Tail ratio: how the right and left tails of a return distribution compare
Two assets can have the same standard deviation but very different return distributions. Some compensate investors with frequent small gains and rare large losses (negative skew); others provide rare large gains alongside frequent small losses (positive skew). The tail ratio captures this asymmetry directly by comparing the magnitudes of the two tails.
What the tail ratio is
The tail ratio is defined as the absolute value of a high return percentile divided by the absolute value of the corresponding low return percentile: tail ratio = |R(95%)| / |R(5%)|, where R(p) denotes the p-th percentile of the return distribution. Higher percentiles (e.g., 99% / 1%) emphasise the extremes more; lower (e.g., 90% / 10%) emphasise more typical fluctuations. The 95% / 5% convention is the most common.
A tail ratio above 1 means the right tail is larger in magnitude than the left—characteristic of positively skewed assets that produce occasional large gains alongside more frequent small losses. A tail ratio below 1 means the left tail dominates—characteristic of negatively skewed assets that deliver frequent small gains and rare large losses. A tail ratio close to 1 indicates approximate symmetry.
The metric is closely related to skewness but has different practical properties. Skewness is sensitive to extreme outliers because it weights observations by their cubed deviation from the mean. The tail ratio uses fixed percentiles and is therefore robust to a single very large observation; the price for that robustness is that it ignores everything beyond the chosen percentile cut-offs.
How it works
For a return series, the tail ratio is computed by sorting the observations from smallest to largest, taking the values at the 5th and 95th percentile positions, and dividing the absolute values. For a 1,000-observation series, the 5th percentile is the 50th-smallest observation and the 95th percentile is the 50th-largest. For a smaller sample, interpolation between adjacent observations is sometimes used to produce a smoother estimate.
An asset with returns of −5%, −2%, +1%, +3%, and +10% (in a five-observation series, an extreme example) has a tail ratio computed roughly as 10/5 = 2.0—a positively skewed profile. A symmetric asset whose largest gain and largest loss are roughly equal in magnitude has a tail ratio close to 1.
The metric scales with the chosen confidence level. Tail ratio at 99% / 1% will typically be larger or more extreme than at 95% / 5% for the same asset, because the more extreme percentiles capture rarer events that often have stronger asymmetry.
What the evidence shows
Equity buy-and-hold portfolios typically have tail ratios below 1 at the 95% / 5% level—the long left tail of equity returns dominates the right tail. The 5th percentile of monthly S&P 500 returns over multi-decade samples is typically around −7%, while the 95th percentile is around +6%; tail ratio approximately 0.85, mildly negatively skewed.
Diversified trend-following strategies typically have tail ratios above 1. The CTA / managed futures literature documents this property as one of the strategy's structural features: the strategy by construction rides positive trends and exits negative ones, producing a return distribution with a long right tail (large gains during sustained bull or bear trends) and a tighter left tail (losses cut by stop-out or signal reversal). Tail ratios above 1.5 are common for well-diversified trend-following implementations.
Volatility-selling strategies show the opposite pattern. The structural feature of selling options or volatility is collecting premium reliably while accepting the rare large loss; the resulting return distribution has a tight right tail and a long left tail, with tail ratios well below 1 in most realistic samples. The metric quickly identifies this profile when applied to an unfamiliar return series.
Limitations and trade-offs
The tail ratio depends on the chosen percentiles. A 95% / 5% ratio and a 99% / 1% ratio for the same asset can produce very different numbers, and comparisons across strategies should use a consistent convention. The 95% / 5% ratio is the most common and produces the most stable estimates from finite samples; more extreme percentiles produce more dramatic ratios but with much wider confidence intervals.
The metric is also insensitive to the magnitude of returns beyond the chosen percentile. Two strategies with the same 95% / 5% ratio can have very different behaviour at the 99% / 1% level—one might cut losses systematically while the other does not. Tail ratio paired with skewness, kurtosis, and expected shortfall provides a more complete picture than any single metric in isolation.
Like all distribution-based metrics, tail ratio assumes the historical sample is representative of the forward-looking distribution. A strategy that has not yet experienced a tail event will appear to have a benign tail ratio in its backtest; the metric does not anticipate previously unseen extremes.
Tail ratio in pfolio
The tail ratio is not currently displayed in pfolio Insights. Skewness, kurtosis, expected shortfall, and the full return distribution are available; the tail ratio (the ratio of the right-tail to left-tail magnitudes) can be computed externally from the same return series.
Related articles
Disclaimer
Get started now
