Market MicrostructureFactor Investing

The Liquidity Premium: How Bid-Ask Spreads Drive Stock Returns

Summary by Robert Gorak · Published June 19, 2026 · Last reviewed June 19, 2026

Yakov Amihud and Haim Mendelson·1986·Journal of Financial Economics
Sample: 49 (7×7) spread-and-beta portfolios; 980 portfolio-year observationsData: CRSP monthly returns and Fitch's Stock Quotations on the NYSE (bid-ask spreads)Period: 1961–1980

The liquidity premium is the extra return investors require to hold assets with higher trading costs, where the bid-ask spread measures illiquidity. In "Asset Pricing and the Bid-Ask Spread," Amihud and Mendelson (1986) studied NYSE stocks from 1961 to 1980. Using 49 portfolios built from CRSP returns and Fitch quotes, they found a 1% spread increase raised monthly risk-adjusted excess return by 0.211% (t=6.83). The correlation between portfolio excess return and spread was 0.239, roughly twice the 0.123 correlation with beta.

What the Study Found

Excess returns increased in both beta and spread, with a spread coefficient of 0.211 (t=6.83) and a beta coefficient of 0.00672 (t=6.18). The difference in monthly mean excess return between the highest- and lowest-spread groups was 0.857% under OLS and 0.681% under GLS. The beta coefficient of 0.00672 closely matched 0.00671, the average monthly excess return on common stocks over the period. Adding firm size as log(SIZE) left the spread effect intact, while the size coefficient stayed insignificant (t=1.56). A formal F-test for the spread variables given log(SIZE) produced F=2.02, significant at the 0.01 level.

Methodology

The data combined CRSP monthly returns with relative bid-ask spreads from Fitch's Stock Quotations on the NYSE. Stocks were sorted into 49 (7×7) spread-and-beta portfolios, yielding 980 portfolio-year observations across twenty overlapping eleven-year periods. Each eleven-year period used a five-year beta estimation period, a five-year portfolio formation period, and a one-year cross-section test period within 1961–1980. The pooled cross-section/time-series regressions controlled for relative risk (beta) and year fixed effects, estimated by both OLS and GLS.

Key Statistics

Metric Finding Context
Spread coefficient 0.211 (t=6.83) 1% spread rise → +0.211% monthly risk-adjusted excess return, 1961–1980
High vs. low spread group 0.857% (OLS); 0.681% (GLS) Monthly mean excess-return difference, extreme spread groups
Return–spread correlation 0.239 vs. 0.123 Correlation with spread vs. correlation with beta, full sample
Spread-adjusted return r_ij = d_j/V_j − μ_i·S_j Gross return minus expected liquidation cost per unit time (Eq. 4)
Equilibrium gross return d_j/V_j* = min_i {r_i* + μ_i·S_j} Increasing, concave, piecewise-linear return–spread relation (Eq. 6)

Why This Matters

Liquidity is not a friction to be ignored but a priced characteristic of every security. Firms can lower their cost of capital by improving the marketability of their shares. Investors accept lower gross returns for assets that are cheaper to trade. For portfolio managers, expected trading costs should be weighed directly against expected returns when selecting assets. The result also offers a rational, microstructure-based reading of the small-firm effect, framing part of it as compensation for illiquidity rather than an anomaly.

Frequently Asked Questions

0.211% is the rise in monthly risk-adjusted excess return Amihud and Mendelson (1986) linked to each 1% increase in a stock's bid-ask spread. Their NYSE sample ran from 1961 to 1980. The liquidity premium is the extra return investors demand to hold assets that cost more to trade.

0.857% was the monthly excess-return gap between the highest- and lowest-spread portfolio groups under OLS. Investors demand higher returns for wider spreads, and firms that improve share liquidity reduce their cost of capital. The increasing, concave relation means the premium per unit of spread shrinks as holding periods lengthen.

t=1.56 was the insignificant size coefficient when Amihud and Mendelson (1986) added log firm size alongside the spread, while the spread effect held. An F-test for spread variables given size returned F=2.02, significant at the 0.01 level. They argued the size effect may partly reflect a spread effect.

0.681% monthly separated the highest- and lowest-spread groups under GLS, a gap that narrows per unit of spread because of the clientele effect. Investors with longer holding periods amortize trading costs over more time and hold higher-spread assets, which makes the return–spread relation concave.

Source

Yakov Amihud and Haim Mendelson (1986). Asset Pricing and the Bid-Ask Spread. Journal of Financial Economics.

Read the full paper