Phase 14 - Lesson 14.3

The Fundamental Theorems of Asset Pricing

No arbitrage \(\Leftrightarrow\) an equivalent martingale measure exists; completeness \(\Leftrightarrow\) it is unique.

⏱ 55 min● Advanced🔗 Prereqs: 14.1–14.2
↖ Phase 14 hub
Builds on: 14.2 built the martingale measure by hand; here we state the general theorems that govern its existence and uniqueness.

Learning Objectives

Click a status chip to cycle: Not started → In progress → Studied → Practiced → Needs review → Mastered.

Key Vocabulary

Equivalent martingale measure (EMM)
A probability \(\mathbb{Q}\) equivalent to \(\Prob\) (same null sets) under which discounted asset prices are martingales; also called a risk-neutral measure.
Equivalent measures
Two measures with the same events of probability zero; they agree on what is possible, disagree on how likely.
First Fundamental Theorem (FTAP1)
No arbitrage \(\Leftrightarrow\) at least one EMM exists.
Second Fundamental Theorem (FTAP2)
An arbitrage-free market is complete \(\Leftrightarrow\) the EMM is unique.
Complete market
One in which every contingent claim is replicable by a self-financing strategy; then prices are unique.
Incomplete market
Some claims cannot be replicated; multiple EMMs exist and no-arbitrage pins a claim only to a price interval.

Intuition & Motivation

Intuition
Two theorems organize all of arbitrage pricing. The first says the economically meaningful condition ‘no free lunch’ is exactly the mathematically clean condition ‘there is a measure making discounted prices fair games’. The second says the market is rich enough to hedge everything precisely when that measure is unique. Together they explain when a claim has one price (complete market, unique \(\mathbb{Q}\)) and when it only has a range of arbitrage-free prices (incomplete market, many \(\mathbb{Q}\)). Equivalence of measures is the fine print: \(\mathbb{Q}\) and \(\Prob\) must agree on which outcomes are possible - you cannot price by pretending an event that can happen has probability zero.

Equivalence of measures

Two probability measures \(\Prob\) and \(\mathbb{Q}\) are equivalent (written \(\mathbb{Q}\sim\Prob\)) if they have the same null sets: \(\Prob(A)=0\iff\mathbb{Q}(A)=0\). They may assign very different probabilities to events, but they agree on what is impossible. Equivalence is essential - an EMM that killed a genuinely possible payoff state would let you construct an arbitrage.

First Fundamental Theorem

Theorem - FTAP1 (no arbitrage \(\Leftrightarrow\) EMM exists)
A (finite-horizon, frictionless) market model admits no arbitrage if and only if there exists at least one equivalent martingale measure \(\mathbb{Q}\sim\Prob\) under which every discounted traded asset price is a martingale.

One direction is easy intuition: if an EMM \(\mathbb{Q}\) exists, then any zero-cost self-financing portfolio has discounted value a \(\mathbb{Q}\)-martingale starting at 0, so its terminal expectation is 0; a nonnegative payoff with zero expectation (under a measure equivalent to \(\Prob\)) must be zero - no arbitrage. The converse (no arbitrage \(\Rightarrow\) an EMM exists) is a separating-hyperplane argument in finite models and the deep Dalang–Morton–Willinger / Delbaen–Schachermayer results in general.

Key Idea
FTAP1 converts an economic hypothesis (no free lunch) into an operational tool: to price consistently, exhibit an EMM and take discounted expectations. In our binomial model the EMM is the \(q=(R-d)/(u-d)\) of 14.1–14.2.

Second Fundamental Theorem

Theorem - FTAP2 (completeness \(\Leftrightarrow\) unique EMM)
An arbitrage-free market is complete - every contingent claim is replicable - if and only if the equivalent martingale measure is unique.

Counting intuition (finite one-period model with \(K\) states and \(A\) traded assets including the bond): replicating an arbitrary payoff means solving \(K\) equations in \(A\) unknowns. You can hit every target payoff iff the asset payoff vectors span \(\R^K\), i.e. \(A\ge K\) with full rank. The same rank condition makes the martingale equations pin down \(\mathbb{Q}\) uniquely. Fewer independent assets than states (\(A\lt K\)) leaves both the hedging system and the martingale system underdetermined - incompleteness and non-unique \(\mathbb{Q}\) arrive together.

PropertyComplete marketIncomplete market
EMM \(\mathbb{Q}\)UniqueInfinitely many
Every claim replicable?YesNo (some claims cannot be hedged)
Price of a claimSingle number \(R^{-T}\E_{\mathbb{Q}}[V_T]\)Interval \([\inf_{\mathbb{Q}},\ \sup_{\mathbb{Q}}]\)
ExampleBinomial (2 states, 2 assets)Trinomial (3 states, 2 assets); most real markets

Why incompleteness gives a price interval

If many EMMs \(\mathbb{Q}\) are consistent with the traded prices, an unhedgeable claim has a different discounted expectation under each. No-arbitrage requires only that its price lie between the smallest and largest such expectation:

\[\inf_{\mathbb{Q}\in\mathcal M}R^{-T}\E_{\mathbb{Q}}[V_T]\ \le\ \text{price}\ \le\ \sup_{\mathbb{Q}\in\mathcal M}R^{-T}\E_{\mathbb{Q}}[V_T],\] (14.6)

where \(\mathcal M\) is the set of EMMs. Within the interval, selecting one price requires an extra criterion (a utility, a calibration, a hedging objective) - the market alone does not decide.

