Constructing a Concrete Trinomial Tree Model

A trinomial tree serves as an effective method for numerical computation of option prices within
the Black-Scholes model. The three outcomes for a trinomial tree are either a movement up, down
or static. An example of a trinomial tree is:
 [2]

This trinomial tree is slightly simplified as any of the nodes at t=2 can be achieved by taking one,
two or three different paths. This is due to the up movement being a reciprocal to the down
movement, i.e. u = 1/d. As a result, the calculations are simplified.

Trinomial trees can be built in a similar way to the binomial tree with the jump defined as ‘𝑢’, ‘𝑑’
and ‘𝓂’, with the corresponding probabilities of ‘𝛼’,’𝛽’ and ‘𝛾’. The probabilities can also be
represented as 𝑝𝑢 , 1 − 𝑝𝑢 − 𝑝𝑑 and 𝑝𝑑 .

Applying this concept to finance, let’s assume at each node of the trinomial tree, the underlying
asset price is modelled in three possible paths [1],

𝑢 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑝𝑢
𝑋𝑡+1 = {𝓂 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 1 − 𝑝𝑢 − 𝑝𝑑
𝑑 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑝𝑑
Now, let’s construct a trinomial tree that assumes the evolution of a risk stock price 𝑋 is,

𝑋𝑛+1 = 𝑋𝑛𝑍𝑛+1 , 𝑋0 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡

Where (𝑍𝑛)𝑛=1 𝑇
is a sequence of independent random variables each taking three distinct values (u,
m and d). Below is how one would construct a concrete trinomial tree model for 𝑋 with T = 2 with
one risky stock. First of all, the filtered probability space (Ω, ℱ, 𝔽, ℙ) is defined as
Ω = { 𝑢𝑢, 𝑢𝓂, 𝑢𝑑, 𝓂𝑢, 𝓂𝓂, 𝓂𝑑, 𝑑𝑢, 𝑑𝓂, 𝑑𝑑 }

𝔽 𝑋 = { ℱ𝑋t : 𝑡 ⋲ 𝕀} 𝑤ℎ𝑒𝑟𝑒 𝕀 = {0,1,2}

Where, ′𝑢’ defines the jump up in stock price, ‘𝓂′ is 1 meaning the stock price is static from one-
time period to the next and ‘𝒹′ is the reduction in the stock price. The three possible paths for this
trinomial tree are:

𝑢 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝛼
𝑋𝑡+1 = {𝓂 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝛽
𝑑 ∗ 𝑋𝑡 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝛾

The filtrations are,

ℱ𝑋 = 𝜎 ({X }) = {∅, Ω}
0 0

ℱ𝑋1= 𝜎({X 0 , 𝑋1 }) = {∅, Ω, {𝑢𝑢, 𝑢𝓂, 𝑢𝑑}, {𝓂𝑢, 𝓂𝓂, 𝓂𝑑}, {𝑑𝑢, 𝑑𝓂, 𝑑𝑑}}
ℱ𝑋 = 𝜎 ({X , 𝑋 , 𝑋 }) = 2Ω
2 0 1 2

Below is a table defining the outcomes 𝜔 at t = 0, 1 and 2:

