Skip to content

Why Do Financial Models Use Stochastic Processes?

Original question

Can you explain the intuition behind stochastic processes and why they are so important in financial modelling?
A stochastic process is a collection of random variables indexed by time or another ordered set. In finance it describes how an uncertain quantity, such as a price, interest rate, or volatility, can evolve across many possible paths. Modelling those paths lets you ask questions that a single forecast misses, such as the probability of hitting a barrier or the cost of a payoff based on average prices.

What changes when you model the path

A two-state Markov chain is a small example. Suppose the current volatility state is either calm or stressed, with transition matrix
P=(0.90.10.40.6).P=\begin{pmatrix}0.9&0.1\\0.4&0.6\end{pmatrix}.

Rows are current states, and columns are next states. Starting calm gives tomorrow's state probabilities (0.9,0.1)(0.9,0.1). The probability of being stressed in two days is the sum over the two possible intermediate states.

Pr⁡(X2=stressed∣X0=calm)=0.9(0.1)+0.1(0.6)=0.15.\Pr(X_2=\mathrm{stressed}\mid X_0=\mathrm{calm}) =0.9(0.1)+0.1(0.6)=0.15.

The probability of having seen stress at least once by day two is 1−0.92=0.191-0.9^2=0.19. A path that is stressed tomorrow and calm the following day counts toward the second probability because it passed through stress.

Here is the difference in path terms. Starting calm, the alert fires either when day one is stressed, with probability 0.10.1, or when day one stays calm and day two turns stressed, with probability 0.9(0.1)=0.090.9(0.1)=0.09. The second term excludes paths already counted on day one. To finish in stress, the path that starts by turning stressed must also stay stressed on day two, so that term becomes 0.1(0.6)0.1(0.6). The two questions use the same transition matrix and count different events.

Imagine a risk team whose monitoring process triggers an alert on the first stressed day in a two-day horizon. With the illustrative transition probabilities above, 19% is the probability of triggering the alert, while 15% is the probability of ending the horizon in stress. A terminal-state estimate would miss paths that triggered the alert and recovered.

A continuously monitored option barrier raises the same distinction with a price process. To estimate the knockout probability, the model has to describe whether paths reach the barrier during the life of the option. When using simulation, a coarse time grid can miss crossings between observation times.

References and next steps