The
Incentive
Audit

A surgeon in Lincoln, Nebraska spent years sending bushel baskets full of normal gallbladders down to the pathology lab at the leading hospital in town. Not diseased gallbladders. Normal ones. Healthy organs removed from patients who didn’t need the surgery, by a doctor who was paid per procedure. The hospital, operating with the kind of permissive quality control for which community hospitals have always been famous, let it continue for years before removing him from the medical staff.

Charlie Munger grew up in that town. The surgeon was a family acquaintance. Munger watched the whole thing unfold, and from it he drew a conclusion that became the first principle of his analytical framework: incentive-caused bias is the most powerful force in professional life, and the person it corrupts most completely is the person doing the corrupting. The surgeon didn’t think he was removing healthy organs. He thought he was being thorough. The compensation formula had rewritten his clinical judgment so seamlessly that he couldn’t detect the distortion from the inside. Intelligence offered no defense. Neither did good intentions. Neither, for that matter, did a medical degree. This is the dark magic of misaligned pay: it doesn’t make you dishonest. It makes you wrong, with total confidence.

Ben Franklin put the operating principle more bluntly three centuries earlier: “If you would persuade, appeal to interest, not to reason.” Munger quoted Franklin constantly because Franklin had identified the load-bearing insight in nine words. Reason is what people use to explain their decisions. Interest is what actually drives them. The gap between those two forces is where most organizational failures live, where most negotiation mistakes are born, and where most strategic blind spots fester until they metastasize into something expensive. Evolutionary biologists have a term for organisms that signal one thing while doing another: mimicry. The corporate version is called a mission statement.

This is the Incentive Audit. It runs on a single diagnostic question: What is every party in this situation actually paid to do? Not what they say they value. Not what they intend. What the architecture of payments and penalties makes almost inevitable. The answer to that question will explain conduct that talent, character, and intelligence cannot.

And like the Inversion Stack, it comes in four modes.

When the original Medicare law was passed, various experts, including Ph.D. economists, forecast its costs using simple extrapolations of past healthcare spending. Their projections landed at less than 10% of what actually happened. A tenfold miss, by people whose entire professional identity was built on the ability to predict costs. The error was not arithmetic. It was conceptual. The economists projected costs as if conduct would remain unchanged after the pay design transformed. Once Medicare installed new rules—patients insulated from costs, doctors paid per visit, hospitals paid for volume—conduct shifted to match the new payments, and the numbers became unrecognizable. This is the same mistake a Newtonian physicist makes when ignoring relativistic effects: the model works perfectly at low speeds and catastrophically at high ones. The Medicare economists were doing low-speed projections on a system that had just broken the sound barrier.

Economists have a name for this phenomenon. The Lucas Critique holds that any model calibrated on historical data will break the moment the underlying policy changes, because the policy change rewrites the conduct the model was built to predict. Munger’s version is simpler and sticks better: “Once they put in place all these new incentives, the behavior changed in response to the incentives.” Every operator who has launched a new pay plan and been surprised by the results has collided with this principle. Every one. The surprise is always the same: we changed the rules, and people changed what they did. As if that outcome were somehow unexpected.

Dan Ariely documented the same mechanism in corporate governance. The SHE Index in the United States counts the percentage of women on corporate boards and in senior positions. What you actually want to know is how women feel working at that company. But feelings are hard to measure. Board composition is easy to count. So companies optimize for the easy metric, check the box, and change nothing about the actual experience. The result: the index systematically underperforms the S&P 500 by a significant margin, because the payment created by the metric is to perform diversity rather than practice it. This is the corporate equivalent of studying for the test instead of learning the material. The transcript looks great. The education is hollow.

Charles Goodhart, the British economist, formulated the principle in the context of monetary policy in the 1970s, but it applies with brutal precision to every dashboard, OKR sheet, and performance review in corporate life: when a measure becomes a target, it ceases to be a good measure. The SHE Index measures board seats. It should measure employee experience. The distance between those two things is where the perverse payment takes up residence. If you have ever watched a team hit every KPI while the actual business deteriorated beneath them, you have watched Goodhart’s Law operate in real time. Congratulations. You were standing inside the Surgeon Test and didn’t know it.

The pattern across all three cases is identical. A measurement system creates a payment. The payment changes conduct. The conduct change is invisible to the people inside the system because they’ve adapted to the new gradient without noticing the adaptation. The surgeon believes he’s being thorough. The Medicare forecaster believes costs are predictable. The corporate board believes it’s progressive. Each is responding to an incentive formula they haven’t identified as an incentive formula.

