Work program development
The research program on market design and pragmatics has had four active components over the last two years:
First, it analyses Recommender Systems. Individually tailored recommendation of products and services is a key feature of today’s online markets. In partnership with engineers, this component develops more economically efficient metrics and algorithms. Second, the research program focusses on Online Ad Auctions. This industry is huge – annual revenue worldwide may soon exceed $100 billion – but is understudied. In collaboration with Kevin McLaughlin, then a grad student and now a data scientist at Pandora, this component uses controlled laboratory experiments to compare the economic efficiency and revenue capture of the two leading formats for conducting ad auctions. Third, networks and markets are taken into consideration. This component analyzes a large (approx 1 Terabyte) data set of all transactions in a large virtual economy over about 2 years. It identifies emergent trader networks and goods networks. Fourth, risk preferences and financial market outcomes are studied. This component seeks relevant measures of individual risk preferences, and examines the extent to which measured preferences predict outcomes in laboratory financial markets.
Out of our published results from all components, we highlight findings from the fourth component here, as reported in a forthcoming working paper “Aggregation and Convergence in Experimental General Equilibrium Economies Constructed Using Naturally Occurring Preferences,” by Ryan Oprea, Daniel Friedman and Sean Crockett. A new laboratory experiment measures subjects’ homegrown preferences (as opposed to induced preferences featured in almost all previous market experiments) in a first phase, and uses those preferences to predict general equilibrium behaviour in a second phase of the experiment featuring financial markets. They find that even though the homegrown preferences turn out to be not especially well behaved, the predicted aggregate behaviour in general equilibrium is very well behaved – this is almost the opposite of the famous and very negative theoretical result that even well behaved preferences can aggregate to badly behaved general equilibrium. Even better, the predictions turn out to rather accurately predict market outcomes in the second phase. General equilibrium used to be the centrepiece of economic theory, but it has become neglected in recent decades due to the negative theoretical result and due to a lack of empirical evidence. The authors hope that this paper will help rekindle economists’ faith in general equilibrium theory.
Work planning for the next few years
Two additional components will be developed over the next few years.
1. Imperfect monitoring. Repeated market interaction sometimes involves cooperation (or tacit collusion) even when the actions of other firms can not be observed precisely. In collaboration with the Economics of Change unit, this component is preparing software for new laboratory experiments.
2. High Frequency Trading. Firms spend billions of dollars each year to route orders to financial markets just a tiny fraction of a second faster. Many observers believe that this is wasteful, or unfair, or even dangerously destabilizing. This component is gearing up to test different reform proposals, first in the lab and later in public tournaments. Funding has been secured via a sub award from ERC grant to Axel Ockenfels (University of Cologne).