open-source-modelling / open_source_economic_model

First open-source asset-liability model.

🐍 Open-Source Economic Model (OSEM) 🐍

Financial institutions such as life insurers or pension funds have a fiduciary duty to manage long-term commitments to their customers. They initially collect premiums/contributions from their clients, and in exchange promise to provide a guarantee or a higher return at a further date. A large part of risk management in such companies is focused on making sure that the invested assets are managed in such a way as to guarantee the honouring of all commitments (pensions, annuities, insurance claims etc.). One such tool is an asset-liability model. Such models try to simulate three things simultaneously:

• Evolution and performance of company's investments
• Evolution of liabilities
• The impact on profitability by different trading strategies in response to outflows and inflows throughout time

Description

The aim of this project is to write a collection of models using technologies that are open-source and widely used (Pyhon and specific packages such as Pands, Numpy, Datetime etc.).

Ultimately, OSEM should be able to take as input:

• Description and size of investments
• Description of economic factors such as interest rates, credit spreads etc.
• Description and size of outflows
• Description of the factor influencing the size and timing of outflows
• Description of planned new business that a company wishes to include into the projection

OSEM should be able to produce a projection of what would happen to the balance sheet of this company in 5, 10, 20, 50 years.

Methodology

The development of the OSEM model is still very much an ongoing project. However, a draft of the methodology document can be found in the Documentation folder as OSEM pdf or a OSEM Jupyter notebook.

Specific deep dives into topics related to the methodology are available as Jupyter notebooks or pdf-s. The topics covered so far are:

• Yield-curve calibration/projection Term structure example
• Equity pricing Equity pricing example
• Fixed income pricing Bond pricing example
• Summary charts Summary charts example

Getting started

So far, we have produced a simple example that can be ran using the script main.py. A hypothetical portfolio of 3 equities and 3 corporate bonds and a single fixed liability cash flow profile. This example will grow as more development is committed to the main branch. To run the example:

• Download the OSEM repository
• Open the Anaconda command prompt
• Make the OSEM repository the root folder
• Run the script main.py
• In the folder Output, look at the cash flow time-series

A simple summary of what goes into the run and what are the outputs can be found in the Summary pdf

Using Copilots for development

If you use copilots to further develope the code, we have started assembling a modelingfile that can be used to refine the prompts. The file is located in the llm_modelfile folder.

Ask to the community

Send us an email at [email protected] with feedback, comments, ideas on what we could do better etc. Do you want to help us on this open source project? Check our wiki page also on GitHub Wiki

open_source_economic_model's Issues

Trading function

In the main script, trading is done using code that is written directly into the script. This could be incapsulated and put in a separate function/class for clarity and since this is on the roadmap of further development.

Jupyter notebook examples contain callback logs

Remove them when running the JNs

Unit test updates after new functionality

Due to the change of asset id treatment and other developments, a lot of unit tests now break. Some effort is needed to update the unit tests and add new ones.

function trade without documentation

New ALM class

A lot of functions are written in the main script and called. They should go in a separate script where they can be maintained properly.

Add pricing of corporate bonds

Function that calculates the price of a corporate bond given the term structure and spread

Generate synthetic equities for rebalancing

There might be types of equities that are not present in portfolios but would be bought in future times. To accommodate this, synthetic equities are created. Tasks include:

• Generate input file for all combinations of equities, save in Intermediate
• Read into a separate portfolio.
• Add to cash flow matrix
• Include in trading algorithm

function process_expired_liab without documentation

function unique_dates_profile in LiabilityClasses without documentation

Add Units to CorporateBond input

Asset IDs are not pulled into the main calculation

Current code dose not take into account asset IDs. This should be the case since later on they might play a bigger role

EquityClasses price_share() was fixed but does not have a unit test

Write a unit test for price_share() in EquityClasses

Equity POC Example out of date

Jupyter notebook showing how equity is treated in OSEM is out of date.

function cash_flow_profile_list_to_matrix in EquityClasses without documentation

Function set_dates_of_interest() inside MainLoop without documentation

ImportData.py has undocumented functions

function create_liabilities_df() without documentation

The Jupyter notebook describing how Corporate bonds are treated in OSEM is not current

Unit test test_output_path for PathCLasses fails

Using Gordon growth formula for terminal values leads to unintuitive results

The idea was that the final price of an equity could be calculated using the Gordon Growth model. In practice, this method is very sensitive to the ultimate forward rate assumption and the growth rate assumption. The proposed new method is to set the terminal value equal to market price at the time.

Add spread as extra argument to RetrieveRates

When pricing fixed income assets, an extra spread needs to be taken into account.

Script MainLoop is missing documentation

Functions:

• set_dates_of_interest()
• calculate_expired_dates()
• create_liabilities_df()
• process_expired_cf()
• process_expired_liab()
• trade()

function price_share in EquityClasses without documentation

Add example of spread parameter when using price_share function for Equity

Example of the new spread parameter is not shown in the Jupyter Notebooks examples

Write documentation for price_share in EquityClasses

The function price_share that calculates the present value of an equity position. The function has not been documented yet

function process_expired_cf without documentation

New dataframe Units

Until now, market value is used for each equity. This is simple to manage, however it does not scale and it limits the options for further development.

The proposal is to create a ledger of "units" of each asset. So that the market value of each position is calculated as market value * units held.

This allows to bring a separation between the movement of price of an asset and the trading activities of a each portfolio.

function init_equity_portfolio_to_dataframe in EquityClasses without documentation

Refine the modelfile