Quant Memo
Advanced

Liquidity-Adjusted VaR

Ordinary VaR assumes you can exit a position at today's quoted price. Liquidity-adjusted VaR adds back the cost of actually getting out, which can dwarf the price-risk component for a large or illiquid position.

Prerequisites: Value at Risk (VaR), Bond Market Liquidity and Dealer Balance Sheets

Ordinary Value at Risk (VaR) answers "how much could the mark-to-market value of my position fall," implicitly assuming that if you needed to sell, you could do so at that same market price, instantly, in full. For a position in a deep, liquid market like large-cap equities, that assumption is close enough to true. For a large corporate bond position, a concentrated small-cap stake, or a crowded trade during a stress event, it is badly wrong: getting out at all requires walking the price against yourself, and the resulting execution cost can be a bigger source of loss than the price move itself. Liquidity-Adjusted VaR (L-VaR) puts a number on that extra, usually ignored, cost.

The analogy

Think of the difference between a house's "listed value" and what you would actually pocket if you needed to sell it by next week. The listed value assumes a patient sale at the going rate. A forced, fast sale means accepting a discount, sometimes a steep one, just to guarantee a buyer shows up in time. Ordinary VaR is the "listed value" risk. Liquidity-adjusted VaR adds in the "forced, fast sale" discount, because in a real crisis you rarely get to sell patiently.

Building the formula

The standard, widely used add-on (Bangia, Diebold, Schuermann & Stroughair) decomposes the bid-ask spread itself as a risky, time-varying quantity, not a fixed constant. Let SS be the relative bid-ask spread (spread divided by mid-price), with mean Sˉ\bar{S} and standard deviation σS\sigma_S estimated from data. The liquidity cost term added to ordinary VaR is

LC=12(Sˉ+zασS)×P,\text{LC} = \frac{1}{2}\left(\bar{S} + z_\alpha \sigma_S\right) \times P,

where PP is the position's market value. In words: half the typical spread (since you only cross half the spread going one direction), plus a cushion for the spread being wider than usual right when you need to sell (scaled by the same confidence multiplier used elsewhere in the VaR calculation), all applied to the position size. Total liquidity-adjusted VaR is then the ordinary price-risk VaR plus this liquidity cost:

L-VaR=VaR+LC.\text{L-VaR} = \text{VaR} + \text{LC}.

In words: the loss you could suffer just from the price moving, plus the extra cost of actually crossing the spread to get out, and that spread itself is treated as something that can widen unpredictably, not as a fixed, known number.

Distribution · normal
-2.000.002.00μvalue →
Within ±1σ 68.3%mean μ 0.00std σ 1.00

Think of this bell curve as describing the bid-ask spread itself, not the return. Most days the spread sits near its average (the bell's center); on a bad day, it can widen into the right tail, exactly the risk this add-on is trying to price in.

liquid stock illiquid bond
Grey is ordinary price-risk VaR, roughly equal for both positions here. The liquidity-cost add-on (top) is a sliver for the liquid stock but a large slice for the illiquid bond, nearly half again as much total risk.

Ordinary VaR prices the risk of the market moving against you. Liquidity-adjusted VaR adds the risk of the spread widening against you at the same time you need to exit, and both risks tend to spike together in a genuine crisis, which is exactly when the difference between the two numbers matters most.

Worked example 1: a liquid large-cap position

A $10 million position in a large, liquid stock. Ordinary 95% VaR (price risk only) is $300,000. The stock's relative spread averages Sˉ=0.05%\bar{S} = 0.05\% with σS=0.02%\sigma_S = 0.02\%. At z=1.65z=1.65:

LC=12(0.0005+1.65×0.0002)×10,000,000=12(0.0005+0.00033)×10,000,00012(0.00083)×10,000,0004,150.\text{LC} = \tfrac{1}{2}(0.0005 + 1.65 \times 0.0002) \times 10{,}000{,}000 = \tfrac{1}{2}(0.0005 + 0.00033) \times 10{,}000{,}000 \approx \tfrac{1}{2}(0.00083)\times 10{,}000{,}000 \approx 4{,}150.

That is, roughly $4,150. Liquidity-adjusted VaR: $300,000 + $4,150 \approx $304,150. The liquidity add-on is barely over 1% of the total, in a genuinely liquid market, this correction is close to a rounding error.

Worked example 2: an illiquid corporate bond position

A $10 million position in a thinly traded high-yield bond. Ordinary 95% VaR is a comparable $300,000. But this bond's relative spread averages Sˉ=1.5%\bar{S} = 1.5\% with σS=0.8%\sigma_S = 0.8\% (spreads on stressed credit can gap out violently). At z=1.65z=1.65:

LC=12(0.015+1.65×0.008)×10,000,000=12(0.015+0.0132)×10,000,000=12(0.0282)×10,000,000=141,000.\text{LC} = \tfrac{1}{2}(0.015 + 1.65 \times 0.008) \times 10{,}000{,}000 = \tfrac{1}{2}(0.015+0.0132)\times 10{,}000{,}000 = \tfrac{1}{2}(0.0282)\times 10{,}000{,}000 = 141{,}000.

That is, $141,000. Liquidity-adjusted VaR: $300,000 + $141,000 = $441,000, a 47% increase over the price-risk-only figure. Two positions with an identical ordinary VaR of $300,000 have liquidity-adjusted VaRs of $304,150 and $441,000, a gap of well over $130,000 that ordinary VaR completely misses, purely because one asset trades in a deep market and the other does not.

What this means in practice

L-VaR is standard in fixed income and structured-credit risk management, where bid-ask spreads are both wide and volatile, and in any risk book with concentrated or crowded positions, where "market depth" itself can vanish exactly when a large holder needs to sell (see Cascading Liquidity Withdrawal In A Selloff). Regulatory capital frameworks have increasingly incorporated liquidity horizons, effectively a related idea: assuming different holding periods for liquidating different asset classes rather than assuming instant exit for everything.

The Bangia et al. formula uses Sˉ\bar{S} and σS\sigma_S estimated from historical, mostly calm spread data, but bid-ask spreads on illiquid assets do not just widen in a crisis, they can become effectively undefined as market makers step away entirely and quoted sizes shrink to nothing. A liquidity-adjusted VaR built on historical spread statistics will still understate crisis-period exit costs, sometimes severely, because it assumes a spread exists to be crossed at all; a full liquidity stress test should be run alongside L-VaR, not instead of it.

Related concepts

Practice in interviews

Further reading

  • Bangia, Diebold, Schuermann & Stroughair (1999), Liquidity on the Outside
  • Jorion, Value at Risk: The New Benchmark for Managing Financial Risk (Ch. 15)
ShareTwitterLinkedIn