Downside deviation
Standard deviation computed using only the months where returns fell below a target — isolating the risk that actually hurts investors.
Downside deviation refines volatility by ignoring the upside. Investors are not bothered when their fund rises unexpectedly — only when it falls. Downside deviation measures precisely that: the dispersion of returns on the bad side of the target return floor.
What it measures
Standard deviation treats a +10% month and a −10% month as equally "risky" because both deviate from the mean. Downside deviation asymmetrically discards the positive deviations, retaining only the negative ones. The result is a purer measure of investor discomfort and is the denominator in the Sortino Ratio.
How it is computed
Given monthly returns R_i and a target return T (typically the monthly equivalent of the risk-free rate, i.e. R_f / 12):
For each month i:
d_i = min(R_i − T, 0) // zero out months above target
Downside variance = Σ(d_i²) / n
Downside deviation (monthly) = √( Σ(d_i²) / n )
Downside deviation (annual) = monthly_dd × √12
Note: the denominator is n (total months), not just the count of below-target months. This penalises a fund that has many bad months relative to one that has few but severe ones in a proportionally fair way.
Example: 36-month fund return series. 14 months fell below T. Sum of squared deviations = 0.0084.
- Monthly downside deviation = √(0.0084 / 36) = √0.000233 = 1.53%
- Annual downside deviation = 1.53% × √12 = 5.3%
Compare to the fund's full annualised σ of 9.8% — the downside-only risk is considerably lower, reflecting that many of the volatile months were upside months.
How to interpret
- Lower downside deviation = fewer/smaller losses below target. Better for risk-averse investors.
- A large gap between annualised σ and annualised downside deviation indicates a positively skewed return distribution — the fund has many more upside surprises than downside ones. This is attractive.
- A small gap (downside deviation ≈ σ) indicates symmetric risk — gains and losses are equally distributed, which is less favourable.
Typical annual downside deviations for Indian equity funds range 7–15% for large-cap, 10–20% for mid-cap, and 12–22% for small-cap.
Limitations + caveats
Downside deviation depends on the target return T. A higher target floor (e.g. 8% annualised) will classify more months as "below target" and produce a larger downside deviation than a lower floor (e.g. 4%). When comparing across funds, ensure the same T is used — MintByte standardises on the 91-day T-bill rate. The metric is as backward-looking as standard deviation and does not predict future downside.
Related metrics
- Volatility — total standard deviation; downside deviation is its asymmetric counterpart.
- Sortino Ratio — uses downside deviation as its risk denominator; the two metrics are always read together.
- Max Drawdown — a complementary downside measure that captures the single worst episode rather than aggregate dispersion.
Sources
Monthly NAV: AdvisorKhoj API. Target return T = monthly equivalent of 91-day RBI T-bill rate. Downside deviation computed monthly over trailing 36 months; annualised using √12.