The Market for Lemons: Don't Just Reward Good and Punish Evil, Make Quality Verifiable

Have you noticed that when many top-tier companies recruit now, they pay special attention to where the applicant went for their undergraduate studies—whether the school is a top-tier university (such as a 985 or 211 in China) determines the caliber of the talent. This is called “first degree worship.” As for master’s and doctoral degrees, their gold standard has significantly declined.

Once upon a time, master’s and doctoral degrees were highly respected titles. A graduate student, by definition, is someone who does research. That someone attained a master’s degree meant they had received advanced training and possessed research capabilities. Why are these degrees no longer fully recognized as proof of capability?

On one hand, there are simply too many people with master’s and doctoral degrees now; high degrees are no longer a scarce signal. But on the other hand, people indeed no longer fully trust the standard these degrees represent.

A master’s can be coasted through, a doctorate can be drifted through, advisors can be nominal, and papers can be bought. What does such a degree prove, other than having the endurance to delay employment?

Look at today—even official regulations acknowledge that graduation theses go through a “plagiarism check” (查重) process, which means searching the web to see if the paper was copied. But if you think about it, isn’t this absurd? As long as you wrote the paper yourself, even if it is not particularly ground-breaking research, it shouldn’t trigger plagiarism alerts. Furthermore, some universities even stipulate that as long as the plagiarism rate does not exceed a certain percentage, the paper is deemed acceptable—which is equivalent to tacitly permitting a certain amount of copying. Isn’t this a collapse of academic decency?

Now with AI, it is even more convenient: students use AI to write papers, teachers use AI to review them, universities use AI to check for plagiarism, and students use AI to bypass detection… Everyone is putting on a show. Who are you putting this show on for? What is the point of pursuing such a degree?

No wonder some people say that the reason for “first degree worship” is that people trust the selection capability of the national college entrance exam (Gaokao), but they do not trust the cultivation capability of higher education institutions.

This is not just an educational issue; it is an economic one. Degree certificates and academic publications are turning into a “market for lemons.”

Information Asymmetry and the Death Spiral #

The “market for lemons” is a concept proposed by American economist George A. Akerlof in 1970, for which he was awarded the 2001 Nobel Prize in Economic Sciences.

Akerlof originally studied the used car market [1]. In American slang, a defective used car is called a “lemon.” Buyers of used cars dread purchasing a lemon. Yet used car transactions are inherently like this: the seller knows the car’s condition, but the buyer does not.

This is “information asymmetry.” Let us deduce what information asymmetry does to a market:

Buyers are not stupid. They know there are both good and bad cars in the market. Since they cannot distinguish between them, they will not pay the price of a good car. They will lower their bids so that if they end up with a bad car, their loss is minimized.

But because of this, owners of good cars will refuse to sell. My car is excellent, yet you only offer the average price of a good and bad car—wouldn’t I be losing out? Of course, owners of bad cars remain eager to sell; they are more than happy to be averaged with good cars.

Consequently, good cars exit, and bad cars remain.

Over time, the proportion of bad cars grows higher, and buyers trust the market even less, pushing prices down further.

As a result, more good cars exit, and more bad cars remain. This is a self-reinforcing cycle that will not stop on its own, which economists call a “death spiral.” Bad money drives out good; with every turn, the market deteriorates further—until the good goods disappear entirely.

Eventually, the market becomes a lemon market.

With this perspective, you will find that many markets are lemon markets. In the job market, truly capable people cannot prove their abilities, while those who can package resumes and grind interview questions multiply. On content platforms, click-through rates and emotional triggers are the easiest to see and score, leaving readers surrounded by clickbait, conflicts, and extreme expressions. In healthcare, treatments require long-term verification; since patients do not know how much advertising is exaggerated, you cannot expect genuine medicine to defeat snake oil.

