Skip to content
All library documents

Proving Incompleteness with Multiple Equivalent Martingale Measures

Article Quant Q&A · Author: Strike

Summary

The document considers a market with one risky asset and a zero-rate risk-free asset, where the stock’s volatility depends on a second Brownian motion independent of the one driving its returns. It constructs a change of measure that removes the stock’s drift, using a bounded Girsanov kernel for the return-driving process and an arbitrary constant kernel for the independent process.

The argument applies the first fundamental theorem to establish absence of arbitrage, then observes that varying the arbitrary kernel produces multiple equivalent martingale measures. By the second fundamental theorem, that non-uniqueness implies market incompleteness. The reasoning relies on the stated model and standard theorem assumptions; it is a conceptual demonstration rather than an empirical trading strategy, and it does not discuss practical estimation or implementation.

Key ideas

  • A risk-free asset with zero interest remains constant under the model.
  • The return-driving Brownian motion’s drift can be removed with a bounded Girsanov kernel.
  • The independent Brownian motion permits additional choices of equivalent martingale measure.
  • Non-unique equivalent martingale measures imply market incompleteness under the stated framework.

Tags

Full text
# Incomplete market


# Incomplete market












How to prove that market with one risky asset $S_t$ and interest rate $r = 0$ is incomplete: $$dS_t = S_t (\mu dt + \sigma_t dW_t^{1}), \quad S_0 = 1,$$ $$\sigma_t = 1 + |W_t^{2}|,$$

$W_t^{1}$ and $W_t^{2}$ are independent Wiener processes.

## Answer by Viktor Nilsson (score 0, accepted)

https://quant.stackexchange.com/a/77238

I edit my answer due to a probable misunderstanding. Since you reference an interest rate, I must assume you mean that you also have access to a risk-free asset $B_t$. Then we have $B_t = 1$ for all $t$, since the interest rate is zero.

The model can be shown to be free of arbitrage, using the first fundamental theorem of asset pricing, as follows. Let $\varphi^1_t = -\frac{\mu}{\sigma_t}$, $\varphi^2$ an arbitrary constant, and let $\mathbb{Q}$ be the measure given by taking the Girsanov kernel $\varphi_t = (\varphi^1_t, \varphi^2)$, i.e., $$ \left.\frac{d\mathbb{Q}}{d\mathbb{P}}\right\vert_{\mathcal{F}_t} = \exp\left\{\int_0^t \varphi_s dW_s - \frac{1}{2}\int_0^t||\varphi_t||^2ds\right\}. $$ Since $|\varphi^1_t| = \frac{|\mu|}{1 + |W^2_t|} \leq |\mu|$ and $\varphi^2$ is constant, we get that $\varphi_t$ satisfies the Novikov condition. Thus $\left.\frac{d\mathbb{Q}}{d\mathbb{P}}\right\vert_{\mathcal{F}_t}$ is a $\mathbb{P}$-martingale and the change of measure is valid.

By Girsanov's theorem, we have that $$ W^{1, \mathbb{Q}}_t := W^1_t - \int_0^t \varphi^1_s ds $$ is a Brownian motion under $\mathbb{Q}$. The stock dynamics under $\mathbb{Q}$ then become $$ \begin{split} dB_t &= 0 \\ dS_t &= (\mu + \sigma_t \varphi_t)dt + \sigma_t dW^{1, \mathbb{Q}}_t = \sigma_t dW^{1, \mathbb{Q}}_t, \end{split} $$ and the $B_t$-normalized versions of course the same. Thus $\mathbb{Q}$ is an equivalent martingale measure and by the first fundamental theorem of asset pricing the model is free of arbitrage. But $\varphi^2$ can be chosen arbitrarily in $\mathbb{R}$, yielding different equivalent martingale measures $\mathbb{Q}$. Hence, the martingale-measure is not unique. By the second fundamental theorem of asset pricing, the market in incomplete.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.