Greg Leonard's August 2026 SSRN working paper, Revealing the Illiquidity Premium: The Role of Short Sale Constrained Stocks, makes a case that the weak illiquidity premium in U.S. equities may be a measurement problem rather than a disappearing effect. The premium might still be there, hidden by stocks that are illiquid and also expensive to short. The paper hasn't been peer reviewed, so treat the numbers as provisional.
The practical takeaway is that illiquidity and short-sale constraints shouldn't be analyzed as separate effects. Stocks that are costly to borrow tend to be illiquid, and they also tend to earn low expected returns. Lump them in with every other illiquid stock and the measured premium shrinks, sometimes well below what's actually being paid for illiquidity.
What the paper examined
Basic theory says investors should demand higher expected returns for holding illiquid assets. If a security is hard to sell quickly without moving its price, holders should be compensated for bearing that cost.
That logic has held up across many markets. Still, a body of recent work, including research on the illiquidity premium and work on illiquidity and stock returns, has found the equity illiquidity premium, meaning the excess return of illiquid stocks over liquid ones, has shrunk noticeably or disappeared. Leonard goes back and tests that conclusion directly. His sample covers U.S. common stocks listed on the NYSE, Amex, and Nasdaq. He uses three tools: - **Amihud's illiquidity measure**, which estimates price impact by dividing absolute daily returns by dollar trading volume. - **Equity loan fees**, the cost of borrowing shares to open and hold a short position. - **Monthly portfolio sorts** that go long the most illiquid quintile and short the most liquid quintile. He calls this the illiquidity-minus-liquidity, or "IML," portfolio. The broad illiquidity-premium analysis runs from July 1964 through March 2022. Loan-fee data only exist from July 2013 through March 2022, so the short-sale work covers a much shorter window. Stocks in the top loan-fee quintile are labeled "special." They're expensive to borrow and therefore more constrained on the short side. The remaining 80% are "general collateral" stocks.
The core argument
Leonard's logic runs in four steps: 1. Illiquid stocks should earn higher expected returns, because investors want payment for higher trading costs and price impact. 2. Hard-to-short stocks tend to earn lower expected returns. Pessimistic investors struggle to express negative views when shorting is expensive or unavailable. 3. High-loan-fee stocks cluster among the illiquid. 4. So an illiquid portfolio can pick up hard-to-short stocks whose low returns cancel out the premium you're trying to capture. The asymmetry is stark. In the paper's recent sample, 28% of stocks in the illiquid leg of the IML portfolio were short-sale constrained. In the liquid leg, that figure was under 1%. The high-illiquidity portfolio isn't just a bundle of liquidity risk. It carries a concentrated dose of short-sale constraints too.Key findings
**1. The premium looks weak before you control for short-sale constraints** From July 2013 through March 2022, the conventional illiquidity premium was modest or negative, depending on the sample and risk adjustment. For all stocks, the equal-weighted long-illiquid/short-liquid portfolio produced these monthly results: | Measure | Monthly result | |---|---| | Raw IML return | -0.32% | | FF5* alpha | 0.17% | | FF5* alpha excluding short-sale-constrained stocks | 0.49% | "FF5*" is a Fama-French five-factor-style adjustment that drops the size factor. Leonard leaves size out because size and the Amihud measure are closely tied together. Size is largely a proxy for illiquidity, so including it would soak up the very effect being measured. Here's the headline: removing the highest-loan-fee stocks lifts the all-stock FF5* alpha from 0.17% to 0.49% a month. That 0.33% monthly increase is the result that matters most. The t-statistic on the level moves from 0.87 to 2.42, and the t-statistic on the 0.33% change is 3.53. The shift is the finding. The level itself isn't impressive.
