Between Institutions and Networks

Arnold Kling

Winter 2023

Many of our most important institutions are in a state of disrepute. Americans doubt that our political parties are committed to solving problems, that government agencies are transparent and account for the economic consequences of their actions, that media outlets report accurately and fairly, that universities seek to help their students pursue truth, and that corporations focus on their business rather than other concerns. These institutions increasingly strike political poses and strive for self-preservation and self-aggrandizement while neglecting their basic missions.

It was not always thus. In the aftermath of the Second World War, American institutions were basking in glory — and for good reason. Our democracy had conscripted, trained, and equipped over 10 million servicemen. Our factories had produced overwhelming amounts of war matériel. Our military leaders and diplomats had developed a strategy and coordinated a complex alliance for fighting in two major theaters and on various minor fronts simultaneously. And we had met unprecedented challenges in achieving the ambitious goals of D-Day and the Manhattan Project.

Some of this large-scale institutional success carried over into the post-war period. We built the interstate highway system and landed a man on the moon. Another post-war project under the auspices of the U.S. military — which initially attracted scant attention — was an attempt to build a communication network that would connect all computers. The network was designed to be capable of surviving an attack that took down any single node.

The success of this project, which evolved into the internet, ironically coincided with, and probably contributed to, the decline of institutions in the 21st century. Today, some commentators contrast networks with institutions, and they see the comparison favoring networks. One prominent network enthusiast, Balaji Srinivasan, has gone so far as to proclaim that existing government functions could be better performed by a "network state," as he termed it in a book by that title.

Will networks replace institutions in the coming decades? The answer might vary by the type of function we are considering. Institutions and networks differ in their intrinsic strengths and weaknesses. It's plausible that a network could emerge to replace institutions of higher education, for example, but much less likely that a network will supplant government's role in securing property rights, enforcing contracts, resolving disputes, or bolstering confidence in the financial system. By defining and evaluating networks and institutions, we can come to a better understanding of the best uses of both, and how they might complement each other.


To begin, it's important to clarify the terms involved. How should we distinguish a network from an institution?

For one, an institution can be characterized as a set of roles. "Assistant manager of a grocery store" is a role, for example. Institutional roles are explicitly defined, often in writing: The store has a job description for the assistant manager. This description gives the assistant manager a set of responsibilities as well as authority concerning some decisions and employees, but it also sets limits on his authority. Putting someone into a role or removing him is costly; the store puts time and effort into hiring and training the assistant manager. Once the assistant manager is on board, the store does not want him to leave or to have to fire him.

Institutional roles are also regular and persistent. We expect the assistant manager to be at a particular place doing specific work at a regular time, and we rely on him to stay on the job until the workday has officially ended. Moreover, the assistant manager is expected to return the next workday, the workday after that, and so on. It is sometimes relatively easy to obtain a role — a potential assistant manager need only apply for the job and be hired — but oftentimes it is difficult and expensive; it takes years of school and training to become a doctor, for instance.

Networks, by contrast, are a set of interpersonal connections. Network connections are not defined in terms of responsibility and authority; there is no written agreement stating what a Twitter user owes a follower or vice versa. Such connections are easily added and dropped: All it takes to follow or unfollow someone on Twitter is a click of a mouse or a tap of a phone screen. One can sign up for Twitter, Facebook, and LinkedIn in a few minutes, though building up a useful set of connections does take more effort.

Network connections are also intermittent and ad hoc. Suppose you want to arrange a tennis match with your friend Bob because you have a connection with him. Bob may be available to play tennis at 3 p.m. on Tuesday, but you cannot just count on him to be there; you have to contact him to arrange a meetup. In this sense, a tennis match in the physical world resembles an institution. An internet-based computer game, on the other hand, is not subject to the same limits: Users can sign on and begin playing at any time, without having to coordinate with friends.

Another way to differentiate between networks and institutions is to think about different types of human capital. Gary Becker, the economist whose theory of human capital helped earn him a Nobel Prize, drew a distinction between what he called "specific human capital" and "general human capital." Specific human capital is knowledge that is useful within a particular firm, while general human capital is knowledge that is helpful in any setting. One of my professors once joked that "specific human capital is knowing the location of the bathroom. General human capital is being able to read the sign on the door."

Consider an office assistant in a medical practice. This assistant plays a particular role, which requires familiarity with the patient-records system, scheduling and billing practices, the way other duties are performed in that office, and even the preferences and personalities of the doctors and nurses. This is specific human capital that has value only in that medical practice. On the other hand, the assistant's understanding of computers and medical terminology is general human capital that could be useful in many organizations, including hospitals and insurance companies.

