Market Risk with Python

Section 1: Python as a Risk Computational Tool (12 hours)

Python has established itself in the financial industry as a powerful programming

language with an elaborate ecosystem of tools and libraries. Advantages of Python
are:
1) open source: Python and most available libraries are completely open source;
this allows an entry to this technology at no cost, something particularly
important for students, academics, or other individuals
2) syntax: Python programming is easy to learn, the code is quite compact and in
general highly readable; at universities it is increasingly used as an
introduction to programming in general; when it comes to numerical or
financial algorithm implementation, the syntax is close to the mathematics in
general (e.g., due to code vectorization approaches)
3) multi-paradigm: Python is as good for procedural programming as well as at
object-oriented programming (which is necessary in more complex/professional
contexts); it also has some functional programming features to offer
4) interpreted: Python is an interpreted language which makes rapid prototyping
and development in general a bit more convenient, especially for beginners;
tools like IPython Notebook and libraries like pandas for time series analysis
allow for efficient and productive interactive analytics workflows
5) libraries: nowadays, there is a wealth of powerful libraries available, and the
supply grows steadily; there is hardly a problem that cannot be easily tackled
with an existing library, be it a numerical problem, a graphical one or a data-
related problem
6) speed: a common prejudice regarding interpreted languages—compared to
compiled ones like C++ or C, is the slow speed of code execution; however,
financial applications are all about matrix and array manipulations and
operations which can be done at the speed of C code with the essential Python
library NumPy for array-based computing; other performance libraries, like
Numba for dynamic code compiling, can also be used to improve performance
7) market: Large financial institutions like Bank of America, Merrill Lynch and
J.P. Morgan have millions of lines of Python code in production, mainly in risk
management; Python is also popular in the hedge fund industry
The things we will cover in Python are as follows:
• Python in a Nutshell
• Python Fundamentals
• Installing Python Packages
• First Steps with Python
• Array Operations
• Random Numbers
• Plotting
• Selected Financial Topics
• Approximation
• Optimization
• Numerical Integration
• Advanced Python Topics
• Classes and Objects
• Basic Input-Output Operations
• Interacting with Spreadsheets
• Rapid Financial Engineering

• Introduction to quantitative risk analysis and Monte Carlo

o Core ideas of risk analysis
o What is a probability distribution
o How scenarios are generated, outputs produced and analysed
o Distributions
o Most common univariate distributions in finance
o Introduction to statistical descriptors-mean, mode, standard deviation,
skewness, kurtosis
o Example financial model and exercise in Python
• Forecasting variables over time (time-series)
o Conceptual underpinning of time series, trend, volatility, seasonality,
autocorrelation, cyclicity, mean reversion
o Various techniques for forecasting future sales
o GBM, GBM with mean reversion, jump diffusion, both, seasonality
o Autoregressive models: AR, ARCH, GARCH
o Markov chains
o Discussion of attributes and application of stochastic time series
o Example model and exercise in Python
• How to Deal with Correlations
o Understanding why correlations are important
o Various techniques of including correlations:
o (Spearman) Rank order
o Lookup tables
o Envelop methods
o Logical relationships
o Copulas
o Example financial model and exercise in Python
• Model Fitting and Conclusion
o Fitting distributions, time series and copulas to historical data
o Distributions (MLE)
o Time series (MLE)
o Copulas (MLE)
o Fit comparisons with information criteria (i.e. AIC, BIC)
o Example model and exercise in Python
• Emphasis on examples model and practical case
o VAR, expected shortfall examples
o Some time series examples (including fitting to past financial datasets)
o Analyzing correlations between stochastic variables, fitting copulas and
applying then in a simulation model
o Basel II example with operational risk
 Section 2: Introduction to Market Risk : Analytical VaR (8 hours)

This section covers how to calculate the return of a portfolio of securities as well
as quantify the market risk of that portfolio, an important skill for financial
market analysts in banks, hedge funds, insurance companies, and other financial
services and investment firms. Using the Python programming language we will
use the two main tools for calculating the market risk of stock portfolios: Value-
at-Risk (VaR) and Expected Shortfall (ES).
This module goes over the versions of Python, the data source (FRED at the
Federal Reserve Bank of St. Louis), and the calculation of returns.
▪ Retrieving Data from FRED
▪ Calculating Daily Returns
▪ Calculating Longer Returns
▪ A Simple Example
▪ Retrieving data from FRED
▪ Calculating Returns on Gold
▪ Longer Horizon Returns of Gold

Risk Management under Normal Distributions

This module covers how to calculate value-at-risk (VaR) and expected shortfall
(ES) when returns are normally distributed.
▪ Distribution of Returns
▪ Value-at-Risk (VaR)
▪ Expected Shortfall (ES)
▪ Using Simulation to Estimate VaR and ES
▪ Estimating Parameters of the Normal Distribution
▪ Estimating VaR of the Normal Distribution
▪ Estimating ES of the Normal Distribution
▪ Estimating VaR and ES via Simulation

