VoxEU Column Education Labour Markets Poverty and Income Inequality

The long-run effects of disruptive peers

Bad behaviour by peers is well-known to worsen educational outcomes in the short run. This column investigates the long-run effects of peers from families marked by domestic violence. Individual-level US data linking middle and high school test scores, college enrolment, and earnings at ages 24–28 show that students exposed to more disruptive peers experience worse adult outcomes. Policies that mitigate exposure to disruptive peers could pay high dividends.

A series of recent papers has examined the extent to which early childhood interventions affect adult outcomes. While some of these papers have examined the impact of early childhood education programs such as Head Start (Garces et al. 2002, Ludwig and Miller 2007) or the Perry Preschool Project (Heckman et al 2013), others have focused on specific inputs to the education process. For example, Chetty et al. (2011) examine the impact of class size and teacher experience on adult earnings using data from Project STAR in Tennessee, where students were randomly assigned across classrooms. They find that while the long-run effects of class size are imprecise, teacher experience and overall classroom quality have significant effects on earnings in young adulthood. Similarly, Chetty et al. (2014) show that being taught by a higher value-added teacher in elementary school has large effects on young adult earnings, even though contemporaneous teacher effects on test scores have been shown to fade out over time.

Thus far, however, little is known about the impact of peers on long-run outcomes such as educational attainment or wages. This contrasts with a large literature that has shown peers affect current educational achievement and misbehaviour in school. The lack of evidence on the long-run effect of peers is important, because it is not clear that one’s peers will affect outcomes years after those peers are gone. For example, peers could primarily affect contemporaneous performance on standardised exams rather than learning, in which case the effects could be short-lived. Similarly, while certain peers may induce some students to commit more disciplinary infractions, those peers may not necessarily put the students onto a long-term path toward worse behaviour. In contrast, those behavioural issues may go away when the student integrates into new and different peer groups in the future.

The long-run effect of peers on test scores and educational attainment

We address the question of whether elementary school peers have long-run effects in recent research (Carrell et al. 2015). Specifically, we focus on peers from families linked to domestic violence, who were shown in previous research to disrupt contemporaneous achievement and behaviour (Carrell and Hoekstra 2010, 2012). To do so, we use administrative individual-level school data from one county in Florida and link it to middle and high school test scores, college enrolment, college degree attainment, and earnings at age 24–28.

The approach we use is to utilise the population variation in the proportion of disruptive children across cohorts in the same school. The intuition of the comparison is that while some cohorts by chance have, say, 3% ‘disruptive’ students, the next cohort in the same school may be composed of 6% disruptive students. We then ask whether the students exposed to an idiosyncratically higher proportion of disruptive peers perform worse years afterward as a result.  

Results indicate that there are persistent effects on both test scores and educational attainment. We estimate that exposure to one disruptive peer in a class of 25 throughout elementary school is associated with a 0.02 standard deviation reduction in test scores during high school, and nearly a one percentage point reduction in the likelihood of receiving a college degree. This suggests that the impact of disruptive peers does persist with respect to educational outcomes years afterward.

The long-run effect of peers on earnings

Estimates of the effect of disruptive peers on earnings are shown in Figure 1. Earnings are shown on the vertical axis, while the horizontal axis shows the extent to which the student was exposed to an idiosyncratically high or low proportion of disruptive peers in their cohort. The red ‘+’ markers indicate average predicted earnings based on student characteristics such as school, race, sex, and family income, while the red dashed line represents the general trend between peer exposure and predicted earnings. Importantly, the slope of predicted earnings is flat, which suggests that absent differential exposure to disruptive peers, we would have expected students to have similar earnings at age 24 to 28.

In contrast, actual earnings, shown in black, vary significantly based on exposure to disruptive peers during elementary school. Thus, Figure 1 shows that while individuals who have idiosyncratically low exposure to disruptive peers (those on the left-hand side) tend to earn more than predicted, those with idiosyncratically high exposure tend to earn somewhat less than predicted. Specifically, we estimate that exposure to one disruptive peer in a class of 25 throughout elementary school reduces earnings by 3–4%, with effects being driven by exposure to disruptive boys.

Figure 1

Mechanisms

There are two major channels through which disruptive peers can affect long-run outcomes. The first is educational achievement and attainment. Several reasons lead us to conclude the effects are unlikely to work through this channel. First and foremost, the effects on test scores and educational achievement, while present, are too small to explain the relatively large effects on earnings. Second, the subgroups most affected in terms of test scores are not those most affected with respect to earnings. For example, the effects of disruptive peers on test scores are largest for the highest-achieving students, while the earnings reductions are largest for the lowest-earning adults. In addition, while the effects on educational achievement and attainment are similar across race, the earnings reductions occur primarily among white students.

In contrast, we conclude that the effects are most likely working through non-cognitive channels such as motivation, work ethic, self-esteem, or behaviour. While it is difficult to test this directly, we do find evidence that disruptive peers lead to increased suspensions during high school, particularly among whites. In addition, the non-cognitive channel is the most common explanation for the effects found in existing research on early childhood interventions such as early childhood education, teacher quality, and class quality (Garces et al. 2002, Heckman et al. 2013, Chetty et al. 2011, 2014).

Conclusion

The question of whether peers have effects that persist into adulthood is an important one for education policy. Many important education policies, such as tracking and school vouchers, affect peer composition. As a result, to the extent peer effects impact adult outcomes, those policies can have important implications. The results of our study suggest that any policy that leaves some students more exposed to disruptive peers will result in significantly worse long-run outcomes for those students.

The results here also highlight the potential impact of policies that mitigate the impact of disruptive peers. To date, there has been relatively little research on the impact of such interventions, and most research has instead been focused on understanding the effects of direct inputs such as teacher quality or class size. But the large effects documented in this study highlight the potential importance of other interventions, such as school counsellors, that are aimed at helping teachers reduce the impact of disruptive students.

Our findings also speak to the extent to which differential exposure to disruptive students can lead to income inequality later in life. We calculate that the increased exposure to disruptive peers by students from low-income families can explain 5–6% of the earnings gap between adults who grew up in low versus high-income households. This is significant given that we have only one particular measure of disruptive peers in our sample, and it highlights the extent to which sorting into schools can lead to the persistence of long-term income inequality across households.

References

Carrell, S E and M Hoekstra (2012) “Family business or social problem? The cost of unreported domestic violence”, Journal of Policy Analysis and Management, 31(4): 861-875.

Carrell, S E and M L Hoekstra (2010) “Externalities in the classroom: How children exposed to domestic violence affect everyone’s kids”, American Economic Journal: Applied Economics, 2(1): 211-228.

Carrell, S E, M Hoekstra and E Kuka (2016) “The long-run effects of disruptive peers”, NBER Working Paper 22042.

Chetty, R, J Friedman, N Hilger, E Saez, D Whitmore Schanzenbach and Yagan (2011) “How does your kindergarten classroom affect your earnings? Evidence from Project STAR”, Quarterly Journal of Economics, 126(4): 1593-1660.

Chetty, R, J Friedman and J E Rockoff (2014) “Measuring the impacts of teachers II: Evaluating bias in teacher value-added estimates”, American Economic Review, 104(9): 2533-2679.

Garces, E, D Thomas and J Currie (2002) “Longer-term effects of Head Start", American Economic Review, 999-1012.

Heckman, J, R Pinto and P Savelyev (2013) “Understanding the mechanisms through which an influential early childhood program boosted adult outcomes", American Economic Review, 103(6): 2052-2086.

525 Reads