Skewness and Kurtosis - Why Returns Aren't Normal
By EC Assets Research Team, Risk Research · Published · Updated
Skewness and Kurtosis: Skewness measures the asymmetry of a return distribution and kurtosis measures its tailedness - the third and fourth moments beyond mean and volatility. Financial returns are negatively skewed and fat-tailed, which is why tools that assume normality (VaR, Sharpe, mean-variance) systematically understate crash risk.
Beyond Mean and Variance
The mean and the standard deviation describe the centre and the spread of a return distribution, but not its shape. Two portfolios can share an average return and a volatility and still behave completely differently in the tails. Skewness and kurtosis, the third and fourth standardised moments, describe that shape: skewness measures asymmetry and kurtosis measures tailedness.
$$\text{Skewness} = \frac{E[(r-\mu)^3]}{\sigma^3}, \qquad \text{Kurtosis} = \frac{E[(r-\mu)^4]}{\sigma^4}$$
A normal distribution has skewness 0 and kurtosis 3. The amount by which kurtosis exceeds 3 is excess kurtosis; positive values mean fatter tails than the normal distribution.
How It Works
Skewness. Negative skew means a longer left tail: frequent small gains and occasional large losses. Equity indices, option selling and carry trades are negatively skewed. Positive skew means the reverse: frequent small losses and occasional large gains, typical of long option positions and trend following.
Kurtosis. High excess kurtosis means a larger share of the variance comes from rare, extreme observations. Mandelbrot (1963) showed that price changes are far more fat-tailed than the normal distribution allows, and Cont (2001) lists heavy tails and gain-loss asymmetry among the stylised facts of asset returns.
Investors are not indifferent between the two kinds of skew. Harvey and Siddique (2000) show that skewness is priced: assets that add negative skew to a portfolio must offer higher expected returns, and those that add positive skew can offer lower ones. That is why strategies that sell crash protection tend to earn a premium, the logic behind the volatility risk premium and the carry trade. What matters for pricing is coskewness, an asset's contribution to the skewness of the whole portfolio, rather than its standalone skewness: an asset that tends to crash when the market crashes is worse than one whose crashes are its own.
Worked Example
S&P 500 log returns from Cboe's daily closes, January 1975 to October 2026:
| Sample | Observations | Skewness | Excess kurtosis |
|---|---|---|---|
| Daily, all days | 13,054 | -1.02 | 24.54 |
| Daily, without 19 Oct 1987 | 13,053 | -0.34 | 10.70 |
| Monthly | 621 | -0.78 | 2.56 |
A normal distribution has 0 and 0 in the last two columns.
With the full-sample daily standard deviation of 1.095 percent, the index fell by more than four standard deviations on 47 days and by more than five on 21 days in 51.8 years. A normal distribution would expect about 0.41 falls beyond four and 0.004 beyond five over the whole period. On 19 October 1987 alone the log return was -22.9 percent, a move of about 21 standard deviations.
Now a hypothetical fund that gains 1.5 percent in eleven months and loses 10 percent in the twelfth. Its mean monthly return is 0.54 percent and its standard deviation 3.18 percent, unremarkable figures. But its skewness is -3.02 and its excess kurtosis 7.09. A normally distributed fund with the same mean and volatility would lose 10 percent in a month about once every 183 years; this one does it every year.
When It Applies (and Limitations)
Estimates are fragile. Higher moments depend on a few extreme observations. One day moved the S&P 500's excess kurtosis from 10.7 to 24.5, and a short sample without a crisis can make a negatively skewed strategy look normal.
Estimators differ. Statistical packages report population or bias-corrected sample versions, and some report kurtosis rather than excess kurtosis. Before comparing figures across sources, check which definition was used; for small samples the differences are material.
Horizon matters. Kurtosis falls as returns are aggregated over longer periods, but it does not vanish: monthly returns are still fat-tailed and negatively skewed.
Volatility clustering inflates kurtosis. Part of the fat tails reflects changing volatility: a calm-period standard deviation makes crisis days look more extreme. Scaling returns by a time-varying volatility forecast reduces, but does not remove, the excess kurtosis.
Moments are not the whole tail. Two distributions can share skewness and kurtosis and still differ in their worst outcomes. Direct tail measures such as expected shortfall complement the moments.
Why It Matters for Institutional Investors
Hidden risk in smooth returns. Strategies that sell insurance, including short volatility, credit and carry, combine attractive Sharpe ratios with negative skew and fat tails. The higher moments are how that exposure shows up before a loss.
Model risk. Parametric value at risk, mean-variance optimisation and the Sharpe ratio behave as if returns were normal. Under fat tails, normal-based value at risk understates losses and the Sharpe ratio flatters negatively skewed strategies.
Due diligence. Asking for the skewness, kurtosis and worst months of a track record, and checking whether the record covers a stress period, reveals a risk profile that average and volatility hide.
Portfolio design. Combining negatively skewed return sources with positively skewed ones, such as trend following or option protection, can improve the shape of the overall distribution even if it costs some average return, which matters most for investors who cannot ride out a deep loss.
References
- Mandelbrot, B. (1963). The Variation of Certain Speculative Prices. Journal of Business, 36(4), 394-419.
- Cont, R. (2001). Empirical Properties of Asset Returns: Stylized Facts and Statistical Issues. Quantitative Finance, 1(2), 223-236.
- Harvey, C. R., & Siddique, A. (2000). Conditional Skewness in Asset Pricing Tests. Journal of Finance, 55(3), 1263-1295.
- McNeil, A. J., Frey, R., & Embrechts, P. (2015). Quantitative Risk Management: Concepts, Techniques and Tools (Revised ed.). Princeton University Press.
Frequently asked questions
What is the difference between skewness and kurtosis?
Skewness measures asymmetry: whether large surprises come mostly from the left or the right tail. Kurtosis measures tailedness: how much of the variance comes from rare, extreme observations rather than ordinary ones. A distribution can be symmetric with very fat tails, or skewed with thin ones.
Why are equity returns negatively skewed?
Markets tend to rise gradually and fall sharply, because selloffs bring forced selling, rising volatility and correlated exits. On S&P 500 daily data since 1975 skewness is -1.02, dominated by the crash of 19 October 1987, and monthly returns are also negatively skewed at -0.78.
What does excess kurtosis mean in practice?
That extreme moves happen far more often than a normal distribution predicts. With the full-sample standard deviation, the S&P 500 had 47 daily falls beyond four standard deviations and 21 beyond five since 1975; a normal distribution would expect fewer than half a fall beyond four and essentially none beyond five.
Why do investors accept negative skew?
Because they are paid for it. Strategies with negative skew, such as selling options, carry trades and much of credit, usually earn a premium precisely because most investors dislike sudden large losses and pay to avoid them. The premium compensates the seller for holding the crash risk others shed.
Are skewness and kurtosis reliable statistics?
Less than mean and variance. Both depend heavily on a few extreme observations: removing the single day of 19 October 1987 cuts the S&P 500's daily excess kurtosis from 24.5 to 10.7. Short samples without a crisis can make a dangerous strategy look normal.
Stay informed
Market commentary, firm news and research from EC Assets - direct to your inbox.