Monte Carlo Methods in Financial Engineering

Citation

Glasserman, P. (2003). Monte Carlo Methods in Financial Engineering. Springer.

Chapter Summary

The book has nine chapters and three appendices over roughly 600 pages, with 99 figures. The preface describes three parts: Chapters 1–3 on the fundamentals of simulation and the models to be simulated, Chapters 4–6 on ways of running a simulation better, and Chapters 7–9 on topics specific to finance.

Chapter 1: Foundations

The chapter has two halves. The first, "Principles of Monte Carlo", introduces simulation through first examples and sets out how to judge the efficiency of a simulation estimator. The second, "Principles of Derivatives Pricing", covers pricing and replication, arbitrage and risk-neutral pricing, change of numeraire and the market price of risk. Together they show how a pricing problem becomes an integration problem that Monte Carlo can attack.

Chapter 2: Generating Random Numbers and Random Variables

This chapter covers the basic machinery of every simulation: random number generation (linear congruential generators, their implementation and lattice structure, combined generators), general sampling methods (inverse transform and acceptance-rejection), and normal random variables and vectors, including generation of univariate and multivariate normals.

Chapter 3: Generating Sample Paths

The longest of the first three chapters (pp. 79–184) goes through the models used in financial engineering and how to simulate each: Brownian motion, geometric Brownian motion (with path-dependent options), Gaussian short rate models, square-root diffusions, processes with jumps, and two families of forward rate models, the continuous-rate HJM framework and the LIBOR market model with simple rates. The preface notes that this chapter and Chapter 1 contain more financial background than a bare simulation needs, because model details are where many simulation issues arise.

Chapter 4: Variance Reduction Techniques

Control variates, antithetic variates, stratified sampling, Latin hypercube sampling, matching underlying assets (moment matching and weighted Monte Carlo) and importance sampling are each treated with finance examples, followed by concluding remarks comparing them.

Chapter 5: Quasi-Monte Carlo

The chapter starts from general principles (discrepancy, van der Corput sequences, the Koksma-Hlawka bound, nets and sequences), then covers specific low-discrepancy constructions (Halton and Hammersley, Faure, Sobol'), lattice rules and randomized QMC, and ends with the finance setting, including numerical examples and strategic implementation.

Chapter 6: Discretization Methods

Simulating a continuous-time model means simulating a discrete-time approximation to it. The chapter starts with the Euler scheme and convergence order, then second-order methods (scalar and vector cases, path dependence, extrapolation), extensions to general expansions, jump-diffusions and mean square error, Brownian interpolation for extremes and barrier crossings, and changes of variables.

Chapter 7: Estimating Sensitivities

Three ways of estimating price sensitivities ("Greeks") are compared: finite-difference approximations (bias, variance and optimal mean square error), pathwise derivative estimates (including conditions for unbiasedness) and the likelihood ratio method, which also treats gamma.

Chapter 8: Pricing American Options

American option pricing requires an optimal stopping problem to be solved inside a simulation. After the problem formulation, the chapter covers parametric approximations, random tree methods (high and low estimators), state-space partitioning, stochastic mesh methods, regression-based methods and weights, and duality.

Chapter 9: Applications in Risk Management

The final chapter turns from pricing to measuring risk: loss probabilities and Value-at-Risk, variance reduction using the delta-gamma approximation, a heavy-tailed setting, and credit risk (default times and valuation, dependent defaults, portfolio credit risk). The preface warns that its models and methods differ from the pricing chapters.

Appendices

Appendix A covers convergence concepts, the central limit theorem and confidence intervals. Appendix B states results from stochastic calculus: Itô's formula, stochastic differential equations, martingales and change of measure. Appendix C gives term structure terminology and interest rate derivatives.

Key Concepts

Chapters 1–3: Foundations, random numbers and sample paths

• Efficiency of a simulation estimator: the criteria Chapter 1 introduces for comparing one simulation estimator with another.
• Risk-neutral pricing and change of numeraire: the pricing principles that turn a derivative's value into an expectation to be estimated by simulation.
• Inverse transform and acceptance-rejection: the two general methods for sampling from nonuniform distributions.
• Model simulation: Brownian motion, short-rate and square-root models, jump processes, HJM and LIBOR market models, each paired with a way of generating paths.

