How could a university create successful peer activities for students?

There is a robust literature on the effects of student peers on learning (see for example this 2011 review by Bruce Sacerdote (ungated version here)). If having successful peers can contribute to a student's success, then it may be attractive for universities to try and set up peer groups. There are various ways to do this, including early group work in classes, social events, and so on. It would be fair to say that the success of these attempts has been mixed.

So, I was interested to read this 2020 article by Thomas Fischer (Lund University) and Johannes Rode (Technische Universität Darmstadt), published in the Journal of Economic Behavior and Organization (ungated earlier version here). To be fair, what attracted me to this article was the question in the title: are persistent peer relationships are formed in the classroom or in the pub? It turns out that isn't really the focus of the paper at all, and that comparison is only of secondary interest to the story in the article.

What Fischer and Rode did was look at the peer effects for industrial engineering students at Technische Universität Darmstadt, which is somewhat of a different learning environment (compared to universities in the US, UK, or New Zealand). As they explain:

It is important to acknowledge that there is little interaction between students in our setting as compared to the Anglo-Saxon case mostly covered in the literature... This is not only a specific case of the institution TU Darmstadt and the study field, but more generally of the German university system. There is no pronounced campus live [sic], nor is there forced interaction in academic studies.

TU Darmstadt is not a campus university. This means above all that students do not usually live in university-provided housing...

The curriculum is built entirely around individual study results that are tested in the form of written exams and thus does not enforce interaction. Even the presence in lectures or lab sessions is not enforced. In fact, the growing online platforms make it more and more obsolete to attend physical meetings.

What this setting ensures (as much as is possible) is that there is unlikely to be any pre-existing peer groups between students, because their education is very individualistic (no group work) and doesn't involve interaction with their peers (inside or outside of class). Fischer and Rode then look at what happens after peer interaction is introduced, in two ways. First:

There is one exception from the focus on non-team work courses. A mandatory group work course labeled Projekt im Bachelor (engl. Project at Bachelor level) is part of the curriculum. The group work course intends to help students build soft skills and gain hands-on experience as group work is essential in future professional life. During the group work course, students have to come up with a business plan for some novel technological idea covering various aspects of marketing, budgeting, and legal implementation. After one week of intensive group work, all groups have to deliver a final report and pitch their results in front of a jury of professors and professionals. The task is deliberately designed with time pressure to induce cooperation between group members. Students are randomly assigned into groups...

Fischer and Rode then look at the effect of the nearest-peer's quality (in terms of pre-university grades) in the same major within these groups (which are quite large groups, of 11 to 13 students) for each student on their own grades in later courses. In other words, they look at the lasting effect of the nearest-peer on later grades. They find that an: 

...increase in the peer quality by one standard deviation (0.5 grade steps) increases the performance of the individual by −0.17 standard deviations (equivalent to a grade step of approx. −0.1). Note that in the German grade system 1.0 is the best grade and 5.0 the worst grade.

Note that last sentence. The negative effect here is an improvement in grades. Having a better nearest-peer significantly improves a student's grades. In contrast, the average quality of the other group members does not have a statistically significant effect on a student's grades. Fischer and Rode attribute this to students having a preference to match up with students who are similar to themselves in terms of academic quality. These peer effects persist into the future because these students continue to work together. And Fischer and Rode document statistically significant peer effects three semesters, five semesters, and seven semesters after the group work project.

Why would students prefer to continue working with other students at a similar level of academic quality? Fischer and Rode use an elaborate mathematical model to explain, but I have a much simpler explanation. If students have to work in pairs, and working with a student of higher quality improves Student A's grade, but working with a student of lower quality improves their grade by less (or decreases their grade), then Student A will want to work with the best student. The top students will pair up together, leaving the next-top students to pair up, and so on down the chain. Essentially, we end up with assortative matching of students into groups. And if working together improves grades, even outside of a group-work setting, those students will continue to work together.

Fischer and Rode then look at a second setting where peer interaction was introduced:

When students enter university, they are invited to an orientation week (OW), which takes place in the very first week of studies... Students are randomly assigned to other students taking the same major with an average group size of eight students to spend one week together. During this week students are given an introduction to the university by senior students. The students’ union organizes the event... The week usually culminates in a big party and includes several gatherings in pubs.

Given that individuals start very likely without social networks in this new setting, this is the very first opportunity in which they can make new contacts.

The orientation week activity is voluntary, but has about 80 percent participation. Fischer and Rode repeat their analysis looking at students matched up in the orientation week groups, and find:

Neither the next-best peer effect... nor the standard mean peer effect... are significant...

They attribute these null results to:

During the social gathering individuals learn little if anything about the academic ability of their potential peers. Nevertheless, they might decide to form learning groups. Once these learning groups are formed, the true ability will be shown also revealing mismatches... As such, these unbalanced learning groups may not last long. It is likely that individuals match with others of similar ability they get to know in other settings such as exercise sessions.

So, since these students cannot easily match up with similar peers in the orientation week exercise, they don't maintain those peer connections and the peer effect doesn't exist. Notice also that this orientation activity happens right at the beginning of the students' university career. Some students will drop out, which will also make it difficult to maintain peer connections.

There are two important learnings from this research for universities looking to generate peer effects. It will likely be more effective to generate these effects through classroom activities, not through social events. And, a focus on peer activities that are too early in the students' university journey is likely to be less effective than peer activities that occur a bit later. In other words, second-year group work will likely have a bigger effect than first-year group work. That is important to know.