Ex post (historical) measures can mislead analysts when trying to estimate ex ante (forward-looking) risk and return, due to two key biases:
- The Peso Problem (Underestimating Ex Ante Risk):
Sometimes, markets price in the possibility of rare, catastrophic events that never happen during the sample period.
Ex post returns will appear high, and measured risk low, since the feared event didn't occur.
Analysts may underestimate true forward-looking risk and overestimate expected returns.
🧠Example: The Argentine peso was pegged 1:1 to the USD from 1992 until the 2002 crisis. Before the collapse, historical data suggested stability, but the market had priced in the risk of a potential collapse. Once it happened, the peso rapidly devalued, showing the real risk was higher than historical returns implied.
- Rare Event Overreaction (Overestimating Risk):
If a rare, extreme event does occur within a short historical sample, it can distort risk estimates like VaR.
These measures may overstate how likely such extreme losses are going forward.
🧠Example: In March 2020, Meta’s stock dropped –14.3% on one day. This led to an inflated daily 5% VaR of –13.4%, implying such a loss might happen every 20 days—a clear overestimate, as no similar drop occurred in the next 19 months.
🔑 Implications for Analysts:
Context matters: Analysts must assess whether risks were priced in, and whether observed outcomes were representative or exceptional.
Use stress testing, scenario analysis, and forward-looking judgment, not just mechanical extrapolation of historical data.
Recognize the limitations of VaR and other risk metrics based solely on historical samples, especially when rare events are involved.