FLASH SALE! - "FINANCIAL MODELING COURSE BUNDLE AT 60% OFF" Enroll Now

# Markov Chain

Updated on April 4, 2024
Article byKumar Rahul
Edited byKumar Rahul
Reviewed byDheeraj Vaidya, CFA, FRM

## What Is Markov Chain?

A Markov chain is a mathematical model that describes a sequence of events where the probability of transitioning from one event to another depends only on the current state and not the previous history. It provides a framework for modeling various systems, such as weather patterns, financial markets, biological processes, etc.

For eg:
Source: Markov Chain (wallstreetmojo.com)

It predicts future states based on the current state. We can expect a system’s behavior by calculating the probabilities of different future conditions. This is particularly useful in weather forecasting, stock market prediction, and speech recognition applications.

### Key Takeaways

• Markov chains are mathematical models representing a sequence of events or states where the future depends only on the current state, following the Markov property.
• Markov chains model and analyze systems with probabilistic transitions in various fields, including finance, biology, natural language processing, and more.
• Its characteristics include a state space, transition probabilities, time homogeneity, stationary distribution, and ergodicity.

### Markov Chain Explained

Markov chains provide a versatile framework for modeling, analyzing, and predicting the behavior of systems with probabilistic transitions. Their applications extend to numerous domains, making them a valuable tool in various scientific, engineering, and computational disciplines.

These originate in the work of the Russian mathematician Andrey Markov in the late 19th and early 20th centuries. Markov was interested in probability theory and stochastic processes, and he developed the concept of a Markov chain as a mathematical model to study sequences of random events.

The critical aspect of IT lies in their assumptions, particularly the memoryless property. While this property simplifies the analysis and computation of probabilities, it may not always hold in real-world scenarios. Natural systems often depend on past events or external factors, limiting the accuracy of predictions made using them.

Furthermore, the complexity of such chains grows exponentially with the number of states, making them computationally challenging for large systems. Approximations and advanced techniques, such as Markov chain Monte Carlo (MCMC) methods, are often employed to overcome these challenges.

###### Financial Modeling & Valuation Courses Bundle (25+ Hours Video Series)

–>> If you want to learn Financial Modeling & Valuation professionally , then do check this ​Financial Modeling & Valuation Course Bundle​ (25+ hours of video tutorials with step by step McDonald’s Financial Model). Unlock the art of financial modeling and valuation with a comprehensive course covering McDonald’s forecast methodologies, advanced valuation techniques, and financial statements.

### Characteristics

Markov chains possess several key characteristics that define their behavior and distinguish them from other mathematical models. These characteristics include:

1. State Space: It operates within a defined set of states. The state space can be discrete, consisting of a finite or countably infinite number of states, or continuous, where the conditions form a constant range.
2. Markov Property: The Markov property states that the probability of transitioning to a future state depends solely on the current state and is independent of past conditions. This property implies that the system’s future behavior is determined only by its present form and not influenced by its previous history. It characterizes the memorylessness of a Markov chain.
3. Transition Probabilities: These represent probabilities of transitioning between states. These probabilities represent a transition matrix, where each element represents the probability of moving from one state to another. The transition probabilities must satisfy certain conditions, such as being non-negative and summing to one for each state.
4. Time Homogeneity: These assume to be time-homogeneous, meaning that the transition probabilities remain constant. This assumption implies that the system’s behavior remains stationary and does not change with time.
5. Stationary Distribution: It may have a fixed distribution, also known as the equilibrium or steady-state distribution. It represents the long-term probabilities of being in each state as the system evolves. The transition probabilities between states stabilize in the stationary distribution, and the system reaches a balance or equilibrium.

### Types

Markov chains can be classified into different types based on various properties and characteristics. Some of the common types include:

1. Discrete-Time Markov Chain (DTMC): The system transitions between states at discrete, evenly spaced time intervals in a discrete-time Markov chain. The transition probabilities are for each time step, and the system evolves.
2. Continuous-Time Markov Chain (CTMC): Continuous-time Markov chains allow for transitions between states to occur at any time rather than being restricted to discrete time intervals.
3. Homogeneous Markov Chain: A homogeneous chain is one in which the transition probabilities remain constant over time. The system’s behavior does not change as time progresses. These are often assumed for simplicity and ease of analysis.
4. Non-homogeneous Markov Chain: Unlike homogeneous chains, non-homogeneous Markov chains allow the transition probabilities to vary with time. Non-homogeneous Markov chains help capture time-varying behaviors.
5. Absorbing Markov Chain: An absorbing chain contains absorbing states. Once the system enters an absorbing state, it cannot transition to any other state. Absorbing states absorb barriers in the chain, influencing long-term behavior and providing information about the system’s ultimate fate.
6. Ergodic Markov Chain: An ergodic chain satisfies the condition of ergodicity. In an ergodic chain, every state is accessible from every other state. This property ensures the chain has a unique stationary distribution, allowing for long-term behavior and convergence properties to be analyzed.