Chapters 4–6: Doing it better

• Variance reduction: control variates, antithetic variates, stratification, Latin hypercube sampling, moment matching and importance sampling.
• Low-discrepancy sequences: deterministic point sets such as Halton, Faure and Sobol' sequences, with randomized QMC as a way to recover error estimates.
• Discretization error: the bias from simulating a discrete-time approximation, governed by the scheme's convergence order (Euler and second-order schemes).

Chapters 7–9: Finance-specific topics

• Pathwise and likelihood ratio estimators: two derivative-estimation methods contrasted with finite differences.
• High and low estimators: bounds on the value of an American option, tied to the duality discussion at the end of Chapter 8.
• Value-at-Risk and delta-gamma approximation: the risk measure and the quadratic portfolio approximation used to speed its simulation.
• Portfolio credit risk: default times and dependent defaults.

Critical Analysis

Strengths

• Simulation and finance explain each other. The preface presents the book as both a Monte Carlo text seen through finance and a financial engineering text seen through simulation, so term-structure and pricing models are learned by implementing them.
• Full coverage of the field in one volume. The contents run from random number generators through variance reduction, quasi-Monte Carlo, discretization, Greeks and American options to risk management. Reviewers in Mathematical Reviews (Benjamin Jourdain), Zentralblatt MATH (Radu Theodorescu) and the Journal of the Operational Research Society (A. Zhigljavsky) describe it as comprehensive and a likely reference for the field.
• Honest about rigor. The preface says that technical conditions such as integrability or the existence of solutions are often passed over, and that they are spelled out only where they matter for the scope of a technique. Appendix B gives precise statements of the main stochastic calculus tools for readers who want them.
• A fixed pattern inside chapters. Most chapters pair methods with examples and end with concluding remarks, which helps with comparing methods and choosing between them.

Limitations

• Prerequisites are real. The preface names familiarity with Itô calculus, stochastic differential equations and martingales as the most important prerequisite, with change of measure and the Girsanov theorem as the most advanced idea used. Appendix B is a reminder of these tools, not a substitute for learning them.
• Informal treatment of technical conditions. Readers who need theorem-level proofs of convergence or unbiasedness will have to consult the cited literature; the book is aimed at implementation.
• Dated in places. The 2003 text predates later work on the topics it covers, so readers who need current practice should supplement it with more recent literature.
• The last chapter stands apart. The preface says the models and methods of Chapter 9 differ from the earlier pricing chapters, so it is a partial introduction to risk management rather than a full treatment.

Where It Fits in the Course

Among the course's books, Hull's Options, Futures, and Other Derivatives and its shorter companion Fundamentals of Futures and Options Markets give the market and pricing background that Glasserman assumes only lightly; they are the place to start for anyone new to derivatives. Björk's Arbitrage Theory in Continuous Time, Musiela and Rutkowski's Martingale Methods in Financial Modelling, Bingham and Kiesel's Risk-Neutral Valuation and Lamberton and Lapeyre's stochastic calculus text are mathematical treatments of the pricing theory that Chapter 1 and Appendix B only summarize, and one of them is the natural prerequisite for a reader who lacks continuous-time finance.

Glasserman supplies the computational side. Platen and Bruti-Liberati's Numerical Solution of Stochastic Differential Equations with Jumps in Finance is a more mathematical, longer monograph on discretization schemes for SDEs with jumps and goes much deeper than Chapter 6 on convergence orders and numerical stability; Glasserman is the broader and more finance-oriented book. Rebonato's LIBOR market model book and the fixed-income texts in the course are useful companions to the term structure material in Section 3.7 and Appendix C.

How to Use This Book

The book suits graduate students in financial engineering, researchers using simulation, and practitioners building models, as the preface says. A reader should arrive with stochastic calculus and some option pricing, and with enough programming experience to implement the algorithms.

For a first pass, read Chapters 1 and 2, then the Brownian motion and geometric Brownian motion sections of Chapter 3, and go straight to Chapter 4 on variance reduction, the most direct way to improve a simulation once it runs. Sections 3.3 to 3.7 on interest rate models can wait until the course reaches them. Chapter 6 on discretization is needed when simulating anything other than exactly solvable models. Chapters 7 and 8 are the core of derivative pricing practice (Greeks and early exercise); Chapter 5 on quasi-Monte Carlo is an optional extension for readers who want deterministic alternatives to random sampling. Chapter 9 suits readers heading for risk management.

Implement as you read: write a simulator for each model in Chapter 3, then add variance reduction and compare standard errors using the confidence interval material in Appendix A. Keep Appendices A and B open for notation and for the precise statements behind the informal arguments.