Incentive Compatibility: A Good System Should Not Rely on Fear or Gratitude

There is a story about a Chinese entrepreneur who opened a factory in Africa and hired many local workers. He treated them with equality and respect, offered generous compensation, and strictly avoided the coercive methods of former colonizers. However, before long, he realized that this approach was not working.

The workers were eager to join his factory, but within a few days of starting, they began arriving late, skipping shifts, and sleeping on the job. The moment supervision relaxed, they would disappear to smoke, and parts in the workshop began to vanish mysteriously.

Out of options, the Chinese owner resorted to a crude workaround: he hired a few local supervisors to manage the workers strictly, using harsh verbal and physical abuse. As a result, the factory got back on track, with stable production and on-time shipments.

Almost every business owner has experienced something similar. How should we interpret this? If you respect workers, they take advantage of you; if you treat them harshly, they work diligently. It seems they "fear authority but do not appreciate virtue." Does this mean these people do not deserve good management?

Do not rush to judgment. Everyone has instinctual desires and self-interests; getting a group of people to work together harmoniously is inherently a miracle. When it comes to managing conflicts and achieving cooperation, traditional Chinese thought spent over two thousand years developing two main solutions:

The Confucian approach is education and moral transformation—elevating people to become better, encouraging them to "appreciate virtue." The Legalist approach is reward and punishment—the carrot and the stick—forcing them to "fear authority."

But in this post, I want to argue that the modern world runs on neither Confucianism nor Legalism. It relies on a third force called "Incentive Compatibility."

The core insight is this: a good system should neither demand that people be saints, nor control them as if they were criminals. Instead, it should allow ordinary people, in the pursuit of their own self-interest, to naturally fulfill the goals of the system.

The highest state of system design is not to defeat human nature, but to make human nature work for the order.

Mechanism Design and the “Inverse Problem” #

The term "Incentive Compatibility" was first systematically proposed in 1972 by Leonid Hurwicz, a Polish-American economist. He received the 2007 Nobel Prize in Economic Sciences for his pioneering work in "mechanism design" [1].

Traditional economics asks the "forward problem": given a set of rules, can you predict how people will play, perhaps using game theory? Mechanism design asks the "inverse problem": given a desired outcome, can you reverse-engineer a set of rules so that people voluntarily play and produce that outcome?

Mechanism design is the engineering of economics. In the past, people often complained that economists were like meteorologists who couldn’t predict the weather. Now, economists can finally build something useful, just like engineers.

And incentive compatibility is the soul of mechanism design. It asks: can we design a game where the most rational choice for self-interested, calculating individuals is to do what benefits the system? Can we create a setup where honest people do not have to make extra sacrifices to do good, and dishonest people have no choice but to act honestly?

This is on a completely different dimension from Confucianism and Legalism. Confucianism relies on the moral extension of the gentleman, assuming human nature can be infinitely elevated—an approach recognized as overly romantic and naive as early as the Warring States period. Legalism assumes human nature can be controlled by rewards and punishments—which is a step above Confucianism—but it still requires an omniscient and omnipotent ruler: one who sees the truth, judges merit, enforces rewards and punishments, and remains uncorrupted. Confucianism depends on the quality of the governed, while Legalism depends on the ruler, neither of which is a fundamental solution.

In contrast, incentive compatibility says: I acknowledge that everyone has their own calculations, and I admit I cannot fully see what you do. But I do not care. The system I design will ensure that however you run your calculations, they will always end up at the number I want.

You might ask, is this really possible? Can such a perfect system exist? In fact, the mainstream of modern government governance and corporate management runs on exactly this kind of system.

The Sales Commission Rule for Software Companies #

Let’s look at an example. Imagine an enterprise software company facing a common issue: to hit their targets, sales representatives frequently promise customers features that do not yet exist. Delighted customers sign contracts on the spot; however, when the product goes live and customers realize those features are missing, they feel deceived. They either file complaints or refuse to renew, causing the company’s reputation to collapse. If you were the boss, what would you do?