𝜔 𝑋0(𝜔) 𝑋1(𝜔) 𝑋2(𝜔) ℙ∗

𝑢𝑢 𝑋0 𝑢 ∗ 𝑋0 𝑢 ∗ 𝑢 ∗ 𝑋0 𝛼∗𝛼
𝑢𝓂 𝑋0 𝑢 ∗ 𝑋0 𝑢 ∗ 𝓂 ∗ 𝑋0 𝛼∗𝛽
𝑢𝑑 𝑋0 𝑢 ∗ 𝑋0 𝑢 ∗ 𝑑 ∗ 𝑋0 𝛼∗𝛾
𝓂𝑢 𝑋0 𝓂 ∗ 𝑋0 𝓂 ∗ 𝑢 ∗ 𝑋0 𝛽∗𝛼
𝓂𝓂 𝑋0 𝓂 ∗ 𝑋0 𝓂 ∗ 𝓂 ∗ 𝑋0 𝛽∗𝛽
𝓂𝑑 𝑋0 𝓂 ∗ 𝑋0 𝓂 ∗ 𝑑 ∗ 𝑋0 𝛽∗𝛾
𝑑𝑢 𝑋0 𝑑 ∗ 𝑋0 𝑑 ∗ 𝑢 ∗ 𝑋0 𝛾∗𝛼
𝑑𝓂 𝑋0 𝑑 ∗ 𝑋0 𝑑 ∗ 𝓂 ∗ 𝑋0 𝛾∗𝛽
𝑑𝑑 𝑋0 𝑑 ∗ 𝑋0 𝑑 ∗ 𝑑 ∗ 𝑋0 𝛾∗𝛾

Arbitrage and Finding Equivalent Martingale Measures

Arbitrage is the ability for a market participant to make a risk-free profit. An arbitrage transaction
involves no negative cash flow at any probabilistic or temporal state and a positive cash flow in at
least one state [4]. More explicitly an arbitrage strategy 𝜑 satisfies the following conditions:

(a) 𝑉0(𝜑) = 0
(b) 𝑉𝑇(𝜑) ≥ 0 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 1
(c) ℙ(𝑉𝑇 (𝜑) > 0) > 0 [5]
Taking into consideration the above definition of arbitrage, there are conditions that need to be
imposed on the trinomial model defined in the previous section. They are that the performance of
the stock price can’t be positive for all outcomes, therefore the down movement must depreciate
the price to enforce a negative return. The conditions which can be imposed on the trinomial model
to ensure an arbitrage free strategy are:

𝑋𝑡+1 < 𝑑 ∗ 𝑋𝑡 ⇒ 𝑑 < 1

𝑋𝑡+1 > 𝑢 ∗ 𝑋𝑡 ⇒ 𝑢 >1
Which can be re-written as

1 − 𝛼 − 𝛾 + 𝛼𝑢 + 𝛾𝑑 = 1 [6]
Applying the Fundamental Theorem of Asset Pricing I (FTAP I) one could also prove the trinomial
model has no arbitrage strategies. The FTAP I states that the following are equivalent:

 The market has no arbitrage strategies

 The set P, of EMMs for X is non-empty

In order to prove an arbitrage free strategy, one needs to find at least a single element of the set 𝒫.
An EMM ℙ∗ is characterized by ℙ∗(𝑢) = 𝛼, ℙ∗(𝑚) = 𝛽, ℙ∗(𝑑) = 𝛾 as well as:

𝛼, 𝛽, 𝛾 > 1 and

𝛼∗𝛼+𝛼∗𝛽+𝛼∗𝛾+𝛽∗𝛼+𝛽∗𝛽+𝛽∗𝛾+𝛾∗𝛼+𝛾∗𝛽+𝛾∗𝛾=

1 (𝛼 + 𝛽 + 𝛾)2 = 1

𝛼+𝛽+𝛾=1

Assuming that 𝑚 = 1, 0 < 𝑑 < 𝑢.

By using the martingale property on X for time t=1:

𝛼𝑢 + 𝛽 + 𝛾𝑑 = 1

And substituting 𝛽 = 1 − 𝛼 − 𝛾 into the above equation yields:

𝛼𝑢 + 1 − 𝛼 − 𝛾 + 𝛾𝑑 = 1
𝛼(𝑢 − 1) − 𝛾(1 − 𝑑) = 0
(1 − 𝑑)
𝛼 = 𝛾
(𝑢 − 1)

𝛼 and 𝛾 are greater than 0 and this can’t hold: 1 – d < 0 because d < 1. That means that:

(1 − 𝑑) > 0 => 𝑑 < 1

(𝑢 − 1) > 0 => 𝑢 > 1 => 0 < 𝑑 < 1 < 𝑢

Let,

(1 − 𝑑)
𝛾=𝑝⟹𝛼= 𝑝
(𝑢 − 1)
Now substituting α and 𝛾 into 𝛼 + 𝛽 + 𝛾 = 1,
(1 − 𝑑)
𝑝+𝛽+𝑝=1
(𝑢 − 1)
(𝒹 − 1)
⟹ 𝛽 = 1 −𝑝 + 𝑝
(𝑢 − 1)

𝑝(1 − 𝑢) + 𝑝(𝒹 − 1)
⟹𝛽=1+
(𝑢 − 1)
𝑝(𝒹 − 𝑢)
⟹𝛽=1+
(𝑢 − 1)
 Therefore, the set of all EMMs is,
(1 − 𝑑 ) 𝑝 (𝒹 − 𝓊 ) (1 − 𝑑 ) 𝑝(𝒹 − 𝓊)
𝒫 = {ℙ∗ ( 𝑝, 1 + , 𝑝) ∶ 0 < 𝒹 < 1 < 𝓊, 0 < 𝑝 < 1, 0 < 𝑝 <1,0<1 + <1}
(𝑢 − (𝓊 − 1) (𝑢 − (𝓊 − 1)
1) 1)

For a market to be complete the set of all EMMs should have only one element. This is inferred from
the Fundamental Theorem of Asset Pricing (II) that states:

For a financial market ((Ω, ℱ, 𝔽, ℙ), 𝑋) with no arbitrage opportunities, the following are equivalent:

1) The market is complete

2) 𝒫 has exactly one element
3) 𝑋 has the predictable representation property with respect to every ℙ∗ ∈ 𝒫 : every (𝔽, ℙ∗ )-
martingale 𝑀 with 𝑀0 = 0 can be written as a martingale transform with respect to X.

[5]

This implies that if |𝒫|=1 then the market is complete. A possibility of making a market of a
trinomial model complete is to consider embedded complete market models inside the incomplete
market model. For example, with reference to [7] at t=1 one can imagine a sub-market were the
price can be 𝑢𝑋0 or 𝑚𝑋0 and this can be shown to be a complete market. Therefore, once can
form three embedded complete markets from the original incomplete market.

In our case we could assume values for p, u and d, this would be enough to have only one element in
set 𝒫.

Conclusion

We have constructed a trinomial model for a market with one risky and one risk-free asset.
For simplicity, up and down movement factors were set such that the sample paths would
recombine at each time step. We established conditions for no-arbitrage of the trinomial tree.
We later attempted to establish additional conditions to turn the market complete.
Various studies have been undertaken to find a way to price options in an incomplete market
as a complete market is usually not encountered in practise. Some ways to deal with the market
incompleteness inherent in the trinomial models is to a) try and find a portfolio which replicates
the portfolio as closely as possible (least square method (Bjorefeldt, Hee and Malmgard), b) add
another risky asset to the portfolio so that there are 2 risky assets and 1 risk-free asset Some
methods include: embedded complete market method (Takahashi), volatility estimation method
(Hu, Guo and Du), etc.

References

[1] H.V.Dedania, V.R.Shah (September 2015). A Note on Exponentially Generated Trinomial Tree

[2] Hu Xiaoping, Guo Jiafeng, Du Tao, Cui Lihua, and Cao Jie (June 2014). Pricing Option Based on
Trinomial Markov Tree

[3] James Chen (October 2018). Link: https://www.investopedia.com/terms/a/arbitrage.asp

[4] https://en.m.wikipedia.org/wiki/Arbitrage
[5] Worldquant University Discrete-time Stochastic Processes Module 4 Notes

[6] Paul Clifford, Yan Wang, Oleg Zaboronski, Kevin Zhang (November 2010). ‘Pricing Options Using
Trinomial Trees’

[7] Takahashi, H. (2000), A Note on Pricing Derivatives in an Incomplete Markets