Currently, the most severely affected area might be academia. Scientific funding and academic titles are distributed by the government, but because the government cannot directly see which research is high-quality, it can only rely on paper counts and impact factors. Thus, doing research has turned into publishing papers. A study [2] analyzed 1,182 papers retracted and identified as originating from paper mills up to June 2022. It found that almost all corresponding authors were affiliated with Chinese institutions, and over three-quarters of the first authors’ affiliations were hospitals.

Isn’t a doctor’s primary duty to treat patients and save lives? Yet our professional evaluation system mandates that they publish papers, so they publish fake ones. The result is a general downgrade in the trustworthiness of papers written by all Chinese doctors.

Does anyone still believe medical advertisements on Baidu? If you are a reputable public hospital with excellent therapeutic effects, would you advertise on Baidu? That market was turned into a lemon market long ago by deceptive private clinics.

The most terrifying thing about a lemon market is not that bad goods exist, but that good goods cannot be identified, forcing them to leave.

This is not a moral issue; it is an information engineering issue.

Signaling Mechanisms and Escrow Transactions #

Lemon markets have solutions. Modern civilization has long invented a series of mechanisms to prevent markets from “lemonizing.” The fundamental idea is to establish signals so that true quality can be recognized.

One way is for the seller to actively send an expensive “signal.” For example, since your product is good, do you dare to offer a warranty? Do you dare to allow returns? This is a contract of accountability.

Another way is to seek third-party “certification,” such as finding a mechanic to inspect a car and write a reliable report before selling it. Today, we have government regulatory bodies in food and medicine, and industry standards and civil ratings in professional fields.

Actually, the oldest and most effective third-party certification is academic peer review. The value of your research is best known by your peers, not by impact factors or citation counts. However, peer review implies autonomy for the academic community, and autonomy implies the ceding of administrative power.

Another way is to implement a measurable “reputation mechanism,” such as transaction histories and user ratings, which are simple and effective in the internet age.

Furthermore, you need dispute resolution and appeal mechanisms to handle bad transactions. Getting deceived once is acceptable, but it must not turn into a total loss of trust in the entire system.

The used car market in reality is much better than the primitive state Akerlof described. This is because we have Vehicle Identification Numbers (VINs), accident records, maintenance logs, warranty systems, Lemon Laws, and vehicle history reports.

The most elegant success story in China is probably Taobao. Imagine if back in 2003, a tiny shop thousands of miles away sold an unfamiliar item at a high price—would you dare buy it? What if you sent the money and they didn’t ship the goods? The seller would also worry: what if they shipped the goods and you didn’t pay? Both sides feared the other was a “lemon,” making transactions impossible.

The key invention of Taobao and Alipay was escrow transactions. The buyer pays a third party first, the seller ships, and once the buyer confirms receipt, the money is released to the seller. Combined with store ratings, buyer reviews, chat records, logistics tracking, return rules, and platform arbitration, a baseline of trust was established among strangers [3].

Please note that this does not mean there are no fake goods on Taobao. On the contrary, Taobao, Pinduoduo, and various online platforms have always hosted counterfeit and knock-off products. But the key difference is: counterfeit goods have not driven out authentic ones.

This is because consumers generally know how to read the signals. If you buy a pair of “designer-style” shoes for 139.9 RMB with free shipping, you know exactly what you are getting. The next time you want the authentic product, you still return to the same platform, but you go to the official flagship store and pay the full price.

As long as the real and fake are not confused, the presence of counterfeits does not make it a lemon market.

None of this is perfect, but we can generally say that internet platforms have made people slightly more honest.

The True Meaning of Procedural Justice #

The antidote to the lemon market—this entire set of mechanisms that “renders the invisible visible, verifiable, and correctable”—is deeply rooted in the concept of “procedural justice.” Procedural justice does not ask beforehand who should win or lose. Instead, it lays out the rules clearly, presents the evidence, opens the process, explains the reasoning, and leaves channels for appeals and corrections.