Risk Management under Non-normal Distribution

This module covers how to test for normality of returns, and how to calculate
value-at-risk (VaR) and expected shortfall (ES) when returns are not normally
distributed.
▪ Non-normal Distributions
▪ Student-t Distribution
▪ Rescaled t Distribution Model
▪ VaR and ES for Multi-day Horizon
▪ Skewness, Kurtosis, Jarque-Bera Test for Normality
▪ Estimate Parameters of the Scaled Student-t Distribution
▪ Estimate VaR and ES at 10-day Horizon
Risk Management under Volatility Clustering
This module covers how to test for the presence of volatility clustering, and how to
calculate value-at-risk (VaR) and expected shortfall (ES) when returns exhibit
volatility clustering.
▪ Future vs Historical Distribution
▪ Volatility Clustering
▪ GARCH
▪ GARCH(1,1)
▪ Diagnostic Tests
▪ Using the Python Functions
▪ Serial Correlation, Volatility Clustering
▪ VaR and ES for GARCH bootstrap

Section 3 : Credit and Counterparty Risk with Python (12 hours)

▪ Probability of Default (PD) Model: Data Preparation

o How is the PD model going to look like?
o Dependent variable: Good/ Bad (default) definition
o Fine classing, weight of evidence, and coarse classing
o Information value
o Data preparation. Splitting data
o Data preparation. An example
o Data preparation. Pre-processing discrete variables: automating
calculations
o Data preparation. Pre-processing discrete variables: visualizing results
o Data preparation. Pre-processing discrete variables: creating dummies
o Data preparation. Pre-processing the test dataset
o PD model: data preparation notebooks
▪ PD model estimation
o The PD model. Logistic regression with dummy variables
o Loading the data and selecting the features
o PD model estimation
o Build a logistic regression model with p-values
o Build a logistic regression model with p-values
o Interpreting the coefficients in the PD model
▪ PD model validation
o Out-of-sample validation (test)
o Evaluation of model performance: accuracy and area under the curve
(AUC)
o Evaluation of model performance: Gini and Kolmogorov-Smirnov
▪ Applying the PD Model for decision making
o Calculating probability of default for a single customer
o Creating a scorecard
o Calculating credit score
o From credit score to PD
o Setting cut-offs
 o PD model: logistic regression notebooks
▪ PD model monitoring
o PD model monitoring via assessing population stability
o Population stability index: pre-processing
o Population stability index: calculation and interpretation
▪ LGD and EAD Models: Preparing the data
o LGD and EAD models: independent variables.
o LGD and EAD models: dependent variables
o LGD and EAD models: distribution of recovery rates and credit
conversion
▪ LGD model
o LGD model: preparing the inputs
o LGD model: testing the model
o LGD model: estimating the accuracy of the model
o LGD model: saving the model
o LGD model: linear regression with comments
o LGD model: linear regression evaluation
▪ EAD model
o EAD model estimation and interpretation
o EAD model validation
o Calculating expected loss

Section 4: Interest Rate Pythons (8 hours)

We will start learning interest rates and interest rate instruments from this week.
Interest rates play a very important role in measuring the future and present
value of financial products. People also use market interest rates to analyse the
economic situation. At the very beginning, we will introduce fundamental interest
rate concepts, including forward rates, spot rates, swap rates and term structures
of interest rates. Then we will apply data-driven analysis to calibrate LIBOR and
swap curves and cross-correlations between these rates. Using the term structure
of these interest rates, we should be able to price market value of bonds, swaps
and other interest rate products. We also provide you with Python codes in order
to show how to obtain the LIBOR curve and how to use it to price bonds. After
learning this module, you will have a brief overview of interest rates and their
applications in bond and swap pricing.

• Zero-Coupon Bond
• Forward Contracts and Simple Forward Rate
• Spot Rate and Instantaneous Spot Rate
• Swap Rates Calculation6m
• LIBOR Curves and Cross-Correlation
• Swap Curves and Cross-Correlation
• Python Code
o Swap Rates
o Interest Rate Instruments Assignment
We will use different models to estimate interest rate processes and implement
regression analysis to calibrate the processes. The models in this week are very
important in practice. For instance, market makers need good models to help
them interpolate or extrapolate market prices of illiquid interest-rate products,
while speculators need models to help them understand the prices of fixed
income securities so that they can bet on interest rates, etc. So in this module,
we will first provide different regression techniques used to fit data in the
market. Then we will introduce Vasicek model and CIR model for bond pricing.
We also show how to use regression to fit data with our models. We provide all
codes needed and also go through the codes to help you know how to apply them.
At the end of the lecture, you will be asked to practice interest rate models by
fitting LIBOR rates in the assignment.
▪ Regression Using Least Squares
▪ Python Code
o Regression using Nelder-Mead
o Regression using Gradient Descent
o Vasicek Model and Calibration
o CIR Model and Calibration
o Python Code Files
▪ Practice exercises
o Interest Rate Instruments Assignment