Institutional roles build specific human capital. In a unique role in a particular organization, an employee can accumulate a lot of specific human capital. This encourages mutual loyalty between the organization and the employee. Over time, the employee will be better paid playing that role in that particular firm than another role at a different firm, and he will be discouraged from upsetting the relationship by violating the firm's norms and expectations. By the same token, the employer has invested a significant amount of time and effort in training and acculturating that individual as an employee. It, too, will likely want to maintain the relationship rather than spending time and money hiring a replacement.

To the extent that network connections extend outside a particular corporation, those connections represent general human capital. Journalists who compose their articles using sources obtained on the internet are relying more on general human capital, while old-fashioned shoe-leather reporters who were trained, equipped, and funded by their outlets to go out into the field to cover a story tended to rely on specific human capital. In addition, a journalist who has built a personal brand on Twitter has more general human capital than an old-fashioned reporter whose credibility is tied to the reputation of the newspaper for which he works.

Yuval Levin has observed that institutions are "formative," while individuals in the internet era tend to be "performative." That is, nowadays many individuals focus on building their own personal followings while paying scant attention to the norms, customs, or even the mission of the organizations to which they belong. Levin's distinction is closely related to the contrast between specific and general human capital. Individuals who shape their behavior according to their roles in institutions are building specific human capital, and in return, the institution rewards those who absorb its lessons and conform to its expectations. Individuals who use their organizational ties as platforms to perform before a larger audience on a network, on the other hand, are scorning specific human capital and instead seeking to obtain and exploit general human capital. The connections and reputation that one acquires using an open network have value that accrues to the individual independently of any specific organization to which he belongs.

Because institutions nurture specific human capital, they are inherently formative. As a set of connections, networks facilitate and reward individuals' investment in general human capital. Networks thus inherently encourage behavior that is performative.

A final divergence between networks and institutions relates to purpose. Institutional roles are part of an overall structure that is designed to accomplish a particular mission. The grocery store's mission, for example, is to supply customers with food products. Network connections, by contrast, are agnostic with respect to mission. On Facebook, I have some friends who share life events, such as the births of grandchildren. Other friends share videos and news relating to mutual hobbies or political commentary. The network itself does not discriminate between these purposes; it simply facilitates users' efforts to pursue their preferred ends.


Considering the differences between the internet and earlier communication methods also helps illuminate the distinction between a network and an institution. The older technologies behaved more like institutions, while the internet is the prime example of a network.

The internet is based on a novel communication method called "packet switching." Previous technologies, including radio, television, and the traditional telephone system, connected the sender and receiver for as long as communication was taking place. Internet connections, on the other hand, open and close many times when sending a single message.

In the traditional phone system, a point-to-point connection, or circuit, is opened between the two parties talking on a call. The circuit remains in use until the parties hang up. If a complete circuit cannot be opened, the call is not completed and the caller receives a busy signal.

With packet switching, there is no ongoing connection between sender and receiver. Instead, a message is digitally encoded and broken into small packets. The packets are relayed to the recipient through devices called "routers," taking whatever path is most efficient at the time. At every step in the relay, the connection lasts only long enough for the packet to be delivered to the next router. Different packets from the same message may pass through different routers to reach their destination. Once the packets reach their final destination, they have to be opened and rearranged in order to form a coherent message.

Because of the many connections that need to be opened and closed when transmitting information over a packet-switched network, the network makes intensive use of routers through dedicated computers. As of 1990, when computer hardware was more expensive than it is today, routers were costly, and packet switching was less affordable relative to other communication technologies. But as the cost of computing power declined exponentially, the price of routers also plummeted, and packet switching became less costly than other technologies. For phone calls, packet switching became cheaper than the old circuit-switched phone system sometime in the early 2000s.

Internet packets of the 1990s also had difficulty going the "last mile" between internet service providers and households. It takes many packets to transmit a voice message or a video, and using copper phone lines largely restricted internet messages to text. Yet as cable companies and local phone companies built out fiber-optic broadband connections, the last-mile problem dissipated. Subsequent improvements in cell-phone technology enabled smartphones to enjoy high-speed internet over the airwaves.

These technological developments ushered in a new era in communication. Before the internet, communication technologies specialized in the type of content they delivered. The traditional phone system delivered conversations; broadcasters transmitted audio; TV stations transmitted video. In contrast to these older methods, packet switching is agnostic about what is inside each packet. Only when the packets are opened on the receiving end does the receiving device learn whether it is acquiring text, audio, or video.

