More Than You Know

by Michael J Mauboussin

  In addition, you must populate your mental system with information from diverse sources. While the ability to simulate may be largely hardwired (although you can improve your skills in this area), the pursuit of diverse ideas is within your control.

  Psychologist Donald Campbell describes the situation in similar terms, referring to the process of creative thinking as “blind variation and selective retention.” In other words, creative thinkers seek a variety of ideas but only choose those that are useful given their current goals.

  Idea diversity allows you to find what Johnson calls “weak signals.” A weak signal may be the start of a trend away from the dominant path (such as new technology or development) or the right piece of information at the right time from an unexpected source. In fact, a recent study suggests that informal learning fulfils up to 70 percent of learning needs inside some organizations.8 It’s often difficult to know where the next beneficial idea will come from. The evidence suggests that exposure to diverse information sources can improve the likelihood of finding a useful idea.

  Creativity and Investing

  In a classic article, former Merrill Lynch Investment Managers president Arthur Zeikel argued that superior investment performance requires key personnel within the firm to be creative.9 He suggested that creative people are:• Intellectually curious

  • Flexible and open to new information

  • Able to recognize problems and define them clearly and accurately

  • Able to put information together in many different ways to reach a solution

  • Antiauthoritarian and unorthodox

  • Mentally restless, intense, and highly motivated

  • Highly intelligent

  • Goal-oriented

  Diversity is the fuel for many natural and cognitive processes. Investors that have investment approaches, or information sources, that are too narrow risk missing out on the power of diversity. The downside, of course, is that entertaining diverse ideas means sorting through lots of potentially useless input. But on balance diversity seems to enrich the investment performance—and the lives—of thoughtful investors.

  29

  From Honey to Money

  The Wisdom and Whims of the Collective

  The most thought-provoking feature of a honey bee colony is its ability to achieve coordinated activity among tens of thousands of bees without central authority.

  Coherence in honey bee colonies depends . . . upon mechanisms of decentralized control which give rise to natural selection processes . . . analogous to those that create order in the natural world and in the competitive market economies of humans.

  —Thomas D. Seeley, The Wisdom of the Hive

  [Decision markets] pool the information that is known to diverse individuals into a common resource, and have many advantages over standard institutions for information aggregation, such as news media, peer review, trials, and opinion polls. Speculative markets are decentralized and relatively egalitarian, and can offer direct, concise, timely, and precise estimates in answer to questions we pose.

  —Robin D. Hanson, Decision Markets

  Smart Ant

  In his wonderful book The Wisdom of the Hive, Cornell biologist Thomas Seeley explains that returning honey bee foragers do a little dance to tell their colony mates where the food is. But what’s remarkable is that the duration of the dance reflects not only the richness of the foraging site the bees are advertising but also the colony’s need for the commodity in question. In other words, each bee’s communication dance considers both the colony’s opportunities and its needs. As a result, a bee colony’s overall allocation pattern is appropriate even though no one bee is in control.1

  Ants also demonstrate remarkable collective behavior. Leading ant researcher Deborah Gordon shows that ants place their cemeteries at the point furthest from the colony. But it gets better, because they place their trash heaps at the spot that maximizes its distance from the cemetery and the colony.2 Without awareness, the ants solve a tricky spatial problem worthy of a standardized intelligence test.

  What makes the behavior of social insects like bees and ants so amazing is that there is no central authority, no one directing traffic. Yet the aggregation of simple individuals generates complex, adaptive, and robust results. Colonies forage efficiently, have life cycles, and change behavior as circumstances warrant. These decentralized individuals collectively solve very hard problems, and they do it in a way that is very counterintuitive to the human predilection to command-and-control solutions.

  I look at three systems that depend on collective behavior—social insects, decision markets, and the stock market—and consider the similarities and differences to gain better insights into how markets work. I conclude that collectives are very effective in a host of circumstances, but that there are substantive differences between these systems.

  Traveling Salesman? Follow the Ant . . .

  After describing the workings of a honey bee colony in some detail, Seeley summarizes the main features of colony organization. When scanning this list, consider your notion of how to optimally allocate resources and the parallels between a colony and a market. Main honey bee colony features include:3 1. Division of labor based on temporary specializations

  2. Absence of physical connections between workers

  3. Diverse pathways of information flow

  4. High economy on communication

  5. Negative feedback

  6. Coordination without central planning

  The comings and goings of bees and ants may be a source of fascination, but what can we humans learn from them? Social insect organization may provide useful insight into how to solve a set of problems that are difficult to tackle deductively.

