There is a recurring argument in popular discourse that goes like this: if you can predict outcomes by watching a small group of people, those people must be in control. The conspiracy theorist and the cynic share this premise. They differ only in their emotional register.
Network theory breaks the premise.
It turns out that real-world networks — financial systems, social hierarchies, information ecosystems — reliably produce a small number of nodes whose behavior predicts almost everything. Not because those nodes are pulling strings. Because structure itself concentrates predictive power, with or without anyone's intention.
Understanding why requires a short walk through five findings in network and complexity theory. Then we can name what actually happens to causality when you take them seriously.
I. The Rich Get Richer: Scale-Free Networks
In the 1990s, physicists Albert-László Barabási and Réka Albert studied how real networks actually grow. What they found violated the intuitions of classical graph theory, which assumed connections distributed more or less evenly.
Real networks don't do that. The internet, citation networks, financial systems, and social graphs all follow a power law: a tiny fraction of nodes hold a wildly disproportionate number of connections. The distribution has no characteristic scale — hence "scale-free." There is no typical node. There is a long tail of peripheral nodes and a short list of hubs that dominate everything.
The mechanism is preferential attachment. When a new node joins a network, it is more likely to connect to nodes that already have many connections. This is not conspiracy. It is probability. Popular nodes become more popular simply because popularity is visible and connection is cheap.
Consider the American airport system. Hub-and-spoke routing — itself a product of deregulation in 1978 — triggered preferential attachment in the airline network. Passengers connecting through Atlanta or Chicago attracted more routes, which attracted more passengers, which attracted more routes. By the 2000s, a handful of hub airports had become so load-bearing that their weather, their labor disputes, their gate configurations shaped the travel experience of the entire country. A snowstorm in Dallas cascades into delays in Portland and Miami. No one decided this. No aviation authority conspired to make Dallas that powerful. Preferential attachment did it, and then the hub became too structurally central to be challenged. Cities without hubs watched their economies slowly drain toward cities with them — not because anyone chose to disadvantage them, but because the geometry made periphery self-reinforcing.
graph TD
ATL["Atlanta<br/><i>hub</i>"]
ORD["Chicago<br/><i>hub</i>"]
DFW["Dallas<br/><i>hub</i>"]
ATL --- ORD
ATL --- DFW
ORD --- DFW
ATL --- P1["Portland"]
ATL --- P2["Miami"]
ATL --- P3["Nashville"]
ATL --- P4["Charlotte"]
ORD --- P5["Minneapolis"]
ORD --- P6["Detroit"]
ORD --- P2
DFW --- P7["Albuquerque"]
DFW --- P8["Tulsa"]
DFW --- P1
style ATL fill:#5a3a3a,color:#fff
style ORD fill:#5a3a3a,color:#fff
style DFW fill:#5a3a3a,color:#fff
style P1 fill:#3a4a5a,color:#fff
style P2 fill:#3a4a5a,color:#fff
style P3 fill:#3a4a5a,color:#fff
style P4 fill:#3a4a5a,color:#fff
style P5 fill:#3a4a5a,color:#fff
style P6 fill:#3a4a5a,color:#fff
style P7 fill:#3a4a5a,color:#fff
style P8 fill:#3a4a5a,color:#fff
II. The Broker's Advantage: Structural Holes
Sociologist Ronald Burt developed a complementary insight from a different angle. In any network, there are clusters of densely connected nodes — friend groups, industry sectors, academic silos — and between those clusters, gaps. Burt called these gaps structural holes.
The nodes that bridge structural holes become brokers. They sit at the junction between worlds that don't otherwise talk. This position confers two automatic advantages: information advantage, because the broker sees signals from multiple clusters before those signals merge; and arbitrage advantage, because the broker can translate, delay, or reshape what moves between groups.
McKinsey & Company is a nearly perfect institutional instantiation of this dynamic. The firm sits in structural holes between industries, between corporations and governments, between the academic world that generates management ideas and the executive world that implements them. No single client sees across McKinsey's full network. McKinsey does. This means the firm doesn't just consult — it propagates. A restructuring framework developed for a telecom gets carried, via the same consultants, into a hospital system, then into a school district. The broker doesn't invent the ideas so much as arbitrage them across gaps that would otherwise slow their spread.