The average American in the private sector holds a job for 3.7 years. That single number, by itself, explains most of what is broken about the American health insurance system. If an insurance company knows it will churn you in 3.7 years on average, its calculus narrows to covering two categories: things that pay themselves back in less than 3.7 years, or things with such overwhelming demand from employees that employers feel they must cover them to remain competitive. Preventive care that pays off over a decade? Not covered. Chronic disease management that compounds value over twenty years? Not their problem. The arithmetic is precise and merciless. Nobody in the system is acting irrationally. Everybody in the system is acting exactly as the time horizon of their pay dictates. The system works perfectly for the people who designed it. It only fails the people trapped inside it, and those people have no voice in the design.

Jamie Dimon encountered the same architecture at J.P. Morgan, but from the other side of the table, as a designer who could fix it. When he arrived, the bank’s traders were paid 20% of profits with access to extreme borrowing ratios. The math was straightforward: if you’re running 30-to-1 and getting 20% of the upside, going to 40-to-1 adds 25% to your bonus. The pay model did not merely tolerate excessive risk. It made excessive risk the rational move for every individual trader, even though it could destroy the entire firm. Dimon eliminated the profit-pool formula and the borrowing allowances. He lost people in the process. But he had identified the architectural defect: when the time horizon of the bonus (one year) is shorter than the time horizon of the risk (decades), the payment will always favor the short bet over the safe one. This is not a character problem. It is a math problem. And math problems have math solutions.

This is the defect that makes quarterly capitalism so reliably destructive. The CEO whose pay vests over three years will make different decisions than the CEO whose pay vests over ten. Not because one is greedier. Because the comp plan has a clock built into it, and every rational actor reads the clock. Thermodynamics has a useful analog: entropy always increases in the direction of time’s arrow. In incentive physics, value always flows in the direction of the shortest payout cycle.

Rome’s political system embedded the same misalignment, and it produced a paradox that persisted for centuries. Roman senators were politicians in constant competition for public support. Any senator who signed a damaging peace treaty could be eviscerated in the Senate as a coward, a fool, or a traitor. The political calculus—survive the next election cycle—meant Roman senators would overwhelmingly vote to continue a war rather than admit defeat. This made Rome ferociously resilient in warfare. It also made Rome incapable of strategic retreat, even when retreat was the correct move. The time horizon of the political calculus (the next election) was shorter than the time horizon of the strategy (the next century of Roman power). The mismatch produced persistence that looked like strength but was often just an inability to quit.

Hannibal destroyed Roman armies repeatedly during the Second Punic War. Any other Mediterranean power would have sued for peace. Rome couldn’t. Not because Roman senators were braver than Carthaginian ones, but because the Roman senator who proposed peace would have been committing career suicide. The architecture made surrender impossible, which made Rome terrifying to fight, because opponents knew that defeating a Roman army did not end the war. Rome would simply raise another army and keep coming. The most powerful strategic commitment of the ancient world was one that nobody had chosen deliberately. It was a byproduct of the election calendar. Game theorists would recognize this as a credible commitment: valuable precisely because it cannot be revoked, and credible precisely because it was never a choice.

Federal Express had one hell of a time getting its night shift to perform. The integrity of the entire system required all packages to be shifted rapidly among airplanes at a single central airport each night. The night shift was the bottleneck. Management tried moral suasion. They tried every motivational technique available. Nothing worked. Then someone changed the pay formula from hourly to per-shift: finish the sort, go home, get paid the same. The problem vanished overnight.

The workers hadn’t been unmotivated. They’d been rationally responding to a pay formula that compensated them for taking longer. Every minute of delay was another minute of wages. The motivational speeches, the managerial oversight, the team-building exercises were all aimed at the wrong target. Management was trying to change psychology when they needed to change the paycheck. Fix the architecture, fix the conduct. This is the lesson that a thousand corporate off-sites per year fail to absorb, probably because the off-site itself is paid for by the hour.

Lee Kuan Yew solved Singapore’s healthcare costs the same way: by redesigning who owns the money. In Singapore, you get a health savings account the day you’re born. If you don’t spend the money, you and your heirs get to spend it eventually. The money is yours. Munger noted that Singapore’s system costs 20% of what the American system costs and works better by most measures. The mechanism is architectural, not cultural. People act more sensibly when they’re spending their own money. The American system insulates patients from costs, which sends a signal to over-consume. The Singaporean system makes every dollar visible, which sends a signal to spend carefully. Same humans. Same biology. Same diseases. Different architecture. Radically different outcomes. Adam Smith would have predicted every detail of this from his armchair in Kirkcaldy. The information was available in 1776. We just keep refusing to use it.