Why this differs from earlier theory
Some classic models argue that short-sale restrictions can improve liquidity. The reasoning: if restrictions mostly keep informed pessimists out of the market, market makers face less adverse selection and can quote tighter spreads. Leonard's point is that this works for a ban on certain traders, not necessarily for a setup where market makers or designated liquidity providers are also restricted. It fits poorly with normal U.S. equity lending, where any trader who stays short eventually has to borrow the stock and pay the fee. The distinction changes the predicted outcome: | Type of constraint | Predicted liquidity effect | |---|---| | Short-sale ban that mainly removes informed short sellers | Liquidity can improve because adverse selection may fall | | Higher cost of borrowing stock that also affects liquidity providers | Liquidity can deteriorate because inventory and adverse-selection costs rise | The paper's focus is the second mechanism: ordinary borrow costs in modern equity markets, not crisis-era outright bans.Historical interpretation
Here's the limitation: loan-fee data for equities are thin before 2013, so Leonard can't run the same analysis back in time. He uses the asymmetry in reversal returns as an indirect signal instead. The idea is that it hints at how exposed illiquid portfolios were to short-sale constraints in earlier decades. The evidence suggests costly-to-borrow stocks probably affected illiquid portfolios back then, though less strongly than after 2013. Leonard also points to Regulation SHO and later amendments, especially those that closed off ways to stay short without borrowing, as a possible reason the link between borrow costs and liquidity grew stronger over time. Treat this part as suggestive. It's an inference from a model-based diagnostic, not a direct test with earlier loan-fee data.Investor takeaways
**1. "Illiquid" isn't a clean standalone factor** An illiquidity screen can pick up stocks that are hard and expensive to short. Those names may carry low expected returns for reasons tied to disagreement, limits to arbitrage, speculative demand, or constrained pessimism. For factor research, a raw illiquidity portfolio can mix two opposing forces: - Positive expected-return compensation for bearing illiquidity. - Negative expected-return exposure tied to short-sale constraints. That mix can make the premium look weak, unstable, or absent, even when the underlying compensation is there. **2. Loan fees are useful conditioning variables** If you're long-only, borrow data still matter, even though you won't short anything yourself. High loan fees can mark a segment where valuation, liquidity, momentum, volatility, or reversal signals behave differently. If you run long-short, borrow fees hit your returns directly. A gross backtest that ignores financing and borrow costs can overstate implementable alpha, and reversal strategies are especially exposed because they short recent winners, which are often the hard-to-borrow names. The special-stock results make the point. The short-leg reversal return is 0.51% a month before borrow costs and 0.20% after the realized fee adjustment. That's a big haircut from a number that looks strong on paper.Amihud illiquidity is a useful rough gauge of price impact, but it doesn't tell you what trading will actually cost you. If you're running a live process, track these separately:
- Bid-ask spreads.
- Market depth and expected price impact.
- Trading volume and capacity.
- Short availability.
- Borrow fees and recall risk.
- Securities-lending supply.
- The potential for crowding or one-sided demand.
The appendix runs bid-ask spreads as a cross-check, and they point the same way. Strip out the high-loan-fee stocks and the illiquidity premium across all stocks gets stronger.
A research result, not a turnkey strategy
The paper builds its premium from equal-weighted, monthly rebalanced academic portfolios. That's the right tool for measuring differences in expected returns. It's a poor blueprint for a portfolio you'd actually run, and the gap between the two is where most of the trouble sits.
Real trading brings frictions the model ignores:
- The most illiquid securities are costly to trade.
- Equal weighting tilts toward smaller firms.
- Capacity may be limited.
- Turnover and market impact can be high.
- Short-sale data can be delayed, incomplete, or unavailable to many investors.
- Borrow conditions can change rapidly, and short availability can disappear entirely.
So the takeaway isn't "buy illiquid stocks and skip the expensive-to-borrow ones." The illiquidity effect has to be judged alongside shorting frictions and real implementation costs, all at once. Pull one out and the numbers stop meaning much.
Bottom line
Leonard's paper gives a credible reason the equity illiquidity premium has looked weak in recent data. The most illiquid portfolios hold a disproportionate number of hard-to-borrow stocks, and those stocks have earned unusually low risk-adjusted returns in this sample. Drop the top quintile of loan fees, and the recent all-stock illiquidity alpha rises from 0.17% to 0.49% per month. The paper also argues that higher borrowing costs degrade liquidity itself, raising both the cost of carrying short inventory and the adverse-selection cost liquidity providers face.
For investors, the lesson is that liquidity risk, short-sale constraints, and trading implementation are tightly linked. Any attempt to harvest an illiquidity premium has to price in what it costs to borrow a stock and whether you can borrow it at all. Volume and spreads alone won't tell you that.