Coordination Without Authority: Zhang Zuolin and the Fengtian Clique

In the summer of 1928, a train carrying Zhang Zuolin threaded the passes northeast of Beijing. He was retreating to Manchuria after another failed bid to hold the capital. The Nationalists were advancing. His coalition was fraying at its seams and its center simultanously, a garment coming apart not at any single stitch but at the idea of stitching. The Japanese, who had financed his ascent with the particular generosity of creditors who expect to collect in territory, were losing patience with his independence.

Somewhere near Huanggutun, the train passed under a bridge. A bomb detonated. Zhang Zuolin died hours later without regaining consciousness.

The man who planted the bomb was a Japanese officer named Komoto Daisaku. He believed Zhang had outlived his usefulness. in a sense he was right, within three years the Japanese would invade Manchuria outright, dissolving the entire substrate Zhang had spent two decades constructing.¹

But for nearly twenty years before that bomb, a former bandit with no formal education had controlled one of the most strategically contested regions in Asia. He commanded 300,000 troops, negotiated with the Japanese as something approaching an equal, and held together a coalition that survived civil wars, betrayals, and the collapse of imperial China.

How?

Zhang Zuolin began his career robbing travelers in rural Liaoning in the 1890s. This was not unusual. The late Qing dynasty was hemorrhaging authority, and the northeast was filling with bandits, refugees, and opportunists of every description. What set Zhang apart was an insight most outlaws missed: legitimacy is cheaper than violence, and violence is only as duarble as the last person willing to absorb it.²

In 1902, rather than fighting the Qing military forces dispatched to suppress him, Zhang offered his services. He rebranded his gang as a local militia. the dynasty, too enfeebled to pacify the region by force, accepted. A bandit became a commander, and the paperwork barely changed. the palimpsest of Chinese governance: scratch the surface of any official title and find the warlord underneath, scratch the warlord and find the bandit, scratch the bandit and find the farmer who ran out of options three harvests ago.

The pattern would recur throughout his career. When the Qing fell, Zhang did not declare independence or join the revolutionaries. He waited, surveyed the new landscape, and allied with whoever seemed most likely to leave him alone. By 1916 he was military governor of Fengtian province. By 1926 he controlled Beijing itself.

The conventional narrative frames this as military conquset, a warlord fighting his way to the top. But Zhang’s army, while large, was never the strongest in China. He won through a different species of warfare entirely.

The Fengtian clique was less a government than a syndicate. Less a syndicate, even, than an ecology with a particular organism at its center, one that the other organisms had not yet found sufficient reason to consume.

The architecture: Zhang at the apex, surrounded by a ring of semi-autonomous governors, generals, and fixers, each administering what amounted to a private fief with a shared postal address. Wu Junsheng controlled the northern railways and the province of Heilongjiang. Tang Yulin held Rehe and managed the Mongolian frontier. Yang Yuting served as chief of staff but also ran his own lattice of Japanese contacts and arms deals, a parrallel foreign policy conducted from the same building. Zhang’s own son, Zhang Xueliang, commanded the modernized units and the fledgling air force, which tells you something about how succession planning worked in organizations like this.³

None of these men worked for Zhang in the way an employee works for a company. They were partners in an enterprise whose articles of incorporation were written in scar tissue and mutual suspicion. They kept most of what they collected in their territories, sometimes 60 or 70 percent of local taxes. They brokered their own deals, cultivated their own clients, built their own power bases. the arrangement had the structural elegance of a protection racket and the emotional valence of a large family at Lunar New Year, everyone performing warmth while keeping careful track of who owes what.

What they could not do alone was survive.

Manchuria in the 1920s was encircled by predators. The Nationalists wanted to reunify China. The Japanese wanted to colonize Manchuria. Soviet Russia wanted influence over the railways. Rival warlords wanted territory. Any single governor, acting independently, would have been consumed in months. The threat environment was omnidirectional, which is exactly the condition under which coordination becomes existential rather than merely useful.

Zhang’s value was that he could hold the center of this, that he could inhabit the liminal position between all these pressures without being crushed by any of them, at least for a while. He could broker deals with the Japanese that no provincial governor could negotiate alone. He could mobilize the collective military strength of the coalition against external threats. He could arbitrate disputes between members before they metastasized into civil war. None of these functions required formal authority. They required something rarer, which is the capacity to make yourself the node through which all information and all trust must pass, and then to survive being that node, which is a form of endurance that looks from the outside like power but from the inside must have felt like a very specific kind of exhaustion.