Similarly, older technologies were limited to specific uses. Phones and radios could only be used to transmit audio. Televisions could transmit audio and video, but viewers could not use them to transmit messages to one another. Internet users, meanwhile, are not constrained to using particular devices for specific purposes. A computer or smartphone does not play as unique a role as a radio, a TV, or a landline phone. Depending on how one uses the device at a particular moment, one can receive phone conversations, audio-based entertainment, text messages, or video.

The use of older communication technologies also followed regular schedules. On television, one would watch the same program at the same time every week. If an individual wanted to watch the show later with a friend, he had to make sure his recording device was on while it was being broadcast. If a phone call was made and the other person did not answer, there was no way to leave a message; the answering machine was exotic technology when it first appeared on the scene.

Internet communication, by contrast, is inherently asynchronous. If a person wants to watch a movie or engage in a live conversation, he can do so at any time. Sophisticated software and high-speed connections are employed to create the illusion of a continuous connection.

The internet also dramatically lowered the barriers to entry for many forms of communication. People today can make an overseas phone call without dreading high long-distance fees, distribute audio or video without owning a radio or TV station, and enjoy the freedom to distribute opinions without having to own, or even rent, a printing press. The packet-switched network thus enables us to achieve much more in the realm of communication by using our general human capital. By the same token, the relative value of specific human capital, such as that which one might obtain while working at a printing press or TV station, has declined.

Finally, with older technologies, central authorities governed use. Early telephones required switchboard operators to function. The manager of a radio station decided what to broadcast to entertain listeners, as did a TV executive with respect to television.

The internet did away with many of the centralized authorities that older technologies required to operate. Relative to the older technologies, the internet allows much wider participation and greater experimentation. It is hard to start a TV station; it is hard to stop someone from sending video over the internet.

The differences between institutions and networks are thus much like the differences between the internet and older technologies. Radio, TV, and traditional phone systems were institutions that featured defined roles of responsibility; high up-front costs; regular, persistent, and mission-oriented communication; development of specific human capital; and governance by central authorities. The internet, by contrast, is a network built on connections that are flexible; low-cost; intermittent and ad hoc, as well as easily added and dropped; agnostic in mission; focused on general human capital; and decentralized, allowing for more participation and innovation.


Given the differences between networks and institutions, how might we evaluate their positive and negative aspects? Will networks supersede institutions in the internet age, or do the latter still have a vital role to play in our politics and society?

When compared to networks, institutions have two advantages: positions of authority and mechanisms of accountability.

Authority is useful for settling disputes and resolving conflicts peacefully, as well as for deterring misbehavior. A dissatisfied customer can speak to a manager to better resolve a conflict; a judge can help opposing parties resolve a civil dispute; a criminal can be prosecuted by the justice system to deter crime. Authorities also have the power to forestall or forbid activities within their purview.

An important complement to authority is accountability. Mechanisms of accountability help ensure that people in authority will endeavor to commit as few errors as possible. Most institutions have accountability mechanisms, which can be internal or external. Examples of internal mechanisms include performance reviews, compensation systems, and scorecards, while external accountability mechanisms include government regulations, external audits, shareholder activism, and the profit-and-loss system of market economies.

Institutions work best when individual accountability and authority are proportional. It is usually unfair to hold someone accountable for accomplishing an assignment without giving him the authority needed to do so. On the other hand, giving someone authority without accountability invites abuse of power.

Unlike institutions, networks do not create positions of authority. No internet courts exist to settle disputes; no police force patrols the internet for cybercrime. This lack of authority creates problems. For example, unsolicited commercial emails — otherwise known as "spam" — have plagued internet users for nearly three decades. Computer malware can strike anyone at any time. Computer-security specialists try to combat these problems, but they are beyond the scope of any single institution to address.

On social-media networks, it is impossible to settle disputes about what speech should be permitted. In the absence of a definitive authority, the most effective censorship in practice comes from spontaneous mob behavior. People will act collectively to damage the reputation of those deemed deserving of punishment. This process is not predictable and follows no formal procedure.

The two advantages that institutions have over networks combine to create a system that is especially conducive to establishing and enforcing property rights. In the physical world, strong property rights are associated with capable institutions governed by authority figures operating under mechanisms of accountability.

Years ago, the economist Hernando de Soto pointed out how the absence of strong property rights makes it difficult to accumulate capital. With secure property rights, someone with a small parcel of land can build improvements on it and borrow against it. But a squatter in a shantytown cannot accumulate capital relative to his dwelling.