Section 5: Pricing Risk of Options Simulation (8 hours)

▪ Valuation of American Options by Simulation

▪ Financial Model
▪ American Option Valuation
o Problem Formulations
o Valuation Algorithms
▪ Numerical Results
o American Put Option
o American Short Condor Spread
▪ Python Scripts
o Binomial Valuation
o Monte Carlo Valuation with LSM
o Primal and Dual LSM Algorithms
▪ Monte Carlo Simulation
▪ Valuation of Zero-Coupon Bonds
▪ Valuation of European Options
▪ Valuation of American Options
o Numerical Results
o Higher Accuracy vs. Lower Speed
▪ Python Scripts
o General Zero-Coupon Bond Valuation
▪ Simulation and Valuation in the General Model Framework
▪ Valuation of Equity Options
o European Options
o American Options
▪ Python Scripts
o Valuation of European Call Options by MCS
o Valuation of American Call Options by MCS

Section 6: Analytical pricing frameworks (8 hours)

▪ Risk-Neutral Valuation
▪ Discrete-Time Uncertainty
▪ Discrete Market Model
o Primitives
o Basic Definitions
▪ Central Results in Discrete Time
▪ Continuous-Time Case
▪ Complete Market Models
▪ Black-Scholes-Merton Model
o Market Model
o The Fundamental PDE
o European Options
▪ Greeks in the BSM Model
▪ Cox-Ross-Rubinstein Model
▪ Proofs and Python Scripts
o Ito’s Lemma
o Script for BSM Option Valuation
o Script for BSM Call Greeks
o Script for CRR Option Valuation
▪ Fourier-Based Option Pricing
▪ The Pricing Problem
▪ Fourier Transforms
▪ Fourier-Based Option Pricing
o Lewis (2001) Approach
o Carr-Madan (1999) Approach
▪ Numerical Evaluation
o Fourier Series
o Fast Fourier Transform
▪ Applications
o Black-Scholes-Merton (1973) Model
o Merton (1976) Model
Discrete Market Model
▪ Python Scripts
o BSM Call Valuation via Fourier Approach
o Fourier Series
o Roots of Unity
o Convolution
o Module with Parameters
o Call Value by Convolution
 o Option Pricing by Convolution
o Option Pricing by DFT
o Speed Test of DFT

Section 7: Complex Valuations with Python- Case Studies (20 hours)

▪ Risk Factor Models
▪ Valuation Models
▪ Excursion: SABR Model
▪ Options Portfolio
▪ Framework Classes and Functions
o Helper Functions
o Discounting Classes
o Market Environment
▪ Model Classes
o geometric_brownian_motion
o jump_diffusion
o stochastic_volatility
o stoch_vol_jump_diffusion
o sabr_stochastic_volatility
o mean_reverting_diffusion
o square_root_diffusion
o square_root_jump_diffusion
o square_root_jump_diffusion_plus
▪ Single-Risk Derivatives Valuation
o Modeling the Risk Factor
o valuation_mcs_european_single
o valuation_mcs_american_single
o Portfolio Valuation
▪ Multi-Risk Derivatives Valuation
o Market Environments
o Valuation Environment
o valuation_mcs_european_multi
o valuation_mcs_american_multi
o More than Two Risk Factors
▪ Multi-Risk Derivatives Portfolios
o Risk Factors
o Derivatives
o Portfolio
▪ Parallel Valuation of Large Portfolios
o Single Risk Factor
o American Put Option
o Large Portfolio
▪ Derivatives Portfolio Risk Statistics
o Risk Factors
o Derivatives Positions
o Portfolio Modeling and Valuation
o Portfolio Risk Reports
▪ Fourier-based Option Pricing
 o Risk Factors
o Plain Vanilla Put and Call Options
o Valuation Benchmarking
o Sources of Errors
▪ Implied Volatilities and Model Calibration
o VSTOXX Futures & Options Data
o Implied Volatilities from Market Quotes
o Market Modeling
o Calibration Functions
o Calibration Results
▪ Interest Rate Swaps
o OIS Data & Discounting
o Libor Market Data
o Model Calibration
o Floating Rate Modeling
o Interest Rate Swap
▪ Mean-Variance Portfolio Class
o Market Environment and Portfolio Object
o Basic Statistics
o Setting Weights
o Optimizing Portfolio Composition
o Efficient Frontier
o Capital Market Line
o More Assets
▪ Stochastic Short Rates
o The Modelling
o Stochstic Drifts
o Visualization of Simulated Stochastic Short Rate
▪ Quite Complex Portfolios
o Multiple Risk Factors
o Options Modeling
o Portfolio Modeling
o Portfolio Valuation
o Risk Analysis
o Visualization of Results