Downside Risk Adjustment for Corporate Operating Risk

Motivation

Traditional downside measures such as semivariance ignore all positive observations. While appropriate in some contexts, they discard useful information.

The proposed approach retains all observations while adjusting total volatility according to whether variability has historically been dominated by downside or upside outcomes.

Step 1. Compute total volatility

$\sigma = \mathrm{StdDev}(g_t) $

where $g_t$ is the detrended growth series.

Step 2. Compute upside and downside deviations

Using deviations from the expected growth path (preferred) or, initially, the sample median:

$d_t=g_t-\hat g_t$

Define

$U= \sum * d_t \ d_t,\qquad$ $D= \sum * d_t|d_t|]$

Step 3. Downside asymmetry factor

$A= \frac{D-U}{D+U}$

Properties:

(A=0): balanced upside and downside
(A>0): downside dominates
(A<0): upside dominates
$(-1 \le A \le1)$

Step 4. Amplification factor

$\text{Amplification Factor} = 1+A$

Step 5. Downside-adjusted volatility

[ \sigma{=tex}_{\rm adj{=tex}}=\sigma{=tex}(1+A) ]

Interpretation

Separate two concepts:

Volatility ((\sigma{=tex})) — magnitude of fluctuations.
Downside asymmetry ((A)) — whether fluctuations are predominantly unfavorable.

Advantages

Uses all observations.
Preserves the interpretation of standard deviation.
Produces a bounded asymmetry measure.
Avoids the instability of the ratio (D/U).
Easy to explain.

Suggested Reporting

For each company report:

Total volatility ((\sigma{=tex}))
Downside asymmetry factor ((A))
Amplification factor ((1+A))
Downside-adjusted volatility ((\sigma{=tex}_{\rm adj{=tex}}))

Use deviations from the expected growth path rather than the sample median:

[ d_t=g_t-\hat {=tex}g_t ]

This compares upside and downside relative to expected performance rather than the historical center of the sample.