Financial Risk Forecasting is a complete introduction to practical quantitative risk management, with a focus on market risk. Derived from the authors teaching notes and years spent training practitioners in risk management techniques, it brings together the three key disciplines of finance, statistics and modeling (programming), to provide a thorough grounding in risk management techniques. Written by renowned risk expert Jon Danielsson, the book begins with an introduction to financial markets and market prices, volatility clusters, fat tails and nonlinear dependence. It then goes on to present volatility forecasting with both univatiate and multivatiate methods, discussing the various methods used by industry, with a special focus on the GARCH family of models. The evaluation of the quality of forecasts is discussed in detail. Next, the main concepts in risk and models to forecast risk are discussed, especially volatility, value-at-risk and expected shortfall. The focus is both on risk in basic assets such as stocks and foreign exchange, but also calculations of risk in bonds and options, with analytical methods such as delta-normal VaR and duration-normal VaR and Monte Carlo simulation. The book then moves on to the evaluation of risk models with methods like backtesting, followed by a discussion on stress testing. The book concludes by focussing on the forecasting of risk in very large and uncommon events with extreme value theory and considering the underlying assumptions behind almost every risk model in practical use – that risk is exogenous – and what happens when those assumptions are violated.
Every method presented brings together theoretical discussion and derivation of key equations and a discussion of issues in practical implementation. Each method is implemented in both MATLAB and R, two of the most commonly used mathematical programming languages for risk forecasting with which the reader can implement the models illustrated in the book.
The book includes four appendices. The first introduces basic concepts in statistics and financial time series referred to throughout the book. The second and third introduce R and MATLAB, providing a discussion of the basic implementation of the software packages. And the final looks at the concept of maximum likelihood, especially issues in implementation and testing.
The book is accompanied by a website - www.financialriskforecasting.com – which features downloadable code as used in the book.
Table of Contents:
Preface xiii
Acknowledgments xv
Abbreviations xvii
Notation xix
1 Financial markets, prices and risk 1
1.1 Prices, returns and stock indices 2
1.1.1 Stock indices 2
1.1.2 Prices and returns 2
1.2 S&P 500 returns 5
1.2.1 S&P 500 statistics 6
1.2.2 S&P 500 statistics in R and Matlab 7
1.3 The stylized facts of financial returns 9
1.4 Volatility 9
1.4.1 Volatility clusters 11
1.4.2 Volatility clusters and the ACF 12
1.5 Nonnormality and fat tails 14
1.6 Identification of fat tails 16
1.6.1 Statistical tests for fat tails 16
1.6.2 Graphical methods for fat tail analysis 17
1.6.3 Implications of fat tails in finance 20
1.7 Nonlinear dependence 21
1.7.1 Sample evidence of nonlinear dependence 22
1.7.2 Exceedance correlations 23
1.8 Copulas 25
1.8.1 The Gaussian copula 25
1.8.2 The theory of copulas 25
1.8.3 An application of copulas 27
1.8.4 Some challenges in using copulas 28
1.9 Summary 29
2 Univariate volatility modeling 31
2.1 Modeling volatility 31
2.2 Simple volatility models 32
2.2.1 Moving average models 32
2.2.2 EWMA model 33
2.3 GARCH and conditional volatility 35
2.3.1 ARCH 36
2.3.2 GARCH 38
2.3.3 The ‘‘memory’’ of a GARCH model 39
2.3.4 Normal GARCH 40
2.3.5 Student-t GARCH 40
2.3.6 (G)ARCH in mean 41
2.4 Maximum likelihood estimation of volatility models 41
2.4.1 The ARCH(1) likelihood function 42
2.4.2 The GARCH(1,1) likelihood function 42
