# Review: Financial Risk Forecasting – The Theory and Practice of Forecasting Market Risk, with Implementation in R and MATLAB by Jon Danielsson

Guest post to R-bloggers by Dr Kris Boudt.

——————–

R has always been my favorite language to forecast financial risk in my research and consulting. But, I have been reluctant to use it in my lectures on financial risk. It is certainly not the absence of appropriate R packages that refrained me. On the contrary, there is a large number of excellent R packages to forecast financial risk, for example, actuar, fPortfolio, QRMlib, VaR and PerformanceAnalytics, reviewed by Bernhard Pfaff at the 2010 R/Finance conference.

However, teaching the practice of forecasting financial risk in R, is more than showing the students how to read data in R and obtain “a number” by applying the function to their time series. It requires students to understand the statistical properties of financial time series, build models that accommodate the statistical features of the data, test the validity of their risk model and interpret the risk forecasts.

The book “Financial Risk Forecasting” by Jon Danielsson will be a very useful reference manual for my course. Let me illustrate this for the learning objective of calculating portfolio expected shortfall using dynamic conditional covariance estimates. Appendix B gives a hands-on introduction to inputting time series in R, work with vectors and matrices, and apply and write functions in R. There is even some attention given to efficient programming by avoiding loops when possible. Chapter 1 presents the statistical techniques used for analyzing prices and returns in financial markets, in particular the tools needed to illustrate the stylized facts of skewness, fat-tails, time-varying volatility and non-linear dependence between multiple return series. Once the properties of the time series have been understood, the models that accommodate the features of the data are introduced. Chapters 2 and 3 give a detailed overview on the specification and applications of univariate (normal and student t GARCH, APARCH) and multivariate GARCH models (in particular, the DCC model) and how to implement these in R. Chapter 4 then derives the formulas of Value-at-Risk and Expected Shortfall, for single assets and portfolios. Chapter 8 shows clearly how to backtest risk models using among others Bernouilli coverage tests.

There are many more interesting topics in the books. Chapters 6-7 focus on the estimation of risk of investing in bonds and options, with analytical methods such as delta-normal VaR and duration-normal VaR but also by simulation. Chapter 8 describes the implementation of stress tests. Some of the stress scenarios correspond to very large and uncommon events, requiring extreme value theory (EVT), which is discussed in Chapter 9. The book concludes with a warning that most risk models assume that financial risk is exogenous, but most financial crises have endogenous risk at their heart, where the behavior of financial agents amplifies the risk. Chapter 10 gives an intuitive explanation of endogenous risk and describes endogenous risk models. Finally, the book is supported by a clearly organized website (www.financialriskforecasting.com) that allows discussions and code downloads.

I find the book pleasant to read. It presents theoretical material in an intuitive way, while still deriving key equations and discussing the issues in practical implementation with many illustrations, both in the form of numerical examples and figures.

In summary, “Forecasting Financial Risk” strikes an excellent balance between the theory and practice of financial risk forecasting. It combines the programming, financial and statistical aspects of forecasting financial risk in an accessible way. As the book moves gradually from financial time series analysis to modeling and forecasting risk in R, I would recommend it for teaching a computational finance oriented class on risk management. Also for experienced risk professionals, the book should be useful, as it covers the latest advances in forecasting risk.

About Dr Kris Boudt:

My background: I hold a PhD in Financial Econometrics from the K. U. Leuven. I am currently assistant professor of finance at the K. U. Leuven, Lessius and VU University Amsterdam and consult on risk management to several investment companies. I have published on the estimation and management of portfolio risk in the Journal of Empirical Finance, the Journal of Financial Econometrics, Journal of Risk and RISK, among others. I have contributed code to the PerformanceAnalytics package and I am a coauthor of the PortfolioAnalytics and RTAQ package.