21 May 2019

Cassette tapes "cause" conflict

Nathan Nunn and Nancy Qian published a very worrying paper in 2014 showing that US food aid causes conflict in recipient countries. Their research design used total US wheat production as a source of quasi-experimental variation in the amount of food aid countries received, to show causality rather than just correlation.

A new paper by Paul Christian and Christopher Barrett apparently debunks the study, showing that the "causal" correlation is spurious. Replace "US wheat production" with "US tape cassette sales" and you can almost exactly replicate the results.

Which reminds me of an earlier paper showing that "average male organ length" is a strong predictor of GDP growth. We only have about 200 countries, which is not a lot of observations to power a robust statistical analysis, so you should take most cross-country empirical analyses with a pinch of salt. These "male organ" and "cassette sales" papers are helpfully colourful reminders.

HT: Jeffrey Bloem

No comments:

Post a Comment