2.4.3 On the importance of σ1 43
2.4.4 Issues in estimation 43
2.5 Diagnosing volatility models 44
2.5.1 Likelihood ratio tests and parameter significance 44
2.5.2 Analysis of model residuals 45
2.5.3 Statistical goodness-of-fit measures 45
2.6 Application of ARCH and GARCH 46
2.6.1 Estimation results 46
2.6.2 Likelihood ratio tests 47
2.6.3 Residual analysis 47
2.6.4 Graphical analysis 48
2.6.5 Implementation 48
2.7 Other GARCH-type models 51
2.7.1 Leverage effects and asymmetry 51
2.7.2 Power models 52
2.7.3 APARCH 52
2.7.4 Application of APARCH models 52
2.7.5 Estimation of APARCH 53
2.8 Alternative volatility models 54
2.8.1 Implied volatility 54
2.8.2 Realized volatility 55
2.8.3 Stochastic volatility 55
2.9 Summary 56
3 Multivariate volatility models 57
3.1 Multivariate volatility forecasting 57
3.1.1 Application 58
3.2 EWMA 59
3.3 Orthogonal GARCH 62
3.3.1 Orthogonalizing covariance 62
3.3.2 Implementation 62
3.3.3 Large-scale implementations 63
3.4 CCC and DCC models 63
3.4.1 Constant conditional correlations (CCC) 64
3.4.2 Dynamic conditional correlations (DCC) 64
3.4.3 Implementation 65
3.5 Estimation comparison 65
3.6 Multivariate extensions of GARCH 67
3.6.1 Numerical problems 69
3.6.2 The BEKK model 69
3.7 Summary 70
4 Risk measures 73
4.1 Defining and measuring risk 73
4.2 Volatility 75
4.3 Value-at-risk 76
4.3.1 Is VaR a negative or positive number? 77
4.3.2 The three steps in VaR calculations 78
4.3.3 Interpreting and analyzing VaR 78
4.3.4 VaR and normality 79
4.3.5 Sign of VaR 79
4.4 Issues in applying VaR 80
4.4.1 VaR is only a quantile 80
4.4.2 Coherence 81
4.4.3 Does VaR really violate subadditivity? 83
4.4.4 Manipulating VaR 84
4.5 Expected shortfall 85
4.6 Holding periods, scaling and the square root of time 89
4.6.1 Length of holding periods 89
4.6.2 Square-root-of-time scaling 90
4.7 Summary 90
5 Implementing risk forecasts 93
5.1 Application 93
5.2 Historical simulation 95
5.2.1 Expected shortfall estimation 97
5.2.2 Importance of window size 97
5.3 Risk measures and parametric methods 98
5.3.1 Deriving VaR 99
5.3.2 VaR when returns are normally distributed 101
5.3.3 VaR under the Student-t distribution 102
5.3.4 Expected shortfall under normality 103
5.4 What about expected returns? 104
5.5 VaR with time-dependent volatility 106
5.5.1 Moving average 106
5.5.2 EWMA 107
5.5.3 GARCH normal 108
5.5.4 Other GARCH models 109
5.6 Summary 109
6 Analytical value-at-risk for options and bonds 111
6.1 Bonds 112
6.1.1 Duration-normal VaR 112
6.1.2 Accuracy of duration-normal VaR 114
6.1.3 Convexity and VaR 114
6.2 Options 115
6.2.1 Implementation 117
6.2.2 Delta-normal VaR 119
6.2.3 Delta and gamma 120
6.3 Summary 120
7 Simulation methods for VaR for options and bonds 121
7.1 Pseudo random number generators 122
7.1.1 Linear congruental generators 122
7.1.2 Nonuniform RNGs and transformation methods 123
7.2 Simulation pricing 124
7.2.1 Bonds 125
7.2.2 Options 129
7.3 Simulation of VaR for one asset 132
7.3.1 Monte Carlo VaR with one basic asset 133
7.3.2 VaR of an option on a basic asset 134
7.3.3 Options and a stock 136
7.4 Simulation of portfolio VaR 137
7.4.1 Simulation of portfolio VaR for basic assets 137
7.4.2 Portfolio VaR for options 139
7.4.3 Richer versions 139
7.5 Issues in simulation estimation 140
7.5.1 The quality of the RNG 140
7.5.2 Number of simulations 140
7.6 Summary 142
8 Backtesting and stress testing 143
8.1 Backtesting 143
8.1.1 Market risk regulations 146
8.1.2 Estimation window length 146
8.1.3 Testing window length 147
8.1.4 Violation ratios 147
8.2 Backtesting the S&P 500 147
8.2.1 Analysis 150
8.3 Significance of backtests 153
8.3.1 Bernoulli coverage test 154
8.3.2 Testing the independence of violations 155
8.3.3 Testing VaR for the S&P 500 157
8.3.4 Joint test 159
8.3.5 Loss-function-based backtests 159
8.4 Expected shortfall backtesting 160
8.5 Problems with backtesting 162
8.6 Stress testing 163
