[Paper Review] Why are quadratic normal volatility models analytically tractable?

Peter Carr, Travis C. Fisher+1 more|RePEc: Research Papers in Economics|Feb 28, 2012

Stochastic processes and financial applications19 citations

TL;DR

This paper explains why Quadratic Normal Volatility (QNV) models are analytically tractable by showing they arise from a transformation of stopped Brownian motion and a measure change dependent only on the terminal value of the process. The key insight is that this structural relationship enables explicit analytic formulas for option prices, even under strict local martingale dynamics.

ABSTRACT

We discuss the class of "Quadratic Normal Volatility" models, which have drawn much attention in the financial industry due to their analytic tractability and flexibility. We characterize these models as the ones that can be obtained from stopped Brownian motion by a simple transformation and a change of measure that only depends on the terminal value of the stopped Brownian motion. This explains the existence of explicit analytic formulas for option prices within Quadratic Normal Volatility models in the academic literature.

Motivation & Objective

• To explain the mathematical origin of analytic tractability in Quadratic Normal Volatility (QNV) models.
• To characterize QNV models as transformations of stopped Brownian motion under a change of measure based on terminal values.
• To establish connections between QNV processes and geometric Brownian motion through alternative transformations.
• To formalize the stability of QNV models under changes of numéraire, relevant for foreign exchange and multi-asset derivatives.
• To interpret strict local martingale dynamics in QNV models as a hyperinflation mechanism under a dominating measure, resolving pricing paradoxes.

Proposed method

• Derives QNV processes as time-changed and measure-transformed stopped Brownian motions, with the change of measure depending solely on the terminal value of the stopped process.
• Uses Feller's test for explosions to determine conditions under which QNV processes are true martingales or strict local martingales.
• Applies a transformation linking a subclass of QNV processes to geometric Brownian motion, enabling replication and numéraire change analysis.
• Introduces a symbolic representation of minimal joint replicating prices for path-dependent claims, using expectations under a transformed measure.
• Employs a technical lemma to prove path independence of time integrals under specific conditions, ensuring consistency in pricing formulas.
• Uses the concept of semistatic hedging and adjusted pricing rules to handle cases where QNV processes are strict local martingales, avoiding paradoxes in put-call parity.

Experimental results

Research questions

• RQ1 Why are Quadratic Normal Volatility models analytically tractable despite their non-linear volatility structure?
• RQ2 How can QNV processes be systematically derived from Brownian motion via transformation and measure change?
• RQ3 Under what conditions do QNV processes behave as true martingales versus strict local martingales?
• RQ4 How does the connection between QNV models and geometric Brownian motion facilitate changes of numéraire in multi-currency or multi-asset settings?
• RQ5 What is the economic interpretation of strict local martingale dynamics in QNV models, particularly in terms of hyperinflation or minimal hedging costs?

Key findings

• QNV models are analytically tractable because they can be constructed from stopped Brownian motion via a deterministic transformation and a measure change that depends only on the terminal value of the process.
• The martingale property of QNV processes is determined by the roots of the quadratic polynomial in the volatility function: they are true martingales if the polynomial has two real roots or if the leading coefficient is zero.
• A subclass of QNV processes can be mapped to geometric Brownian motion via a transformation, enabling the use of standard numéraire change techniques in foreign exchange and interest rate modeling.
• The minimal joint replicating cost for a claim denominated in one currency and settled in another is given by an expectation involving a time-changed and measure-transformed Brownian motion, with explicit formulas derived for such costs.
• When QNV processes are strict local martingales, the pricing paradoxes (e.g., violation of put-call parity) are resolved by using adjusted prices based on sums of expectations, which can be interpreted as minimal hedging costs.
• The symbolic pricing formula (19) consistently yields the correct minimal cost for a claim like one Euro in Dollars, confirming that the adjusted pricing rule is arbitrage-free and economically meaningful.