Establishing and enforcing property rights on the internet has long been a challenge. For digital content, the default is free availability. Music publishers have gone to court to try to stop file-sharing services, while newspaper publishers have put up paywalls to ensure their journalism is protected and compensated. In addition to the trouble and expense involved in trying to limit content use, publishers are put in the unenviable position of deliberately limiting their potential audience.

Software engineers have long promised to make it possible for internet users to create digital content while retaining control over access. Initially this was referred to as "digital-rights management"; more recently, the discussion has focused on non-fungible tokens, or NFTs. The goal of these efforts is to create property-rights systems that run themselves, without requiring an authority to settle disputes.

The ability of networks to supplant institutions may depend on whether property rights can truly be secured on a network. The Web3 project seeks to develop decentralized web-software protocols in the hopes of delivering a "decentralized and fair internet where users control their own data, identity and destiny." It won't fulfill the promise of its boosters, however, if it turns out to be nothing more than a digital shantytown.


Institutions do many beneficial things: They develop specific human capital, create positions of authority with mechanisms of accountability, and facilitate the protection of property rights. Their relative weakness, however, concerns adaptation to new circumstances. Indeed, the institutional virtues of continuity and reliability can become defects when the environment changes rapidly.

As time goes on and circumstances change, institutions suffer from decay. They tend to resist useful innovations because those updates might change or eliminate established roles. They opt instead for innovations that add authority to existing roles, regardless of whether that authority can be put to good use. And as people within institutions learn to "game the system," institutional mechanisms of accountability become less effective over time.

Consider universities. Since the advent of the internet, it is no longer necessary for every university to offer a course teaching calculus; many campuses could simply outsource calculus courses to a specialized teaching organization. But doing so would eliminate internal roles in math departments.

In recent decades, the growth in administrative roles at universities has outpaced the increase in faculty and students. Once a given position — say, a dean of student life, a vice president for diversity, or a director of the Institutional Review Board — is created, the person who holds that role typically identifies problems that require more authority and a larger staff.

As employees figure out how to take advantage of reward systems, the ability of an organization to carry out its mission also deteriorates. A compensation system, to take one example, is supposed to reward effort and results. But the goal of the employee is to acquire the highest pay for the least trouble. As time passes on and employees figure out how to obtain remuneration while avoiding unpleasant work, the incentive system becomes less effective at achieving the results that management desires.

In the case of universities, faculty are rewarded for publications. This goal is best served by working on short-term projects with a high probability of success. But focusing on this type of research also means avoiding investigations that might take longer, or go against political orthodoxy in the field. It means not spending time getting to know students, because this could detract from completing research. The incentive system of the university thus begins to undermine its mission of free inquiry and helping students pursue truth.


As outlined above, the problem of institutional decay is particularly apparent in higher-education institutions. Structuring universities like networks could make them more innovative and more responsive to student needs.

The main reason students compete to attend prestigious colleges is to obtain the certification conferred upon graduation. This high-status degree positions the graduate to do well in the job market. A network university might replace the system of formal credentials — grades and degrees — with letters of introduction and recommendation from faculty members. These referential letters, relayed through the faculty's network of connections, would enable students to obtain jobs.

To be useful to students, faculty would need connections with employers. In fact, many faculty might already be employed elsewhere and teach part time. A firm, meanwhile, might want to have its employees teach in a university network so as to better identify talent and recruit from the pool of students. A banker could teach a finance course, but the banker might also instruct a course in literature based on a side interest. Faculty would put most of their effort into selecting topics for courses and developing exercises for students to participate in. They would put less effort into delivering content, which nowadays is readily available on the internet.

For faculty to provide meaningful letters of introduction and recommendation, they would have to work closely with students. To facilitate this, each faculty member would have only a few students at a time. Courses would work like seminars, not lectures. Tests would not be necessary; student writing, projects, class discussions, and problem sets would provide sufficient opportunities for students to develop and demonstrate their skills.

Of course, students need to connect with one another as well as with faculty. Since most classes at a network university would likely be taught remotely, students would see each other at online seminars and communicate with one another by email and chat. But to enhance connections, network campuses could hold in-person conferences several times a year with, say, a thousand students and 200 faculty members. The conferences would not involve lectures or panel discussions, but would rather be designed to facilitate interaction among the participants.

This network approach would suit students who have the maturity and self-discipline to learn on their own with the guidance of faculty as mentors and advisors. It would also aid students from other countries, who may otherwise find it difficult and expensive to come to the United States for their education. This approach would be less appropriate, however, for young people who attend college to root for sports teams, attend frequent parties, or postpone adulthood.

