This site belongs to UNESCO's International Institute for Educational Planning

How can learning surveys inform policies to close the learning gap due to bullying?



Suggest a resource

How can learning surveys inform policies to close the learning gap due to bullying?

Contributed by: Marcos Delprato, Kwame Akyeampong and Máiréad Dunne, Centre for International Education (CIE), University of Sussex, UK


Learning surveys such as the Third Regional Comparative and Explanatory Study (TERCE) allows an examination of the impact of bullying on learning outcomes in the Latin American (LA) region. Importantly, the breadth of information on family, neighbourhood, school factors included in learning surveys can, with an appropriate analysis, guide effective policies to minimise the negative effects of bullying on learning as well as for comparing what works for each country within the LA region. 


Being bullied is known to significantly lower students’ achievement and it has other long-term negative consequences making this a particularly important social and economic issue all over the world. Studies using the Health Behaviour in School-aged Children (HBSC) and Global School-based Health Survey (GSHS), estimate that 30% of adolescents report being the target of bullying across five regions covering 72 countries. Given the importance of improving the quality of learning in schools as an important part of the post-2015 development agenda, this is now a more pressing issue for less developed regions.


Against this backdrop, we recently published an article in the International Journal of Educational Development with new evidence for the association of bullying with learning scores as well as for non-cognitive outcomes for sixth grade students in 15 Latin American countries using the Third Regional Comparative and Explanatory Study (TERCE) learning survey of 2013.


What insights can TERCE offer?

TERCE permits a more detailed and comparative analysis of the school violence and learning relationship in the LA region:                                                                                                 

. By using distinction of bullying by type: physical or psychological.

. By examining the extent to which bullying impacts on the acquisition of non-cognitive outcomes such as students’ sense of belonging at    school, home study and socialising.

. By using appropriate techniques:

. to find out whether the effect of bullying affects relatively more low, medium or high performers and whether policies are more effective      for any of these specific group of performing students. 

. to find out which policies help to reduce the negative effect of bullying as well as which neighbourhood and school features are useful to    identify the most problematic school in each country after comparing students with the same background.


What do we find?

We find that how bullying translates into poorer achievement varies considerably across countries in the LA region. For math, estimates suggest that bullied students achieve between 9.5 and 18.4 points less than their non-bullied peers, and between 5.8 and 19.4 lower scores for reading (Figure 1).[1]


Figure 1. Impact of bullying on learning scores (matching estimates)


We also find that both physical and psychological bullying are equally damaging to learning (Figure 2), with psychological bullying being a major determinant explaining low degree of socialisation among students (Figure 3). Again, these effects vary substantially across countries.


Figure 2. Impact of bullying types on learning scores (matching estimates)



Figure 3. Impact of bullying types on non-cognitive outcomes (matching estimates)



When analysing the role of school policies to minimise the effect of bullying on learning[2] we find that:

. In general, there is a mismatch of some in-school policies (e.g., on teachers’ skills), but simple measures such as allocating female    teachers to the most problematic classrooms can have wide-ranging positive effects across countries. Hence, school violence      programs and policy in the region should increase recruitment and retention of female educators (see Diagram 1 and Table 1 in         Appendix).

. To boost the success of school violence policies, they should be showcased based on students’ achievements levels (i.e., top      students in the case of math and bottom learning performers of reading).


Diagram 1. Factors/policies effects on the negative bullying-learning gap


Policy implications?

Our analysis also points towards two main policy implications. First, targeting either students from households that receive conditional cash transfers or students living in violent communities could potentially lead to no effect of bullying on learning for half of countries in the LA region (see Table 2). A more promising avenue is to move targeting to the proximal social contexts of students (and their families) by incorporating anti-bullying strategies within social programs, as this can have wide ranging effects.


Second, teaching skills programs seem to be disconnected from the school violence phenomenon.  Quite the opposite, school programs operating beyond internal school factors with a focus on nurturing school-community social capital are very powerful, particularly cultural programs.


There needs to be greater recognition that underachievement among some students could be partly due to the effects of bullying.  Teachers need to be sensitised to this impact and provided with skills to identify students suffering from bullying for the necessary support.


The overreaching message from this blog is for education policy makers as well as those individuals working with disadvantaged populations of students in the LA region to outline policies on how to tackle bullying based on empirical evidence from learning data such as the TERCE. By doing so, future policies that tackle bullying in schools could have a better chance of tackling the problem effectively and improve learning outcomes for the most vulnerable and disadvantaged student in the LA region. 




* p < 0.10, **p < 0.05, ***p < 0.01.




* p < 0.10, **p < 0.05, ***p < 0.01.




[1] The acronym WS denotes the whole sample estimate in all figures.

[2] See Section 4.4 of our article for details.