Fama & French (1992) Cross-Sectional Replication

VCU – Advanced Financial Analytics (FIRE 691) | Weeks 3–5 Assignments

Full replication of Fama & French (1992) Tables 1–3 — the landmark cross-sectional asset pricing study — using 1.04 million CRSP-Compustat stock-month observations from 1962–2025.

---

Overview

This project tests whether size (market cap) and book-to-market (B/M) ratios explain the cross-sectional variation in stock returns, as documented by Fama & French. The analysis replicates their sorting methodology and extends it to modern data.

---

Data

Dataset	Coverage	Size
CRSP_monthly_1962_1989.csv	1962–1989	78.6 MB
CRSP_monthly_1990_2025.csv	1990–2025	283.2 MB
crsp_comp_ff1992.csv	Merged CRSP-Compustat	635.4 MB
compustat_60_25.csv	Compustat financials	38.9 MB

Total: 1.04M+ stock-month observations

---

Methodology

Weeks 3–4: FF1992 Replication

• Merge CRSP daily returns with Compustat annual book equity (6-month lag to avoid look-ahead bias)
• Compute size (ME = price × shares) and B/M = book equity / market equity
• Sort into 25 size × B/M portfolios (5×5) using NYSE-only breakpoints
• Compute value-weighted returns for each portfolio monthly
• Estimate FF3 factor model loadings via OLS

Week 5: Cross-Sectional Fama-MacBeth Regressions

• Monthly cross-sectional regression of returns on lagged size, B/M, and beta
• Fama-MacBeth standard errors for inference
• Replicate Tables 1, 2, and 3 from FF (1992)

---

Key Results

Table	Finding	Match to FF (1992)
Table 1	Small stocks earn higher returns than large stocks	✅ Within tolerance
Table 2	High B/M stocks outperform low B/M stocks	✅ Value premium confirmed
Table 3	Beta alone cannot explain cross-section	✅ Beta flat after controlling for size

Heatmap visualization of 25 portfolio average returns confirms the size and value effects — returns increase monotonically from large-growth to small-value portfolios.

---

Output Files

FF1992_heatmaps.png             # 5×5 size/B-M portfolio return heatmap
Table1_FF1992_1962_1989.png     # Replicated Table 1
ff3_comparison.png              # FF3 factor comparison vs. published series
my_ff3_factors.csv              # Constructed FF3 monthly factors
my_ff3_factors_complete.csv     # Final cleaned FF3 output
six_portfolio_returns.csv       # 6 size-BM portfolio returns

---

Tech Stack

Python pandas numpy statsmodels matplotlib Google Colab WRDS/CRSP Compustat

---

Virginia Commonwealth University · MS Business (Financial Analytics) · FIRE 691 Reference: Fama & French (1992), Journal of Finance