“Global Stock Markets in
the Twentieth Century”
Philippe Jorion And William N.
Goetzmann 1999
, Background: The Equity Premium Puzzle
The Mehra & Prescott (1985) Problem:
• U.S. equities earned about 6% average equity risk premium (1889–1978).
• Standard general equilibrium models cannot explain such a large premium.
• To match the data, the model requires an unrealistically high coefficient of risk aversion (much
higher than the usual value ≈ 2).
• But such high risk-aversion seems unrealistic.
This creates the equity premium puzzle.
, Motivation: Data Limitations vs. Model Failure
Alternative Explanations Focus on Data: Instead of changing preferences, some researchers argue that
the problem lies in the data
1. Rare Disaster Hypothesis (Rietz, 1988)
Assume a small probability of a large crash (e.g., 50% output drop with 0.4% annual probability of a
crash). Even though crashes are very rare, their possibility makes stocks risky enough to justify a high
expected premium before outcomes are realized. But the problem is that such crashes are extremely
rare (one every 250 years), making estimates unreliable based on U.S. data alone.
2. Survivorship Bias (Brown, Goetzmann & Ross, 1995)
Observed ex post returns may be biased upward because we only observe markets that survived. Even
a zero ex-ante premium could look positive if we condition on survival. So, risk aversion cannot be
inferred from data that excludes failed markets.
So, only around 100 years of U.S. data is not enough to solve the puzzle. Therefore, the authors
decided to collect data from 39 countries during the 20th century.
the Twentieth Century”
Philippe Jorion And William N.
Goetzmann 1999
, Background: The Equity Premium Puzzle
The Mehra & Prescott (1985) Problem:
• U.S. equities earned about 6% average equity risk premium (1889–1978).
• Standard general equilibrium models cannot explain such a large premium.
• To match the data, the model requires an unrealistically high coefficient of risk aversion (much
higher than the usual value ≈ 2).
• But such high risk-aversion seems unrealistic.
This creates the equity premium puzzle.
, Motivation: Data Limitations vs. Model Failure
Alternative Explanations Focus on Data: Instead of changing preferences, some researchers argue that
the problem lies in the data
1. Rare Disaster Hypothesis (Rietz, 1988)
Assume a small probability of a large crash (e.g., 50% output drop with 0.4% annual probability of a
crash). Even though crashes are very rare, their possibility makes stocks risky enough to justify a high
expected premium before outcomes are realized. But the problem is that such crashes are extremely
rare (one every 250 years), making estimates unreliable based on U.S. data alone.
2. Survivorship Bias (Brown, Goetzmann & Ross, 1995)
Observed ex post returns may be biased upward because we only observe markets that survived. Even
a zero ex-ante premium could look positive if we condition on survival. So, risk aversion cannot be
inferred from data that excludes failed markets.
So, only around 100 years of U.S. data is not enough to solve the puzzle. Therefore, the authors
decided to collect data from 39 countries during the 20th century.
