Cokurtosis
Cokurtosis measures the co-movement between an asset's returns and the cubed returns of the market, quantifying the asset's contribution to the tail thickness (kurtosis) of the portfolio return distribution. Assets with high cokurtosis amplify extreme outcomes in the portfolio and are riskier than variance-based measures indicate.
Overview
Kurtosis is the fourth standardized moment of a distribution and captures the propensity for extreme outcomes -- fat tails. While univariate kurtosis measures the tail thickness of a single asset's returns, cokurtosis measures how an asset contributes to the tail behavior of the overall portfolio. A portfolio composed of assets with high cokurtosis will experience more frequent and more extreme joint moves than would be expected under a Gaussian model.
Dittmar (2002) and Fang and Lai (1997) showed that systematic kurtosis risk is priced: investors demand higher returns from assets that contribute to portfolio kurtosis because they increase the probability of extreme portfolio losses. This is particularly important for risk management, where the focus is on tail events that can cause catastrophic losses.
Mathematical Formulation
Kurtosis (Univariate)
The fourth standardized moment of a return distribution:
A normal distribution has kurtosis of 3 (or excess kurtosis of 0). Financial returns typically exhibit excess kurtosis (leptokurtosis), meaning more probability mass in the tails than a Gaussian distribution.
Standardized Cokurtosis
The standardized cokurtosis between asset and market measures the asset's contribution to portfolio tail risk:
A high value of indicates that the asset tends to have extreme returns precisely when the market also experiences extreme returns, amplifying the fat tails of the portfolio distribution.
Cokurtosis Tensor
The full cokurtosis structure for assets is captured by a fourth-order tensor:
This tensor has elements in total, though symmetry reduces the number of unique elements to .
Portfolio Kurtosis
The kurtosis of a portfolio with weight vector involves a quadruple sum over the cokurtosis tensor:
In matrix notation: where is the cokurtosis matrix.
Sample Estimation
Given observations, the sample cokurtosis is:
Interpretation
| Cokurtosis Value | Interpretation |
|---|---|
| High positive | Asset amplifies extreme market moves in both directions. Increases portfolio tail risk. Investors demand higher expected returns. |
| Near 3 (standardized) | Consistent with normal co-movement in the fourth moment. No excess tail contribution. |
| Low positive | Asset dampens extreme portfolio outcomes. Desirable for risk management as it reduces tail risk. |
Advantages & Limitations
Advantages
- Tail risk quantification:Directly measures an asset's contribution to portfolio tail risk, which is critical for risk management.
- Priced factor: Empirical evidence suggests systematic kurtosis is priced in expected returns.
- Complete distribution: Together with coskewness, enables four-moment portfolio optimization that accounts for the full shape of the return distribution.
- Crisis detection: High cokurtosis assets serve as early warning indicators for systemic tail risk.
Limitations
- Extreme data requirements: Fourth moments require substantially more data for stable estimation than lower moments.
- Dimensionality curse: The cokurtosis tensor has elements, making it infeasible for large portfolios without strong structural assumptions.
- Sensitivity to outliers: Fourth-moment statistics are extremely sensitive to individual extreme observations.
- Optimization complexity: Incorporating cokurtosis in portfolio optimization leads to fourth-order polynomial programs that are computationally challenging.
- Interpretation difficulty: The economic intuition for fourth moments is less straightforward than for variance or skewness.
References
- Fang, H., & Lai, T.-Y. (1997). "Co-Kurtosis and Capital Asset Pricing." Financial Review, 32(2), 293-307.
- Dittmar, R. F. (2002). "Nonlinear Pricing Kernels, Kurtosis Preference, and Evidence from the Cross Section of Equity Returns." Journal of Finance, 57(1), 369-403.
- Jondeau, E., & Rockinger, M. (2006). "Optimal Portfolio Allocation under Higher Moments." European Financial Management, 12(1), 29-55.