Beyond academia, are there other areas that would benefit from operating more like a network? Many consider government institutions to be decaying, sclerotic, and ripe for a fundamental restructuring. Could a network replace government?

The short answer is "no." Communication networks can help expose government shortcomings and galvanize opposition movements, but we cannot do away with roles, responsibilities, and accountability in the political realm.

In The Revolt of the Public, Martin Gurri pointed out how the internet has made it much more difficult for public officials to cover up mistakes and foibles. We find out when politicians who impose mask mandates go to events unmasked, for instance. Furthermore, social-media platforms facilitate the assembly of mass demonstrations, as seen during the Arab Spring, the Black Lives Matter and Yellow Vest movements, and recent protests by truckers in Canada and farmers in Europe. But as Gurri noted, these movements are merely expressions of anger. Without organizational structure and a specific agenda, they achieve few, if any, tangible reforms.

People who make payments and save money for the future need the security provided by financial regulation. Unregulated financial services are not always used for criminal activity, but they are usually adjacent to it. Reliable banking is bound to be buttressed by government, for reasons of security and capacity. One reason to be cautious about proclaiming that the future of finance is some Bitcoin-like monetary technology is that the transactions capacity of the decentralized Bitcoin ledger is orders of magnitude less than that of today's global financial exchanges.

It turns out that making social policy, including defining and protecting property rights, requires coordination and planning beyond the capabilities of networks. Institutions in general, and government institutions in particular, are essential for modern society.


Networks and institutions operate best in different social environments. Institutions, including the government, are reasonably well adapted to the world of the physical. Networks, on the other hand, may be better suited to the world of the virtual.

Consider two extreme thought experiments: a primarily physical world in which the virtual is unimportant, and a primarily virtual world in which the physical is unimportant. In the physical world, there are no computer networks. Machines might help to calculate and store information, but they cannot communicate with one another or aid in human communication. In that sense, the primarily physical world is like the pre-internet era. A primarily virtual world is somewhat harder to imagine, but in a purely virtual realm, all of our important activities and relationships would be embedded in a computer network — in what science-fiction works (and Meta CEO Mark Zuckerberg) have dubbed the "metaverse."

In the physical world, geographic location and material goods play large roles in people's lives. The people nearby matter much more than individuals farther away. Goods are mostly what economists call "rivalrous" and "excludable": Two people cannot eat the same hamburger; if one person has the hamburger, he can prevent another from eating it. The economy is driven by decisions about investment in physical capital, with these investments overseen by corporations, banks, and government regulators.

Everything is the opposite in the primarily virtual world. Close friends and associates may reside thousands of miles away from where one lives. Digital goods can be duplicated and shared at essentially no cost. Instead of physical capital, the digital economy is driven by creativity and innovation, which emerge out of a decentralized process of trial and error. The borders that define states would be considered anachronistic, while large organizations and cumbersome regulations would be seen as counterproductive.

The challenge our society faces today is that we find ourselves between these two worlds. The virtual world is becoming increasingly important, but the physical world is still very much with us. There are aspects of society in which physical and virtual activity are both present.

The changing balance between the physical and the virtual may prove to be a constant source of tension in our time. There will be mutual resentment between elites who are primarily comfortable in institutions and elites who are primarily comfortable in networks, for example. How might we navigate these tensions and capitalize on the strengths of both networks and institutions?

Network leaders, for their part, should be humble about the promise and limits of their technologies. Networks offer a tremendous opportunity for breaking down barriers to participation and spurring innovation, and they could transform institutions like higher education for the better. But they lack the authority, accountability, and coordinated planning of government institutions. Other institutions, both religious and secular, remain vital for human flourishing.

Institutions have their own temptations to avoid. As the virtual world expands, leaders of institutions can become frustrated. Their first instinct is to regard networks as a threat. Journalists, educators, and political professionals may resent the "outsiders" who are amplifying complaints and sparking protests. They will be tempted to try to shut down the opposition. But they would do better to focus on the problems of institutional decay, to work on improving incentives and accountability within their own organizations, and to form better, mission-oriented leaders. If institutions want to be trusted again, they must become trustworthy themselves.

Arnold Kling is an adjunct scholar with the Cato Institute and a senior affiliated scholar at the Mercatus Center at George Mason University.


from the


A weekly newsletter with free essays from past issues of National Affairs and The Public Interest that shed light on the week's pressing issues.


to your National Affairs subscriber account.

Already a subscriber? Activate your account.


Unlimited access to intelligent essays on the nation’s affairs.

Subscribe to National Affairs.