### Examples

Let us understand it better with the help of examples:

#### Example #1

Let’s consider a simplified model for the stock market with three states: Bull Market (B), Bear Market (R), and Sideways Market (S). The transition probabilities between these states represent the likelihood of the market transitioning from one state to another. We assume that the market behavior follows a first-order Markov chain.

Transition Matrix:

B        R        S

B    0.6     0.3     0.1

R    0.2     0.5     0.3

S    0.3     0.2     0.5

In this example, the transition probabilities represent the following:

• If the market is currently in a Bull Market (B), there is a 60% chance of it remaining in a Bull Market, a 30% chance of transitioning to a Bear Market, and a 10% chance of transitioning to a Sideways Market.
• If the market is currently in a Bear Market (R), there is a 20% chance of transitioning to a Bull Market, a 50% chance of remaining in a Bear Market, and a 30% chance of transitioning to a Sideways Market.
• If the market is currently in a Sideways Market (S), there is a 30% chance of transitioning to a Bull Market, a 20% chance of transitioning to a Bear Market, and a 50% chance of remaining in a Sideways Market.

By simulating the transitions based on the probabilities, one can analyze the long-term chances of being in each market state. It can also estimate the likelihood of future market conditions, helping investors make informed decisions or optimize their portfolios based on different market scenarios.

#### Example #2

In 2022, a financial research firm published a study using a Markov chain model to analyze stock market regimes. Based on historical market data, the study aimed to identify different market states, such as bull markets, bear markets, and sideways markets.

By analyzing the transition probabilities between these states, the researchers provided insights into the likelihood of transitioning from one market regime to another, helping investors make informed decisions about portfolio allocation and risk management strategies. The research received attention from financial news outlets due to its implications for understanding market dynamics and navigating investment opportunities.

### Applications

Some of the critical applications of Markov chains include:

1. Finance and Economics: Markov chains model stock market behavior, asset prices, and economic decision-making. They provide insights into market regimes, risk analysis, option pricing, and portfolio optimization.
2. Biology and Genetics: Markov chains study DNA sequences, protein folding, evolutionary processes, and genetic modeling. They help understand genetic mutations, analyze genomic data, and learn biological systems.
3. Natural Language Processing: Markov chains are applicable in language modeling, text generation, and speech recognition. They aid in predicting the next word in a sentence, generating realistic text, and improving speech synthesis and recognition systems.
4. Computer Science and Artificial Intelligence: Markov chains apply to machine learning algorithms, particularly reinforcement learning and hidden Markov models. They help in decision-making in uncertain environments, sequence prediction, and anomaly detection.
5. Operations Research: Markov chains find applications in optimization problems, queueing theory, and supply chain management. They help analyze and optimize processes with random events and transitions.

### Markov Chain vs Monte Carlo vs Hidden Markov Model

Let’s compare Markov Chains, Monte Carlo simulation, and Hidden Markov Models (HMMs) from a different perspective:

#### #1 – Conceptual Basis

Markov chains are based on the concept of memorylessness, where the future state depends only on the current state and is independent of the past. Monte Carlo methods are based on random sampling and statistical simulation to approximate complex calculations or solve problems that are difficult to solve analytically. HMMs extend Markov chains by incorporating hidden states and visual observations.

#### #2 – Purpose

Markov chains model and analyze systems with probabilistic transitions to understand the behavior, predict future states, and calculate long-term probabilities.

Monte Carlo methods are used for numerical approximation and simulation. They aim to estimate values, solve complex problems, and make statistical inferences by generating random samples. Hidden Markov Models are used for sequence prediction, classification, and pattern recognition tasks where the underlying states affect the observed data. They aim to infer the hidden states based on the observable sequence.

#### #3 – Application Areas

Markov chains find applications in various fields, such as finance, biology, natural language processing, operations research, etc. For example, they are used to model stock market behavior, DNA sequences, language modeling, and optimization problems. Monte Carlo methods have broad applications in finance, physics, engineering, optimization, and computational biology. Hidden Markov Models are primarily applied in speech recognition, bioinformatics, natural language processing, and pattern recognition.

#### #4 – Mathematical Representation

Markov chains are represented by a state space, transition probabilities, and possibly a stationary distribution. They take the help of a transition matrix. Monte Carlo methods involve generating random samples and performing statistical calculations. Hidden states, visual observations, and transition/emission probabilities represent HMMs.

1. Can a Markov chain have multiple absorbing states?

Yes, a Markov chain can have multiple absorbing states. Absorbing states are states from which the system cannot transition to any other state. Once the system enters an absorbing state, it remains indefinitely.

2. What are the limitations of Markov chains?

Markov chains assume that the future depends only on the current state and is independent of the past. This assumption may not hold in all scenarios, significantly when long-term dependencies or external factors influence the system.

3. Can Markov chains be used for optimization problems?

Yes, Markov decision processes (MDPs) extend the concept of Markov chains to include decision-making and optimization. MDPs incorporate actions, rewards, and policies to model decision problems where actions influence state transitions and outcomes.

This has been a guide to what is Markov Chain. We explain its examples, applications, types, characteristics, & comparison with Monte Carlo and Hidden Markov Model. You can learn more about it from the following articles –