Better Together? Social Networks in Truancy and the Targeting of Treatment
Key Finding
Because students skip school with friends, social network maps can help target truancy interventions and reduce class absences. A model of social networks using school absence data shows that students skip class 4.7 times more often with their “strongest tie” peer than any other peer. When parents received weekly text messages about their child’s missed classes, not only did their own child attend more classes–their child’s “strongest tie” peer did, too.
. . .
Other Findings
Partial-day absences are more common than full-day absences, showing the importance of understanding how interventions affect attendance at the class level.
For an individual student, the benefit of skipping class doesn’t come from the act of skipping class itself, but from doing so with a friend. Students are nearly five times more likely to skip class with their “strongest tie” peer than any other peer.
Interventions to reduce truancy can be made twice as effective – and cheaper – if they target students with high truancy rates and many strong connections to other students.
An intervention that consists of sending text messages to parents about missed classes, which costs about $7 per student per year, is 19% more cost-effective on a per-student basis when taking into account the spillover effects throughout the student network.
. . .
Methodology & Data Highlights
Social network mapping of truancy using student-by-class-by-day attendance data.
Field experiment in 22 middle and high schools in Kanawha County Schools (KCS), West Virginia, 2014-2015 school year.
Summary
Reducing truancy rates matters for student outcomes because students who “skip school” more often are more likely to drop out of high school, have substance abuse problems, or engage in criminal activity. [1] Addressing truancy is also important for schools. State funding for schools is often affected by truancy rates, and many states use rates of chronic absenteeism to measure school performance under the 2015 Every Student Succeeds Act.
In a study that contributes to the literature on the interaction between social networks and risky youth behaviors, this paper finds that students are more likely to miss an individual class than a full day of school, that students systematically miss classes together, and that it’s possible to identify specific students that likely influence many other students to skip class.
Understanding the spillover effects of a single truant student can inform policy interventions. Check and Connect, a successful truancy intervention that reduces absences by linking students with trained mentors, is priced at $1,700 per student. However, these costs don’t consider spillover effects and it could be more cost-effective to employ a program like Check and Connect by targeting key students in the social network.
A network map of truancy reveals influential students
Social Network for School 1: Pre-Intervention Period. Each node (circle) corresponds to a student and each interconnecting line or edge corresponds to the number of absences between two students on the same day and same class during the pre-intervention period.
The researchers built a truancy social network by mapping the frequency with which students missed classes at the same time. Mapping the students this way allowed the researchers to measure the strength of connections between students who skip classes together and to identify clusters that indicate a group of students primarily skipping classes with other students in the group. The network can also determine an individual student’s “degree of centrality” – or the student’s number of connections with other students.
The researchers estimate that almost 50% of the students who have a “strongest tie” with another student actively coordinate their absences. Compared to average absences across all their other peers, students are absent 4.7 times more often with the peer who is their “strongest tie.”
To understand the characteristics that students may share with their truancy peers, the researchers reviewed data from the district’s digital gradebook that includes by-class attendance, missed assignments, and grades for 14,000 students (11,000 households). They combined this data with administrative data showing students’ race and gender, as well as each student’s suspension record and English language status from the previous year.
The researchers found that students tended to skip class alongside those whose GPAs, behaviors, and racial characteristics mirror their own. While much of this phenomenon may be driven by the characteristics of students in the overall network, researchers also see that students tend to skip class alongside students of the same gender and with similar academic performance.
An experiment to measure spillover effects of low-cost interventions
Text Message Scripts
To study whether these findings will create spillover effects in existing programs to reduce absences, the researchers incorporated findings from a previous experiment in the same setting. In that experiment, Peter Bergman and Eric Chan used the digital gradebook data to test a system in which parents were randomly selected to receive text messages on missed assignments (weekly), missed classes (weekly), and low grades (monthly). The experiment showed that parent alerts caused significant (39%) reductions in course failures and increases (17%) in class-level attendance.
However, because the researchers had data on missed classes for all students and not just those who participated in the experiment, they were able to investigate how the reduction in truancy of the students in the treatment group influenced the truancy of treated student’s friends. They asked: If Student A skipped classes most often with Student B and Student B’s parents were in the treatment group to receive text messages, did that affect Student A’s attendance, as well?
The short answer is yes. For students in the treatment group whose parents received weekly text messages about missed classes, their “strongest tie” peer attended, on average, 24 more classes from late October through the end of the school year. Most of that increase came from attending classes with the “strongest tie” peer, specifically, reflecting the idea that students are skipping or attending classes to spend time with a particular friend, as opposed to skipping classes for the sake of skipping classes.
Benefits of the methodology
The research methodology used in this paper to build a social network sits between two methods that have been commonly used by researchers in the past to understand network effects on risky behavior. One approach infers an individual’s peers by exploiting random variation within groups. The other method gathers, through surveys of large groups, self-reported information on peers or friendships.
In this paper, researchers used administrative data that build a network based on an observed behavior - i.e. truancy - that is both less subject to self-reporting bias and may be more relevant to risky behaviors than general friendship networks.
Still, a disadvantage of the approach is that the researchers define social networks using class schedules and, despite documenting significant evidence of coordination, cannot be certain students are actually coordinating their absences.
To test for coordination, the researchers run a simulation and find that, in a network where students do not coordinate, joint absences occur less frequently than they observe in their real data. Still, it’s possible that students skip classes simultaneously for other reasons than coordinating (for example to attend a sports-related event), or that students that share similar characteristics tend to skip the same classes without coordination.
With better targeting, low-cost text interventions could be even more effective
These findings suggest that programs aiming to reduce absences could both increase their impact and save money by working more closely with particular students in a network. The researchers estimate that an “optimally-targeted” intervention could result in a five-fold decrease in class absences.
Even without changes, however, the researchers’ own intervention of sending text messages to parents about missed classes, which costs about $7 per student per year, is 19% more cost-effective on a per-student basis when taking into account the spillover effects throughout the student network.