The governors stayed loyal not because Zhang could punish defection, he often couldn’t, but because defection was expensive, not immediately expensive but expensive in ways that compounded, the way cracks in friable stone don’t matter until the season turns. Leave the coalition and you lose access to Japanese loans. You lose the protection of the collective army. You lose your reputation as a reliable partner, which matters enormously in a world where trust is scarce and contracts are unenforceable. But you also lose something harder to name, a kind of legibility. Inside the coalition, your position made sense, you were a figure in a composition. Outside it, you were just another armed man in a province full of armed men, a shape without a ground. The calculus was not loyalty. It was the terror of becoming illegible, of waking up one morning to find that the story explaining why you mattered had simply, without announcement, stopped being told.⁴

Zhang understood something about power that formal political theory often occludes: authority and influence are not the same thing, and the gap between them is where most of the interesting politics actually lives.

He had no legal claim to Manchuria. Beijing, when it functioned at all, considered him a rebel. The international community recognized the Republic of China, not the Fengtian clique. On paper, Zhang was nobody.

But paper didn’t matter. What mattered was that when Zhang suggested a course of action, people followed. When he withdrew favor from someone, their position became precarious. When he signaled support, attending a banquet, sending a gift, arranging a marriage, careers advanced. Power as a crepuscular phenomenon: visible only in the half-light, in the oblique evidence of who moves toward whom and who moves away.

This is soft power in its most distilled form. Zhang rarely issued direct orders. He didn’t need to. The structure of incentives was such that people anticipated his preferences and acted accordingly. A governor who pleased Zhang received access to loans and protection. A governor who displeased him found his rivals suddenly better armed and his Japanese contacts suddenly unavailable. The mechanism was legible to everyone and articulated by no one, which is probably how you know it’s working, or maybe that’s just how you know nobody wants to be the first to describe it out loud.

The marriages were particularly elegant. Zhang’s daughters married into the Manchurian aristocracy. His sons-in-law received governorships. His grandchildren became hostages to cooperation, though no one would have used that word.⁵ If a governor contemplated defection, he had to weigh not merely political consequences but social ones: Would his daughter be disowned? Would his grandchildren lose their inheritance? Would his in-laws turn against him?

Betrayal was not merely risky. It was embarrassing. And in a culture that prized face above almost everything, embarrassment functioned as a deterrent of surprising force, the whole apparatus held together less by guns than by the threat of awkwardness at dinner, which sounds like a joke but is actually (I think this is the thing people miss about patron-client networks) the deepest structural truth about how they operate.

The system worked until it didn’t, but the “didn’t” is more interesting than the formula suggests and I want to stay with it for a moment before letting the narrative do what narratives want to do, which is collapse into endings.

By 1928 the strains were legible. The Nationalists had unified most of southern China and were marching north. Zhang’s coalition had overextended, attempting to hold Beijing against a superior force. The Japanese, who had financed much of his military expansion, were growing frustrated with his attempts to play them against other foreign powers.⁶

Zhang was attempting something genuinely difficult: to maintain independence inside a closing vise. The Japanese wanted Manchuria as a colony. The Nationalists wanted it as a province. Zhang wanted it as his own domain, beholden to neither. The interstitial space he had occupied for two decades was narrowing to a vanishing point, a lacuna in the map that the surrounding powers had finally agreed, without quite agreeing, to fill.

The bomb at Huanggutun was the Japanese answer to this ambiguity. Komoto Daisaku and his fellow officers believed they could manufacture a crisis, attribute it to the Chinese, and use Zhang’s death as a pretext for intervention. They were premature, Tokyo was not ready for war in 1928, but they were directionally correct. Three years later the Kwantung Army would invade Manchuria, and this time Tokyo would accept the fait accompli.

Zhang Xueliang inherited the coalition. He was young, modernizing, and, crucially, incandescent at the Japanese for killing his father. He aligned with the Nationalists, flying the Nationalist flag over Manchuria in December 1928.

The coalition held for three more years, increasingly hollow, a shell whose occupant had already departed. When the Japanese invaded in September 1931, it collapsed in weeks.⁷ Local governors defected or fled. Revenue streams dried up. The structure that had survived civil wars and foreign pressure evaporated the moment the external threat exceeded the coalition’s capacity to absorb it. Five and a half months to conquer a territory larger than France and Germany combined, defended by a quarter-million troops. The “defense” was a network that had already reached its tipping point, and nobody inside it knew because everyone was still showing up to the meetings, still performing the gestures of a coalition whose interior had already gone to powder.

