Institutions

Institutions are the rules and organizational arrangements that shape how people act together: who holds information, how decisions are made, what behavior is rewarded or punished, and who bears the costs of collective undertakings. Their decisive importance in the book lies in this practical design problem. A group may possess abundant knowledge and share an apparent interest, yet fail to make good decisions or cooperate if its institutions suppress independent judgment, block information, or make contribution less attractive than free riding.

Information, independence, and aggregation

The book’s account of institutions begins with the conditions under which a group can turn scattered information into a sound collective judgment. In an experiment devised by economists Angela Hung and Charles Plott, participants drew marbles in sequence from one of two urns: one held twice as many light marbles as dark, the other twice as many dark as light. Each person privately observed a draw but announced only a guess about which urn was in use. When individuals were paid for their own correct guesses, earlier answers could outweigh later participants’ private evidence, and information cascades began in 78 percent of trials.

Hung and Plott changed the incentive rather than the order of the draws. Participants were paid according to whether the group’s majority answer was correct, shifting the goal from personal accuracy to collective accuracy. Because a student’s own draw could now help the group even when it contradicted preceding guesses, participants attended more closely to private information. Individual guesses could be wrong, but the group’s judgment became significantly more accurate than that of cascading groups. The institutional lesson is not that independent judgments are always correct; it is that a decision procedure can preserve varied evidence rather than allowing early choices to drown it out.

That principle also complicates the apparent case for centralizing intelligence. After failures to share information before the September 11 attacks, Senator Richard Shelby criticized the Federal Bureau of Investigation’s decentralized structure for leaving information in largely independent fiefdoms. The broader criticism was that agencies had produced localized snapshots instead of recognizing the collective significance of information already available. Yet the book cautions that replacing fragmentation with a single commanding authority would not necessarily solve the problem: a small group of decision makers could not be assumed to be wiser than the larger intelligence community whose knowledge they were meant to assess.

The proposed alternative is aggregation without simply abolishing institutional autonomy. Linking agency databases could improve the flow of information while leaving agencies distinct; even two years after September 11, the government still lacked one unified watch list drawing on data from across the intelligence community. A more ambitious proposal, FutureMAP, would have created decision markets in which analysts from different agencies could trade contracts based on expectations about events in the Middle East and elsewhere. Its proposed internal markets, perhaps limited to twenty or thirty analysts and outside experts, might have used classified information to forecast specific events. The intended change was in the incentive structure: accuracy, rather than pleasing a boss, agency, or administration, would be rewarded. The proposal’s second component, the Policy Analysis Market, became the focus of fierce political criticism in the summer of 2003. Together, these proposals show the book’s distinction between central control and information aggregation: institutions can connect knowledge without requiring everyone to surrender autonomy.

Local knowledge and the problem of coordination

Drawing on Friedrich Hayek’s account of tacit knowledge, the book emphasizes that valuable information may be difficult to summarize or transmit because it belongs to a particular place, job, or experience. A person close to a problem may be better placed to recognize what will work. The principle has older precedents: decisions about local festivals in ancient Athens were left to local demes, and regional magistrates handled most minor crimes. The book also invokes the advice Moses received to judge only great matters and leave others to local rulers.

Decentralization therefore has a double character. It encourages independence and specialization, increasing the amount of knowledge available across a system, but there is no guarantee that information discovered in one part will reach the rest. The institutional challenge is to keep knowledge specific enough to remain useful where it originates while making it available to the wider group. The book’s example is Google, whose search results draw on the local knowledge of millions of Web-page operators. The point is broader than any one organization: markets, corporations, and intelligence agencies all need ways to make local information collectively useful without erasing its local character.

The same balance appears in companies. Genuine employee involvement remains unusual, but the book describes evidence that autonomy can affect performance even when workers do not exercise an explicit power. In one study, people working on puzzles and proofreading amid loud, random noise were divided into two groups; one group had a button it could press to turn off the sound. That group solved five times as many puzzles and made fewer proofreading errors, although no one pressed the button. The availability of control over their environment mattered in itself.