The result is that McKinsey has shaped the internal organization of more institutions than any single government policy — not through power in any conventional sense, but through structural position. Their influence on how organizations think about headcount, efficiency, and metrics has been total and largely invisible, because it arrived not as a mandate but as advice, mediated through a node that sat between everything.
graph LR
subgraph Academic World
MT["Management<br/>Theory"]
end
subgraph Corporate Sectors
TEL["Telecom"]
HOSP["Hospitals"]
SD["School<br/>Districts"]
end
subgraph Government
GOV["Public Policy"]
end
MCK["McKinsey<br/><i>(structural hole broker)</i>"]
MT ---|"ideas"| MCK
MCK ---|"framework"| TEL
MCK ---|"same framework"| HOSP
MCK ---|"same framework"| SD
MCK ---|"advisory"| GOV
style MCK fill:#5a4a3a,color:#fff
style MT fill:#4a4a6a,color:#fff
style TEL fill:#3a5a3a,color:#fff
style HOSP fill:#3a5a3a,color:#fff
style SD fill:#3a5a3a,color:#fff
style GOV fill:#3a4a5a,color:#fff
III. The Rich Club: Elites Find Each Other
In scale-free networks, hubs don't just accumulate connections — they preferentially connect to each other. This is called the rich-club phenomenon: high-degree nodes form a densely interconnected subnetwork inside the larger network.
The consequences are structural and automatic. Communication within the rich club is faster, more redundant, and less noisy than communication in the periphery. Signals that enter the club propagate quickly throughout it. Signals from outside reach it slowly if at all.
The Basel Committee on Banking Supervision illustrates this cleanly. Formed in 1974 by the central bank governors of ten wealthy nations, the committee has no formal legal authority over any country. It cannot compel legislation. It issues standards — capital requirements, liquidity ratios, risk-weighting frameworks — that emerge from deliberation among central bankers who are, by definition, the rich-club nodes of global finance. These standards then propagate globally, because any country that wants access to international capital markets must be legible to institutions that use Basel frameworks. Basel III's capital requirements shaped the lending behavior of banks in Nigeria and Vietnam, whose regulators had no seat at the table and no vote on the outcome. The rich club did not impose its will through force. The network made the club's outputs the default, and the default became the rule.
graph TD
FED["Federal<br/>Reserve"]
BOE["Bank of<br/>England"]
ECB["European<br/>Central Bank"]
BOJ["Bank of<br/>Japan"]
SNB["Swiss Natl.<br/>Bank"]
BUBA["Deutsche<br/>Bundesbank"]
FED --- BOE
FED --- ECB
FED --- BOJ
FED --- BUBA
BOE --- ECB
BOE --- BUBA
BOE --- SNB
ECB --- BUBA
ECB --- BOJ
BOJ --- SNB
SNB --- FED
FED -.->|"Basel III"| NIG["Nigerian<br/>Banks"]
ECB -.->|"Basel III"| VN["Vietnamese<br/>Banks"]
BOJ -.->|"Basel III"| BR["Brazilian<br/>Banks"]
style FED fill:#6a4a6a,color:#fff
style BOE fill:#6a4a6a,color:#fff
style ECB fill:#6a4a6a,color:#fff
style BOJ fill:#6a4a6a,color:#fff
style SNB fill:#6a4a6a,color:#fff
style BUBA fill:#6a4a6a,color:#fff
style NIG fill:#3a4a5a,color:#fff
style VN fill:#3a4a5a,color:#fff
style BR fill:#3a4a5a,color:#fff
IV. Zealots and the Fragility of Consensus
Here the findings become genuinely strange. In computational models of opinion dynamics — where simulated agents update their beliefs based on the beliefs of their neighbors — researchers introduced a category called zealots: nodes that never update their own state, regardless of what their neighbors believe.
A small fraction of zealots, in an otherwise normal network, can drag the entire system toward their position. Not through force. Not through majority. Simply through stubborn structural persistence while everything around them shifts.