Dee Hock, who built Visa, designed what may be the most elegant alignment architecture in corporate history. Visa is a for-profit, non-stock membership corporation. Your ownership consists of irrevocable, non-transferable rights of participation. You own it by participating in it. The percentage of ownership you hold is proportional to the volume you contribute to the network. You can’t sell it. If you stop participating, you stop owning. This single design choice prevents a cascade to the exits—the problem that destroys most networks when early participants cash out—while aligning every member’s self-interest with the network’s growth. Hock didn’t appeal to his members’ altruism. He built a vehicle where self-interest and network interest converged into the same action. No speeches required.

Capital Group, which has managed money for nearly a century, applied the same logic to portfolio management. At Capital Group, a manager running $50 billion and a manager running $5 billion receive the same bonus if their investment results and tenure are the same. The firm doesn’t pay for a money grab. You don’t have to manage a lot of money to do well for yourself. You have to do well on the assets you’re asked to manage. By decoupling pay from assets under management, Capital Group eliminated the perverse gravitational pull that plagues most of the asset management industry: the pull toward gathering assets at the expense of returns. The manager’s interest and the investor’s interest point in the same direction. That alignment has survived for a hundred years. Most fund shops don’t survive twenty. The industry thinks the problem is talent retention. Capital Group suggests the problem is pay design. They’ve been right for longer than most of their competitors have existed.

On July 18, 1963, President Kennedy proposed an Interest Equalization Tax to throttle the outflow of American dollars. The tax penalized the sale of foreign securities to American investors. Henry Alexander, then running J.P. Morgan, assembled his officers that afternoon and made a prediction that proved exactly right: “This is a day you will all remember forever. It will change the face of American banking and force all the business off to London. It will take years to get rid of this legislation.”

Kennedy’s intention was to keep capital at home. The actual effect was to build London into a global financial center, because the payment created by the tax was for banks to move operations offshore, outside the tax’s reach. The entire Eurodollar market, which today dwarfs American domestic banking in daily transaction volume, traces its origins to this single miscalculation. Alexander saw it the afternoon the proposal was announced. Kennedy’s advisors never saw it at all. The difference came down to a single cognitive habit: asking the second question. And then what happens? Newton’s third law applies to economics with the same inexorability it applies to physics. Every policy action generates an equal and opposite market reaction. The only variable is whether the policymaker bothered to compute it.

Sam Altman identified a contemporary version of the same cascade in the architecture of search advertising. ChatGPT consistently ranks as the most trusted technology product from a major tech company, which is odd, because AI is the technology that hallucinates. Altman’s explanation: the business model. Google’s ad revenue depends on search results being imperfect. If Google gave you the perfect answer immediately, there would be no reason to buy the ad that sits above it. Users sense this misalignment even if they can’t articulate it. ChatGPT, by contrast, is paid by the user, which means its commercial interest is to give the best possible answer. Trust follows alignment, not accuracy. The second-order effect of Google’s ad model extends well beyond the existence of ads. The entire product is constrained from being as good as it could be, and hundreds of millions of users feel the drag without being able to name it. This is Upton Sinclair’s famous observation updated for the platform era: it is difficult to get a search engine to give you the right answer when its revenue depends on giving you an almost-right answer surrounded by sponsored alternatives.

Apple’s App Store produced the same pattern from the platform side. As Apple discovered the toll-booth economics of taking 30% of every transaction on the most important computer most people own, the drive to innovate on the platform itself eroded. Why improve the product when the revenue comes from taxing other people’s products? DHH described what happened next: “There was a rot that crept in to the foundation.” The 30% cut didn’t just redistribute profits. It changed what Apple cared about. A platform that once defined innovation gradually became a rent-collection operation, because the economics paid more generously for collecting rent than for building. Resistance to improvement is the natural second-order effect of any toll-booth model: every improvement to the platform increases the value flowing through it, which increases the value of the toll, which increases the motivation to protect the toll, which decreases the motivation to do anything that might disrupt it. The feedback loop tightens until the company’s entire strategic posture is defensive. Apple didn’t decide to stop innovating. The economics decided for them.

Munger named the compound version of this pattern: Serpico Syndrome. Frank Serpico joined a near-totally corrupt New York police division. The corruption was driven by two forces operating simultaneously: social proof (everyone around you is doing it) and financial pressure (doing it pays better than not doing it). Either force alone might be resistible. Together, they create a system so powerful that the person who resists it is nearly murdered for his trouble. The second-order cascade goes beyond the spread of corruption. The system develops antibodies against the people trying to stop it, because everyone inside the system now shares the same payment for maintaining the status quo. Whistleblowing becomes the irrational act. Compliance becomes the rational one. The cascade is complete when even the honest people can’t afford to stay honest. If you have ever sat in a meeting where everyone knew the initiative was failing and nobody said a word, you have visited a mild suburb of Serpico Syndrome.

Two scenarios show what that looks like in real time.