Millennium Interview 2027: Local Volatility Model Limitations
The Millennium local volatility model limitations answer: local vol fits today's vanilla smile perfectly but has no volatility dynamics — it predicts unrealistic forward smiles and misprices forward-start exotics. Calibration via Dupire is also numerically fragile. Commonly reported by candidates.
What This Question Assesses
This tests whether you understand models as tools with domains of validity. The interviewer wants to see that you know what local volatility assumes (deterministic volatility as a function of spot and time, complete market, no jumps), what it gets right (exact calibration to the vanilla smile), and where it breaks (exotics, forward-starting products, anything volatility-dynamic). Explaining the reasons, not just listing them, is what counts.
Millennium Local Volatility Model Limitations: How to Answer
- Step 1 — State what the model is. "Dupire local volatility: dS/S = μdt + σ(S,t)dW — volatility is a deterministic function of spot and time, calibrated via Dupire's formula to match all vanilla option prices exactly."
- Step 2 — Name the core limitation. "Because volatility is deterministic given the spot path, the model has no independent volatility dynamics — no vol-of-vol, no volatility clustering. It is a one-factor model wearing a smile."
- Step 3 — Give the practical consequences. "It predicts the forward smile flattens unrealistically, so it misprices forward-starting options, cliquets, and other forward-vol-dependent exotics. Hedging volatility exposure is also conceptually off, since vega hedges assume volatility moves that the model denies."
- Step 4 — Mention calibration fragility. "Dupire's formula needs second derivatives of call prices in strike and first derivatives in maturity — differentiating noisy market data twice is numerically unstable, so the surface requires heavy regularisation."
An example line: "Local vol is a perfect static photograph of today's smile with no theory of how the smile moves — so it prices vanillas exactly but forward-vol exotics wrongly, which is why desks use stochastic or local-stochastic volatility where forward dynamics matter."
Millennium Local Volatility Model Limitations: Common Mistakes
- Saying "it doesn't fit the smile." It fits the smile perfectly by construction — that is its defining feature. The limitations are about dynamics, not static fit.
- Confusing local vol with implied vol. Implied vol is a quoting convention per option; local vol is a model of instantaneous volatility as a function of (S, t). Mixing them up is a fundamental error.
- No mention of alternatives. A strong answer names what replaces it where it fails: stochastic volatility (Heston), local-stochastic vol, or jump models, matched to the product.
Model-critique questions are where quant interviews separate users of models from understanders of models — always answer with the mechanism, not just the list.
Keep Reading
- societe generale client price urgent
- Millennium Caliper Test Questions: Personality Section Breakdown
- Millennium Caliper Test Tips: Surviving the 1.5-Hour Test
FAQ
What is Dupire's formula in one sentence? It expresses local volatility σ(K,T) in terms of market call prices' derivatives: essentially the risk-neutral expectation linking the smile's shape to instantaneous volatility as a function of strike and maturity.
Why does local vol misprice forward-start options? Because its deterministic structure forces the forward smile to flatten over time, while markets typically preserve forward skew — the model understates the value of forward volatility exposure.
What is local-stochastic volatility (LSV)? A hybrid: a stochastic volatility process multiplied by a leverage function calibrated to the vanilla smile — it keeps the exact static fit of local vol while adding realistic volatility dynamics.
Preparing for Millennium's interview? Our 2027 Millennium Caliper Assessment and Quantitative Assessment Exact Questions and Answers has practice questions and answers — $79 one-time, instant download.













