So the obvious reading is fragility. The system worked until a bomb and an invasion destroyed it. Coordination-based coalitions are resilient until they aren’t. You could write that sentence on a napkin and it would pass for insight at most dinner parties.

But the fragility reading, if you actually hold it up to the light, has to account for the fact that the Fengtian clique lasted nearly two decades. It outlived the Qing dynasty, survived the warlord era’s constant warfare, and produced something resembling stability in one of the most volatile regions on earth. Zhang Zuolin, a bandit with no education and no legitimacy, died controlling more territory than most national governments. Whatever fragility means here, it has to be compatible with that durability, and the people who reach for “fragile” as their verdict are usually reaching for it too quickly, before they’ve reckoned with what they’re trying to explain.

The more honest formulation is that all power is fragile, that formal authority collapses too, ask the Qing, or the Republic of China, or the Soviet Union. The question is not whether your system can survive any conceivable shock, no system can, but whether it is adapted to the shocks it actually faces. Which is a question about ecological fitness rather than structural engineering, though by the time you learn the answer you have usually already learned it the hard way.

Zhang’s system was adapted to civil war, to factional competition, to the intricate choreography of warlord politics. It was not adapted to a modern army backed by an industrial economy, one willing to dispense with negotiation entirely and simply occupy the board.

But then, what would have been? The criticism of the Fengtian clique’s fragility implies the existence of a more robust alternative, and I’m not sure one was available. I keep turning this over and what I find is not a better option Zhang should have taken but a closing of options, a room with fewer and fewer doors that were ever real doors and not just, you know, painted on the wall.

The architecture of the Fengtian clique rhymes with startups, DAOs, any organization that runs on coordination rather than command.⁸ A central figure who does not control resources directly but controls access. A network of semi-autonomous operators who stay aligned because the alternative is worse. Soft power, reputation, the careful cultivation of mutual dependence. The whole thing held together by a penumbra of shared interest rather than a spine of formal authority. The load-bearing structure in organizations like this is invisible until it fails, and that is partly by design, partly something stranger, a kind of occlusion where the very success of the coordination makes the coordination impossible to see. Everyone assumes the coordination is a natural property of the group rather than the continuous output of a specific person’s labor. The coordinator’s job is to make coordination look effortless, which means that when the coordinator disappears, or falters, or simply gets tired, everyone is genuinely surprised that things fall apart. They shouldn’t be, but they always are, and this surprise is itself a symptom of how well the thing was working.

These structures can outcompete hierarchies in volatile environments. They attract talent that would never submit to traditional employment. They adapt faster, move lighter, route around obstacles that would paralyze a bureaucracy. They are also, in a specific and technical sense, more alive than hierarchies, more responsive to signal, more capable of local intelligence. The trade-off is that living systems die, and the manner of their dying is not the slow institutional senescence of a bureaucracy but something more like the way a flock of birds disperses, all at once, a pattern that was there and then simply is not. A hierarchy can survive bad leadership through sheer institutional momentum, the organism outliving its nervous system for a while. A coordination-based coalition lives and dies with the coordinator’s ability to keep everyone at the table. The moment the table stops being the best option, people leave. They do not announce their departure; they just stop showing up, and by the time you notice the room is half-empty the coalition is already a memory, or maybe it was always a memory, the coalition was only ever the story the coordinator told about why everyone was in the same room and once the story stopped being convincing the room was just a room.

Zhang understood incentives better than almost anyone in his era; but understanding is not the same as control, and he did not control them. The Japanese had their own incentives. The Nationalists had theirs. When those external incentives shifted, the internal logic of the Fengtian clique stopped mattering. You can build a system of extraordinary sophistication, a system that accounts for every internal variable, that anticipates defection and prices loyalty and routes around betrayal, and it can still be rendered irrelevant by a change in the external conditions it was never designed to read. The failure is not a bug in the system. It is the system’s boundary condition, the place where its model of the world ends and the actual world, unmodeled and immoderate, begins.

Power is what people willingly give you because the alternative is worse. The catch, the thing that makes this a tragedy and not just a mechanism, is that you don’t control the alternatives, you only ever rented the space between them, and the lease was always revocable by parties who were not signatories and did not care about your terms.

Zhang Zuolin figured this out, eventually. By then his train was already passing under the bridge.