The kosher food certification system offers a compact version of this. A small fraction of American consumers — observant Jewish communities — held fixed, non-negotiable dietary requirements. Food manufacturers who wanted access to that market had to reformulate products to meet certification standards. Reformulating a product line is expensive; maintaining separate lines for certified and uncertified versions is often more expensive still. The economically rational move, once you've reformulated, is to run a single certified line for everyone. The result: the dietary commitments of a small committed minority restructured the supply chains, ingredient sourcing, and manufacturing processes of the American food industry. Today, roughly 40 percent of all packaged food in the United States carries kosher certification — consumed overwhelmingly by people who have no idea what the symbol means and made no choice in the matter. The zealots held. The network reorganized around what held.
A starker version: the Mont Pelerin Society, founded in 1947 by Friedrich Hayek, gathered a small group of economists and intellectuals committed to free-market liberalism at a moment when that position was nearly marginal in mainstream policy discourse. They did not update. For three decades, as Keynesianism dominated, they continued producing papers, funding think tanks, training students, and holding conferences. By 1980, the policy environment had shifted enough — stagflation had discredited the prevailing consensus — that the zealot positions suddenly had traction. Reagan and Thatcher didn't invent the policy framework they implemented. They inherited it from nodes that had simply refused to move while everything else did. The structural stubbornness paid off across a thirty-year time horizon.
V. The Map of Maximum Influence
The final finding is perhaps the most philosophically arresting. In 2003, computer scientists Kempe, Kleinberg, and Tardos posed what they called the influence maximization problem: given a network and a cascade model, find the minimal set of initial seed nodes that maximizes eventual spread through the network.
This set exists. It is computable. It is specific.
For any given network, there is a real answer to the question "which actors are most predictive of outcomes?" And that answer often names a small, nameable group — not because that group controls things, but because the network's structure makes their behavior causally dense relative to everyone else.
VI. The Collapse
Take these five findings together and something strange happens to the concept of causality. The best way to see it is through a single case where all five mechanisms fired in sequence, and where the consequences landed on people who had never heard of any of the nodes involved.
In 1970, Milton Friedman published an essay in the New York Times Magazine arguing that the sole social responsibility of a business is to increase its profits. The claim was not new — but its timing and placement were. Friedman was a zealot node in the technical sense: a position held with total conviction, immune to updating, sitting inside the Chicago School network at the precise moment when the Keynesian consensus was beginning to crack under stagflation. The position had been held for years without traction. Then the environment shifted, and the held position became the available answer.
The Chicago School itself occupied a structural hole. It sat between academic economics and the practical world of corporate governance, law schools, and business education — clusters that didn't speak the same language and didn't read the same journals. Chicago translated. The law and economics movement, seeded largely by Chicago-trained scholars, carried the shareholder primacy framework into legal academia, then into judicial reasoning, without most lawyers or judges understanding they were importing a network theory dressed as neutral principle.
Business schools were the preferential attachment hubs. Harvard Business School, Wharton, Stanford GSB — institutions that already concentrated the most connected students, the most sought-after faculty, the most influential alumni networks. When shareholder primacy arrived in the curriculum, it propagated not as ideology but as the obvious vocabulary of finance. Students who graduated into consulting, banking, and executive roles carried it as a cognitive default. The more graduates used the framework, the more legible it became to the institutions they entered, which attracted more prestige to the schools that taught it, which attracted more students, which produced more graduates carrying the same default.
Michael Jensen and William Meckling's 1976 paper on agency theory — which formalized the idea that managers should be compensated in stock to align their interests with shareholders — was the influence maximization seed. It was not the most-read paper in economics. It became the most-cited paper in business literature, not because it was most correct but because it seeded the nodes most likely to cascade: compensation consultants, board members, proxy advisory firms, and the legal frameworks governing executive pay. Each of those nodes connected to thousands of others. The cascade was structurally overdetermined once the seed took.
The Business Roundtable, the lobbying organization of America's largest corporations, was the rich club. In 1978 it issued a statement endorsing the idea that corporations had obligations to multiple stakeholders — employees, communities, suppliers. By 1997 it had reversed this entirely, declaring that the principal obligation of corporations was to their shareholders. No external force compelled this. The club's internal communication was fast and dense. The updated position propagated through the corporate governance network — board interlocks, shared law firms, shared compensation consultants — faster than any countervailing signal could organize.
