You are here: Home Projects A: Family and education policies A04: Peer effects in school

A04: Peer effects in school

We provide empirical evidence on peer effects and class size effects in school across awide range of countries using a so-called birth cohort approach. This approach exploits "quasi-natural experiments" that take place every year in every school in school systems with birth-date-based school enrollment rules. The approach is used to estimate gender peer effects, peer effects related to the age and socioeconomic composition of birth
cohorts, and peer effects related to class size. The data requirements for the approach are only slightly more stringent than for existing approaches exploiting variation at the grade level.


Project Members


Discussion papers (A04)



Estimation of peer effects is challenging. State-of-the-art research exploiting quasinatural experiments does not appear to point to general conclusions. A reason could be the idiosyncrasies of quasi-natural experiments combined with data mining. Hoxby’s and Lavy and Schlosser’s grade level approach could be implemented in a variety of settings, but does not work in school systems with grade retention. Ciccone and Garcia-Fontes propose an alternative: the birth cohort approach to peer effects. This approach exploits within-school differences in student characteristics across birth cohorts, defined by the school system’s enrollment rule. Peer effects estimated using the birth cohort approach are causal intention-to-treat effects if within-school differences in the characteristics of birth cohorts are random.  

Policy relevance

Peer effects limit the equality of opportunity across children born into different ethnicities, socio-economic groups, neighborhoods, etc. Peer effects require rethinking school policies on joint/separate learning. Should single-sex schools be permitted because of gender peer effects? Should boys and girls or the youngest and oldest in a birth cohort be taught separately in some subjects? The birth cohort approach can also be used to re-evaluate the effect of class size on learning, a main determinant of the optimal class size. 

Project Plan

Work package 1 - Gender peer effects in school  
  • The birth cohort approach to gender peer effects can be implemented in a wide variety of countries, whether or not their school system allows for grade retention.  
  • The data requirements are only slightly more stringent than for the Hoxby/LavySchlosser within-school grade level approach.  
  • We will assemble a database that permits estimation of gender peer effects using the birth cohort approach, starting with the three to four countries where the data is most easily available.  
  • We start with gender peer effects as this is where drawbacks of Hoxby/LavySchlosser’s within-school grade level approach are especially evident. This could explain why Hoxby/Lavy-Schlosser’s results stand in contrast to the evidence from idiosyncratic quasi-experiments.  
Work package 2 - Peer effects in school: Cross-country evidence based on the same natural experiment  
  • We use the within-school birth cohort approach to estimate externalities related to three other types of characteristics of peers:  
  1.        the average age of peers within a birth cohort;  
  2.        the country of origin of peers within a birth cohort;
  3.        and the education and profession of peers’ parents within a birth cohort.  
  • A key challenge is data availability and trends that are driven by changes in the quality of schools.  
Work package 3 - The effect of class size on learning  
  • The existing evidence on the effect of class size on learning is partly based on the influential grade level approach of Hoxby (2000b). This approach is likely to yield upward-biased estimates of the effect of class size on learning in school systems with grade retention.  
  • This is because exceptionally large birth cohorts in a school will result in grades with a smaller share of retained students, which could mask a true negative effect of class size on learning.  
  • The within-school birth cohort approach will be used to re-evaluate the effect of class size on learning.  
  • A main challenge is that changes in the size of a cohort may reflect trends that are driven by changes in the quality of schools. 


Document Actions