The Confucian solution would be to hire a top-tier speaker to run a "values workshop," preaching the culture of honesty, customer-first mindset, and the importance of not tarnishing the company’s reputation. The sales reps might listen with tears in their eyes, but once they return to work, they will continue to oversell—after all, commissions are real, while values workshops are not.

The Legalist solution would be to declare that anyone making false promises will have half their annual bonus deducted, and severe violators will be fired. As a result, sales reps start playing word games, writing oral promises as "actively exploring this direction," shifting the blame to the product department, and "accidentally" deleting customer call recordings. You make examples of a few, but the remaining sales reps only become more cunning.

The incentive-compatible solution is to redesign the sales commission structure.

From now on, sales reps no longer receive their full commission upfront upon signing. Instead, it is distributed in phases: a portion at signing, a portion after three months of active usage, and a portion upon renewal. If a customer churns early due to "unfulfilled promises," the already distributed commission is clawed back proportionally.

Under this system, you need neither moral training nor constant surveillance of your employees. The sales reps will naturally choose the behavior that is most beneficial to the company.

Confucianism demands people be saints; Legalism threatens people not to be villains. Incentive compatibility allows people to be smart, ordinary individuals.

Incentive compatibility does not eliminate rewards and punishments—it still uses bonuses, penalties, and dismissals. The difference lies in the fact that Legalist rewards and punishments are external sticks, whereas incentive-compatible ones are components of the mechanism. Legalism deals with people after the act; incentive compatibility designs the game before the act.

Legalism asks: How do we make people fear doing wrong?

Incentive compatibility asks: How do we make doing wrong unprofitable, and doing right no longer disadvantageous?

Revisiting the Lessons of the African Factory #

Let us revisit the factory in Africa.

The Chinese owner’s first mistake was confusing "good working conditions" with "proper incentives," failing to realize that the two are entirely different concepts [2].

Whether the compensation you offer is good belongs to the "participation constraint," which answers the question: "Is this job worth taking?" The "incentive constraint," on the other hand, answers the question: "Once I am here, why should I work hard?" Wages only buy attendance; mechanisms ensure effort.

Economic historian Sidney Pollard studied factories during the early British Industrial Revolution and found that early workers did not naturally adapt to punctuality, repetitive labor, external pacing, and assembly line discipline. The "modern worker" was trained by an entire system of rules [3]. You must have mechanism design.

Harsh management was the owner’s second mistake because it is not an effective incentive constraint.

Workers have ways to adapt to physical and verbal abuse. When the supervisor is present, production is maintained; the moment the supervisor leaves, workers find ways to slough off. Furthermore, abusive management systematically weeds out workers with dignity, competence, and other options, leaving only those who are most submissive, good at reading faces, and adept at evading punishment. While production may stabilize temporarily, you invite lawsuits, strikes, brand damage, high turnover of key personnel, and quality issues down the line—creating endless trouble.

The Four Pillars: Authority, Benefit, Trust, and Virtue #

Drawing from academic research, an incentive-compatible factory system can be summarized in four words: Authority, Benefit, Trust, and Virtue.

First, establish Authority (威). This authority is not the fear of a supervisor’s stick, but the certainty that rules will be enforced.

How to handle tardiness, absenteeism, quality defects, and safety violations must be written clearly beforehand, recorded transparently during the process, and penalized predictably. The rules must apply to everyone equally, including supervisors. The effectiveness of a rule depends on the certainty of punishment, not its severity, and certainly not on personal degradation.

Second, add Benefit (利). As long as tasks are simple and outcomes are measurable, effort should be translated into money as quickly as possible.

According to research by personnel economist Edward Lazear [4], when an American auto glass installation company switched from hourly wages to piece-rate pay, productivity surged by 44%. Half of this increase came from existing workers putting in more effort, while the other half came from less productive workers, who had coasted on hourly wages, self-selecting out of the company. When tasks are measurable, monetary incentives are highly effective.

Third, ensure Trust (信). A major problem in many developing countries’ factories is not that workers do not understand incentives, but that they do not trust them.

The owner promises a year-end or month-end bonus, but the worker wonders: Will you claim the quality was subpar at the end of the month? Will you find excuses to deduct hours? Will you delay it to next month?

