pbenson.github.io by pbenson

Contributors

Modeling portfolio P&L distribution at some horizon is a key tool in portfolio risk management. A typical horizon is one day, but it could be longer. Forecasting the P&L distribution typically is done by first creating a market model of the factors that drive changes in portfolio value. Generating a large number of market scenarios gives a distribution of P&L.

In this project, writing in C++, student researchers have built a Monte Carlo simulation of financial markets relevant to a client portfolio. They use this simulation to analyze the forecasted distribution of P&L. The simulation is based on a method used by MSCI/RiskMetrics, originally described in A general approach to calculating VaR without volatilities and correlations (1997), and again in Monte Carlo Simulation using the Benson-Zangari Approach (2013).

The project currently has these characteristics/features:

MarketFactor class representing random variables used to value securities
MarketData class containing MarketFactors. Can be populated from a file of historical prices.
MarketScenario class that represents a change in price levels of all MarketFactors
Security types: Stocks, ETFs, FX (any simple security with an available source of observed prices)
Position class that associates a quantity to a Security
Portfolio class that contains positions
Securities can be prices, and Positions and Portfolio valued given a base set of MarketFactor prices and a MarketScenario. This produces P&L for each Position and the Portfolio.
A set of classes for statistically analyzing the simulated P&L. Statistics include VaR and Expected Shortfall (aka Average VaR).

Additional resources used:

Obtained market data from Yahoo Finance.
Read CSV market data from a stream following EECS 381 guidelines
For date arithmetic, used the boost date_time library.

Want to contribute?

The code for the project is on GitHub. See how to contribute.