Policy Nov 2, 2016
Why Sending Your Kid to the Best Possible School May Backfire
Being surrounded by smarter peers can hurt test scores and incite disruptive behavior.
Parents may assume that moving smart kids from a subpar school to a top school is an unequivocal win. After all, these students will probably have better teachers, get access to more resources, and have the opportunity to learn from their able classmates.
But the kids also have to deal with a shift in status. At their old school, they were at the top of their class; now, they might rank in the middle or bottom, even though their intelligence level has not changed.
According to new research from the Kellogg School, this downgrading in academic rank can significantly increase the risk of academic and behavioral problems—indeed, exactly the types of issues parents were likely trying to avoid at the old school. Students may end up performing worse on standardized tests and misbehaving more in school than they would otherwise. All else being equal, kids in a class with smarter peers are more likely to have these problems than if they are surrounded by less skilled students.
Students who have dropped in academic rank might struggle because it is not as fun to work on a task when you are lagging behind your peers. “It may be inherently more enjoyable to do something at which you’re good,” he says. “And ‘good’ is defined relative to the people you observe.”
“It seems to be good to be a big fish in a small pond.”
The researchers are not saying that high-achieving students should not move to better schools. With the positive changes that come from such a move, such as better teachers or newer facilities, “it might still be a good idea at the end of the day,” Spenkuch says. But the research suggests that parents and teachers should keep in mind that the transition also will bring some challenges.
Academic Rank Influences Test Scores
We all know that we are influenced by our peers.
For example, “if all the people are wearing red sneakers, I want to wear red sneakers too,” Spenkuch says. But we don’t interact with all of our peers; we usually choose a smaller group to befriend. At school, a kid could decide to either hang out with studious, hard-working classmates or with students who struggle academically and often break the rules.
How do students make these choices? Spenkuch and his collaborators, Steve Cicala at the University of Chicago and Roland Fryer, Jr. at Harvard University, devised a mathematical model to represent how students with varying levels of intelligence and physical strength select cliques.
The model included two groups: “nerds,” who focus on their studies, and “troublemakers,” who misbehave. Kids in their model do not know their academic ranking, but they know how much they would enjoy being part of each group. The model suggested that a kid performing relatively well would associate with nerds. But if that student moves to a class where she ranked near the bottom, she would begin associating with the troublemakers.
Spenkuch and his colleagues wanted to test this prediction in the real world. So they analyzed data from an earlier study by other researchers in Kenya. In 2005, students from 60 primary schools were randomly assigned to classes. With the shuffling, a high-ranking student might find herself lower on the totem pole in her new class, or vice versa. Spenkuch’s team could then compare the performance of kids with similar skill levels who ended up with different rankings.
An increase in class rank was linked to higher scores on standardized tests 18 months later, the team found. For example, consider kids who were initially in the 25th percentile—that is, they performed as well as or better than one-quarter of their peers. If some of those students moved to classes where they were in the 75th percentile, while others stayed at their original rank, the first group’s test scores were an average of 0.2 standard deviations higher than the second group’s scores. (For comparison, the test gap in reading between white and black children entering kindergarten in the United States is about 0.4 standard deviations.)
“It seems to be good to be a big fish in a small pond,” Spenkuch says.
Misbehaving in Class
Spenkuch’s team also wondered if changes in academic rank affected the kids’ tendency to misbehave.
To test that, they turned to data from New York City public schools, which tracked students as they switched from elementary to middle school in 2003–04 and 2008–09. The administrators had recorded disruptive incidents such as horsing around in class, being rude to teachers, and fighting with other kids. About nine percent of the students had at least one such record during their first year of middle school.
As the team had predicted, dropping in class rank appeared to make students act out more. If one kid fell from the 75th to 25th percentile upon entering middle school and another rose from the 25th to 75th percentile, the first student was 3 to 5 percentage points more likely to have at least one record of misbehavior than the second student.
Spenkuch and his colleagues then turned to a different dataset to investigate whether the same kid would behave differently in classes depending on his ranking in that particular subject—say, English versus math.
The team analyzed data on more than 24,000 students from the National Educational Longitudinal Study, collected during eighth and tenth grades. The data set included reports of problems such as absences, tardiness, not finishing homework, not paying attention in class, and disrupting class. About 44% of the students had at least one record.
Again, the team found that ranking mattered. If a kid was in the 75th percentile in English but in the 25th percentile in math, the teachers were 10 percentage points more likely to have reported a problem for that student in his math class than in his English class.
“Sliming” Other Students
So why was this happening? Perhaps teachers were partly responsible: if they geared their instruction toward the top students, lower-ranking kids would be less engaged. But maybe a decrease in status also prompted students to neglect their studies or act out, regardless of what the teacher did.
To find out, Spenkuch’s team ran an experiment with 573 middle-school students in Houston. In groups of about 20 kids each, the students played a computer game to solve mazes and were paid 25 cents for each completed maze. The software then displayed how all the students ranked in their group, based on how quickly they had finished.
Because students were grouped with different sets of peers, it was possible for the researchers to see how students who had similar scores, but different peer rankings, acted next.
In the last stage of the game, students were paid $3 per finished maze. But just beforehand, they were given a set period of time where they could practice additional mazes, if they were willing to pay a small amount of their earnings. The researchers found that lower-ranked students paid more to practice before the final round than did students who had similar scores but happened to rank high.
But there was a twist. Some groups were given an additional option: disrupting others. In these groups, students were allowed to pay to “slime” another student’s screen during the practice period. This made a green blob appear on that person’s maze, making it more difficult for them to practice.
In groups that allowed sliming, the lower-ranked students became more disruptive. If two kids had the same scores, and one was ranked first in her group and the other was ranked last, the low-ranking student paid an average of 39 cents more on sliming and about 6 to 8 cents less on practicing than the top student did.
In other words, a relatively poor performer would rather disrupt other classmates’ work than try to improve at the task. “It’s not fun to do something where others are better,” Spenkuch speculates. “If it’s not fun to solve mazes, then I might as well slime them and disturb them.”
The Real World Is More Complicated
So should we all rush to move our children into poorer-performing schools?
Spenkuch cautions that the results do not translate into clear policy recommendations.
The current research analyzed the effect of ranking alone, deliberately isolating it from other factors that might influence students’ performance and behavior. For example, the researchers did not investigate whether smart peers might also act as good role models for other kids. The study suggests that having a low ranking would counter that positive effect, Spenkuch says, but “it’s not clear whether it would outdo it.”
Researchers will need to perform randomized trials to test whether placing kids with smarter peers is a good idea overall, he says: “The best way to do it is to try the actual policy on a small scale and see what happens.”
Cicala, Steve, Roland G. Fryer, Jr., and Jörg L. Spenkuch. 2016. “Comparative Advantage in Social Interactions.” Working paper.