OverviewAnomalies Data

Chapter 2 — Market Efficiency & Arbitrage

Market Anomalies Data Viewer

95 years of monthly return data (1926–2020) from the Fama-French Data Library. Explore the Size, Value, and Momentum anomalies that challenge the Efficient Market Hypothesis.

Source: Prof. Ioanid Rosu · HEC Paris1,134 monthly observations4 anomaly factors
Filter period:
1927 – 2020 · 94 years

Market Premium (Mkt-RF)

10.6%

Average annual excess return

Size Premium (SMB)

3.0%

Small minus Big — avg annual

Value Premium (HML)

3.7%

High minus Low BE/ME — avg annual

Momentum Premium (MOM)

7.6%

Winners minus Losers — avg annual

01

Cumulative Wealth

$1 Invested in Each Factor (Year-End)

Starting with $1 at the beginning of the selected period, this chart shows how each Fama-French factor would have grown. The Market factor represents the equity risk premium; SMB, HML, and MOM are zero-cost long-short portfolios.

19271939195119631975198719992011$0.0$650.0$1300.0$1950.0$2600.0
  • Market (Mkt-RF)
  • Momentum (MOM)
  • Value (HML)
  • Size (SMB)
02

Decade Analysis

Average Annual Factor Returns by Decade

Factor premiums are not stable across time. The value premium was strong in the 1940s–1980s but has weakened since 2000. Momentum has been the most consistent factor, though it crashed spectacularly in 2009.

1920s1930s1940s1950s1960s1970s1980s1990s2000s2010s2020s-40%-20%0%20%40%
  • Market
  • Value (HML)
  • Momentum
  • Size (SMB)

What do these anomalies mean for Market Efficiency?

If markets were perfectly efficient (EMH), no systematic strategy should earn excess returns after adjusting for risk. Yet the data shows that small stocks have historically outperformed large stocks (Size effect), value stocks (high book-to-market) have outperformed growth stocks (Value effect), and past winners have continued to outperform past losers over 3–12 month horizons (Momentum). These patterns are either (1) compensation for risk factors not captured by CAPM, (2) behavioral biases (overreaction, underreaction), or (3) data-mining artifacts. The debate remains unresolved — and is central to the Chapter 2 discussion in class.