Economics is often called the science of human choice. As economists, we spend a great deal of time observing, explaining and predicting the choices people make. Let’s assume (and I hope this is true!) that economists also care about how ‘important’ a choice is in some economic sense, such as the number of people affected or the amount of other people’s money involved. We will see that many of such important choices are made in groups. Think about how many important decisions are taken by political parties and committees, corporate boards, central banks, families or groups of friends.
Funnily enough, most of what scientists know about choices made by groups is from research in psychology, not economics. But things have been changing since the late 1990s. Much of the ‘new’ research on choices by groups is done by experimental and behavioural economists.
I first got interested in groups at the start of my PhD. I was planning to work on financial herding (not a very behavioural topic – many of the models assume good old common rationality) but instead I was drawn to the research on groups. Looking back through my old notes, my choice was strongly influenced by a couple of experimental papers at the time: “Are Groups More (or Less) Consistent Than Individuals?”by John Bone, John Hey & John Suckling (1999, JRU) and “Are Two Heads Better Than One?” by Blinder & Morgan (2005, JMCB).
Blinder & Morgan looks at how groups perform compare to individuals with respect to adjusting their decisions based on information they receive. They do an experiment on how well groups respond to information in a ‘monetary policy game’. The objective of the game is to stabilise a (simulated) economic system by distinguishing everyday fluctuations from permanent economic shocks. If you think about it, distinguishing information from noise is an essential skill in many environments. Inspired by Blinder’s time on the US Monetary Policy Committee (and his apparent frustrations due to slow group decision making) they wonder how groups compare to individuals. Contrary to their expectations, they find that groups do significantly better than individuals. Although groups and individuals take the same amount of time to decide, groups are much more accurate in the task.
Bone and co-authors look at whether groups make different decisions when it comes to risky outcomes, or more precisely, whether group choices are more consistent with Expected Utility Theory than individual choices. They are interested in finding out whether, even though there is plenty of experimental evidence that individuals do not act as “expected utility maximisers”, this might change when people choose in groups. Maybe group interaction makes people see things differently, bringing them more in line with the theory. They find, however, that groups are just as inconsistent as individuals. One thing that struck me about their paper is that groups choose options with greater expected outcomes than individuals do. The authors don’t make much of this result but I think it is something rather fundamental about groups: they are more comfortable taking ‘calculated risks’ than individuals.
I think it is something rather fundamental about groups: they are more comfortable taking ‘calculated risks’ than individuals.
Some of my own work – with Spiros Bougheas and Martin Sefton, my thesis supervisors at the University of Nottingham – connects quite nicely to the results of the papers mentioned above. But instead of focussing on the decisions of groups themselves, we looked at how people respond when they choose individually, but in a group environment. This brings out some subtle and surprising results. One of our conclusions is that one of the key elements of deciding in a group – communicating with others – has some unexpected consequences.
Sharing information with others
Just like Blinder & Morgan, we find that groups are generally good environments for sharing and interpreting information. In our experiments, people in groups are good at incorporating different pieces of information about probability into their decisions. But there is more: more than half of the people in our experiments do not equally weight information they discover themselves and information they learn from other group members.
Our results would make perfect sense if people were simply suspicious about the information reported by others. But interestingly, when we let people in groups communicate freely with one another, the majority of people actually put more weight on the information received from others than on their own information. There are two ways of interpreting our results: (1) people either put ‘too much’ weight on the information received from others, or (2) group communication teaches them to adjust for having put too much weight on their own information in the first place. In either case, our experiment is one of the first papers showing such a strong effect of sharing information with others.
Consulting with others
In another experiment, we looked at the difference between groups that make consensus decisions and groups whose only function is consultation – letting people talk to one another about their individual decisions. Our experiment was based on two papers by Matthias Sutter (2007 Econ Letters, 2009 AER) in which groups achieve higher expected earnings by taking more risk than individuals take (as in Bone, Hey and Suckling’s paper I described above).
We confirmed the results of Sutter, and we even found evidence in the chat logs of the communication between group members that talking about expected earnings and higher investment are related. Next, we tested whether consultation had the same effect on risk taking as group discussion – and found it doesn’t. It turns out that, when people have no financial incentive to talk to others, they don’t talk about expected earnings. Even when we gave people specific round-by-round feedback on how the others in their group were doing, consultation did not lead to more risk taking. The only effect of consultation that we found is a slight convergence between the decisions of individuals – talking brings your decisions closer together.
The experiments I have talked about paint a fairly positive picture of groups, provided group members have an incentive to exchange information with each other. But many of us still have plenty of frustrating experiences with group decision-making in real life. What can research tell us about those? One important thing that we know very little about is how groups and teams learn. Under what circumstances do they improve their choices over time? The experiments above don’t tell us much about that. We are currently working on experiments that investigate these types of questions. As always, more research is needed..