Furthermore, humans naturally discount future rewards. Workers often struggle with self-control, and a month-long wait for a bonus can feel far too distant [5].

If mutual trust is not yet established in an environment, wages and bonuses should have short cycles—even daily payouts.

Fourth, allow Virtue (德) to emerge naturally within the system.

We have often discussed that the ultimate incentive is identity. You can design a clear promotion path for the factory: apprentice, operator, skilled worker, inspector, team leader, with pay differentials for each level. Short-term bonuses buy effort; long-term identity buys loyalty.

Another approach is the "team system." Instead of having one supervisor watch twenty unrelated workers, organize them into teams of 5 to 10 with a small allowance for team leaders. The key is that the team shares a joint production target, quality standard, and attendance review. When the target is met, the entire team receives a bonus. The beauty of this setup is that workers monitor one another: since collective gains benefit everyone, anyone who slacks off for long will be pressured or pushed out by their peers. Studies show that when a garment factory switched from individual piece rates to team-based rates, worker productivity increased by 14% on average [6]. Experiments also demonstrate that team incentives simultaneously alter both worker effort and team composition [7].

The Authority and Benefit discussed here are not the arbitrary punishments of Legalism; Trust and Virtue are not the moralizing lectures of Confucianism. Authority is rule certainty; Benefit is the payoff function; Trust is commitment reliability; Virtue is identity formation. The key is that these are requirements placed upon the system, not demands on the character of employees or the emotional energy of managers.

The Dawn of Modern Institutional Enlightenment #

While incentive compatibility is a modern term, the underlying concepts are not unique to the twentieth century. Ancient Chinese sages recognized long ago that rewards, punishments, and moral education could not solve fundamental societal problems.

The Daodejing states: "The more laws and edicts are made prominent, the more thieves and robbers there will be," and "Not exalting the worthy keeps the people from competing; not valuing rare goods keeps the people from stealing." The Analects remarks: "The hypocrite is the thief of virtue," and "When the crowd hates a man, one must examine him; when the crowd likes a man, one must examine him." These quotes reflect the realization that punishment only forces bad actors to upgrade their cheating techniques, while moral exhortation allows hypocrites to pose as virtuous, leaving truly honest people at a disadvantage.

The engineering solution to these problems represents a fundamental breakthrough of modern institutional design.

The earliest dawn of this idea can be traced back to 1741, when philosopher David Hume wrote in Of the Independency of Parliament: "In designing any system of government… every man ought to be supposed a knave, and to have no other end, in all his actions, than private interest."

In 1788, James Madison, a core figure of the American Constitutional Convention and later the fourth president of the United States, wrote even more directly in Federalist No. 51: "If men were angels, no government would be necessary." He added, "Ambition must be made to counteract ambition."

We must remember that in those days, the mainstream Western ideology still highly exalted virtue. The insights of Hume and Madison were truly groundbreaking. This was not cynicism, nor did they truly believe everyone was a villain. Rather, they argued that system design must not stake public welfare on the goodwill of those in power.

Going beyond Confucianism and Legalism, a modern system is one where people do not have to be saints, and where being honest does not force one to suffer.

The Boundary Between Strong and Weak Incentives #

This raises a question: What if the participants in a system are genuinely selfless, high-caliber individuals? If you insist on a mechanism design that pays them piece rates like assembly-line workers, isn’t that insulting?

That is a misunderstanding. Piece-rate pay is a "strong incentive," and incentive compatibility theory does not advocate for strong incentives in all contexts. Sometimes, you must use "weak incentives."

Specifically, if a job is an "algorithmic task"—meaning it has a single outcome, a clear process, and is easy to measure, such as food delivery or assembly line work—it is suited for strong incentives. However, for "heuristic tasks"—characterized by multiple dimensions, requiring judgment, long-term focus, and deep collaboration, such as scientists, officials, and doctors—using strong incentives can be counterproductive [8].

This is because whatever you strongly incentivize is exactly what people will optimize for. For multi-dimensional tasks, optimizing for a single metric inevitably distorts the entire work. Put simply, if you pay scientists bonuses based solely on the number of papers published, they will flood the field with meaningless papers. This involves the "multi-task principal-agent problem" and "Goodhart’s Law," which we will discuss in detail later.