By the 1980s, the doctrine was functionally hegemonic. Stock buybacks, which had been restricted as market manipulation, were deregulated in 1982. Executive compensation decoupled from worker wages and tracked share price instead. Corporate planning horizons shortened. Supply chains were hollowed in favor of quarterly returns. A factory worker in Ohio in 1995 experienced the consequences: plant closure, wage stagnation, loss of benefits, community contraction. The closure was not caused by any conspiracy. It was caused by a board applying a framework it had absorbed as obvious, recommended by consultants who had learned it as neutral, enforced by analysts who had inherited it as metric, all of them downstream of a network cascade that had begun with a zealot holding a position in Chicago in 1946.
graph TD
Z["<b>Zealot</b><br/>Friedman / Chicago School<br/><i>1946–1970</i>"]
Z --> CHICAGO["Chicago School<br/><i>broker → law & business</i>"]
CHICAGO --> BSCHOOL["Elite Business Schools<br/>HBS · Wharton · Stanford<br/><i>hub</i>"]
CHICAGO --> LEGAL["Law & Economics<br/>Movement<br/><i>hub</i>"]
BSCHOOL --> JENSEN["Jensen & Meckling<br/>Agency Theory<br/><i>seed · 1976</i>"]
JENSEN --> COMP["Compensation<br/>Consultants"]
JENSEN --> BOARDS["Boards &<br/>Proxy Advisors"]
BSCHOOL --> CLUB["Rich Club<br/>Business Roundtable<br/><i>reversal · 1997</i>"]
LEGAL --> CLUB
COMP --> CLUB
BOARDS --> CLUB
CLUB --> BUYBACKS["Buyback<br/>Deregulation"]
CLUB --> EXEC["Exec Pay Tied<br/>to Share Price"]
CLUB --> SUPPLY["Supply Chain<br/>Hollowing"]
BUYBACKS --> WORKER["Factory Worker<br/>in Ohio<br/><i>1995</i>"]
EXEC --> WORKER
SUPPLY --> WORKER
style Z fill:#4a4a6a,color:#fff
style CHICAGO fill:#5a4a3a,color:#fff
style BSCHOOL fill:#3a5a3a,color:#fff
style LEGAL fill:#3a5a3a,color:#fff
style JENSEN fill:#5a3a5a,color:#fff
style CLUB fill:#6a4a6a,color:#fff
style WORKER fill:#3a5a5a,color:#fff
This is the collapse. The conspiracy theorist, watching the factory close, identifies the board and calls them the cause. The naive structuralist says it's globalization, it's technology, it's nobody. Network theory says something harder: there are identifiable nodes whose behavior predicts this outcome with high accuracy, and none of them were in control.
Friedman didn't control Jensen. Jensen didn't control the Business Roundtable. The Business Roundtable didn't control the compensation consultants. No one sent instructions. The cascade moved through structural position — through zealot persistence, broker arbitrage, hub amplification, rich-club propagation, and a seed paper placed in the highest-leverage nodes. The outcome was as determined as a flood, and as leaderless.
Statisticians distinguish two kinds of causal claims. Predictive causality asks: does knowing X help me predict Y? Interventional causality asks: if I do X, does Y change? These come apart in network systems. The nodes named above are maximally predictive — watch them and you forecast the outcome. They are often minimally interventional — remove Friedman and the network, already primed by stagflation and the structural hole between Chicago and corporate law, finds another zealot. Remove Jensen and another paper seeds the same cascade. The structure outlives any particular occupant, and keeps producing the same kinds of occupants.
The conspiracy theorist confuses predictive power with intentional control. The naive structuralist dismisses predictive concentration because no one intended it. The network theorist holds both: these nodes matter enormously and no one is in charge. That is not a paradox. It is a description of how complex systems actually work.
VII. The Symmetry Test
A structural account earns its credibility by applying across political valence. If the mechanism only illuminates cascades we are already inclined to criticize, it is not a structural account — it is post-hoc rationalization wearing structural clothing. The shareholder primacy case is relatively safe ground: its critics are distributed across the political spectrum and its consequences are legible as harm without much ideological loading. So it is worth asking whether the identical five mechanisms can be traced in a cascade whose political charge runs in a completely different direction.
