Last updated on March 18, 2014
A few weeks ago, I published a post in which I announced that my paper titled “Rising Food Prices, Food Price Volatility, and Social Unrest” (click here for the version that was accepted; a version devoid of typos and which includes missing prepositions will be published) had been conditionally accepted for publication at the American Journal of Agricultural Economics, the top journal in my field.
As has been the case every single time I have posted on the topic, this generated a number of tweets, retweets, direct messages, and so on. On this more than anything else I have worked and written on, it was fairly easy to tell what kinds of prior belief people were bringing to the table when they commented, and it was interesting to see confirmation bias do its thing.
The reactions I got to my core findings that “rising food price levels cause social unrest, increases in food price volatility do not appear to do so” were usually of the following varieties:
- The Blasé: Someone sent me a direct message saying “We already knew food prices caused food riots. So what are we learning from your study, really?” This was also the reaction of a reviewer at a journal other than the one in which the paper will be published (“Why do you do so much maths [sic] to tell me something I already know?”). Of all the reactions I get when I present or write about this paper, this is assuredly the most disheartening, because it usually comes from people who should know better, and who should have the critical thinking skills required to not mistake correlation for causation. If we can afford the luxury of scientific investigation, we would be fools to be satisfied with mere belief.
- The Sees-the-Trees-but-not-the-Forest 1: A number of people inevitably responded with “Yes, but in [insert their favorite country], food prices went up, and there were no riots!” Sure, but if you remember what a regression does, you’ll remember that a regression is right “on average,” so while rising food prices might not have caused riots in, say, China, my estimates still say that they can cause them elsewhere in the world. On this, I had a post a few weeks ago about failure to acknowledge that treatment effects have a whole distribution about their mean.
- The Sees-the-Trees-but-not-the-Forest 2: Others responded with “Yes but income shocks/lack of human rights/unemployment/etc. can also cause food riots!” Sure, but nowhere in my paper do I claim that food prices are the only cause of social unrest; I just claim that they can cause social unrest. Economists have pretty good tools to study whether changes in a variable X cause outcome Y to change. As far as I can tell, we have not yet developed tools that allow us to rule out all other causes. In other words, finding that smoking causes lung cancer does not invalidate the fact that breathing in asbestos dust might cause lung cancer, too!
- The “Skeptic by Default”: This is probably the reaction I am most sympathetic to, as it usually comes from graduate students, recent PhDs, and assistant professors. Usually, this comes in the shape of “Have you tried [additional robustness check]?” Of all four reactions I usually get to this work, this is the most useful, as it makes for better social science.
I am perfectly willing to concede that the fault my entirely be my own, and that bad writing might be what leads people to have the aforementioned misconceptions. But after writing on the same topic for three years and saying the same things over and over, I just thought I would try to put all of my responses to common criticisms in one post.