I was reading the old, but still interesting paper "The volatility smile and its implied tree" by Derman and Kani. I have a two questions about the derivation of the $2n+1$ equations, both of them regarding Arrow-Debreu price. On page 6, is shown in figure 4 how the model is set up. On the next page the authors writes down:

$$C(K,t_{n+1})=\exp{(-r\Delta t)}\sum_{j=1}^n\{\lambda_jp_j+\lambda_{j+1}(1-p_{j+1})\}\max{\{S_{j+1}-K,0\}}$$

where $C(K,t_{n+1}$ denotes the price of a call option with strike $K$ and expiry and expiry $t_{n+1}$. The first term in the sum are the probabilities. However, since $p_j$ is already the risk neutral probability why do the authors multiply them by $\lambda_j$? $\lambda_j$ is the known Arrow-Debreu price at node $(n,i)$. I've never heard of an Arrow-Debreu price. After checking the web it is still unclear to me, what the reason is for this equation.

Moreover using the above equation, there should be a $\lambda_{n+1}$, which is not the case! So is it just set to $0$?


1 Answer 1


We use Derman and Kani's notations.

Arrow-Debreu prices

The Arrow-Debreu price $\lambda_i$ is the price of the security $\Lambda_i$ paying \$1 in node $(n, i)$, and \$0 in all other states $(n, j)$, for $j \neq i$.

Let $\mathbb{P}_{n,j}$ be the risk-neutral probability of getting to state $(n,j)$, from state $(1,1)$.

The price of $\Lambda_i$ is the risk-neutral expectation of its discounted payoff, which is simply $$\lambda_i = e^{-rt_n}\mathbb{P}_{n,i}$$

Call price

The call price is the risk-neutral expectation of its discounted payoff: \begin{eqnarray} \tag{1} C(K, t_{n+1}) & =& e^{-rt_{n+1}}\mathbb{E}(S_{t_{n+1}} - K)^+ \\ &=& e^{-rt_{n+1}} \sum\limits_{j=1}^{n} \mathbb{P}_{n+1,j+1} (S_{j+1} - K)^+ \end{eqnarray}

Now in the binomial tree, there are two ways to get to state $(n+1, j+1)$:

  • either you were in state $(n,j)$ (probability $\mathbb{P}_{n,j} $) and you went up (probability $p_{j}$),
  • or you were in state $(n,j+1)$ (probability $\mathbb{P}_{n,j+1}$) and you went down (probability $1-p_{j+1}$)

Therefore: \begin{eqnarray} \mathbb{P_{n+1,j+1}} &=& p_j\mathbb{P_{n,j}} + (1-p_{j+1})\mathbb{P}_{n,j+1} \\&=& e^{rt_n}\left[p_j \lambda_j + (1-p_{j+1})\lambda_{j+1} \right] \tag{2} \end{eqnarray}

The only case where (2) doesn't hold is when you are in an extremal node, you can resolve this issue by setting $\lambda_{n+1}=0$.

You get Derman and Kani's formula by plugging (2) back into (1).

The intuition here is that the Arrow-Debreu prices capture all the information you need from $t_0$ to $t_n$, so that all you have to worry about is what happens between $t_n$ and $t_{n+1}$

  • $\begingroup$ Many thanks for the great answer. I will accept it. One additional question. So $C(K,t_{n+1})$ is the price at time $0$, not at time $t_n$? Then I understand the use of Arrow-Debreu prices. Otherwise not. $\endgroup$
    – math
    Dec 13, 2014 at 8:19
  • $\begingroup$ From the article: "Let $C(s_i,t_{n+1})$ and $P(s_i,t_{n+1})$, respectively, be the known market values for a call and put struck today at $si$ and expiring at $t_{n+1}$. We know the values of each of these calls and puts from interpolating the smile curve at time $t_{n+1}$." So yes it is from time $t=0$, since at the time of pricing, you only have access to the smile at $t=0$. $\endgroup$ Dec 15, 2014 at 17:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.