It Is Not Illiquidity That Scares You, It Is Liquidity Risk: Pastor and Stambaugh

Amihud and Mendelson showed that illiquid assets pay a premium because they are costly to trade. That is a story about cost, and it is a good story.

Pastor and Stambaugh, in 2003, told a different and sharper one. Their claim is about risk, and the distinction turns out to be the whole ballgame.

The problem: two different fears

Suppose you own a stock with a wide spread. There are two separate things you might be afraid of.

Fear one: the toll. Getting in and out costs you money. That is unpleasant, but it is predictable. You can budget for it. It is a cost of doing business, and, as Amihud and Mendelson explained, the market compensates you for it.

Fear two: the trapdoor. The market seizes up. Not just your stock, the whole market. Spreads blow out everywhere, depth vanishes, and the thing you own becomes untradeable at any sane price, at exactly the moment you most desperately need to sell.

Fear two is a completely different animal. It is not a cost, it is a contingency. And crucially, it is a contingency that arrives when your other investments are also cratering, when your fund is facing redemptions, when your leverage is being called. It hurts you when you are already down.

Standard asset pricing has a name for a risk that hurts you when you are already hurting: a priced risk factor. Investors will pay to avoid it, which means assets that expose you to it must offer higher returns.

Pastor and Stambaugh's question is therefore not "does illiquidity cost money?" but "is the risk of a market-wide liquidity drought a systematic risk that stocks get compensated for bearing?"

The key idea via analogy: whose umbrella breaks in the storm

Imagine everyone in a city carries an umbrella. Most days it is fine. But there is such a thing as a city-wide storm, a day when the wind is so bad that a lot of umbrellas simply fail.

Now: not every umbrella fails equally. Some are flimsy and turn inside out at the first gust. Others hold up fine. What you want to know about an umbrella is not just "is it a nice umbrella" but "how badly does it fail on the days when the whole city's umbrellas are failing?"

That is liquidity beta. Pastor and Stambaugh's procedure has two steps and both are essential:

Build a market-wide liquidity index. They needed a single number saying "how liquid was the entire market this month?" Their construction is clever: they looked for the tell-tale sign of illiquidity, which is that order flow pushes prices around temporarily and then the price bounces back. In a deeply liquid market, a wave of buying barely moves the price and there is nothing to bounce back from. In an illiquid market, a wave of buying shoves the price up and it reverts the next day. So they measured, stock by stock, how strongly today's volume-driven price move gets reversed tomorrow. Strong reversals mean illiquidity. Aggregate that across stocks and you have a monthly index of how liquid the market as a whole was.
Measure each stock's sensitivity to that index. Some stocks get hammered whenever aggregate liquidity dries up. Others shrug. That sensitivity is the stock's liquidity beta.

Then comes the test. Do the stocks with high liquidity betas, the flimsy umbrellas, earn higher average returns than the sturdy ones?

They do. Pastor and Stambaugh found a large and economically meaningful spread in returns between high and low liquidity-beta stocks, and it survived controlling for market, size, value and momentum exposure. In other words, this is not the size effect or the value effect in disguise. It is a genuinely additional source of compensated risk.

Why it mattered

It separated a cost from a risk, and priced the risk. This is the conceptual leap. The level of a stock's illiquidity is a cost. The covariance of a stock's liquidity with market-wide liquidity is a risk. Those are different objects, they require different data, and they demand different compensation. Almost everything written about liquidity afterwards respects this distinction.
It gave the profession a liquidity factor. The traded liquidity factor built from this paper sits alongside market, size, value and momentum in the standard factor toolkit, and fund performance is routinely evaluated against it. If a manager's returns evaporate once you control for liquidity beta, they were being paid for holding flimsy umbrellas, not for skill.
It explains crises better than the cost story does. Why did seemingly unrelated assets all collapse together in 1998, in 2008, in March 2020? Because they shared a common exposure: they were all things that become impossible to sell when funding dries up. Pastor and Stambaugh gave that intuition a measurable form, and it connects directly to the later literature on funding liquidity and fire sales.
It reframed a large class of "alpha" as risk premium. Many strategies that look like free money in normal times, from merger arbitrage to carry to selling volatility, are quietly short liquidity. This paper is a big part of why sophisticated allocators now ask, of any attractive return stream, "what does this do in a liquidity event?"

The honest limitations

The liquidity index is a construction, not an observation. Nobody publishes "market liquidity" as a number. Pastor and Stambaugh had to build it from return reversals, which is an inference resting on assumptions. Other researchers building liquidity factors different ways get indices that do not always agree, and the estimated premium moves around depending on the construction.
Liquidity betas are noisy. Estimating how sensitive a single stock is to an aggregate factor that is itself estimated, using monthly data over rolling windows, produces very imprecise numbers. Sorting stocks on a noisy estimate means your "high liquidity beta" bucket contains a lot of stocks that are just measurement error.
The factor is not always there. Replication work has found that the strength of the liquidity risk premium is sensitive to the sample period, the exact index construction, and the treatment of small stocks. It is not as robust across specifications as one would like, and some researchers regard it as considerably weaker than the headline result suggests.
The events that matter are rare. Liquidity crises are, thankfully, uncommon. That means the whole premium is compensation for a tail event, and estimating the price of a tail event from a few decades of data is intrinsically hard. Your standard errors are large and your sample of actual disasters is tiny.
Cause and effect are murky. Do prices fall because liquidity dries up, or does liquidity dry up because prices are falling and everyone is panicking? Both, probably, and the model cannot cleanly separate them.

The one-line takeaway

Pastor and Stambaugh showed that the risk that matters is not that your stock is illiquid, but that it becomes illiquid exactly when the whole market does, and stocks most exposed to market-wide liquidity droughts have earned meaningfully higher returns as payment for that exposure.