement in all the other problems that afflict low-income kids when they turn into low-income adults would more than repay such an investment. But politically, this will be a heavy lift.
Starting cognitive skills don't matter; starting noncognitive skills do. In fact, the kids who started with below-average cognitive scores benefited more than those on the other side of the distribution. But kids with higher-than-average noncognitive skills experienced large gains, while the kids who started out below the curve gained nothing from the intervention.
This is really important. Our best measures of educational performance are cognitive because that's what's easiest to test. If the things that are harder to test matter more, that presents something of a conundrum for people trying to formulate educational policy.
The outcome was pretty much the same whether parents received the money in cash or in a college fund. I find this a little bit surprising. All else being equal, people generally prefer cash to in-kind benefits. If you think that incentives matter -- and that is, after all, the finding of the study -- then a better incentive should produce a better performance. If the parents in the study are indifferent about money that can be used for college or the electric bill versus money that can be accessed only when their children enroll in college, that tells you something very interesting about them. However, there's a caveat: The study experienced substantial attrition by the time of the end-of-year assessments, and parents in the college group had higher attrition rates than those who received cash, so the ones who stayed may have been especially motivated about college.
The researchers can't explain the racial differences in success. Whites and Hispanics experienced large positive effects, while the effects on black children were slightly negative but not statistically significant. There's a long section where the researchers look for possible explanations, but they don't really come up with any. I'm afraid I can't, either. Those households may be systematically different from the white and Hispanic households in the neighborhood in some way that's hard for the researchers to observe, or it may be a fluke. As you break down your sample into subgroups, you inevitably run a larger risk that statistical noise will swamp your result.
Overall, it's an extremely important experiment. If the results hold, it suggests that you can do something to help people prepare their children for school. On the other hand, it also shows that such a program will be somewhat limited and relatively expensive. Incentives do matter -- but they can't work miracles.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author on this story:
Megan McArdle at firstname.lastname@example.org
To contact the editor on this story:
Daniel Niemi at email@example.com