The spread of gender identity ideology from academic theory into elementary school curricula maps the framework almost perfectly. Judith Butler's Gender Trouble (1990) is the Friedman essay: a zealot position placed into a network primed for it, held with total conviction through decades of mainstream non-adoption, waiting for the surrounding consensus to destabilize. The structural hole brokers were WPATH — translating academic queer theory into clinical language — and the ACLU, translating it into enforceable civil rights frameworks. The preferential attachment hubs were elite education schools and social work programs, followed by major pediatric hospital gender clinics that set protocol for the entire field once they adopted affirmative care. The rich club was dense and fast: Human Rights Campaign, PFLAG, the American Psychological Association, the American Academy of Pediatrics, and the Arcus Foundation connecting them through shared grant relationships, board memberships, and mutually reinforcing citations. The influence maximization seeds were the DSM-5 reclassification in 2013 — which entered the most connected node in mental health practice — and the Obama administration's 2016 Title IX guidance letters, which entered the most connected node in K-12 education: federal funding dependency.
The cascade from there to drag story time in public libraries and gender identity curricula in elementary schools was not directed. Teacher certification programs updated requirements. School library associations updated acquisition guidance. Publishers sought recognition from those hubs. Individual schools, seeking to avoid perceived legal exposure after the 2016 guidance, revised policies without any ideological mandate from above. A parent in Columbus, Ohio encountering a gender identity picture book on their third-grader's shelf in 2019 was downstream of a network cascade that had begun in a Berkeley philosophy department thirty years earlier. None of the nodes knew about the Columbus parent. None of them were in control. All of them were predictive.
The political valences of the two cascades are opposite. The structural logic is identical. The mechanisms don't care about the content — they care about connectivity, commitment, and position. This is what a genuine structural account predicts, and it is the most uncomfortable version of the argument: not that your side was structurally captured while the other side conspired, but that both outcomes arrived by the same route, and neither route required a driver.
VIII. Why Conspiracy Is Comforting
The conspiracy framing is seductive not because it is stupid, but because it is almost right and emotionally complete.
If an identifiable group is controlling things, then the problem is bounded. You know who to confront. You can imagine removing them and having a different world. Agency — theirs, and in response, yours — is preserved.
The network account is harder to inhabit. The hubs are real. Their predictive power is real. The outcomes are real. But the control is diffuse, structural, regenerative. Remove the current occupants of the hub positions and the network, shaped by the same dynamics, will produce new ones. The problem is not the people. The problem is the topology. And you cannot arrest a topology.
This is not fatalism. Topology can change. Networks can be restructured — through antitrust policy, through deliberate platform design, through cultural shifts that alter the attachment dynamics themselves. But those interventions require targeting the structure, not the occupants. And they require accepting a kind of agency that is slower, less satisfying, and harder to narrate than the confrontation the conspiracy framing promises.
Coda: Phenomenological Directionlessness
When causality becomes structural rather than linear — when A produces B which produces C which maintains A — the normal experiential sense of causality dissolves. We experience causes as arrows: something earlier makes something later happen. In complex networks, the experienced arrow is an artifact of where you enter the loop.
If you enter at the ambitious young banker, the arrow points toward the institution. If you enter at the institution, it points toward the political structure. If you enter at the political structure, it points toward the regulatory culture. If you enter at the regulatory culture, it points back toward the ambitious young banker who embodies it.
The system is not directionless in the sense of being random. It is directionless in the phenomenological sense: there is no privileged entry point from which an honest observer can say here is where it starts. The conspiracy theorist picks an entry point and calls it the origin. The structuralist refuses all entry points and ends up saying nothing useful.
The network theorist maps the loop, identifies the load-bearing nodes, and says: this is where intervention is possible, and this is how hard it will be. That is not the same as control. But it is closer to the truth than the alternatives.
And it suggests that asking whether anyone is responsible for systemic outcomes may be the wrong question entirely — not because responsibility is meaningless, but because it was never designed to operate at this scale of structural complexity. Responsibility is a concept built for arrows. Networks don't have arrows. They have topology.
conversation
Comments
Sign in to join the conversation.
No comments yet. Start the thread.