8.6.1 Scenario analysis 163
8.6.2 Issues in scenario analysis 165
8.6.3 Scenario analysis and risk models 165
8.7 Summary 166
9 Extreme value theory 167
9.1 Extreme value theory 168
9.1.1 Types of tails 168
9.1.2 Generalized extreme value distribution 169
9.2 Asset returns and fat tails 170
9.3 Applying EVT 172
9.3.1 Generalized Pareto distribution 172
9.3.2 Hill method 173
9.3.3 Finding the threshold 174
9.3.4 Application to the S&P 500 index 175
9.4 Aggregation and convolution 176
9.5 Time dependence 179
9.5.1 Extremal index 179
9.5.2 Dependence in ARCH 180
9.5.3 When does dependence matter? 180
9.6 Summary 181
10 Endogenous risk 183
10.1 The Millennium Bridge 184
10.2 Implications for financial risk management 184
10.2.1 The 2007–2010 crisis 185
10.3 Endogenous market prices 188
10.4 Dual role of prices 190
10.4.1 Dynamic trading strategies 191
10.4.2 Delta hedging 192
10.4.3 Simulation of feedback 194
10.4.4 Endogenous risk and the 1987 crash 195
10.5 Summary 195
Appendices
A Financial time series 197
A.1 Random variables and probability density functions 197
A.1.1 Distributions and densities 197
A.1.2 Quantiles 198
A.1.3 The normal distribution 198
A.1.4 Joint distributions 200
A.1.5 Multivariate normal distribution 200
A.1.6 Conditional distribution 200
A.1.7 Independence 201
A.2 Expectations and variance 201
A.2.1 Properties of expectation and variance 202
A.2.2 Covariance and independence 203
A.3 Higher order moments 203
A.3.1 Skewness and kurtosis 204
A.4 Examples of distributions 206
A.4.1 Chi-squared (χ2) 206
A.4.2 Student-t 206
A.4.3 Bernoulli and binomial distributions 208
A.5 Basic time series concepts 208
A.5.1 Autocovariances and autocorrelations 209
A.5.2 Stationarity 209
A.5.3 White noise 210
A.6 Simple time series models 210
A.6.1 The moving average model 210
A.6.2 The autoregressive model 211
A.6.3 ARMA model 212
A.6.4 Random walk 212
A.7 Statistical hypothesis testing 212
A.7.1 Central limit theorem 213
A.7.2 p-values 213
A.7.3 Type 1 and type 2 errors and the power of the test 214
A.7.4 Testing for normality 214
A.7.5 Graphical methods: QQ plots 215
A.7.6 Testing for autocorrelation 215
A.7.7 Engle LM test for volatility clusters 216
B An introduction to R 217
B.1 Inputting data 217
B.2 Simple operations 219
B.2.1 Matrix computation 220
B.3 Distributions 222
B.3.1 Normality tests 223
B.4 Time series 224
B.5 Writing functions in R 225
B.5.1 Loops and repeats 226
B.6 Maximum likelihood estimation 228
B.7 Graphics 229
C An introduction to Matlab 231
C.1 Inputting data 231
C.2 Simple operations 233
C.2.1 Matrix algebra 234
C.3 Distributions 235
C.3.1 Normality tests 237
C.4 Time series 237
C.5 Basic programming and M-files 238
C.5.1 Loops 239
C.6 Maximum likelihood 242
C.7 Graphics 243
B Maximum likelihood 245
D.1 Likelihood functions 245
D.1.1 Normal likelihood functions 246
D.2 Optimizers 247
D.3 Issues in ML estimation 248
D.4 Information matrix 249
D.5 Properties of maximum likelihood estimators 250
D.6 Optimal testing procedures 250
D.6.1 Likelihood ratio test 251
D.6.2 Lagrange multiplier test 252
D.6.3 Wald test 253
Bibliography 255
Index 259
About the Author :
Jón Daníelsson has a PhD in the economics of financial markets and is a reader in finance at the London School of Economics. His research interests include financial stability, extreme market movements, risk, market liquidity and financial crisis. He has published extensively in both academic and practitioner journals, has consulted with a variety of private sector and public institutions, frequently gives executive education courses and has presented his work in a number of universities and institutions. In addition, he has been a frequent commentator of issues in financial markets in the media, appearing on CNN, the BBC, and many other TV and radio stations, with comments and op-ed pieces in newspapers like the Financial Times.
