Liquidity

Illiquid stocks earn a premium for being hard to trade.

Amihud illiquidityLiquidity risk premiumBid-ask spread factorTrading cost factor

Typical IS Sharpe

0.4 – 0.7

Typical OOS Sharpe

0.2 – 0.4

Capacity

Small-cap

Signal decay

Persistent

Medium turnoverLong-only viable

Overview

Investors demand compensation for holding illiquid assets because illiquidity increases transaction costs and the risk of not being able to exit a position in adverse conditions. Amihud (2002) proposed a simple illiquidity measure — average daily |return|/volume — that predicts cross-sectional and time-series stock returns. Pastor and Stambaugh (2003) showed that stocks with high sensitivity to aggregate liquidity shocks earn a premium of about 7.5% per year. Liu (2006) constructed a multi-period measure of trading continuity that also predicts returns. Together, these papers establish that liquidity risk is a priced factor distinct from size.

Economic Intuition

The risk-based story is clean: liquidity risk is systematic because it spikes in market downturns when investors most need to sell. Assets with high liquidity betas lose value precisely when the marginal investor's wealth and risk tolerance are lowest — a bad combination that warrants a risk premium. Transaction costs provide a direct economic mechanism: if trading an illiquid stock costs 1–3% round-trip, the gross return premium must exceed this threshold for net returns to be positive. This creates a natural floor on how much the liquidity premium can be arbitraged.

Out-of-Sample Evidence

Moderate OOS survival

The illiquidity premium is real but inherently hard to harvest for large investors. Transaction costs in illiquid stocks often consume the entire gross premium, leaving nothing net. The premium survives post-publication precisely because it is self-limiting: any large position in illiquid stocks destroys its own premium through market impact. For small accounts (as in this platform's simulation environment), liquidity signals are implementable, but the lesson is to always model transaction costs explicitly.