Definition
Tail conditional expectation is the expected loss in a specified upper tail, usually beyond a loss quantile, under a stated convention for atoms at the threshold. In continuous distributions it is commonly written as expected loss conditional on exceeding VaR. In discrete samples, a quantile-integral expected shortfall is often the less ambiguous definition.
How it is calculated
Tail conditional expectation = expected loss L conditional on L being at least its alpha VaR, for the stated continuous-tail convention.
| Symbol | Meaning | Unit |
|---|---|---|
| TCE | tail conditional expectation | loss units |
| alpha | confidence level | probability |
| L | loss variable | loss units |
| VaR | loss quantile | loss units |
| E | conditional expectation | loss units |
Interpretation
A threshold tells you where the tail begins; tail conditional expectation summarizes what lies there. The phrase 'given beyond' becomes ambiguous when many observations land exactly on the threshold. That is why discrete implementations need a convention.
Assumptions and conventions
-
The loss convention and tail event are explicit.
-
The confidence level satisfies 0<alpha<1, and the upper-tail expectation is finite.
-
For atoms at VaR, the allocation convention is specified or ES is defined by quantile integration.
Loss units over the stated horizon.
Limitations and common misreadings
-
TCE and ES can differ under atom conventions despite similar names.
-
Tail conditional expectation is not a probability of loss.
References
- Carlo Acerbi, Dirk Tasche. (2002). Expected Shortfall: A Natural Coherent Alternative to Value at Risk. Economic Notes. Whole article. doi:10.1111/1468-0300.00091. Source

