Study affirms: Two heads are better than one
Yvonne Chiu Hays
It's widely believed that groups make decisions more slowly than individuals. Anyone who has ever sat on a committee might agree.
Two Princeton economists, however, say that maxim is simply not true. They conducted two sets of experiments and found that groups were just as quick as individuals to reach decisions. In addition, they discovered the groups made better decisions on average than individuals.
The results from both experiments are reported in a paper for the National Bureau of Economic Research titled "Are Two Heads Better Than One?"
The question's answer surprised even the researchers. "We thought that there would be disagreement over the right course of action in group settings. We figured that the way the groups would resolve this disagreement would be to wait on making a decision until more data came in," said John Morgan, assistant professor of economics and public affairs at Princeton. "Thus, our (assumptions) were that groups would be slower to make decisions. We also suspected that they would not be significantly better than individuals once they made a decision."
Morgan and Alan Blinder, the Gordon Rentschler Memorial Professor of Economics at Princeton, initiated the study because of the scarcity of economic literature on group decision making. They point out in their paper that the field of economics tends to dote on individual choices: a consumer deciding what to purchase or a banker choosing an interest rate. But many decisions in the real world are made collectively rather than individually, including setting monetary policy in many countries.
When Blinder served as vice chairman of the Federal Reserve Board from 1994 to 1996, he said he came to believe economic models might be missing something important by treating monetary policy decisions as if they were made by an individual. He thought the nature of committees bogged groups down, preventing them from recognizing and reacting efficiently to changes in the economy.
The first experiment was devoid of economic content so prior knowledge or experience could not be construed as a factor. The researchers wanted to "isolate the pure effect of individual vs. group decision making." But the design was meant to evoke the problems faced by monetary policy makers.
Subjects were placed in front of computers programmed with electronic urns consisting initially of 50 percent blue balls and 50 percent red balls. Their task was to sample from the urn and then estimate its composition. They were told that the chance of drawing a blue or red ball would change sometime in the first 10 draws from 50-50 to 70-30. Subjects had to figure out when a change took place and determine the majority color. They were rewarded for their accuracy and speed, which was not measured by clock time but by the number of draws (or additional information) needed before a decision was rendered.
"For decisions of the sort that Alan and I are interested in, such as setting monetary policy, the clock time in meetings is really not the key component of decision delay. Since these meetings are held infrequently, the time of a meeting is dwarfed by the time between meetings," Morgan said.
Because accuracy is vastly more important in setting monetary policy, the experiment was designed to give correct predictions more value than speed in the score. This also discouraged subjects from jumping the gun.
Each session consisted of 30 rounds played individually and 60 played in groups of five. A total of 100 subjects participated in 20 sessions, yielding data on 1,200 group rounds and 3,000 individual rounds.
Communication between subjects, mostly Princeton undergraduates, was permitted only during the group sessions, when students gathered around a single computer. In half of those sessions, decisions were made by majority rule; in the other half, decisions were made unanimously.
Speed the same
The results indicated that there was no systematic difference in speed. Groups did not wait for more information in reaching their decisions. Groups did take more clock time, which was observed by the experimenters but not explicitly measured.
"Part of the reason groups took more clock time is that before each round of the game, there was a bit of discussion about the joint decision. Even if this discussion was only cursory, it was present in the group regime and not in the individual regime. Also, in the groups there would be a little vote on the decision to be taken. This also took clock time," Morgan explained.
The researchers also were struck by how much the groups outperformed the individuals. Groups guessed the wrong color about 11 percent of the time, while individuals made a mistake nearly 16 percent of the time.
Also contrary to prior expectations, the researchers were surprised to find that the rules of interaction didn't slow down the decision process. "If you observed the game without having heard the instructions, it was hard to tell whether the game was being played under the unanimity principle or majority rule," they write in their paper. "Perhaps it was peer group pressure, or perhaps it was simply a desire to be cooperative. But, for whatever reason, majority decisions quickly evolved into unanimous decisions."
No paralysis of analysis
The second experiment required its student subjects to have had at least one course in macroeconomics. In these tests, subjects were asked to steer an electronic model of an economy by manipulating the interest rate.
The results were the same. Groups were not any slower than individuals in reaching a decision and, once again, they produced the superior results.
"The main implication of the study is to call into question the conventional wisdom about the 'paralysis of analysis' that is often thought to afflict group decisions," Morgan said. "Our results suggest that even in complex environments, groups are not paralyzed by indecision. Moreover, groups use the data they have on hand to make timely -- and better -- decisions. If I were running a company, I think this would be a good thing to keep in mind when thinking about how to structure task forces for handling various key projects."
Another implication of the study is the significance of unanimous decisions. "The results for groups where decisions were made by majority are (statistically) identical to those where decisions were made unanimously. This suggests that efforts within companies and within government to carefully specify the rules and protocols of interactions in group settings may sometimes be a waste of time. Moreover, it appears that the cost, in terms of inaction, of seeking consensus is not all that great and may be negligible," Morgan said.
While this study can be broadly applied across industries, there are limits. "For decisions with a strong real-time component, such as directing military operations, I do not think our results apply," Morgan said.
For more information on the study, visit this Web site: <www.princeton.edu/~rjmorgan/>.