It’s a serious problem…

This week, a paper from outside of my field caught my attention. Miguel Sarzosa and Sergio Urzúa study the effect of bullying among teenagers in Korea and find a substantially negative impact on wealth, educational outcomes and life satisfaction. Teenagers that report to have been “severely teased or bantered, threatened, collectively harassed, severely beaten, or robbed” at the age of 15, are 75% more likely to be sick, have a 50% higher chance to suffer from mental health issues and experience 20% higher stress levels in their relationships with friends and parents at the age of 18. Bullying is especially harmful for students with low non-cognitive skills (self-esteem, perseverance, positive attitude towards life, etc.). For them bullying increases the risk of depression, decreases the likelihood of going to college and makes them more likely to smoke.The age cohort under study, 15 to 18 years old, is particularly interesting because these years are a “decisive period of development” for every teenager.

The methodology is noteworthy here. You cannot simply regress the outcome variables of interest (life satisfaction, going to college, or smoking) on a measure of bullying while controlling for other measures of cognitive and non-cognitive skills. There are at least two problems with this approach: (1) Bullies pick their targets very specifically. It’s the quiet and unpopular kid that gets harassed and teased. Thus, the likelihood of being bullied is correlated with other individual characteristics which researcher might not all observe. (2) Cognitive and non-cognitive skills are usually measured in discrete categories. Think of a survey item that states “I think that I am a worthy person” and respondents are asked to rate their agreement on a 5-point Likert scale ranging from “I strongly disagree” to “I strongly agree”. However, these categories are only a proxy for the underlying (so-called latent) skill which we believe to be a continuous variable. An estimation strategy has to take this problem of proxy variables into account otherwise it will lead to incorrect conclusions.

The study shows that teenage bullying is very harmful already in the medium-run. And it’s plausible that the effects are even more severe in the long-run. The main mechanism at play is that bullying creates a dislike for going to school and other educational institutions because they provide an ideal environment for bullying to occur. Victims enroll less often in college because they associate class rooms with the faces of their harassers. Lower education then means lower wages later in life which leads to lower life satisfaction and perception of self-worth–a perfect vicious circle.

But the researchers also show ways to fight the problem. I already mentioned that the consequences of bullying are most severe for teenagers with low non-cognitive skills. Moreover, low non-cognitive skills make it much more likely (37% for a one-standard-deviation reduction in skills) to become the victim of bullying in the first place. That means that raising the skill level would reduce both the occurrence of bullying as well as its dangerous consequences. On the other hand, and somehow contrary to the cliché, cognitive skills (measured by grades) show no impact on bullying. It’s not the smart kid that gets bantered more often.

Schools can try to develop the non-cognitive skills of their students in class. These positive character traits are essential in many different situations in life. They deserve greater emphasis compared to cognitive skills which school education is traditionally concerned with.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s