open-risk/solstice

Solstice is an economic network simulation framework

Solstice

Solstice is an economic network simulation framework

Overview

A brief overall description of the economic network simulation engine Solstice (NB: much more in the manuals / docs).

Solstice is an economic network simulator. The primary outcomes are quantitative analyses of the behavior of economic systems under uncertainty. It can be used both as a production tool in a portfolio / risk management context or as a research tool.

The objective is to provide a performant, easily usable, extensible simulation framework to support economic network analysis.

What is included in the repository

• Set of c++ library objects / methods implement the Solstice framework
• Assorted auxiliary code / scripts
• Documentation
• Illustrative implementation of toy problems
• Sample data sets

Dependencies / Requirements

• c++17
• cmake
• conan
• poco++
• eigen
• statslib (including gcem dependency)
• catch2

Installing these dependencies is system dependent, please follow instructions as per your situation. (In the future we will have a Docker based installation that can simplify this process)

Computational Design

• Solstice is written in C++17
• The framework is "network ready". I/O can be file based or over http.
• It uses Poco++ for many of the common app functionalities
• It uses Eigen as the core container of numerical data (vectors / tensors) and linear algebra algorithms

ECS inspired Design

Solstice adopts in its implementation a number of features of recent entity-component-system C++ frameworks. This favors composition over inheritance in certain critical objects. Runtime polymorphism allows the flexible construction and extension of Solstice to enable the analysis of a variety new models and network structures

Documentation

• Introduction
• Analytic Framework

Econometric Models

An indicative list of econometric models and associated financial concepts implemented

• Multiperiod - Macro Scenario Generator (VAR type)
  • Single factor
  • Equity type multi-factor
  • Macro-economic multi-factor
• Single Period - Markov Scenario Generator (Graph type)
  • Conditional independence
  • Contagion / network models
• Collateral Value Simulation
• Regulatory Capital Calculation

Calculation Methodologies

• monte carlo - simple
• monte carlo - with importance sampling
• asymptotic limit (large N)
• analytic functions
  • moments / analytic approximations
  • regulatory capital (ASFR)

Risk Metrics / Outputs

• rating distributions at different timepoints
• quantile loss result at [99.XX] / other distribution statistics
• results statistical errors / confidence levels
• expectations at future timepoints
• risk capital allocation