Factor InvestingMarket Efficiency

The Fama-French Two-Factor Model: Size and Value Drive Stock Returns

Summary by Robert Gorak · Published June 9, 2026 · Last reviewed June 9, 2026

Eugene F. Fama and Kenneth R. French·1992·Journal of Finance
Sample: approximately 2,267 stocks per month on averageData: CRSP / COMPUSTAT merged files (NYSE, AMEX, NASDAQ)Period: July 1963 to December 1990

Fama and French (1992) studied "The Cross-Section of Expected Stock Returns" and found that size (ME) and book-to-market equity (BE/ME) capture average stock returns. They analyzed approximately 2,267 NYSE, AMEX, and NASDAQ stocks per month from July 1963 to December 1990. The highest BE/ME portfolio returned 1.83% per month; the lowest returned 0.30% — a spread of 1.53% per month. Market beta's univariate slope was 0.15% per month (t = 0.46) — not reliably different from zero.

What the Study Found

The average FM slope on ln(BE/ME) alone was 0.50% per month (t = 5.71). The average FM slope on ln(ME) alone was -0.15% per month (t = -2.58). In the bivariate regression, the BE/ME slope was 0.35% per month (t = 4.44). The size slope was -0.11% per month (t = -1.99). The size-return spread was 0.74% per month (1.64% vs. 0.90%). Within size deciles, the BE/ME spread averaged 0.99% per month (1.63% vs. 0.64%).

Methodology

The study used CRSP and COMPUSTAT data for all non-financial NYSE, AMEX, and NASDAQ stocks. The sample averaged approximately 2,267 stocks per month across 330 monthly observations from July 1963 to December 1990. Monthly Fama-MacBeth regressions used beta, ln(ME), ln(BE/ME), ln(A/ME), ln(A/BE), and E/P as explanatory variables. Post-ranking betas from 100 size-beta portfolios were assigned to individual stocks to reduce measurement error.

Key Statistics

Metric Finding Context
BE/ME return spread 1.53% per month Highest BE/ME (1.83%) vs. lowest (0.30%); 1963-1990
Size return spread 0.74% per month Smallest ME (1.64%) vs. largest (0.90%); 1963-1990
FM slope on ln(BE/ME) alone 0.50%/mo (t = 5.71) Univariate regression; 1963-1990
FM slope on ln(ME) alone -0.15%/mo (t = -2.58) Univariate regression; 1963-1990
FM slope on beta alone 0.15%/mo (t = 0.46) Not reliably different from zero; 1963-1990
FM slope on beta with size -0.37%/mo (t = -1.21) beta adds no power once size is included
FM slope on ln(BE/ME) with size 0.35%/mo (t = 4.44) Bivariate regression with ln(ME); 1963-1990
Within-decile BE/ME spread 0.99% per month Average across size deciles (1.63% - 0.64%); Table V
ln(BE/ME) = ln(A/ME) - ln(A/BE) Identity BE/ME captures difference between market and book leverage

Why This Matters

Beta-based portfolio evaluation underweights size and value exposures. Strategies targeting small-cap or high-BE/ME stocks embed measurable risk premia that beta alone does not capture. Factor-based benchmarking — using size- and BE/ME-matched benchmarks rather than broad indices — follows directly from this framework.

Frequently Asked Questions

1.53% per month separated the highest and lowest BE/ME portfolios over 1963-1990 in Fama and French (1992). The relation held in the 1963-1976 subperiod (b3 = 0.36, t = 2.96) and in 1977-1990 (b3 = 0.35, t = 3.30). High-BE/ME stocks consistently earned higher average returns across both halves of the sample.

The smallest ME portfolio returned 1.64% per month vs. 0.90% for the largest in Fama and French (1992). That is a spread of 0.74% per month across NYSE, AMEX, and NASDAQ stocks over 1963-1990. The FM slope on ln(ME) was -0.15% per month (t = -2.58) and remained negative in every multivariate specification.

The FM slope on beta alone was 0.15% per month (t = 0.46) — not reliably different from zero. Fama and French (1992) found this across the full 1963-1990 period and both subperiods. The 50-year NYSE appendix (1941-1990) confirmed that once size is controlled, beta adds no explanatory power.

0.35% per month (t = 4.44) was the FM slope on ln(BE/ME) after controlling for size in Fama and French (1992). Portfolios should be evaluated against benchmarks with matched size and BE/ME characteristics. Subperiod t-statistics were 2.96 (1963-1976) and 3.30 (1977-1990) for the BE/ME slope, confirming its reliability.

Source

Eugene F. Fama and Kenneth R. French (1992). The Cross-Section of Expected Stock Returns. Journal of Finance.

Read the full paper