From natural variation to optimal policy? The importance of endogenous peer group formation.

How to cite

Sacerdote, Bruce; West, James E.; Carrell, Scott E.. 2016. "From natural variation to optimal policy? The importance of endogenous peer group formation.." AEA RCT Registry. AEARCTR-0001446. https://doi.org/10.1257/rct.1446-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Bruce Sacerdote* , James E. West , Scott E. Carrell

* Primary investigator

Registry ID
AEARCTR-0001446
DOI
https://doi.org/10.1257/rct.1446-1.0
First registered
Last updated
Status
Completed
Trial dates
2007-08-01 – 2012-12-31
Countries
United States of America (North America)
Registry record
http://www.socialscienceregistry.org/trials/1446

Abstract

We take cohorts of entering freshmen at the United States Air Force Academy and assign half to peer groups designed to maximize the academic performance of the lowest ability students. Our assignment algorithm uses nonlinear peer effects estimates from the historical pre-treatment data, in which students were randomly assigned to peer groups. We find a negative and significant treatment effect for the students we intended to help. We provide evidence that within our “optimally” designed peer groups, students avoided the peers with whom we intended them to interact and instead formed more homogeneous subgroups. These results illustrate how policies that manipulate peer groups for a desired social outcome can be confounded by changes in the endogenous patterns of social interactions within the group.

Intervention

The United States Air Force Academy organizes its freshman class into squadrons of approximately 30 students each according to an algorithm that creates demographically balanced peer-groups. Researchers conducted a randomized evaluation to test the impact of an alternate algorithm that optimizes the performance of lower-ability students instead.

Incoming students in 2007 and 2008 were randomly assigned to either a control or treatment group. The control group squadrons were assigned using the existing algorithm. The treatment group squadrons were assigned as specially engineered peer groups. Researchers analyzed the squadron composition and academic performance of first-year students from 2001-2006 and, based on these data, generated a formula that sorted students into groups predicted to maximize the performance of the lower-ability students. The application of this formula resulted in the creation of two types of squadrons: one type composed largely of lower-ability and high-ability students only (bimodal squadron), and the other composed largely of middle-ability students (homogenous squadrons). Students remained in their squadrons for the first year.

Experimental design

A sorting algorithm was created using pre-treatment data that sorted incoming freshmen into squadrons meant to optimize the grade point of the bottom third of low-ability students. Freshmen all take the same required courses during their first-year, take common exams to measure achievement, and have no choice as to their professors thus ensuring there is no self-selection of students into courses or certain professors. Incoming students in the fall of 2007 and 2008 were randomly allocated to control or treatment arms. Control students were assigned to squadron based on the existing algorithm that oprtimized demographic diversity. Treatment students were sorted using the new algorithm.

Researchers used administrative data from the Academy track academic performance. Researchers also surveyed students in the spring of their sophomore and junior years about their study partners and friends.

Design details

Randomization method
The random selection of the treatment and control squadrons was stratified across the four cadet “groups” which contain 10 squadrons each. It was also stratified with respect to new and returning “Air Officers Commanding” or AOCs, the officer in charge of military training within each squadron. This was done to eliminate any potential group or AOC level common shocks to academic performance. We flipped the treatment and control squadrons after the first year of the experiment. Additionally, the random division was subject to the constraint that siblings were split between the treatment and control groups.
Randomization unit
Individual students
Clusters
Study not clustered
Observations
2653 students
Arms
about 80 squadrons from two cohorts at the US Air Force Academy
Primary outcomes
GPA at the end of freshman year

Papers from this trial

  1. FROM NATURAL VARIATION TO OPTIMAL POLICY? THE IMPORTANCE OF ENDOGENOUS PEER GROUP FORMATION
  2. Carrell, Scott E., Bruce I. Sacerdote, and James E. West. 2013. "From Natural Variation to Optimal Policy? The Importance of Endogenous Peer Group Formation." Econometrica 81(3), 855-882. https://www.povertyactionlab.org/sites/default/files/publications/990%20Endogenous%20Peer%20Group%20Formation%202013_0.pdf

Cited by

No citation data available for this registration yet.

Keywords