The lemon market asks: Why should I believe you are selling good goods?

Procedural justice asks: Why should I believe you are not rigging the game in a black box?

The answer to both is to transform invisible qualities, motives, and judgments into visible public signals.

Just as managing a lemon market is not about eliminating all bad goods but about retaining the good ones, procedural justice is not about pursuing absolute equality of outcomes. It is about ensuring that even those who lose acknowledge that the game is worth playing.

The primary mission of a system is not to punish bad people, but to retain good ones.

You offer a high-quality product, but perhaps because of its high price, it doesn’t sell—so you don’t win. That is fine, as long as you believe the system can distinguish you from bad products. You will choose to stay, perhaps lowering your price or refining your product. But if you feel that no matter how good you are, no one can see it—that ratings are manipulated, promotions depend on connections, judgments are black boxes, and the market is dominated by frauds—you will exit.

You can even tolerate a certain rate of errors. The greatest danger is not an occasional misjudgment, but the losers feeling that “the outcome was rigged from the start.”

Why People Obey the Law #

Yale social psychologist Tom R. Tyler has long researched a question: Why do people obey the law? Is it out of fear of punishment?

Tyler conducted extensive research. In 1984, he interviewed 1,575 Chicago residents by phone, and a year later re-interviewed 804 of them selected at random. He asked them questions like: “Why do you obey the police? Is it because they have guns?” He compiled his findings into a classic book, Why People Obey the Law [4].

The answer is that people do not obey the law solely because they fear punishment. More importantly, they must believe that the legal process is fair, neutral, transparent, and respectful.

But what kind of judicial system can satisfy these demands? Courts cannot directly know the objective truth, but you can ensure that the process of arriving at a verdict is fair.

The entire effort of modern justice is not focused on “instantly declaring the truth,” but on ensuring that the process of approaching the truth is not contaminated: rules of evidence dictate what can be presented; prosecution and defense challenge each other; judges must explain the reasoning behind their verdicts; and errors can be appealed and reviewed.

The most counter-intuitive aspect of this process is that it can sometimes be frustrating. You might feel certain that someone is guilty, yet the system demands rigorous evidence. You might want immediate punishment, yet the system insists on hearing the defense. Public opinion might have already condemned them, yet the system demands a trial.

This is because letting a guilty person go free harms the system far less than wrongfully punishing an innocent one [5]. You must give good people a sense of security so they will stay. Security comes from the fact that procedural justice is an engineered process: it may occasionally err, but it is operational, verifiable, and correctable.

Procedures in Workplace Reviews and Academic Peer Evaluation #

The same logic applies to corporate promotions and appraisals.

An executive declaring, “We reward those who make real contributions,” is not operational.

Operational questions include: What constitutes a contribution? Who evaluates it? Is the evaluator biased? How do we balance short-term and long-term contributions? How do we distinguish individual contributions from team ones? How do we balance visible contributions with invisible ones? Can the reviewee challenge an incorrect evaluation?

Without these parameters, “rewarding contributors” in practice turns into “rewarding those who are good at projecting contributions.” That is how a company lemonizes. The true builders silently resign; the credit-takers multiply; the boss believes they are rewarding good and punishing bad, when in reality they are rewarding signal manipulation.

A good company must have values, but daily management cannot rely on preaching values. Good management relies on establishing clear, preemptive promotion criteria, verifiable evidence of contributions, multi-source feedback, transparent reasoning for decisions, and an appeal process for employees. The goal is to make “contribution,” which is inherently subjective, partially verifiable.

Academic review operates on the same logic.

The scientific community does not proclaim, “We reward truth and punish error.” Insiders know that truth is not immediately visible, and today’s “error” might be tomorrow’s truth, while today’s “breakthrough” might be tomorrow’s retraction.

Science runs on a set of procedures: replicability, open methodology, data transparency, peer review, falsifiability, citation, and retractions for errors. These are not truth itself, but verifiable mechanisms to approach truth.