  One example is the famous traveling salesman problem, which researchers consider a benchmark challenge in combinatorial optimization. The goal is to figure out how to route the salesman from city to city using the shortest path possible. Scientists have demonstrated that the ant algorithm—based on ant-foraging patterns—provides as good or better results than more standard approaches.4

  Delphic Decision Markets

  One lesson we can draw from social insects is that the whole is often greater than the sum of the parts. Yet we humans often rely on experts in all sorts of fields, including medicine, politics, finance, and public policy. Do experts give us the best answers, or is there a way to tap the collective knowledge of many individuals?

  Recent years have seen a rise in decision markets, where individuals bet on the outcomes of questions of interest and make or lose money based on whether or not they’re right. These decision markets have proven to be uncannily accurate and, like the social insect colonies, their success relies on distributed intelligence.

  The best known decision market is the Iowa Electronic Markets, founded in 1988.5 The IEM allows for bets on what percentage of the election vote individual candidates will receive. The market’s record is enviable: in the four presidential elections, the IEM’s market price was a better predictor of the election results than the polls (nearly 600 of them) three-quarters of the time. The IEM also hosts other markets.6

  Decision markets have proliferated well beyond the political sphere. Want to gauge the opening weekend box office receipts for a movie? Check the Hollywood Stock Exchange, where traders have been consistently more accurate than movie-industry pundits. This exchange also does a good job of predicting Oscar nominations and currently allows you to bet on future stars from the TV show American Idol.7

  One of the most liquid markets is BetFair, which allows bets on everything from sports to politics to the stock market. Investors can observe—and bet on—the market’s assessment of specific outcomes across a wide range of domains.8

  Why do decision markets work so well? First, individuals in these markets think they have some edge, so they self-select to participate. Second, traders have an incentive to be right—they can take money from less insightful traders. Third, these markets prov
ide continuous, real-time forecasts—a valuable form of feedback. The result is that decision markets aggregate information across traders, allowing them to solve hard problems more effectively than any individual can.

  The Stock Market—the Ultimate Hive?

  Stock markets share many of the same features as social insect colonies and decision markets. Markets emerge from the interaction of many individual investors. We’ve seen that both colonies and decision markets solve problems effectively. To gain better insight into the workings of these systems, we need to look at the differences as well as the similarities.

  Perhaps the biggest differences between the hive and the market are incentives and the role of prices. In a colony, each bee acts not to maximize its own well-being but rather the well-being of the colony (evolution shaped this behavior). In markets, each trader seeks to maximize his own utility. This difference may make colonies more robust than markets because colonies are not as susceptible to the positive feedback that creates market fragility.

  Also, hives do not have prices. Prices are important in a free-market economic system because they help individuals determine how to allocate resources. Bees convey information through their dances, but prices in markets often go beyond informing investors to influencing them, spurring unhealthy imitative behavior.

  Decision markets are also very different than stock markets because they have finite time horizons and defined outcomes. This specificity creates outcome boundaries that effectively limit speculative imitation. In other words, momentum strategies don’t work. Further, in stock markets, the performance of the stock can influence the company’s fundamental outlook.9 In decision markets, the outcome and the market are independent.

  Swarm Smarts

  Investors can draw a few messages from this discussion. First, decentralized systems, even with parts of limited intelligence, are often very effective at solving complex problems. The significance of distributed smarts will continue to rise as we create cheaper ways to harness the wisdom of the collective.10

  Next, while we may be tempted to lump together all decentralized problem-solving systems, important distinctions exist—and those distinctions shape system performance. For example, stock prices tend to be efficient when investors are heterogeneous. But when heterogeneity does not prevail and investor errors become nonindependent, markets become subject to excesses.11 Markets are more prone to excesses than colonies and decision markets.

  Finally, decentralized systems tend to be robust. Despite episodic excesses, markets adapt well to change. This perspective shifts the onus of rationality away from individual investors and suggests that allocative efficiency arises from the structure of the market itself. Market smarts are the result of the aggregation of local information. That’s why it is so hard to beat well-functioning markets.

  30

  Vox Populi

  Using the Collective to Find, Solve, and Predict

  We must show how a solution is produced by interactions of people each of whom possesses only partial knowledge. To assume all the knowledge to be given to a single mind in the same manner in which we assume it to be given to us as the explaining economists is to assume the problem away and to disregard everything that is important and significant in the real world.

  —Freidrich Hayek, “The Use of Knowledge in Society”

  Even when traders are not necessarily experts, their collective judgment is often remarkably accurate because markets are efficient at uncovering and aggregating diverse pieces of information. And it doesn’t seem to matter much what markets are being used to predict.

  —James Surowiecki, “Damn the Slam PAM Plan!”

  The Accuracy of Crowds

  Most investors do not associate group behavior with sparkling outcomes. An Amazon book review crows that Charles MacKay’s classic Extraordinary Popular Delusions and the Madness of Crowds “shows that the madness and confusion of crowds knows no limits, and has no temporal bounds.” Throwing vital problems at a diverse group doesn’t look like an obvious way to generate satisfactory solutions.

