The Paradox at the Heart of Efficient Markets: Grossman and Stiglitz

Most critiques of efficient markets are empirical: here is a pattern, the theory says it shouldn't exist, checkmate. The theory usually survives by absorbing the pattern as a risk premium.

Grossman and Stiglitz took a completely different route in 1980. They did not go looking for anomalies. They took the efficient market hypothesis at its word, followed its logic all the way to the end, and showed that it destroys itself. The strong version of the theory is not merely false. It is logically incoherent.

This is the rare paper in finance that wins by pure argument, and every quant should have its punchline tattooed somewhere accessible.

The problem: who is paying for the efficiency?

The efficient market hypothesis says prices fully reflect all available information. Fine. But stop and ask an awkward question: how does the information get into the price?

It does not teleport there. It gets in because some human being, somewhere, spent money and time to find it. They read the filing. They flew to the factory. They built the model. They hired the analyst. All of that costs real resources.

Now ask why anyone would bother.

If prices already fully reflect all information, then knowing the information gives you no edge at all. The price has already moved. You spent 10 million dollars on research and earned exactly the same return as the guy who bought the index fund and went to the beach. You are 10 million dollars poorer for your trouble.

So no rational person gathers information. So nobody gathers information. So the information never gets into the price. So prices are not efficient.

But wait: if prices are not efficient, then information is now very valuable, so people rush to gather it, so it flows into the price, so prices become efficient, so gathering it becomes pointless again, so everyone stops.

The theory has eaten its own tail. Perfect informational efficiency cannot be an equilibrium, because it removes the incentive for the very activity that produces it. That is the Grossman-Stiglitz paradox.

The key idea via analogy: nobody checks the ATM

Imagine a legend spreads that a certain ATM sometimes spits out extra cash. At first, everybody detours past it. The bug gets found and fixed instantly. Word gets around: the ATM is always correct now, checking is a waste of time. So people stop checking.

And now, of course, the ATM can be broken for weeks and nobody would know.

The ATM's reliability was never a property of the machine. It was produced by the crowd of people who kept checking it, and the crowd only checked because it was sometimes wrong. Fix it perfectly and the crowd disperses and it stops being fixed.

Markets are that ATM. The accuracy of prices is a product manufactured by traders hunting for inaccuracy. Remove the inaccuracy entirely and you remove the hunters, and then there is nobody left to manufacture the accuracy.

So what does Grossman and Stiglitz's actual model produce? A beautiful compromise they called an equilibrium degree of disequilibrium:

• Some traders pay a cost to become informed. Others stay uninformed and just look at the price.
• The informed traders' buying and selling pushes the price partway toward the truth. Uninformed traders can therefore learn something by watching the price, which is a genuine free ride.
• But the price only reveals things partially, because it is muddied by noise, meaning random supply and demand from people trading for reasons unrelated to information (they need cash, they are rebalancing, they are gambling). The uninformed can never quite tell whether the price is up because someone knows something or because someone's grandmother liquidated a portfolio.
• That fog is what keeps information valuable. The informed earn just enough extra return to cover their research costs, and not a penny more. In equilibrium, the reward for being informed exactly equals the cost of becoming informed, and both types of trader are equally happy.

Prices are therefore almost efficient, never fully. There is a permanent, self-sustaining layer of mispricing, precisely large enough to pay the people whose job is to remove it.

Why it mattered

• It gives active management a coherent reason to exist. The strict efficient market hypothesis says active management is pointless. Grossman and Stiglitz say active management is what creates efficiency, and must therefore be compensated. This is the intellectual license under which every hedge fund operates. Note the sting in the tail: your excess return is supposed to be exactly your research cost. Efficiency does not promise fund managers profits, only reimbursement.
• It sets a hard ceiling on how much alpha can exist. Total available alpha is bounded by the total cost of the research needed to find it, plus what the noise traders are willing to lose. This is a genuinely useful sanity check when someone pitches you a strategy that supposedly prints money forever with no competition.
• It explains why noise traders are load-bearing. If everyone traded only on information, nobody could hide, so nobody would trade at all (a related result: rational traders should not trade with each other, since anyone willing to take the other side probably knows something you don't). The market needs a supply of people trading for non-informational reasons. Fischer Black would make exactly this argument in his 1986 essay Noise, and it is the reason liquidity exists.
• It predicts where inefficiency lives. Follow the logic: mispricing should persist wherever information is expensive to gather and edges are small. Illiquid corners, small caps, obscure jurisdictions, complicated structures, anything a large fund cannot be bothered to research. Everyone knows this in their bones. Grossman and Stiglitz derived it.

The honest limitations

• The model is a stylized toy. It runs on one risky asset, normally distributed everything, exponential utility, and a single round of trading. Real markets have thousands of assets, fat tails, and continuous trading. The intuition generalizes. The specific equations do not.
• It assumes information has a price tag and a payoff you can compute. In practice you do not know what your research will be worth until after you have done it, and much of the cost is a fixed, sunk investment in people and infrastructure. The clean "pay X, get information worth Y" trade is not how a real research budget works.
• It says nothing about how long the correction takes. The model reaches equilibrium instantly. In reality, informed traders can be right and still be underwater for years, which is exactly the gap that Shleifer and Vishny would later fill with the limits of arbitrage. Grossman and Stiglitz say a profit opportunity must exist. They do not say you will survive long enough to collect it.
• Costless index investing complicates the story. The paper's uninformed trader is a passive free rider, and today free riding costs almost nothing. If the free ride keeps getting cheaper, the informed crowd should keep shrinking until mispricing grows large enough to pay them again. Whether that has actually happened, and how big the passive share can get before prices start to wobble, is one of the genuinely open questions in modern finance, and Grossman and Stiglitz is the paper you reach for when you want to argue about it.

The one-line takeaway

Grossman and Stiglitz proved that a perfectly efficient market is impossible, because if prices already contained all information nobody would be paid to find any, so every real market must carry just enough mispricing to keep its information hunters in business.