Worked Example - A trinomial market is incomplete
1
One period, \(r=0\) so \(R=1\); stock \(S_0=100\) can go to \(120,\,100,\,80\) (three states).
2
EMM condition: \(q_u120+q_m100+q_d80=100\) and \(q_u+q_m+q_d=1\), \(q_\bullet\ge0\).
3
Solve: \(q_u=q_d=t\), \(q_m=1-2t\) for \(t\in[0,\tfrac12]\) - a whole family, so the EMM is not unique.
4
A call with \(K=100\) pays \(20,0,0\); its \(\mathbb{Q}\)-price is \(20t\in[0,10]\). No-arbitrage bounds are \([0,10]\), not a single number.

Interactive: no-arbitrage price bounds in an incomplete market

Common Mistakes to Avoid
  • Saying ‘risk-neutral measure’ as if it were always unique; uniqueness is FTAP2 and holds only in complete markets.
  • Using a \(\mathbb{Q}\) that assigns probability 0 to a possible state; then \(\mathbb{Q}\not\sim\Prob\) and an arbitrage can hide in that state.
  • Expecting a single price in an incomplete market; no-arbitrage delivers only bounds, and a model/utility choice is needed to narrow them.
  • Confusing ‘no arbitrage’ (EMM exists) with ‘complete’ (EMM unique) - they are two different theorems.
Quant Practitioner Tips
  • Rank rule of thumb (finite one-period): complete \(\iff\) independent traded payoffs span the state space \(\iff\) \(\#\text{assets}\ge\#\text{states}\) with full rank.
  • Real markets are incomplete (stochastic volatility, jumps, transaction costs); quoting a single option price implicitly selects one \(\mathbb{Q}\) via calibration.
  • Stochastic-volatility and jump models (Phase 15, Gatheral) are the canonical incomplete markets where EMM choice is a modeling decision.
  • Keep the logic crisp: FTAP1 is about existence (pricing at all), FTAP2 about uniqueness (one price vs a range).

Knowledge Check

Q1 Medium
The First Fundamental Theorem of Asset Pricing states that no arbitrage is equivalent to:
Market completeness
Existence of at least one equivalent martingale measure
Uniqueness of the martingale measure
The stock being a martingale under \(\Prob\)
Q2 Medium
An arbitrage-free market is complete if and only if:
It has two states
The equivalent martingale measure is unique
The stock has zero drift
There are more states than assets
Q3 Hard
In an incomplete market, a non-replicable claim has:
No arbitrage-free price
Exactly one arbitrage-free price
A range of arbitrage-free prices, one per EMM
A negative price

Practical Exercise

Consider a one-period market with three equally-possible states and only a bond and a stock. (a) Argue from FTAP2 whether it can be complete. (b) If the stock can go to \(110,\,100,\,90\) with \(R=1\), find the family of EMMs. (c) Give the no-arbitrage price bounds for a claim paying \((1,0,0)\).

▶ Show full solution

(a) With 3 states but only 2 traded assets, the payoff vectors cannot span \(\R^3\), so some claim is unhedgeable - by FTAP2 the market is incomplete and the EMM is not unique.

(b) EMM needs \(110q_u+100q_m+90q_d=100\) and \(q_u+q_m+q_d=1\), \(q\ge0\). Subtracting \(100\times\) the second: \(10q_u-10q_d=0\Rightarrow q_u=q_d=t\), \(q_m=1-2t\), \(t\in[0,\tfrac12]\).

(c) The claim pays 1 in the up state, so its price is \(q_u=t\in[0,\tfrac12]\) (with \(R=1\)). No-arbitrage bounds: \([0,\tfrac12]\). Any price strictly inside is consistent with some EMM; the market does not single one out.

After the reveal, answer for yourself: What extra information (e.g. a traded second option) would complete this market and collapse the interval to a point?

Lesson Summary

The two Fundamental Theorems anchor arbitrage pricing. FTAP1: no arbitrage \(\Leftrightarrow\) an equivalent martingale measure exists, so pricing = discounted \(\mathbb{Q}\)-expectation. FTAP2: the market is complete (every claim replicable) \(\Leftrightarrow\) the EMM is unique, giving a single price. When assets cannot span the states the market is incomplete, many EMMs coexist, and no-arbitrage delivers only a price interval - narrowing it requires a modeling or preference choice. \(\mathbb{Q}\) must stay equivalent to \(\Prob\).

Retrieval Practice

Close the lesson and answer from memory before checking. This is deliberate, effortful recall - the single highest-yield study action.

▶ Show retrieval prompts & answers
Q: State FTAP1 and FTAP2 precisely.
A: FTAP1: no arbitrage \(\Leftrightarrow\) an equivalent martingale measure exists. FTAP2: an arbitrage-free market is complete \(\Leftrightarrow\) the EMM is unique.
Q: Why does an incomplete market give a price interval?
A: Multiple EMMs are consistent with traded prices; an unhedgeable claim has a different discounted expectation under each, and no-arbitrage only forces the price between the inf and sup over EMMs.

Completion Checklist

Confidence / mastery rating
Personal notes