  But in recent years social scientists have started to gain a greater appreciation for the information-aggregation ability of markets. Recognition of this ability, when combined with the Internet’s connectivity, has opened up new ways to find solutions to hard-to-answer questions, to solve complex problems, and to improve on predictions.

  Of course, the stock market is no panacea. There’s no question that markets periodically zoom to excesses when investors correlate their behavior. Yet on the whole, most investors and executives don’t realize how and why markets are so good at generating accurate answers.

  Not all collectives operate in the same way. In some situations, the challenge is to find a specific solution—typically knowledge or expertise held by an individual—to a specific problem. Other cases tap groups as information aggregators, where the group’s collective judgment solves a problem or predicts an outcome better than almost any individual can.

  Investors should take note of the accuracy of crowds for two reasons. First, information aggregation lies at the core of market efficiency. Here, I define efficiency as the inability of an individual to systematically exploit the market for superior returns. Second, companies that take advantage of the information embedded in collectives might be able to gain a competitive edge. I’ll describe a few companies that are trying to do just that.

  Needle in a Haystack

  A recent McKinsey Quarterly article opened with the story of a manager at a biotechnology company seeking technical knowledge of a particular protein. After scouring internally for weeks, the manager concluded the expert didn’t exist. Three days later, while in an elevator explaining the problem to a coworker, a woman next to the manager interjected, “I wrote my doctoral thesis on that protein. What do you need to know?”1

  An ability to cost-effectively solve specific research questions is increasingly critical in our knowledge-based economy. Consider the pharmaceutical industry as an example. Research and development investment has nearly doubled as a percentage of sales over the past twenty years, and it costs roughly $800 million to shepherd a drug from development to full FDA approval and rollout.

  Part of the challenge of an R&D-intensive company is to find experts in the lab who can solve tough research problems. Now think for a moment what would happen if a pharmaceutical company could present some of its tricky research problems to all capable scientists in the world, not just those on the company’s payroll. Could the company solve its problems faster? Cheaper? With less risk?

  In mid-2001, executives at Eli Lilly tried to answer these questions by launching a new company, Innocentive (see www.innocentive.com). In 2006, Innocentive had nearly forty “seeker” companies and a community of 95,000 scientist “solvers.” After paying a membership fee, seeker companies post research problems along with a cash reward for the solution. Solvers come from all over, half from outside the United States.

  Does Innocentive work? It may be too early to say, but some early results are encouraging. Take Procter & Gamble, which in 2002 had a $1.7 billion R&D budget and roughly 9,500 R&D employees, including 1,200 Ph.D.’s. Larry Huston, head of R&D, explains that he uses Innocentive because “these are difficult problems we cannot solve inside [the company].” At Innocentive, P&G enjoyed about a 45 percent solution success rate for the first set of its problems, above the one-third target it set.

  P&G’s success underscores the importance of a diverse solver group. Says Huston, “Our first problem was solved by a patent attorney in North Carolina who does patent law by day and chemistry at home by night while his wife reads romance novels—at least that’s what he told us. Our second problem was solved by a graduate student in Spain, the third by a person in Bangalore (India), the fourth by a freelance chemist-consultant.”2

  It’s not hard to imagine other areas where matching problems and solutions might be useful. In spite of some real hurdles like intellectual property rights and improper
dissemination of inside or competitive information, the Innocentive model looks like a great step toward finding the idea needle in the diverse haystack.3

  Weighing the Ox with the Vox

  Creating a market from a collective is another powerful way to aggregate information and solve problems. Here, rather than matching a problem with a unique solution, the group solves a problem better than any single individual—even an expert.

  Victorian polymath Francis Galton was one of the first to thoroughly document this group-aggregation capability. In a 1907 Nature article, “Vox Populi,” Galton describes a contest to guess the weight of an ox at the West of England Fat Stock and Poultry Exhibition in Plymouth. He collected 787 participants who each paid a sixpenny fee to participate. (A small cost to deter practical joking.) According to Galton, some of the competitors were butchers and farmers, likely expert at guessing the weight. He surmised that many others, though, were guided by “such information as they might pick up” or “by their own fancies.”

  Galton calculated the median estimate—the vox populi—as well as the mean. He found that the median guess was within 0.8 percent of the correct weight, and that the mean of the guesses was within 0.01 percent. To give a sense of how the answer emerged, Galton showed the full distribution of answers. Simply stated, the errors cancel out and the result is distilled information.4

  Subsequently, we have seen the vox populi results replicated over and over. Examples include solving a complicated maze, guessing the number of jellybeans in a jar, and finding missing bombs.5 In each case, the necessary conditions for information aggregation to work include an aggregation mechanism, an incentive to answer correctly, and group heterogeneity.