Skip to content

Leverage effect

The equity return–volatility asymmetry, its balance-sheet explanation and why a fitted volatility response does not establish an economic cause.

The leverage effect is the tendency for negative equity returns to be associated with higher subsequent volatility, often with a stronger volatility response than positive returns of similar size. For a quant forecasting risk, the sign of a return surprise may therefore contain information that its squared magnitude alone misses.
The name also describes a proposed explanation: a fall in equity value raises a company's debt relative to equity, making the remaining equity more sensitive to asset-value changes. The observed asymmetry and this corporate-finance mechanism are related ideas, but observing the first does not prove the second. Bekaert and Wu (2000) explicitly separate asymmetric volatility from its potential leverage and volatility-feedback explanations.

A smaller equity cushion

Consider an illustrative company with asset value AA, debt value DD and equity value E=A−DE=A-D. All values below use the same arbitrary value unit. Assume debt has fixed value and is locally riskless, while assets have unchanged annualized return volatility σA=20%\sigma_A=20\%. These assumptions isolate the balance-sheet mechanism.

With debt locally unchanged, dE=dAdE=dA. An asset-value movement of one unit changes equity by one unit, but represents a larger percentage of the smaller equity base:

dEE=AEdAA,σE=AEσA.\frac{dE}{E}=\frac{A}{E}\frac{dA}{A}, \qquad \sigma_E=\frac{A}{E}\sigma_A.

Here σE\sigma_E is the local annualized equity-return volatility implied by these assumptions, not a forecast of the standard deviation of the next year's realised return.

Balance-sheet stateAsset valueFixed debt valueEquity valueAssets / equityLocal equity volatility
Before decline10040605/35/333.33%33.33\%
After decline9040501.81.836.00%36.00\%
Assets fall by 10%10\%, while equity falls by 16.67%16.67\%. Without any change in asset-return volatility, the local equity-volatility estimate rises by about 2.67 percentage points. This is corporate balance-sheet leverage; it does not require the shareholder to borrow personally or trade on margin.
Illustrative balance sheets: assets fall from 100 to 90, debt stays 40 and equity falls from 60 to 50. At unchanged annualized asset volatility of 20%, local equity volatility rises from 33.33% to 36.00%.
Illustrative balance sheets: assets fall from 100 to 90, debt stays 40 and equity falls from 60 to 50. At unchanged annualized asset volatility of 20%, local equity volatility rises from 33.33% to 36.00%.
Original illustrative balance-sheet mechanism in arbitrary value units. Debt value is fixed and locally riskless; annualized asset-return volatility stays at 20%. These are model calculations, not observed market data.

The relation is local. In a full structural equity model, debt value can change with asset value and equity has limited liability. Close to default, extending the fixed-debt formula mechanically is especially misleading. The example demonstrates amplification under stated assumptions, rather than measuring how much of an observed market effect corporate leverage explains.

How a volatility model uses the idea

A symmetric GARCH model reacts equally to positive and negative return surprises of the same magnitude because it uses their squares. To represent an asymmetric response, a threshold variance model can add a term that is active only after negative surprises:

ht+1=ω+αεt2+γ1{εt<0}εt2+βht.h_{t+1}=\omega+\alpha\varepsilon_t^2 +\gamma\mathbf{1}_{\{\varepsilon_t<0\}}\varepsilon_t^2+\beta h_t.
Here εt\varepsilon_t is the return minus its conditional mean, hth_t is conditional return variance, and the indicator is one for a negative surprise and zero otherwise. Under the sufficient nonnegativity restrictions ω>0\omega>0 and α,β,γ≥0\alpha,\beta,\gamma\geq0, choosing γ>0\gamma>0 gives a larger next-period variance response to a negative surprise than to an equally large positive surprise, holding current variance fixed. This is a statistical representation of asymmetry, not an estimate of the company's debt ratio. McNeil, Frey and Embrechts develop this distinction and threshold specification in §4.2.3.
For the symmetric baseline, SelfMadeQuant's ARCH and GARCH lesson explains the variance recursion and its news-impact curve. The full lesson requires membership; the linked page has a public preview.

What the label does—and does not—tell you

Several distinctions matter when reading results:

  • Causality: volatility feedback offers another channel. If greater expected risk raises the return investors require, prices can fall when volatility expectations rise. News can also affect prices and uncertainty together. In their Nikkei-stock application, Bekaert and Wu find evidence against a pure leverage explanation and in favour of volatility feedback; this is a study-specific result, not a universal ranking of mechanisms.
  • Timing and measurement: a same-period return–volatility association, a lagged return–future-volatility relation and an asymmetric fitted variance response are different objects. State the horizon, volatility measure and conditioning information when estimating an effect.
  • Other stylised facts: volatility clustering concerns persistence in return magnitudes. Negative skewness concerns the marginal return distribution. Neither property alone establishes the leverage effect.

For forecasting, the useful question is whether return sign improves the conditional risk forecast for the instrument and horizon being modelled. A fitted asymmetry parameter describes that model's response; identifying an economic cause requires further evidence.

Sources

  • Alexander J. McNeil, Rüdiger Frey and Paul Embrechts, Quantitative Risk Management: Concepts, Techniques and Tools, revised edition (2015), §4.2.3, pp. 122–123: leverage terminology, asymmetric variance responses and threshold GARCH.
  • Geert Bekaert and Guojun Wu, “Asymmetric Volatility and Risk in Equity Markets”, The Review of Financial Studies 13(1), 2000, pp. 1–42: empirical asymmetry and competing leverage/volatility-feedback explanations. The linked university research page provides the authors' abstract and citation.
Last updated
  • volatility
  • equity risk
  • leverage
  • asymmetric volatility
  • garch
Keep connecting the ideas
← Back to the glossaryBack to top ↑