Last updated on May 31, 2015
That is the topic of a new working paper of mine, written with my PhD student Yu Na Lee and my collaborator David Just, and which we titled “Was Sandmo Right? Experimental Evidence on Attitudes to Price Risk and Ambiguity.”
This is probably the most exciting research project I have ever had a chance to work on, and I wish we were releasing a more polished and less rough draft, but the submission deadline for presented papers at this summer’s AAEA meetings in San Francisco and the association’s policy of posting papers online also being what they are, the paper is now available online, so we might as well talk about it. So a caveat is in order: This is a rough and preliminary version of a paper that will almost surely look very different once it is published. It has not yet gone through the peer-review process. Keep that in mind as you read this post.
In this paper, Yu Na (who will be on the market in two years), David, and I decided to build on the results in my award-winning 2013 AJAE article with Chris Barrett and David Just, where we estimated the effects of price volatility on the welfare of rural Ethiopian households. The results in the 2013 paper, however, relied on survey data, which are both noisy and not conducive to the cleanest of identification strategies. So this time around, we decided to gun for the gold standard and turn to the lab.
Thanks to generous funding from the University of Minnesota’s Office of the Vice President for Research, we conducted experiments in December 2014 and March 2015 to test how producers behaved in the face of two types of price uncertainty: (i) price risk, i.e., uncertain prices whose distribution is known, and (ii) price ambiguity, i.e., uncertain prices whose distribution is unknown. Proceeding this way allowed us to test the claim made in a seminal paper in microeconomic theory by Agnar Sandmo (here is an ungated copy), where he showed that the presence of output price risk uncertainty (in this case, price risk, or uncertainty of a known distribution) would lead a risk-averse firm manager or producer to decrease how much he produces in an effort to hedge against price risk relative to a situation where the price is known with certainty. (Of course, this assumes no insurance against price risk in the form of contracts of futures and options markets.)
Sandmo’s finding is actually pretty important for both agricultural and development economics. For agricultural economics, Sandmo’s finding means that the presence of uncertain prices for agricultural commodities leads producers to produce less than they otherwise would, and so to the provision of insurance programs for farmers. For development economics, Sandmo’s finding means the same thing, but crop insurance program usually being nonexistent in developing countries, it means that all those small producers who individually decide to produce less aggregate into there being less food to go around, and thus as a cause of food insecurity.
Was Sandmo right? What we find, if it holds up to the additional experimental sessions and treatments we wish to conduct in 2015-2016, is very surprising. Here is the abstract of our paper:
In his seminal 1971 article, Sandmo showed that when faced with an uncertain output price, a risk-averse firm manager would hedge by producing less than he would have when faced with a certain output price. We take Sandmo’s prediction, among other things, to the lab. We study in turn the effects of price risk (i.e., uncertain prices whose distribution is known) and price ambiguity (i.e., uncertain prices whose distribution is not known, but whose range is known) while controlling for our subjects’ income risk preferences. Our experimental protocol closely mimics Sandmo’s theoretical model. For price risk, we use a two-stage randomization strategy aimed first at studying the effect of price uncertainty relative to price certainty, and then the effect of increases in price uncertainty conditional on there being price uncertainty. For price ambiguity, we use the same randomization strategy to study the effect of price ambiguity relative to price certainty while preventing our subjects from guessing the shape of the price distribution. For price risk, we find that, in stark contradiction to Sandmo’s theoretical result, the presence of price uncertainty causes subjects to produce more than they do under price certainty, but that increases in price uncertainty makes them decrease their production monotonically. For price ambiguity, results are mixed and depend on whether the portion of the experiment aimed at eliciting our subjects’ income risk aversion is played before or after the price uncertainty game. Lastly, we use our price risk data to study the problem structurally, in order to get at preference heterogeneity, and find that our structural results are consistent with our reduced-form results.
At a conference in March, Yu Na was asked by someone in the audience whether we had undertaken this research project to “confirm” the findings in Bellemare, Barrett, and Just (2013). Um, no. That is not how social science proceeds, at least not the kind of social science I want to be known for. Rather, the lab gives us a clean test of the theory, and that is why we turned to experimental methods for this project. At any rate, no matter what we find in this paper–a positive, null, or negative effect of price uncertainty on output choice–we will have an interesting finding that might help the design of better policies.