Peer Effects in Education

February 26, 2011

Who would deny that friends, especially those in the same classroom, influence how much each other learn.  This seems like a really important process to understand.  It’s one of my research interests.  Unfortunately it’s really hard!

Mark, over at Observational Epidemiology, links us to VOXEU for a paper attempting to shed light on the topic.  The authors are economists, and they offer sophisticated econometrics in an attempt to estimate a truly causal effect.  Unfortunately, the “treatment” effect, they attempt to identify is rather odd… the sum of one’s peers’ future educational attainment:

“We find that a standard deviation increase in peers’ aggregate years of education (roughly two more high-school graduate friends) translates into roughly a 10 percent increase of a standard deviation in the individual’s education attainment (roughly 3.5 more months of education).”

One of the key generalizations the authors would like to make is that improving one student’s education, improves their friends’ educations.  It seems that to inform that generalization, a better independent variable would be the mean of friend’s educational attainment, rather than the sum.  There’s a lot more that could be said about the methodology, but I’ll hold off for now.  Get the full paper, by Eleonora Pattachini and Yves Zenou, here.


Eleonora Pattachini and Yves Zenou (2011). Dynamic Aspects of Teenage Friendships and Educational Attainment Center for Economic Policy Research (February) Other: DP8223


Ngrams of Social Science Disciplines: Part 2

February 4, 2011

I have a few more thoughts about the graph of Ngrams of social science disciplines that I posted about last week.

The trends in economics, sociology, and anthropology appear, to my eye, to be correlated.  What cultural factors, cause this?  General interest in the formal study of social behavior?  What causes that?  Perhaps the correlated short-term, i.e. year-to-year variation, is best understood as having a different cause from the correlated trends over longer periods?  In particular, why did each term decline since the 1990s, with a particularly big drop-off from 2007 to 2008?

Perhaps the last year of data is unreliable, but should we be worried about the trend?

I plugged in the name of a practitioner of each discipline: “sociologist, economist, political scientist, anthropologist, psychologist” to see if I noticed any differences.  The basic trends look the same, but there are some differences.  e.g. in the last two decades, “anthropologist” is more popular than you’d expect given “anthropology” and “economist” is less popular than you’d expect given, “economics.”  I can easily spin a story: anthropology is more subjective, so the identity of the researcher is more relevant.  I can think of more possibilities, but none of them explains why this difference only becomes notable in the last twenty years.

Looking over a longer time period, we can see that each of the human sciences has become more prominent since 1900 (at least by this measure.  In addition to other concerns, consider that economics was once known as “political economy.”)

The “harder” sciences have also done well.  Physics and mathematics growing faster in the first half of the century, but then stagnating as biology (tracking sociology amazingly closely) catches up.


I’m having a lot of fun fooling around with these ngrams, so I’ll probably do at least one more post using them soon.  I highly encourage interested parties to check out the paper published in Science which will introduce you to a far greater range of types of analysis that this dataset allows.