Post-Randomization Peer Compositional Stability

In Simple Words

If participants in a fitness study invite outside friends to join their workouts, researchers must verify that the incoming friends do not give certain workout groups an artificial boost compared to others.

Research question

Post-assignment student mobility and the entry of non-sample peers into study classrooms can alter peer composition, potentially confounding treatment effect estimates if high-effectiveness teachers systematically receive better non-sample peers.

Participants

N = 27,255 randomized students in grades 4–8 across 619 randomization blocks in 6 urban school districts.

Study Design

SURE joint hypothesis regression of actual spring classroom peer characteristics (baseline math/ELA scores, special education %, ELL %, race, absences, suspensions) on assigned teacher effectiveness.

Results

Assigned teacher effectiveness showed no statistically significant relationship with peer baseline math scores (coefficient = 0.246, SE = 0.164) or other peer traits; joint hypothesis p-values were 0.405 for grades 4–8 and 0.321 for high school, proving post-randomization peer stability.

Takeaway

Assess the baseline characteristics of actual end-of-year classroom peers and regress them on assigned treatment intensity to rule out peer compositional contamination.

Read alongside the research

Plain-English glossary

Original explanations, informed by the references below. Examples are illustrations, not data from the studies above. The research text stays unchanged.

Baseline

The starting measurements in a study. Researchers use these to compare what participants were like at the start with what happens later.

ExampleBefore a training programme starts, researchers measure workers’ earnings. That first measurement is the baseline.

GET-IT glossary · Baseline characteristics ↗Read more about baseline →

Statistically significant

A label used when a result passes a chosen statistical threshold, often p < 0.05. It does not show how large or useful an effect is. A result that is not statistically significant does not prove there is no effect.

ExampleA study reports p = 0.03 against a 0.05 threshold. It passes that test, but the effect could still be too small to matter.

American Statistical Association · Statement on p-values ↗Read more about statistically significant →

Related findings

• Attrition Independence from Assigned Treatment Effectiveness →Education

  In a drug study, if patients assigned to less effective doctors quit the trial far more frequently than those assigned to top doctors, comparing the patients who stayed creates a skewed picture. Verifying that missing data rates are equal regardless of doctor rating confirms the experiment remains fair.

• Class Size Reduction Ineffectiveness →Education

  Reducing the number of audience members in a lecture hall does not help the remaining listeners understand a presentation if the speaker continues delivering the material in a language that the audience cannot comprehend.

• Heterogeneous Peer Preference Sorting →Education

  Imagine two groups of students—one group wants intensive French immersion and the other wants computer programming. Allowing each group to gather in specialized academies lets both groups enjoy their ideal learning environment without depriving the other group of what they value.

• Linear-in-Means Peer Externalities Model →Education

  Think of a team relay race where swapping a fast runner onto Team A makes Team A faster but makes Team B equally slower, leaving the combined average time of all runners completely unchanged. Unless the swap somehow inspires individual runners to run faster on their own regardless of who runs next to them, overall race performance remains identical.