When a new finding is published, no one can guarantee it is absolutely correct. But this procedure allows science as a whole to move forward, keeping good conclusions and filtering out erroneous ones.

The greatness of science lies not in the nobility of scientists, but in the fact that scientific errors can be exposed.

Also, why do employers trust the Gaokao? Because the Gaokao is an exam of procedural justice: questions are sealed, timing is unified, grading is standardized, marking is anonymous, and results are public.

The Gaokao is certainly not perfect. There are many random factors. For two students separated by a single point, we have no reason to believe one is fit for a top university while the other is not. But only by letting scores speak can people accept the outcome.

In contrast, the credibility of master’s and doctoral degrees is much weaker; any official can obtain an “executive doctorate.” Every watered-down doctorate makes those with true scholarship suffer.

Verifiability: The Technology of Modern Institutions #

Combining this with the “incentive compatibility” we discussed in the previous lecture, you arrive at an insight: for modern society to improve, we do not need to demand that everyone be good people, but rather that we have a good system.

You can demand that you yourself be a good person, but you cannot directly demand that “everyone be good.” Rewarding good and punishing bad directly only turns many into hypocrites and forces bad actors to hide deeper. Moreover, who are you to distribute rewards and punishments? Just because you are powerful? Can you not misjudge? Can you not corrupt?

A good system does not directly demand moral perfection from people. Instead, it ensures that good people do not suffer and are willing to stay, allows quality to automatically manifest, and makes it rational for bad actors to learn to act like good ones.

The genius of modern institutions is that they do something highly counter-intuitive: they abandon a direct obsession with the moral outcomes of actions, and instead build procedures that allow good and bad to be automatically exposed.

The SEC does not directly judge which company is good; it only mandates information disclosure and lets the market decide. The FDA does not directly declare which drug is good; it demands “substantial evidence” of efficacy, letting trials and data speak. The court does not directly declare who is a good person; it establishes procedures for evidence to be cross-examined.

Therefore, strictly speaking, mature systems are not machines for “rewarding good and punishing evil,” but machines for “making good and evil verifiable.”

Rewarding good and punishing evil is the desire of pre-modern societies; verifiability is the technology of modern institutions.

Modernization teaches us to stop fantasizing about finding all good people and putting them in the right places, or destroying all bad people to make the world instantly better. We acknowledge that human nature is difficult to reform, information is limited, and judges themselves are biased participants. We then design procedures that allow genuine quality to reveal itself, make deception unprofitable, and allow errors to be discovered and corrected.

We have learned not to directly reform people, but to change the information structure. This is perhaps one of the deepest wisdoms of modernization.

The Self-Verifiability of Modern Individuals #

The more I write, the more I feel that “engineering” and “operability” are the hidden themes of this course. In fact, they are the themes of modernization.

A modern person should not simply claim “I am excellent,” “I work hard,” or “I am kind.” To make it operable for others, you must make yourself verifiable. You must let your work, records, credit, deliveries, long-term partnerships, reviewable results, and the responsibilities others willingly entrust to you speak for themselves.

Modern people do not say: “Trust me.”

Modern people say: “Here is how you verify me.”

References #

[1] Akerlof, George A. “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism.” The Quarterly Journal of Economics 84, no. 3 (1970): 488-500.

[2] Candal-Pedreira, Cristina, et al. “Retracted Papers Originating from Paper Mills: Cross Sectional Study.” BMJ 379 (2022): e071517.

[3] “Alipay Releases Social Responsibility Report: 15 Years of Building Trust Mechanisms.” Sina Finance, May 20, 2019.

[4] Tyler, Tom R. Why People Obey the Law. Princeton, NJ: Princeton University Press, 2006. Original edition, 1990.

[5] A famous legal principle is Blackstone’s ratio: “It is better that ten guilty persons escape than that one innocent suffer.”