Toyota’s production system provides a more operational example. Rather than isolate each worker on a narrow task in a conventional assembly line, it trains frontline workers in multiple skills and organizes them into teams responsible for their production process. Any worker can pull a cord to stop the line when something needs fixing; the cord is rarely pulled. Like the unused button, it gives workers a real avenue for action. Decentralized responsibility can help workers coordinate and find more efficient methods without constant orders and threats, reducing the need for supervision and leaving managers more time for other work.

Incentives, trust, and collective contribution

Institutions also determine whether people have reason to contribute to shared goods. A public-goods experiment by Ernst Fehr and Simon Gächter makes the conflict explicit. Four participants each begin with twenty tokens and play four rounds. A token put into the common pot costs its contributor one token, while returning 0.4 tokens to each of the four players: the group gains 1.6 tokens, but the contributor personally receives only part of that return. If everyone contributes everything, each finishes with thirty-two tokens; if no one contributes, each keeps twenty. Yet each individual has an incentive to hold back and benefit from other people’s contributions. Participants generally begin by contributing about half their tokens, but as they see others free ride, contributions fall; by the end, 70 to 80 percent are free riding, leaving the group poorer than it could have been.

Taxation presents a related institutional problem. Public benefits such as roads and national defense are nonexcludable: people can enjoy them whether or not they personally paid. Even if someone values government services, that structure creates an incentive to avoid contributing. In 2001, the Internal Revenue Service audited only 0.5 percent of returns, so legal prohibition alone could not explain compliance. The book describes taxpayers as contingent consenters: many are willing to pay their fair share if they believe others are doing so and that those who cheat may be caught and punished. When people think others are breaking the rules with impunity, compliance can begin to feel like being taken advantage of. A workable institution must therefore support both the provision of common benefits and confidence that obligations are shared.

Trust has likewise been built through institutional arrangements that made commercial promises more credible. In the eleventh century, Moroccan Maghribi traders used collective sanctions against members who violated their commercial codes. Trade between communities relied on group-wide consequences: if a Genoese trader cheated a merchant in France, Genoese traders generally could face the resulting costs. Merchant guilds, including the Hanseatic League, used collective trade embargoes to protect members against city-states that seized merchant property. These systems were not necessarily fair in every instance, but they could make cross-community trade more practicable by giving trading groups reasons to discipline their own members. Over time, the book contrasts such group-based enforcement with the development in England of contract law and a broader emphasis on individual responsibility for agreements.

Institutional failure and unintended behavior

A policy can benefit an entire industry while remaining too costly for any single participant to improve. Television ratings gathered through paper diaries produced the data used in sweeps, but replacing them with people meters would require substantial investment. The book estimates that equipping homes across the country’s 210 television markets would cost at least hundreds of millions of dollars, while about $25 billion in advertising each year was spent on the basis of sweeps data. More accurate ratings might benefit networks, stations, and advertisers collectively, yet no one participant stood to gain enough to justify paying for the alternative. If each waited for others to fund the change, all could free ride and the inferior arrangement would persist. The case illustrates how institutional inertia can arise not from a lack of shared benefit but from a mismatch between collective gains and individual incentives.

Institutions can also reward concealment rather than honest reporting. In the early 1950s, sociologist Donald Roy worked as a lathe operator in a machine shop where piece-rate pay rose when workers passed production targets and rose again at a second threshold before being capped. Workers knew that unusually high output could lead the company to raise the hurdles. They therefore restricted their pace and concentrated on managing the pay system instead of producing as much as possible. The same logic can operate in corporate targets: when managers receive disproportionate rewards for surpassing a threshold, they may set expectations low and manipulate accounts to meet them, even at the expense of future performance. A rule intended to spur effort can thus make the information on which planning depends less trustworthy.

The book’s institutional account is consequently neither a blanket defense of decentralization nor a call for unified command. Centralized authority may gather information into one place but leave it less independent; decentralization can protect local knowledge while allowing important discoveries to remain isolated. Effective arrangements must connect dispersed information, give participants reason to report and contribute, and ensure that collective benefits do not depend on each person finding it advantageous to wait for someone else. Institutions matter because they shape not only what a group knows, but whether its members can use that knowledge together.