The key point is that such complex, multi-dimensional tasks are best driven by intrinsic motivation. According to "Self-Determination Theory," an incentive-compatible system should provide a free and flexible environment.

Yet, weak incentives do not mean no incentives. Weak incentives motivate people in softer, longer-term, and more multi-dimensional ways: such as high base salaries, long-term contracts, autonomy, peer review, honorifics, growth paths, and a clear sense of mission.

This is exactly how high-end talent is managed in reality. High-tech companies do not use piece-rate systems or issue fines; they offer a sufficiently high annual salary so employees do not have to calculate their earnings daily. Similarly, universities and research institutions grant tenure once you prove your capability, removing job insecurity so you can boldly explore whatever interests you.

Incentive compatibility theory suggests that the more reliable the indicators, the more direct the bonuses can be; the coarser the indicators, the more crucial culture and reputation become.

The Revelation Principle and Daily Order #

Can system design solve every problem? Perhaps there is hope.

The fundamental difficulty of mechanism design is "information asymmetry"—the fact that managers cannot see exactly how much effort workers put in. However, in 1979, American economist Roger Myerson and others mathematically proved a theorem known as the "Revelation Principle" [10]. It suggests that for any outcome achievable by a complex mechanism, there exists an equivalent "direct mechanism" where everyone truthfully reporting their private information is their optimal strategy.

In plain terms, it is theoretically always possible to reframe a problem into a system design where speaking the truth becomes everyone’s best strategy.

Myerson was also awarded the 2007 Nobel Prize in Economics for his contribution.

Thus, at least mathematically, there is hope for this world to run better.

Confucianism can only lament when "honest people suffer," and Legalism is helpless against "information asymmetry." Modern industrial civilization was built neither on moral lecturing nor on raw coercion, but on executable daily order.

We must not say that China is limited to only Confucianism and Legalism. As early as 1980, Deng Xiaoping made a very modern statement [9]: "…A good system can prevent bad people from running amok at will, while a bad system can prevent good people from doing good deeds fully, or even lead them to the opposite."

This statement inverted the traditional Chinese approach of "selecting good people first, then doing good deeds." It is, in essence, the core of incentive compatibility.

Epilogue Poem #

We ask not that all be saints, Nor guard against all as beasts. Just align the path of gain with desire, And order shall rise like the morning sun.

References #

[1] Nobel Prize Outreach. “The Prize in Economic Sciences 2007: Popular Information.” NobelPrize.org, 2007.

[2] Laffont, Jean-Jacques, and David Martimort. The Theory of Incentives: The Principal-Agent Model. Princeton: Princeton University Press, 2002.

[3] Pollard, Sidney. “Factory Discipline in the Industrial Revolution.” Economic History Review 16, no. 2 (1963): 254–271.

[4] Lazear, Edward P. “Performance Pay and Productivity.” American Economic Review 90, no. 5 (2000): 1346–1361.

[5] Kaur, Supreet, Michael Kremer, and Sendhil Mullainathan. “Self-Control at Work.” Journal of Political Economy 123, no. 6 (2015): 1227–1277.

[6] Hamilton, Barton H., Jack A. Nickerson, and Hideo Owan. “Team Incentives and Worker Heterogeneity: An Empirical Analysis of the Impact of Teams on Productivity and Participation.” Journal of Political Economy 111, no. 3 (2003): 465–497.

[7] Bandiera, Oriana, Iwan Barankay, and Imran Rasul. “Team Incentives: Evidence from a Firm Level Experiment.” Journal of the European Economic Association 11, no. 5 (2013): 1079–1114.

[8] Pink, Daniel H. Drive: The Surprising Truth About What Motivates Us. New York: Riverhead Books, 2009.

[9] 邓小平：《党和国家领导制度的改革》，1980 年 8 月 18 日。

[10] Myerson, Roger B. “Incentive Compatibility and the Bargaining Problem.” Econometrica 47, no. 1 (1979): 61–73.